From d02b74a5dcfed4bfc8f2f8e545bca4d2afabb296 Mon Sep 17 00:00:00 2001 From: Googler Date: Fri, 27 Apr 2018 10:37:02 -0700 Subject: Check in gVisor. PiperOrigin-RevId: 194583126 Change-Id: Ica1d8821a90f74e7e745962d71801c598c652463 --- pkg/sentry/kernel/BUILD | 234 +++++ pkg/sentry/kernel/README.md | 106 +++ pkg/sentry/kernel/abstract_socket_namespace.go | 108 +++ pkg/sentry/kernel/auth/BUILD | 73 ++ pkg/sentry/kernel/auth/auth.go | 22 + pkg/sentry/kernel/auth/capability_set.go | 61 ++ pkg/sentry/kernel/auth/context.go | 36 + pkg/sentry/kernel/auth/credentials.go | 227 +++++ pkg/sentry/kernel/auth/id.go | 121 +++ pkg/sentry/kernel/auth/id_map.go | 283 ++++++ pkg/sentry/kernel/auth/id_map_functions.go | 45 + pkg/sentry/kernel/auth/user_namespace.go | 130 +++ pkg/sentry/kernel/context.go | 135 +++ pkg/sentry/kernel/epoll/BUILD | 52 ++ pkg/sentry/kernel/epoll/epoll.go | 466 ++++++++++ pkg/sentry/kernel/epoll/epoll_state.go | 51 ++ pkg/sentry/kernel/epoll/epoll_test.go | 54 ++ pkg/sentry/kernel/eventfd/BUILD | 46 + pkg/sentry/kernel/eventfd/eventfd.go | 172 ++++ pkg/sentry/kernel/eventfd/eventfd_test.go | 78 ++ pkg/sentry/kernel/fd_map.go | 340 +++++++ pkg/sentry/kernel/fd_map_test.go | 134 +++ pkg/sentry/kernel/fs_context.go | 172 ++++ pkg/sentry/kernel/futex/BUILD | 48 + pkg/sentry/kernel/futex/futex.go | 405 +++++++++ pkg/sentry/kernel/futex/futex_test.go | 500 +++++++++++ pkg/sentry/kernel/g3doc/run_states.dot | 99 ++ pkg/sentry/kernel/ipc_namespace.go | 43 + pkg/sentry/kernel/kdefs/BUILD | 10 + pkg/sentry/kernel/kdefs/kdefs.go | 20 + pkg/sentry/kernel/kernel.go | 957 ++++++++++++++++++++ pkg/sentry/kernel/memevent/BUILD | 31 + pkg/sentry/kernel/memevent/memory_events.go | 98 ++ pkg/sentry/kernel/memevent/memory_events.proto | 25 + pkg/sentry/kernel/pending_signals.go | 126 +++ pkg/sentry/kernel/pipe/BUILD | 68 ++ pkg/sentry/kernel/pipe/buffers.go | 50 ++ pkg/sentry/kernel/pipe/device.go | 20 + pkg/sentry/kernel/pipe/node.go | 175 ++++ pkg/sentry/kernel/pipe/node_test.go | 308 +++++++ pkg/sentry/kernel/pipe/pipe.go | 335 +++++++ pkg/sentry/kernel/pipe/pipe_test.go | 138 +++ pkg/sentry/kernel/pipe/reader.go | 37 + pkg/sentry/kernel/pipe/reader_writer.go | 91 ++ pkg/sentry/kernel/pipe/writer.go | 37 + pkg/sentry/kernel/ptrace.go | 1054 ++++++++++++++++++++++ pkg/sentry/kernel/rseq.go | 118 +++ pkg/sentry/kernel/sched/BUILD | 20 + pkg/sentry/kernel/sched/cpuset.go | 105 +++ pkg/sentry/kernel/sched/cpuset_test.go | 44 + pkg/sentry/kernel/sched/sched.go | 16 + pkg/sentry/kernel/seccomp.go | 205 +++++ pkg/sentry/kernel/semaphore/BUILD | 62 ++ pkg/sentry/kernel/semaphore/semaphore.go | 473 ++++++++++ pkg/sentry/kernel/semaphore/semaphore_test.go | 172 ++++ pkg/sentry/kernel/sessions.go | 462 ++++++++++ pkg/sentry/kernel/signal.go | 69 ++ pkg/sentry/kernel/signal_handlers.go | 79 ++ pkg/sentry/kernel/syscalls.go | 305 +++++++ pkg/sentry/kernel/syscalls_state.go | 29 + pkg/sentry/kernel/syslog.go | 100 +++ pkg/sentry/kernel/table_test.go | 108 +++ pkg/sentry/kernel/task.go | 606 +++++++++++++ pkg/sentry/kernel/task_acct.go | 111 +++ pkg/sentry/kernel/task_block.go | 207 +++++ pkg/sentry/kernel/task_clone.go | 475 ++++++++++ pkg/sentry/kernel/task_context.go | 179 ++++ pkg/sentry/kernel/task_exec.go | 240 +++++ pkg/sentry/kernel/task_exit.go | 1139 ++++++++++++++++++++++++ pkg/sentry/kernel/task_identity.go | 557 ++++++++++++ pkg/sentry/kernel/task_log.go | 137 +++ pkg/sentry/kernel/task_net.go | 35 + pkg/sentry/kernel/task_resources.go | 126 +++ pkg/sentry/kernel/task_run.go | 346 +++++++ pkg/sentry/kernel/task_sched.go | 329 +++++++ pkg/sentry/kernel/task_signals.go | 1056 ++++++++++++++++++++++ pkg/sentry/kernel/task_start.go | 252 ++++++ pkg/sentry/kernel/task_stop.go | 226 +++++ pkg/sentry/kernel/task_syscall.go | 434 +++++++++ pkg/sentry/kernel/task_test.go | 69 ++ pkg/sentry/kernel/task_usermem.go | 298 +++++++ pkg/sentry/kernel/thread_group.go | 269 ++++++ pkg/sentry/kernel/threads.go | 443 +++++++++ pkg/sentry/kernel/time/BUILD | 32 + pkg/sentry/kernel/time/context.go | 44 + pkg/sentry/kernel/time/time.go | 649 ++++++++++++++ pkg/sentry/kernel/timekeeper.go | 270 ++++++ pkg/sentry/kernel/timekeeper_state.go | 41 + pkg/sentry/kernel/timekeeper_test.go | 156 ++++ pkg/sentry/kernel/timer.go | 282 ++++++ pkg/sentry/kernel/uts_namespace.go | 100 +++ pkg/sentry/kernel/vdso.go | 145 +++ pkg/sentry/kernel/version.go | 33 + 93 files changed, 19474 insertions(+) create mode 100644 pkg/sentry/kernel/BUILD create mode 100644 pkg/sentry/kernel/README.md create mode 100644 pkg/sentry/kernel/abstract_socket_namespace.go create mode 100644 pkg/sentry/kernel/auth/BUILD create mode 100644 pkg/sentry/kernel/auth/auth.go create mode 100644 pkg/sentry/kernel/auth/capability_set.go create mode 100644 pkg/sentry/kernel/auth/context.go create mode 100644 pkg/sentry/kernel/auth/credentials.go create mode 100644 pkg/sentry/kernel/auth/id.go create mode 100644 pkg/sentry/kernel/auth/id_map.go create mode 100644 pkg/sentry/kernel/auth/id_map_functions.go create mode 100644 pkg/sentry/kernel/auth/user_namespace.go create mode 100644 pkg/sentry/kernel/context.go create mode 100644 pkg/sentry/kernel/epoll/BUILD create mode 100644 pkg/sentry/kernel/epoll/epoll.go create mode 100644 pkg/sentry/kernel/epoll/epoll_state.go create mode 100644 pkg/sentry/kernel/epoll/epoll_test.go create mode 100644 pkg/sentry/kernel/eventfd/BUILD create mode 100644 pkg/sentry/kernel/eventfd/eventfd.go create mode 100644 pkg/sentry/kernel/eventfd/eventfd_test.go create mode 100644 pkg/sentry/kernel/fd_map.go create mode 100644 pkg/sentry/kernel/fd_map_test.go create mode 100644 pkg/sentry/kernel/fs_context.go create mode 100644 pkg/sentry/kernel/futex/BUILD create mode 100644 pkg/sentry/kernel/futex/futex.go create mode 100644 pkg/sentry/kernel/futex/futex_test.go create mode 100644 pkg/sentry/kernel/g3doc/run_states.dot create mode 100644 pkg/sentry/kernel/ipc_namespace.go create mode 100644 pkg/sentry/kernel/kdefs/BUILD create mode 100644 pkg/sentry/kernel/kdefs/kdefs.go create mode 100644 pkg/sentry/kernel/kernel.go create mode 100644 pkg/sentry/kernel/memevent/BUILD create mode 100644 pkg/sentry/kernel/memevent/memory_events.go create mode 100644 pkg/sentry/kernel/memevent/memory_events.proto create mode 100644 pkg/sentry/kernel/pending_signals.go create mode 100644 pkg/sentry/kernel/pipe/BUILD create mode 100644 pkg/sentry/kernel/pipe/buffers.go create mode 100644 pkg/sentry/kernel/pipe/device.go create mode 100644 pkg/sentry/kernel/pipe/node.go create mode 100644 pkg/sentry/kernel/pipe/node_test.go create mode 100644 pkg/sentry/kernel/pipe/pipe.go create mode 100644 pkg/sentry/kernel/pipe/pipe_test.go create mode 100644 pkg/sentry/kernel/pipe/reader.go create mode 100644 pkg/sentry/kernel/pipe/reader_writer.go create mode 100644 pkg/sentry/kernel/pipe/writer.go create mode 100644 pkg/sentry/kernel/ptrace.go create mode 100644 pkg/sentry/kernel/rseq.go create mode 100644 pkg/sentry/kernel/sched/BUILD create mode 100644 pkg/sentry/kernel/sched/cpuset.go create mode 100644 pkg/sentry/kernel/sched/cpuset_test.go create mode 100644 pkg/sentry/kernel/sched/sched.go create mode 100644 pkg/sentry/kernel/seccomp.go create mode 100644 pkg/sentry/kernel/semaphore/BUILD create mode 100644 pkg/sentry/kernel/semaphore/semaphore.go create mode 100644 pkg/sentry/kernel/semaphore/semaphore_test.go create mode 100644 pkg/sentry/kernel/sessions.go create mode 100644 pkg/sentry/kernel/signal.go create mode 100644 pkg/sentry/kernel/signal_handlers.go create mode 100644 pkg/sentry/kernel/syscalls.go create mode 100644 pkg/sentry/kernel/syscalls_state.go create mode 100644 pkg/sentry/kernel/syslog.go create mode 100644 pkg/sentry/kernel/table_test.go create mode 100644 pkg/sentry/kernel/task.go create mode 100644 pkg/sentry/kernel/task_acct.go create mode 100644 pkg/sentry/kernel/task_block.go create mode 100644 pkg/sentry/kernel/task_clone.go create mode 100644 pkg/sentry/kernel/task_context.go create mode 100644 pkg/sentry/kernel/task_exec.go create mode 100644 pkg/sentry/kernel/task_exit.go create mode 100644 pkg/sentry/kernel/task_identity.go create mode 100644 pkg/sentry/kernel/task_log.go create mode 100644 pkg/sentry/kernel/task_net.go create mode 100644 pkg/sentry/kernel/task_resources.go create mode 100644 pkg/sentry/kernel/task_run.go create mode 100644 pkg/sentry/kernel/task_sched.go create mode 100644 pkg/sentry/kernel/task_signals.go create mode 100644 pkg/sentry/kernel/task_start.go create mode 100644 pkg/sentry/kernel/task_stop.go create mode 100644 pkg/sentry/kernel/task_syscall.go create mode 100644 pkg/sentry/kernel/task_test.go create mode 100644 pkg/sentry/kernel/task_usermem.go create mode 100644 pkg/sentry/kernel/thread_group.go create mode 100644 pkg/sentry/kernel/threads.go create mode 100644 pkg/sentry/kernel/time/BUILD create mode 100644 pkg/sentry/kernel/time/context.go create mode 100644 pkg/sentry/kernel/time/time.go create mode 100644 pkg/sentry/kernel/timekeeper.go create mode 100644 pkg/sentry/kernel/timekeeper_state.go create mode 100644 pkg/sentry/kernel/timekeeper_test.go create mode 100644 pkg/sentry/kernel/timer.go create mode 100644 pkg/sentry/kernel/uts_namespace.go create mode 100644 pkg/sentry/kernel/vdso.go create mode 100644 pkg/sentry/kernel/version.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD new file mode 100644 index 000000000..62794cff5 --- /dev/null +++ b/pkg/sentry/kernel/BUILD @@ -0,0 +1,234 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_generics:defs.bzl", "go_template_instance") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_stateify( + name = "kernel_state", + srcs = [ + "abstract_socket_namespace.go", + "fd_map.go", + "fs_context.go", + "ipc_namespace.go", + "kernel.go", + "pending_signals.go", + "pending_signals_list.go", + "process_group_list.go", + "ptrace.go", + "rseq.go", + "session_list.go", + "sessions.go", + "signal.go", + "signal_handlers.go", + "syscalls.go", + "syscalls_state.go", + "syslog.go", + "task.go", + "task_clone.go", + "task_context.go", + "task_exec.go", + "task_exit.go", + "task_list.go", + "task_resources.go", + "task_run.go", + "task_sched.go", + "task_signals.go", + "task_start.go", + "task_syscall.go", + "thread_group.go", + "threads.go", + "timekeeper.go", + "timekeeper_state.go", + "timer.go", + "uts_namespace.go", + "vdso.go", + "version.go", + ], + out = "kernel_state.go", + imports = ["gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs"], + package = "kernel", +) + +go_template_instance( + name = "pending_signals_list", + out = "pending_signals_list.go", + package = "kernel", + prefix = "pendingSignal", + template = "//pkg/ilist:generic_list", + types = { + "Linker": "*pendingSignal", + }, +) + +go_template_instance( + name = "process_group_list", + out = "process_group_list.go", + package = "kernel", + prefix = "processGroup", + template = "//pkg/ilist:generic_list", + types = { + "Linker": "*ProcessGroup", + }, +) + +go_template_instance( + name = "seqatomic_taskgoroutineschedinfo", + out = "seqatomic_taskgoroutineschedinfo.go", + package = "kernel", + suffix = "TaskGoroutineSchedInfo", + template = "//pkg/sync:generic_seqatomic", + types = { + "Value": "TaskGoroutineSchedInfo", + }, +) + +go_template_instance( + name = "session_list", + out = "session_list.go", + package = "kernel", + prefix = "session", + template = "//pkg/ilist:generic_list", + types = { + "Linker": "*Session", + }, +) + +go_template_instance( + name = "task_list", + out = "task_list.go", + package = "kernel", + prefix = "task", + template = "//pkg/ilist:generic_list", + types = { + "Linker": "*Task", + }, +) + +go_library( + name = "kernel", + srcs = [ + "abstract_socket_namespace.go", + "context.go", + "fd_map.go", + "fs_context.go", + "ipc_namespace.go", + "kernel.go", + "kernel_state.go", + "pending_signals.go", + "pending_signals_list.go", + "process_group_list.go", + "ptrace.go", + "rseq.go", + "seccomp.go", + "seqatomic_taskgoroutineschedinfo.go", + "session_list.go", + "sessions.go", + "signal.go", + "signal_handlers.go", + "syscalls.go", + "syscalls_state.go", + "syslog.go", + "task.go", + "task_acct.go", + "task_block.go", + "task_clone.go", + "task_context.go", + "task_exec.go", + "task_exit.go", + "task_identity.go", + "task_list.go", + "task_log.go", + "task_net.go", + "task_resources.go", + "task_run.go", + "task_sched.go", + "task_signals.go", + "task_start.go", + "task_stop.go", + "task_syscall.go", + "task_usermem.go", + "thread_group.go", + "threads.go", + "timekeeper.go", + "timekeeper_state.go", + "timer.go", + "uts_namespace.go", + "vdso.go", + "version.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel", + visibility = ["//:sandbox"], + deps = [ + "//pkg/abi", + "//pkg/abi/linux", + "//pkg/amutex", + "//pkg/binary", + "//pkg/bits", + "//pkg/bpf", + "//pkg/cpuid", + "//pkg/eventchannel", + "//pkg/log", + "//pkg/refs", + "//pkg/secio", + "//pkg/sentry/arch", + "//pkg/sentry/context", + "//pkg/sentry/fs", + "//pkg/sentry/fs/lock", + "//pkg/sentry/fs/timerfd", + "//pkg/sentry/hostcpu", + "//pkg/sentry/inet", + "//pkg/sentry/kernel/auth", + "//pkg/sentry/kernel/epoll", + "//pkg/sentry/kernel/futex", + "//pkg/sentry/kernel/kdefs", + "//pkg/sentry/kernel/sched", + "//pkg/sentry/kernel/semaphore", + "//pkg/sentry/kernel/time", + "//pkg/sentry/limits", + "//pkg/sentry/loader", + "//pkg/sentry/memmap", + "//pkg/sentry/mm", + "//pkg/sentry/platform", + "//pkg/sentry/safemem", + "//pkg/sentry/socket/netlink/port", + "//pkg/sentry/time", + "//pkg/sentry/uniqueid", + "//pkg/sentry/usage", + "//pkg/sentry/usermem", + "//pkg/state", + "//pkg/state/statefile", + "//pkg/sync", + "//pkg/syserror", + "//pkg/tcpip", + "//pkg/tcpip/stack", + "//pkg/tcpip/transport/unix", + "//pkg/waiter", + ], +) + +go_test( + name = "kernel_test", + size = "small", + srcs = [ + "fd_map_test.go", + "table_test.go", + "task_test.go", + "timekeeper_test.go", + ], + embed = [":kernel"], + deps = [ + "//pkg/abi", + "//pkg/sentry/arch", + "//pkg/sentry/context/contexttest", + "//pkg/sentry/fs/filetest", + "//pkg/sentry/kernel/kdefs", + "//pkg/sentry/kernel/sched", + "//pkg/sentry/limits", + "//pkg/sentry/platform", + "//pkg/sentry/time", + "//pkg/sentry/usage", + "//pkg/sentry/usermem", + "//pkg/syserror", + ], +) diff --git a/pkg/sentry/kernel/README.md b/pkg/sentry/kernel/README.md new file mode 100644 index 000000000..3306780d6 --- /dev/null +++ b/pkg/sentry/kernel/README.md @@ -0,0 +1,106 @@ +This package contains: + +- A (partial) emulation of the "core Linux kernel", which governs task + execution and scheduling, system call dispatch, and signal handling. See + below for details. + +- The top-level interface for the sentry's Linux kernel emulation in general, + used by the `main` function of all versions of the sentry. This interface + revolves around the `Env` type (defined in `kernel.go`). + +# Background + +In Linux, each schedulable context is referred to interchangeably as a "task" or +"thread". Tasks can be divided into userspace and kernel tasks. In the sentry, +scheduling is managed by the Go runtime, so each schedulable context is a +goroutine; only "userspace" (application) contexts are referred to as tasks, and +represented by Task objects. (From this point forward, "task" refers to the +sentry's notion of a task unless otherwise specified.) + +At a high level, Linux application threads can be thought of as repeating a "run +loop": + +- Some amount of application code is executed in userspace. + +- A trap (explicit syscall invocation, hardware interrupt or exception, etc.) + causes control flow to switch to the kernel. + +- Some amount of kernel code is executed in kernelspace, e.g. to handle the + cause of the trap. + +- The kernel "returns from the trap" into application code. + +Analogously, each task in the sentry is associated with a *task goroutine* that +executes that task's run loop (`Task.run` in `task_run.go`). However, the +sentry's task run loop differs in structure in order to support saving execution +state to, and resuming execution from, checkpoints. + +While in kernelspace, a Linux thread can be descheduled (cease execution) in a +variety of ways: + +- It can yield or be preempted, becoming temporarily descheduled but still + runnable. At present, the sentry delegates scheduling of runnable threads to + the Go runtime. + +- It can exit, becoming permanently descheduled. The sentry's equivalent is + returning from `Task.run`, terminating the task goroutine. + +- It can enter interruptible sleep, a state in which it can be woken by a + caller-defined wakeup or the receipt of a signal. In the sentry, interruptible + sleep (which is ambiguously referred to as *blocking*) is implemented by + making all events that can end blocking (including signal notifications) + communicated via Go channels and using `select` to multiplex wakeup sources; + see `task_block.go`. + +- It can enter uninterruptible sleep, a state in which it can only be woken by a + caller-defined wakeup. Killable sleep is a closely related variant in which + the task can also be woken by SIGKILL. (These definitions also include Linux's + "group-stopped" (`TASK_STOPPED`) and "ptrace-stopped" (`TASK_TRACED`) states.) + +To maximize compatibility with Linux, sentry checkpointing appears as a spurious +signal-delivery interrupt on all tasks; interrupted system calls return `EINTR` +or are automatically restarted as usual. However, these semantics require that +uninterruptible and killable sleeps do not appear to be interrupted. In other +words, the state of the task, including its progress through the interrupted +operation, must be preserved by checkpointing. For many such sleeps, the wakeup +condition is application-controlled, making it infeasible to wait for the sleep +to end before checkpointing. Instead, we must support checkpointing progress +through sleeping operations. + +# Implementation + +We break the task's control flow graph into *states*, delimited by: + +1. Points where uninterruptible and killable sleeps may occur. For example, +there exists a state boundary between signal dequeueing and signal delivery +because there may be an intervening ptrace signal-delivery-stop. + +2. Points where sleep-induced branches may "rejoin" normal execution. For +example, the syscall exit state exists because it can be reached immediately +following a synchronous syscall, or after a task that is sleeping in `execve()` +or `vfork()` resumes execution. + +3. Points containing large branches. This is strictly for organizational +purposes. For example, the state that processes interrupt-signaled conditions is +kept separate from the main "app" state to reduce the size of the latter. + +4. `SyscallReinvoke`, which does not correspond to anything in Linux, and exists +solely to serve the autosave feature. + +![dot -Tsvg -Goverlap=false -orun_states.svg run_states.dot](g3doc/run_states.dot "Task control flow graph") + +States before which a stop may occur are represented as implementations of the +`taskRunState` interface named `run(state)`, allowing them to be saved and +restored. States that cannot be immediately preceded by a stop are simply `Task` +methods named `do(state)`. + +Conditions that can require task goroutines to cease execution for unknown +lengths of time are called *stops*. Stops are divided into *internal stops*, +which are stops whose start and end conditions are implemented within the +sentry, and *external stops*, which are stops whose start and end conditions are +not known to the sentry. Hence all uninterruptible and killable sleeps are +internal stops, and the existence of a pending checkpoint operation is an +external stop. Internal stops are reified into instances of the `TaskStop` type, +while external stops are merely counted. The task run loop alternates between +checking for stops and advancing the task's state. This allows checkpointing to +hold tasks in a stopped state while waiting for all tasks in the system to stop. diff --git a/pkg/sentry/kernel/abstract_socket_namespace.go b/pkg/sentry/kernel/abstract_socket_namespace.go new file mode 100644 index 000000000..014c4a3bf --- /dev/null +++ b/pkg/sentry/kernel/abstract_socket_namespace.go @@ -0,0 +1,108 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "sync" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/refs" + "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" +) + +type abstractEndpoint struct { + ep unix.BoundEndpoint + wr *refs.WeakRef + name string + ns *AbstractSocketNamespace +} + +// WeakRefGone implements refs.WeakRefUser.WeakRefGone. +func (e *abstractEndpoint) WeakRefGone() { + e.ns.mu.Lock() + if e.ns.endpoints[e.name].ep == e.ep { + delete(e.ns.endpoints, e.name) + } + e.ns.mu.Unlock() +} + +// AbstractSocketNamespace is used to implement the Linux abstract socket functionality. +type AbstractSocketNamespace struct { + mu sync.Mutex `state:"nosave"` + + // Keeps mapping from name to endpoint. + endpoints map[string]abstractEndpoint +} + +// NewAbstractSocketNamespace returns a new AbstractSocketNamespace. +func NewAbstractSocketNamespace() *AbstractSocketNamespace { + return &AbstractSocketNamespace{ + endpoints: make(map[string]abstractEndpoint), + } +} + +// A boundEndpoint wraps a unix.BoundEndpoint to maintain a reference on its +// backing object. +type boundEndpoint struct { + unix.BoundEndpoint + rc refs.RefCounter +} + +// Release implements unix.BoundEndpoint.Release. +func (e *boundEndpoint) Release() { + e.rc.DecRef() + e.BoundEndpoint.Release() +} + +// BoundEndpoint retrieves the endpoint bound to the given name. The return +// value is nil if no endpoint was bound. +func (a *AbstractSocketNamespace) BoundEndpoint(name string) unix.BoundEndpoint { + a.mu.Lock() + defer a.mu.Unlock() + + ep, ok := a.endpoints[name] + if !ok { + return nil + } + + rc := ep.wr.Get() + if rc == nil { + delete(a.endpoints, name) + return nil + } + + return &boundEndpoint{ep.ep, rc} +} + +// Bind binds the given socket. +// +// When the last reference managed by rc is dropped, ep may be removed from the +// namespace. +func (a *AbstractSocketNamespace) Bind(name string, ep unix.BoundEndpoint, rc refs.RefCounter) error { + a.mu.Lock() + defer a.mu.Unlock() + + if ep, ok := a.endpoints[name]; ok { + if rc := ep.wr.Get(); rc != nil { + rc.DecRef() + return syscall.EADDRINUSE + } + } + + ae := abstractEndpoint{ep: ep, name: name, ns: a} + ae.wr = refs.NewWeakRef(rc, &ae) + a.endpoints[name] = ae + return nil +} diff --git a/pkg/sentry/kernel/auth/BUILD b/pkg/sentry/kernel/auth/BUILD new file mode 100644 index 000000000..7f0680b88 --- /dev/null +++ b/pkg/sentry/kernel/auth/BUILD @@ -0,0 +1,73 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_generics:defs.bzl", "go_template_instance") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_stateify( + name = "auth_state", + srcs = [ + "credentials.go", + "id.go", + "id_map_range.go", + "id_map_set.go", + "user_namespace.go", + ], + out = "auth_state.go", + package = "auth", +) + +go_template_instance( + name = "id_map_range", + out = "id_map_range.go", + package = "auth", + prefix = "idMap", + template = "//pkg/segment:generic_range", + types = { + "T": "uint32", + }, +) + +go_template_instance( + name = "id_map_set", + out = "id_map_set.go", + consts = { + "minDegree": "3", + }, + package = "auth", + prefix = "idMap", + template = "//pkg/segment:generic_set", + types = { + "Key": "uint32", + "Range": "idMapRange", + "Value": "uint32", + "Functions": "idMapFunctions", + }, +) + +go_library( + name = "auth", + srcs = [ + "auth.go", + "auth_state.go", + "capability_set.go", + "context.go", + "credentials.go", + "id.go", + "id_map.go", + "id_map_functions.go", + "id_map_range.go", + "id_map_set.go", + "user_namespace.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/abi/linux", + "//pkg/bits", + "//pkg/log", + "//pkg/sentry/context", + "//pkg/state", + "//pkg/syserror", + ], +) diff --git a/pkg/sentry/kernel/auth/auth.go b/pkg/sentry/kernel/auth/auth.go new file mode 100644 index 000000000..c49a6b852 --- /dev/null +++ b/pkg/sentry/kernel/auth/auth.go @@ -0,0 +1,22 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package auth implements an access control model that is a subset of Linux's. +// +// The auth package supports two kinds of access controls: user/group IDs and +// capabilities. Each resource in the security model is associated with a user +// namespace; "privileged" operations check that the operator's credentials +// have the required user/group IDs or capabilities within the user namespace +// of accessed resources. +package auth diff --git a/pkg/sentry/kernel/auth/capability_set.go b/pkg/sentry/kernel/auth/capability_set.go new file mode 100644 index 000000000..5b8164c49 --- /dev/null +++ b/pkg/sentry/kernel/auth/capability_set.go @@ -0,0 +1,61 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package auth + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/bits" +) + +// A CapabilitySet is a set of capabilities implemented as a bitset. The zero +// value of CapabilitySet is a set containing no capabilities. +type CapabilitySet uint64 + +// AllCapabilities is a CapabilitySet containing all valid capabilities. +var AllCapabilities = CapabilitySetOf(linux.MaxCapability+1) - 1 + +// CapabilitySetOf returns a CapabilitySet containing only the given +// capability. +func CapabilitySetOf(cp linux.Capability) CapabilitySet { + return CapabilitySet(bits.MaskOf64(int(cp))) +} + +// CapabilitySetOfMany returns a CapabilitySet containing the given capabilities. +func CapabilitySetOfMany(cps []linux.Capability) CapabilitySet { + var cs uint64 + for _, cp := range cps { + cs |= bits.MaskOf64(int(cp)) + } + return CapabilitySet(cs) +} + +// TaskCapabilities represents all the capability sets for a task. Each of these +// sets is explained in greater detail in capabilities(7). +type TaskCapabilities struct { + // Permitted is a limiting superset for the effective capabilities that + // the thread may assume. + PermittedCaps CapabilitySet + // Inheritable is a set of capabilities preserved across an execve(2). + InheritableCaps CapabilitySet + // Effective is the set of capabilities used by the kernel to perform + // permission checks for the thread. + EffectiveCaps CapabilitySet + // Bounding is a limiting superset for the capabilities that a thread + // can add to its inheritable set using capset(2). + BoundingCaps CapabilitySet + // Ambient is a set of capabilities that are preserved across an + // execve(2) of a program that is not privileged. + AmbientCaps CapabilitySet +} diff --git a/pkg/sentry/kernel/auth/context.go b/pkg/sentry/kernel/auth/context.go new file mode 100644 index 000000000..914589b28 --- /dev/null +++ b/pkg/sentry/kernel/auth/context.go @@ -0,0 +1,36 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package auth + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/context" +) + +// contextID is the auth package's type for context.Context.Value keys. +type contextID int + +const ( + // CtxCredentials is a Context.Value key for Credentials. + CtxCredentials contextID = iota +) + +// CredentialsFromContext returns a copy of the Credentials used by ctx, or a +// set of Credentials with no capabilities if ctx does not have Credentials. +func CredentialsFromContext(ctx context.Context) *Credentials { + if v := ctx.Value(CtxCredentials); v != nil { + return v.(*Credentials) + } + return NewAnonymousCredentials() +} diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go new file mode 100644 index 000000000..b832b28fe --- /dev/null +++ b/pkg/sentry/kernel/auth/credentials.go @@ -0,0 +1,227 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package auth + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// Credentials contains information required to authorize privileged operations +// in a user namespace. +type Credentials struct { + // Real/effective/saved user/group IDs in the root user namespace. None of + // these should ever be NoID. + RealKUID KUID + EffectiveKUID KUID + SavedKUID KUID + RealKGID KGID + EffectiveKGID KGID + SavedKGID KGID + + // Filesystem user/group IDs are not implemented. "... setfsuid() is + // nowadays unneeded and should be avoided in new applications (likewise + // for setfsgid(2))." - setfsuid(2) + + // Supplementary groups used by set/getgroups. + // + // ExtraKGIDs slices are immutable, allowing multiple Credentials with the + // same ExtraKGIDs to share the same slice. + ExtraKGIDs []KGID + + // The capability sets applicable to this set of credentials. + PermittedCaps CapabilitySet + InheritableCaps CapabilitySet + EffectiveCaps CapabilitySet + BoundingCaps CapabilitySet + // Ambient capabilities are not introduced until Linux 4.3. + + // KeepCaps is the flag for PR_SET_KEEPCAPS which allow capabilities to be + // maintained after a switch from root user to non-root user via setuid(). + KeepCaps bool + + // The user namespace associated with the owner of the credentials. + UserNamespace *UserNamespace +} + +// NewAnonymousCredentials returns a set of credentials with no capabilities in +// any user namespace. +func NewAnonymousCredentials() *Credentials { + // Create a new root user namespace. Since the new namespace's owner is + // KUID 0 and the returned credentials have non-zero KUID/KGID, the + // returned credentials do not have any capabilities in the new namespace. + // Since the new namespace is not part of any existing user namespace + // hierarchy, the returned credentials do not have any capabilities in any + // other namespace. + return &Credentials{ + RealKUID: NobodyKUID, + EffectiveKUID: NobodyKUID, + SavedKUID: NobodyKUID, + RealKGID: NobodyKGID, + EffectiveKGID: NobodyKGID, + SavedKGID: NobodyKGID, + UserNamespace: NewRootUserNamespace(), + } +} + +// NewRootCredentials returns a set of credentials with KUID and KGID 0 (i.e. +// global root) in user namespace ns. +func NewRootCredentials(ns *UserNamespace) *Credentials { + // I can't find documentation for this anywhere, but it's correct for the + // inheritable capability set to be initially empty (the capabilities test + // checks for this property). + return &Credentials{ + RealKUID: RootKUID, + EffectiveKUID: RootKUID, + SavedKUID: RootKUID, + RealKGID: RootKGID, + EffectiveKGID: RootKGID, + SavedKGID: RootKGID, + PermittedCaps: AllCapabilities, + EffectiveCaps: AllCapabilities, + BoundingCaps: AllCapabilities, + UserNamespace: ns, + } +} + +// NewUserCredentials returns a set of credentials based on the given UID, GIDs, +// and capabilities in a given namespace. If all arguments are their zero +// values, this returns the same credentials as NewRootCredentials. +func NewUserCredentials(kuid KUID, kgid KGID, extraKGIDs []KGID, capabilities *TaskCapabilities, ns *UserNamespace) *Credentials { + creds := NewRootCredentials(ns) + + // Set the UID. + uid := kuid + creds.RealKUID = uid + creds.EffectiveKUID = uid + creds.SavedKUID = uid + + // Set GID. + gid := kgid + creds.RealKGID = gid + creds.EffectiveKGID = gid + creds.SavedKGID = gid + + // Set additional GIDs. + creds.ExtraKGIDs = append(creds.ExtraKGIDs, extraKGIDs...) + + // Set capabilities. If capabilities aren't specified, we default to + // all capabilities. + if capabilities != nil { + creds.PermittedCaps = capabilities.PermittedCaps + creds.EffectiveCaps = capabilities.EffectiveCaps + creds.BoundingCaps = capabilities.BoundingCaps + creds.InheritableCaps = capabilities.InheritableCaps + // // TODO: Support ambient capabilities. + } else { + // If no capabilities are specified, grant the same capabilites + // that NewRootCredentials does. + creds.PermittedCaps = AllCapabilities + creds.EffectiveCaps = AllCapabilities + creds.BoundingCaps = AllCapabilities + } + + return creds +} + +// Fork generates an identical copy of a set of credentials. +func (c *Credentials) Fork() *Credentials { + nc := new(Credentials) + *nc = *c // Copy-by-value; this is legal for all fields. + return nc +} + +// InGroup returns true if c is in group kgid. Compare Linux's +// kernel/groups.c:in_group_p(). +func (c *Credentials) InGroup(kgid KGID) bool { + if c.EffectiveKGID == kgid { + return true + } + for _, extraKGID := range c.ExtraKGIDs { + if extraKGID == kgid { + return true + } + } + return false +} + +// HasCapabilityIn returns true if c has capability cp in ns. +func (c *Credentials) HasCapabilityIn(cp linux.Capability, ns *UserNamespace) bool { + for { + // "1. A process has a capability inside a user namespace if it is a member + // of that namespace and it has the capability in its effective capability + // set." - user_namespaces(7) + if c.UserNamespace == ns { + return CapabilitySetOf(cp)&c.EffectiveCaps != 0 + } + // "3. ... A process that resides in the parent of the user namespace and + // whose effective user ID matches the owner of the namespace has all + // capabilities in the namespace." + if c.UserNamespace == ns.parent && c.EffectiveKUID == ns.owner { + return true + } + // "2. If a process has a capability in a user namespace, then it has that + // capability in all child (and further removed descendant) namespaces as + // well." + if ns.parent == nil { + return false + } + ns = ns.parent + } +} + +// HasCapability returns true if c has capability cp in its user namespace. +func (c *Credentials) HasCapability(cp linux.Capability) bool { + return c.HasCapabilityIn(cp, c.UserNamespace) +} + +// UseUID checks that c can use uid in its user namespace, then translates it +// to the root user namespace. +// +// The checks UseUID does are common, but you should verify that it's doing +// exactly what you want. +func (c *Credentials) UseUID(uid UID) (KUID, error) { + // uid must be mapped. + kuid := c.UserNamespace.MapToKUID(uid) + if !kuid.Ok() { + return NoID, syserror.EINVAL + } + // If c has CAP_SETUID, then it can use any UID in its user namespace. + if c.HasCapability(linux.CAP_SETUID) { + return kuid, nil + } + // Otherwise, c must already have the UID as its real, effective, or saved + // set-user-ID. + if kuid == c.RealKUID || kuid == c.EffectiveKUID || kuid == c.SavedKUID { + return kuid, nil + } + return NoID, syserror.EPERM +} + +// UseGID checks that c can use gid in its user namespace, then translates it +// to the root user namespace. +func (c *Credentials) UseGID(gid GID) (KGID, error) { + kgid := c.UserNamespace.MapToKGID(gid) + if !kgid.Ok() { + return NoID, syserror.EINVAL + } + if c.HasCapability(linux.CAP_SETGID) { + return kgid, nil + } + if kgid == c.RealKGID || kgid == c.EffectiveKGID || kgid == c.SavedKGID { + return kgid, nil + } + return NoID, syserror.EPERM +} diff --git a/pkg/sentry/kernel/auth/id.go b/pkg/sentry/kernel/auth/id.go new file mode 100644 index 000000000..37522b018 --- /dev/null +++ b/pkg/sentry/kernel/auth/id.go @@ -0,0 +1,121 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package auth + +import ( + "math" +) + +// UID is a user ID in an unspecified user namespace. +type UID uint32 + +// GID is a group ID in an unspecified user namespace. +type GID uint32 + +// In the root user namespace, user/group IDs have a 1-to-1 relationship with +// the users/groups they represent. In other user namespaces, this is not the +// case; for example, two different unmapped users may both "have" the overflow +// UID. This means that it is generally only valid to compare user and group +// IDs in the root user namespace. We assign distinct types, KUID/KGID, to such +// IDs to emphasize this distinction. ("k" is for "key", as in "unique key". +// Linux also uses the prefix "k", but I think they mean "kernel".) + +// KUID is a user ID in the root user namespace. +type KUID uint32 + +// KGID is a group ID in the root user namespace. +type KGID uint32 + +const ( + // NoID is uint32(-1). -1 is consistently used as a special value, in Linux + // and by extension in the auth package, to mean "no ID": + // + // - ID mapping returns -1 if the ID is not mapped. + // + // - Most set*id() syscalls accept -1 to mean "do not change this ID". + NoID = math.MaxUint32 + + // OverflowUID is the default value of /proc/sys/kernel/overflowuid. The + // "overflow UID" is usually [1] used when translating a user ID between + // namespaces fails because the ID is not mapped. (We don't implement this + // file, so the overflow UID is constant.) + // + // [1] "There is one notable case where unmapped user and group IDs are not + // converted to the corresponding overflow ID value. When viewing a uid_map + // or gid_map file in which there is no mapping for the second field, that + // field is displayed as 4294967295 (-1 as an unsigned integer);" - + // user_namespaces(7) + OverflowUID = UID(65534) + OverflowGID = GID(65534) + + // NobodyKUID is the user ID usually reserved for the least privileged user + // "nobody". + NobodyKUID = KUID(65534) + NobodyKGID = KGID(65534) + + // RootKUID is the user ID usually used for the most privileged user "root". + RootKUID = KUID(0) + RootKGID = KGID(0) + RootUID = UID(0) + RootGID = GID(0) +) + +// Ok returns true if uid is not -1. +func (uid UID) Ok() bool { + return uid != NoID +} + +// Ok returns true if gid is not -1. +func (gid GID) Ok() bool { + return gid != NoID +} + +// Ok returns true if kuid is not -1. +func (kuid KUID) Ok() bool { + return kuid != NoID +} + +// Ok returns true if kgid is not -1. +func (kgid KGID) Ok() bool { + return kgid != NoID +} + +// OrOverflow returns uid if it is valid and the overflow UID otherwise. +func (uid UID) OrOverflow() UID { + if uid.Ok() { + return uid + } + return OverflowUID +} + +// OrOverflow returns gid if it is valid and the overflow GID otherwise. +func (gid GID) OrOverflow() GID { + if gid.Ok() { + return gid + } + return OverflowGID +} + +// In translates kuid into user namespace ns. If kuid is not mapped in ns, In +// returns NoID. +func (kuid KUID) In(ns *UserNamespace) UID { + return ns.MapFromKUID(kuid) +} + +// In translates kgid into user namespace ns. If kgid is not mapped in ns, In +// returns NoID. +func (kgid KGID) In(ns *UserNamespace) GID { + return ns.MapFromKGID(kgid) +} diff --git a/pkg/sentry/kernel/auth/id_map.go b/pkg/sentry/kernel/auth/id_map.go new file mode 100644 index 000000000..6adb33530 --- /dev/null +++ b/pkg/sentry/kernel/auth/id_map.go @@ -0,0 +1,283 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package auth + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// MapFromKUID translates kuid, a UID in the root namespace, to a UID in ns. +func (ns *UserNamespace) MapFromKUID(kuid KUID) UID { + if ns.parent == nil { + return UID(kuid) + } + return UID(ns.mapID(&ns.uidMapFromParent, uint32(ns.parent.MapFromKUID(kuid)))) +} + +// MapFromKGID translates kgid, a GID in the root namespace, to a GID in ns. +func (ns *UserNamespace) MapFromKGID(kgid KGID) GID { + if ns.parent == nil { + return GID(kgid) + } + return GID(ns.mapID(&ns.gidMapFromParent, uint32(ns.parent.MapFromKGID(kgid)))) +} + +// MapToKUID translates uid, a UID in ns, to a UID in the root namespace. +func (ns *UserNamespace) MapToKUID(uid UID) KUID { + if ns.parent == nil { + return KUID(uid) + } + return ns.parent.MapToKUID(UID(ns.mapID(&ns.uidMapToParent, uint32(uid)))) +} + +// MapToKGID translates gid, a GID in ns, to a GID in the root namespace. +func (ns *UserNamespace) MapToKGID(gid GID) KGID { + if ns.parent == nil { + return KGID(gid) + } + return ns.parent.MapToKGID(GID(ns.mapID(&ns.gidMapToParent, uint32(gid)))) +} + +func (ns *UserNamespace) mapID(m *idMapSet, id uint32) uint32 { + if id == NoID { + return NoID + } + ns.mu.Lock() + defer ns.mu.Unlock() + if it := m.FindSegment(id); it.Ok() { + return it.Value() + (id - it.Start()) + } + return NoID +} + +// allIDsMapped returns true if all IDs in the range [start, end) are mapped in +// m. +// +// Preconditions: end >= start. +func (ns *UserNamespace) allIDsMapped(m *idMapSet, start, end uint32) bool { + ns.mu.Lock() + defer ns.mu.Unlock() + return m.SpanRange(idMapRange{start, end}) == end-start +} + +// An IDMapEntry represents a mapping from a range of contiguous IDs in a user +// namespace to an equally-sized range of contiguous IDs in the namespace's +// parent. +type IDMapEntry struct { + // FirstID is the first ID in the range in the namespace. + FirstID uint32 + + // FirstParentID is the first ID in the range in the parent namespace. + FirstParentID uint32 + + // Length is the number of IDs in the range. + Length uint32 +} + +// SetUIDMap instructs ns to translate UIDs as specified by entries. +// +// Note: SetUIDMap does not place an upper bound on the number of entries, but +// Linux does. This restriction is implemented in SetUIDMap's caller, the +// implementation of /proc/[pid]/uid_map. +func (ns *UserNamespace) SetUIDMap(ctx context.Context, entries []IDMapEntry) error { + c := CredentialsFromContext(ctx) + + ns.mu.Lock() + defer ns.mu.Unlock() + // "After the creation of a new user namespace, the uid_map file of *one* + // of the processes in the namespace may be written to *once* to define the + // mapping of user IDs in the new user namespace. An attempt to write more + // than once to a uid_map file in a user namespace fails with the error + // EPERM. Similar rules apply for gid_map files." - user_namespaces(7) + if !ns.uidMapFromParent.IsEmpty() { + return syserror.EPERM + } + // "At least one line must be written to the file." + if len(entries) == 0 { + return syserror.EINVAL + } + // """ + // In order for a process to write to the /proc/[pid]/uid_map + // (/proc/[pid]/gid_map) file, all of the following requirements must be + // met: + // + // 1. The writing process must have the CAP_SETUID (CAP_SETGID) capability + // in the user namespace of the process pid. + // """ + if !c.HasCapabilityIn(linux.CAP_SETUID, ns) { + return syserror.EPERM + } + // "2. The writing process must either be in the user namespace of the process + // pid or be in the parent user namespace of the process pid." + if c.UserNamespace != ns && c.UserNamespace != ns.parent { + return syserror.EPERM + } + // """ + // 3. (see trySetUIDMap) + // + // 4. One of the following two cases applies: + // + // * Either the writing process has the CAP_SETUID (CAP_SETGID) capability + // in the parent user namespace. + // """ + if !c.HasCapabilityIn(linux.CAP_SETUID, ns.parent) { + // """ + // * Or otherwise all of the following restrictions apply: + // + // + The data written to uid_map (gid_map) must consist of a single line + // that maps the writing process' effective user ID (group ID) in the + // parent user namespace to a user ID (group ID) in the user namespace. + // """ + if len(entries) != 1 || ns.parent.MapToKUID(UID(entries[0].FirstParentID)) != c.EffectiveKUID || entries[0].Length != 1 { + return syserror.EPERM + } + // """ + // + The writing process must have the same effective user ID as the + // process that created the user namespace. + // """ + if c.EffectiveKUID != ns.owner { + return syserror.EPERM + } + } + // trySetUIDMap leaves data in maps if it fails. + if err := ns.trySetUIDMap(entries); err != nil { + ns.uidMapFromParent.RemoveAll() + ns.uidMapToParent.RemoveAll() + return err + } + return nil +} + +func (ns *UserNamespace) trySetUIDMap(entries []IDMapEntry) error { + for _, e := range entries { + // Determine upper bounds and check for overflow. This implicitly + // checks for NoID. + lastID := e.FirstID + e.Length + if lastID <= e.FirstID { + return syserror.EINVAL + } + lastParentID := e.FirstParentID + e.Length + if lastParentID <= e.FirstParentID { + return syserror.EINVAL + } + // "3. The mapped user IDs (group IDs) must in turn have a mapping in + // the parent user namespace." + // Only the root namespace has a nil parent, and root is assigned + // mappings when it's created, so SetUIDMap would have returned EPERM + // without reaching this point if ns is root. + if !ns.parent.allIDsMapped(&ns.parent.uidMapToParent, e.FirstParentID, lastParentID) { + return syserror.EPERM + } + // If either of these Adds fail, we have an overlapping range. + if !ns.uidMapFromParent.Add(idMapRange{e.FirstParentID, lastParentID}, e.FirstID) { + return syserror.EINVAL + } + if !ns.uidMapToParent.Add(idMapRange{e.FirstID, lastID}, e.FirstParentID) { + return syserror.EINVAL + } + } + return nil +} + +// SetGIDMap instructs ns to translate GIDs as specified by entries. +func (ns *UserNamespace) SetGIDMap(ctx context.Context, entries []IDMapEntry) error { + c := CredentialsFromContext(ctx) + + ns.mu.Lock() + defer ns.mu.Unlock() + if !ns.gidMapFromParent.IsEmpty() { + return syserror.EPERM + } + if len(entries) == 0 { + return syserror.EINVAL + } + if !c.HasCapabilityIn(linux.CAP_SETGID, ns) { + return syserror.EPERM + } + if c.UserNamespace != ns && c.UserNamespace != ns.parent { + return syserror.EPERM + } + if !c.HasCapabilityIn(linux.CAP_SETGID, ns.parent) { + if len(entries) != 1 || ns.parent.MapToKGID(GID(entries[0].FirstParentID)) != c.EffectiveKGID || entries[0].Length != 1 { + return syserror.EPERM + } + // It's correct for this to still be UID. + if c.EffectiveKUID != ns.owner { + return syserror.EPERM + } + // "In the case of gid_map, use of the setgroups(2) system call must + // first be denied by writing "deny" to the /proc/[pid]/setgroups file + // (see below) before writing to gid_map." (This file isn't implemented + // in the version of Linux we're emulating; see comment in + // UserNamespace.) + } + if err := ns.trySetGIDMap(entries); err != nil { + ns.gidMapFromParent.RemoveAll() + ns.gidMapToParent.RemoveAll() + return err + } + return nil +} + +func (ns *UserNamespace) trySetGIDMap(entries []IDMapEntry) error { + for _, e := range entries { + lastID := e.FirstID + e.Length + if lastID <= e.FirstID { + return syserror.EINVAL + } + lastParentID := e.FirstParentID + e.Length + if lastParentID <= e.FirstParentID { + return syserror.EINVAL + } + if !ns.parent.allIDsMapped(&ns.parent.gidMapToParent, e.FirstParentID, lastParentID) { + return syserror.EPERM + } + if !ns.gidMapFromParent.Add(idMapRange{e.FirstParentID, lastParentID}, e.FirstID) { + return syserror.EINVAL + } + if !ns.gidMapToParent.Add(idMapRange{e.FirstID, lastID}, e.FirstParentID) { + return syserror.EINVAL + } + } + return nil +} + +// UIDMap returns the user ID mappings configured for ns. If no mappings +// have been configured, UIDMap returns nil. +func (ns *UserNamespace) UIDMap() []IDMapEntry { + return ns.getIDMap(&ns.uidMapToParent) +} + +// GIDMap returns the group ID mappings configured for ns. If no mappings +// have been configured, GIDMap returns nil. +func (ns *UserNamespace) GIDMap() []IDMapEntry { + return ns.getIDMap(&ns.gidMapToParent) +} + +func (ns *UserNamespace) getIDMap(m *idMapSet) []IDMapEntry { + ns.mu.Lock() + defer ns.mu.Unlock() + var entries []IDMapEntry + for it := m.FirstSegment(); it.Ok(); it = it.NextSegment() { + entries = append(entries, IDMapEntry{ + FirstID: it.Start(), + FirstParentID: it.Value(), + Length: it.Range().Length(), + }) + } + return entries +} diff --git a/pkg/sentry/kernel/auth/id_map_functions.go b/pkg/sentry/kernel/auth/id_map_functions.go new file mode 100644 index 000000000..889291d96 --- /dev/null +++ b/pkg/sentry/kernel/auth/id_map_functions.go @@ -0,0 +1,45 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package auth + +// idMapFunctions "implements" generic interface segment.Functions for +// idMapSet. An idMapSet maps non-overlapping ranges of contiguous IDs in one +// user namespace to non-overlapping ranges of contiguous IDs in another user +// namespace. Each such ID mapping is implemented as a range-to-value mapping +// in the set such that [range.Start(), range.End()) => [value, value + +// range.Length()). +type idMapFunctions struct{} + +func (idMapFunctions) MinKey() uint32 { + return 0 +} + +func (idMapFunctions) MaxKey() uint32 { + return NoID +} + +func (idMapFunctions) ClearValue(*uint32) {} + +func (idMapFunctions) Merge(r1 idMapRange, val1 uint32, r2 idMapRange, val2 uint32) (uint32, bool) { + // Mapped ranges have to be contiguous. + if val1+r1.Length() != val2 { + return 0, false + } + return val1, true +} + +func (idMapFunctions) Split(r idMapRange, val uint32, split uint32) (uint32, uint32) { + return val, val + (split - r.Start) +} diff --git a/pkg/sentry/kernel/auth/user_namespace.go b/pkg/sentry/kernel/auth/user_namespace.go new file mode 100644 index 000000000..0980aeadf --- /dev/null +++ b/pkg/sentry/kernel/auth/user_namespace.go @@ -0,0 +1,130 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package auth + +import ( + "math" + "sync" + + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// A UserNamespace represents a user namespace. See user_namespaces(7) for +// details. +type UserNamespace struct { + // parent is this namespace's parent. If this is the root namespace, parent + // is nil. The parent pointer is immutable. + parent *UserNamespace + + // owner is the effective UID of the namespace's creator in the root + // namespace. owner is immutable. + owner KUID + + // mu protects the following fields. + // + // If mu will be locked in multiple UserNamespaces, it must be locked in + // descendant namespaces before ancestors. + mu sync.Mutex `state:"nosave"` + + // Mappings of user/group IDs between this namespace and its parent. + // + // All ID maps, once set, cannot be changed. This means that successful + // UID/GID translations cannot be racy. + uidMapFromParent idMapSet + uidMapToParent idMapSet + gidMapFromParent idMapSet + gidMapToParent idMapSet + + // TODO: Consider supporting disabling setgroups(2), which "was + // added in Linux 3.19, but was backported to many earlier stable kernel + // series, because it addresses a security issue" - user_namespaces(7). (It + // was not backported to 3.11.10, which we are currently imitating.) +} + +// NewRootUserNamespace returns a UserNamespace that is appropriate for a +// system's root user namespace. +func NewRootUserNamespace() *UserNamespace { + var ns UserNamespace + // """ + // The initial user namespace has no parent namespace, but, for + // consistency, the kernel provides dummy user and group ID mapping files + // for this namespace. Looking at the uid_map file (gid_map is the same) + // from a shell in the initial namespace shows: + // + // $ cat /proc/$$/uid_map + // 0 0 4294967295 + // """ - user_namespaces(7) + for _, m := range []*idMapSet{ + &ns.uidMapFromParent, + &ns.uidMapToParent, + &ns.gidMapFromParent, + &ns.gidMapToParent, + } { + if !m.Add(idMapRange{0, math.MaxUint32}, 0) { + panic("Failed to insert into empty ID map") + } + } + return &ns +} + +// Root returns the root of the user namespace tree containing ns. +func (ns *UserNamespace) Root() *UserNamespace { + for ns.parent != nil { + ns = ns.parent + } + return ns +} + +// "The kernel imposes (since version 3.11) a limit of 32 nested levels of user +// namespaces." - user_namespaces(7) +const maxUserNamespaceDepth = 32 + +func (ns *UserNamespace) depth() int { + var i int + for ns != nil { + i++ + ns = ns.parent + } + return i +} + +// NewChildUserNamespace returns a new user namespace created by a caller with +// credentials c. +func (c *Credentials) NewChildUserNamespace() (*UserNamespace, error) { + if c.UserNamespace.depth() >= maxUserNamespaceDepth { + // "... Calls to unshare(2) or clone(2) that would cause this limit to + // be exceeded fail with the error EUSERS." - user_namespaces(7) + return nil, syserror.EUSERS + } + // "EPERM: CLONE_NEWUSER was specified in flags, but either the effective + // user ID or the effective group ID of the caller does not have a mapping + // in the parent namespace (see user_namespaces(7))." - clone(2) + // "CLONE_NEWUSER requires that the user ID and group ID of the calling + // process are mapped to user IDs and group IDs in the user namespace of + // the calling process at the time of the call." - unshare(2) + if !c.EffectiveKUID.In(c.UserNamespace).Ok() { + return nil, syserror.EPERM + } + if !c.EffectiveKGID.In(c.UserNamespace).Ok() { + return nil, syserror.EPERM + } + return &UserNamespace{ + parent: c.UserNamespace, + owner: c.EffectiveKUID, + // "When a user namespace is created, it starts without a mapping of + // user IDs (group IDs) to the parent user namespace." - + // user_namespaces(7) + }, nil +} diff --git a/pkg/sentry/kernel/context.go b/pkg/sentry/kernel/context.go new file mode 100644 index 000000000..261ca6f7a --- /dev/null +++ b/pkg/sentry/kernel/context.go @@ -0,0 +1,135 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" +) + +// contextID is the kernel package's type for context.Context.Value keys. +type contextID int + +const ( + // CtxCanTrace is a Context.Value key for a function with the same + // signature and semantics as kernel.Task.CanTrace. + CtxCanTrace contextID = iota + + // CtxKernel is a Context.Value key for a Kernel. + CtxKernel + + // CtxPIDNamespace is a Context.Value key for a PIDNamespace. + CtxPIDNamespace + + // CtxTask is a Context.Value key for a Task. + CtxTask + + // CtxUTSNamespace is a Context.Value key for a UTSNamespace. + CtxUTSNamespace + + // CtxIPCNamespace is a Context.Value key for a IPCNamespace. + CtxIPCNamespace +) + +// ContextCanTrace returns true if ctx is permitted to trace t, in the same sense +// as kernel.Task.CanTrace. +func ContextCanTrace(ctx context.Context, t *Task, attach bool) bool { + if v := ctx.Value(CtxCanTrace); v != nil { + return v.(func(*Task, bool) bool)(t, attach) + } + return false +} + +// KernelFromContext returns the Kernel in which ctx is executing, or nil if +// there is no such Kernel. +func KernelFromContext(ctx context.Context) *Kernel { + if v := ctx.Value(CtxKernel); v != nil { + return v.(*Kernel) + } + return nil +} + +// PIDNamespaceFromContext returns the PID namespace in which ctx is executing, +// or nil if there is no such PID namespace. +func PIDNamespaceFromContext(ctx context.Context) *PIDNamespace { + if v := ctx.Value(CtxPIDNamespace); v != nil { + return v.(*PIDNamespace) + } + return nil +} + +// UTSNamespaceFromContext returns the UTS namespace in which ctx is executing, +// or nil if there is no such UTS namespace. +func UTSNamespaceFromContext(ctx context.Context) *UTSNamespace { + if v := ctx.Value(CtxUTSNamespace); v != nil { + return v.(*UTSNamespace) + } + return nil +} + +// IPCNamespaceFromContext returns the IPC namespace in which ctx is executing, +// or nil if there is no such IPC namespace. +func IPCNamespaceFromContext(ctx context.Context) *IPCNamespace { + if v := ctx.Value(CtxIPCNamespace); v != nil { + return v.(*IPCNamespace) + } + return nil +} + +// TaskFromContext returns the Task associated with ctx, or nil if there is no +// such Task. +func TaskFromContext(ctx context.Context) *Task { + if v := ctx.Value(CtxTask); v != nil { + return v.(*Task) + } + return nil +} + +// AsyncContext returns a context.Context that may be used by goroutines that +// do work on behalf of t and therefore share its contextual values, but are +// not t's task goroutine (e.g. asynchronous I/O). +func (t *Task) AsyncContext() context.Context { + return taskAsyncContext{t: t} +} + +type taskAsyncContext struct { + context.NoopSleeper + t *Task +} + +// Debugf implements log.Logger.Debugf. +func (ctx taskAsyncContext) Debugf(format string, v ...interface{}) { + ctx.t.Debugf(format, v...) +} + +// Infof implements log.Logger.Infof. +func (ctx taskAsyncContext) Infof(format string, v ...interface{}) { + ctx.t.Infof(format, v...) +} + +// Warningf implements log.Logger.Warningf. +func (ctx taskAsyncContext) Warningf(format string, v ...interface{}) { + ctx.t.Warningf(format, v...) +} + +// IsLogging implements log.Logger.IsLogging. +func (ctx taskAsyncContext) IsLogging(level log.Level) bool { + return ctx.t.IsLogging(level) +} + +// Value implements context.Context.Value. +func (ctx taskAsyncContext) Value(key interface{}) interface{} { + return ctx.t.Value(key) +} diff --git a/pkg/sentry/kernel/epoll/BUILD b/pkg/sentry/kernel/epoll/BUILD new file mode 100644 index 000000000..04651d961 --- /dev/null +++ b/pkg/sentry/kernel/epoll/BUILD @@ -0,0 +1,52 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_stateify( + name = "epoll_autogen_state", + srcs = [ + "epoll.go", + "epoll_state.go", + ], + out = "epoll_autogen_state.go", + package = "epoll", +) + +go_library( + name = "epoll", + srcs = [ + "epoll.go", + "epoll_autogen_state.go", + "epoll_state.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/ilist", + "//pkg/refs", + "//pkg/sentry/context", + "//pkg/sentry/fs", + "//pkg/sentry/fs/anon", + "//pkg/sentry/fs/fsutil", + "//pkg/sentry/kernel/kdefs", + "//pkg/sentry/kernel/time", + "//pkg/sentry/usermem", + "//pkg/state", + "//pkg/waiter", + ], +) + +go_test( + name = "epoll_test", + size = "small", + srcs = [ + "epoll_test.go", + ], + embed = [":epoll"], + deps = [ + "//pkg/sentry/context/contexttest", + "//pkg/sentry/fs/filetest", + "//pkg/waiter", + ], +) diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go new file mode 100644 index 000000000..b572fcd7e --- /dev/null +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -0,0 +1,466 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package epoll provides an implementation of Linux's IO event notification +// facility. See epoll(7) for more details. +package epoll + +import ( + "fmt" + "sync" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/ilist" + "gvisor.googlesource.com/gvisor/pkg/refs" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/anon" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// Event describes the event mask that was observed and the user data to be +// returned when one of the events occurs. It has this format to match the linux +// format to avoid extra copying/allocation when writing events to userspace. +type Event struct { + // Events is the event mask containing the set of events that have been + // observed on an entry. + Events uint32 + + // Data is an opaque 64-bit value provided by the caller when adding the + // entry, and returned to the caller when the entry reports an event. + Data [2]int32 +} + +// EntryFlags is a bitmask that holds an entry's flags. +type EntryFlags int + +// Valid entry flags. +const ( + OneShot EntryFlags = 1 << iota + EdgeTriggered +) + +// FileIdentifier identifies a file. We cannot use just the FD because it could +// potentially be reassigned. We also cannot use just the file pointer because +// it is possible to have multiple entries for the same file object as long as +// they are created with different FDs (i.e., the FDs point to the same file). +type FileIdentifier struct { + File *fs.File + Fd kdefs.FD +} + +// pollEntry holds all the state associated with an event poll entry, that is, +// a file being observed by an event poll object. +type pollEntry struct { + ilist.Entry + file *refs.WeakRef `state:"manual"` + id FileIdentifier `state:"wait"` + userData [2]int32 + waiter waiter.Entry `state:"manual"` + mask waiter.EventMask + flags EntryFlags + + epoll *EventPoll + + // We cannot save the current list pointer as it points into EventPoll + // struct, while state framework currently does not support such + // in-struct pointers. Instead, EventPoll will properly set this field + // in its loading logic. + curList *ilist.List `state:"nosave"` +} + +// WeakRefGone implements refs.WeakRefUser.WeakRefGone. +// weakReferenceGone is called when the file in the weak reference is destroyed. +// The poll entry is removed in response to this. +func (p *pollEntry) WeakRefGone() { + p.epoll.RemoveEntry(p.id) +} + +// EventPoll holds all the state associated with an event poll object, that is, +// collection of files to observe and their current state. +type EventPoll struct { + fsutil.PipeSeek `state:"zerovalue"` + fsutil.NotDirReaddir `state:"zerovalue"` + fsutil.NoFsync `state:"zerovalue"` + fsutil.NoopFlush `state:"zerovalue"` + fsutil.NoMMap `state:"zerovalue"` + fsutil.NoIoctl `state:"zerovalue"` + + // Wait queue is used to notify interested parties when the event poll + // object itself becomes readable or writable. + waiter.Queue + + // files is the map of all the files currently being observed, it is + // protected by mu. + mu sync.Mutex `state:"nosave"` + files map[FileIdentifier]*pollEntry + + // listsMu protects manipulation of the lists below. It needs to be a + // different lock to avoid circular lock acquisition order involving + // the wait queue mutexes and mu. The full order is mu, observed file + // wait queue mutex, then listsMu; this allows listsMu to be acquired + // when readyCallback is called. + // + // An entry is always in one of the following lists: + // readyList -- when there's a chance that it's ready to have + // events delivered to epoll waiters. Given that being + // ready is a transient state, the Readiness() and + // readEvents() functions always call the entry's file + // Readiness() function to confirm it's ready. + // waitingList -- when there's no chance that the entry is ready, + // so it's waiting for the readyCallback to be called + // on it before it gets moved to the readyList. + // disabledList -- when the entry is disabled. This happens when + // a one-shot entry gets delivered via readEvents(). + listsMu sync.Mutex `state:"nosave"` + readyList ilist.List + waitingList ilist.List + disabledList ilist.List +} + +// cycleMu is used to serialize all the cycle checks. This is only used when +// an event poll file is added as an entry to another event poll. Such checks +// are serialized to avoid lock acquisition order inversion: if a thread is +// adding A to B, and another thread is adding B to A, each would acquire A's +// and B's mutexes in reverse order, and could cause deadlocks. Having this +// lock prevents this by allowing only one check at a time to happen. +// +// We do the cycle check to prevent callers from introducing potentially +// infinite recursions. If a caller were to add A to B and then B to A, for +// event poll A to know if it's readable, it would need to check event poll B, +// which in turn would need event poll A and so on indefinitely. +var cycleMu sync.Mutex + +// NewEventPoll allocates and initializes a new event poll object. +func NewEventPoll(ctx context.Context) *fs.File { + // name matches fs/eventpoll.c:epoll_create1. + dirent := fs.NewDirent(anon.NewInode(ctx), fmt.Sprintf("anon_inode:[eventpoll]")) + return fs.NewFile(ctx, dirent, fs.FileFlags{}, &EventPoll{ + files: make(map[FileIdentifier]*pollEntry), + }) +} + +// Release implements fs.FileOperations.Release. +func (e *EventPoll) Release() { + // We need to take the lock now because files may be attempting to + // remove entries in parallel if they get destroyed. + e.mu.Lock() + defer e.mu.Unlock() + + // Go through all entries and clean up. + for _, entry := range e.files { + entry.id.File.EventUnregister(&entry.waiter) + entry.file.Drop() + } +} + +// Read implements fs.FileOperations.Read. +func (*EventPoll) Read(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, syscall.ENOSYS +} + +// Write implements fs.FileOperations.Write. +func (*EventPoll) Write(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, syscall.ENOSYS +} + +// eventsAvailable determines if 'e' has events available for delivery. +func (e *EventPoll) eventsAvailable() bool { + e.listsMu.Lock() + + for it := e.readyList.Front(); it != nil; { + entry := it.(*pollEntry) + it = it.Next() + + // If the entry is ready, we know 'e' has at least one entry + // ready for delivery. + ready := entry.id.File.Readiness(entry.mask) + if ready != 0 { + e.listsMu.Unlock() + return true + } + + // Entry is not ready, so move it to waiting list. + e.readyList.Remove(entry) + e.waitingList.PushBack(entry) + entry.curList = &e.waitingList + } + + e.listsMu.Unlock() + + return false +} + +// Readiness determines if the event poll object is currently readable (i.e., +// if there are pending events for delivery). +func (e *EventPoll) Readiness(mask waiter.EventMask) waiter.EventMask { + ready := waiter.EventMask(0) + + if (mask&waiter.EventIn) != 0 && e.eventsAvailable() { + ready |= waiter.EventIn + } + + return ready +} + +// ReadEvents returns up to max available events. +func (e *EventPoll) ReadEvents(max int) []Event { + var local ilist.List + var ret []Event + + e.listsMu.Lock() + + // Go through all entries we believe may be ready. + for it := e.readyList.Front(); it != nil && len(ret) < max; { + entry := it.(*pollEntry) + it = it.Next() + + // Check the entry's readiness. It it's not really ready, we + // just put it back in the waiting list and move on to the next + // entry. + ready := entry.id.File.Readiness(entry.mask) & entry.mask + if ready == 0 { + e.readyList.Remove(entry) + e.waitingList.PushBack(entry) + entry.curList = &e.waitingList + + continue + } + + // Add event to the array that will be returned to caller. + ret = append(ret, Event{ + Events: uint32(ready), + Data: entry.userData, + }) + + // The entry is consumed, so we must move it to the disabled + // list in case it's one-shot, or back to the wait list if it's + // edge-triggered. If it's neither, we leave it in the ready + // list so that its readiness can be checked the next time + // around; however, we must move it to the end of the list so + // that other events can be delivered as well. + e.readyList.Remove(entry) + if entry.flags&OneShot != 0 { + e.disabledList.PushBack(entry) + entry.curList = &e.disabledList + } else if entry.flags&EdgeTriggered != 0 { + e.waitingList.PushBack(entry) + entry.curList = &e.waitingList + } else { + local.PushBack(entry) + } + } + + e.readyList.PushBackList(&local) + + e.listsMu.Unlock() + + return ret +} + +// readyCallback is called when one of the files we're polling becomes ready. It +// moves said file to the readyList if it's currently in the waiting list. +type readyCallback struct{} + +// Callback implements waiter.EntryCallback.Callback. +func (*readyCallback) Callback(w *waiter.Entry) { + entry := w.Context.(*pollEntry) + e := entry.epoll + + e.listsMu.Lock() + + if entry.curList == &e.waitingList { + e.waitingList.Remove(entry) + e.readyList.PushBack(entry) + entry.curList = &e.readyList + + e.Notify(waiter.EventIn) + } + + e.listsMu.Unlock() +} + +// initEntryReadiness initializes the entry's state with regards to its +// readiness by placing it in the appropriate list and registering for +// notifications. +func (e *EventPoll) initEntryReadiness(entry *pollEntry) { + // A new entry starts off in the waiting list. + e.listsMu.Lock() + e.waitingList.PushBack(entry) + entry.curList = &e.waitingList + e.listsMu.Unlock() + + // Register for event notifications. + f := entry.id.File + f.EventRegister(&entry.waiter, entry.mask) + + // Check if the file happens to already be in a ready state. + ready := f.Readiness(entry.mask) & entry.mask + if ready != 0 { + (*readyCallback).Callback(nil, &entry.waiter) + } +} + +// observes checks if event poll object e is directly or indirectly observing +// event poll object ep. It uses a bounded recursive depth-first search. +func (e *EventPoll) observes(ep *EventPoll, depthLeft int) bool { + // If we reached the maximum depth, we'll consider that we found it + // because we don't want to allow chains that are too long. + if depthLeft <= 0 { + return true + } + + e.mu.Lock() + defer e.mu.Unlock() + + // Go through each observed file and check if it is or observes ep. + for id := range e.files { + f, ok := id.File.FileOperations.(*EventPoll) + if !ok { + continue + } + + if f == ep || f.observes(ep, depthLeft-1) { + return true + } + } + + return false +} + +// AddEntry adds a new file to the collection of files observed by e. +func (e *EventPoll) AddEntry(id FileIdentifier, flags EntryFlags, mask waiter.EventMask, data [2]int32) error { + // Acquire cycle check lock if another event poll is being added. + ep, ok := id.File.FileOperations.(*EventPoll) + if ok { + cycleMu.Lock() + defer cycleMu.Unlock() + } + + e.mu.Lock() + defer e.mu.Unlock() + + // Fail if the file already has an entry. + if _, ok := e.files[id]; ok { + return syscall.EEXIST + } + + // Check if a cycle would be created. We use 4 as the limit because + // that's the value used by linux and we want to emulate it. + if ep != nil { + if e == ep { + return syscall.EINVAL + } + + if ep.observes(e, 4) { + return syscall.ELOOP + } + } + + // Create new entry and add it to map. + // + // N.B. Even though we are creating a weak reference here, we know it + // won't trigger a callback because we hold a reference to the file + // throughout the execution of this function. + entry := &pollEntry{ + id: id, + userData: data, + epoll: e, + flags: flags, + waiter: waiter.Entry{Callback: &readyCallback{}}, + mask: mask, + } + entry.waiter.Context = entry + e.files[id] = entry + entry.file = refs.NewWeakRef(id.File, entry) + + // Initialize the readiness state of the new entry. + e.initEntryReadiness(entry) + + return nil +} + +// UpdateEntry updates the flags, mask and user data associated with a file that +// is already part of the collection of observed files. +func (e *EventPoll) UpdateEntry(id FileIdentifier, flags EntryFlags, mask waiter.EventMask, data [2]int32) error { + e.mu.Lock() + defer e.mu.Unlock() + + // Fail if the file doesn't have an entry. + entry, ok := e.files[id] + if !ok { + return syscall.ENOENT + } + + // Unregister the old mask and remove entry from the list it's in, so + // readyCallback is guaranteed to not be called on this entry anymore. + entry.id.File.EventUnregister(&entry.waiter) + + // Remove entry from whatever list it's in. This ensure that no other + // threads have access to this entry as the only way left to find it + // is via e.files, but we hold e.mu, which prevents that. + e.listsMu.Lock() + entry.curList.Remove(entry) + e.listsMu.Unlock() + + // Initialize new readiness state. + entry.flags = flags + entry.mask = mask + entry.userData = data + e.initEntryReadiness(entry) + + return nil +} + +// RemoveEntry a files from the collection of observed files. +func (e *EventPoll) RemoveEntry(id FileIdentifier) error { + e.mu.Lock() + defer e.mu.Unlock() + + // Fail if the file doesn't have an entry. + entry, ok := e.files[id] + if !ok { + return syscall.ENOENT + } + + // Unregister from file first so that no concurrent attempts will be + // made to manipulate the file. + entry.id.File.EventUnregister(&entry.waiter) + + // Remove from the current list. + e.listsMu.Lock() + entry.curList.Remove(entry) + entry.curList = nil + e.listsMu.Unlock() + + // Remove file from map, and drop weak reference. + delete(e.files, id) + entry.file.Drop() + + return nil +} + +// UnregisterEpollWaiters removes the epoll waiter objects from the waiting +// queues. This is different from Release() as the file is not dereferenced. +func (e *EventPoll) UnregisterEpollWaiters() { + e.mu.Lock() + defer e.mu.Unlock() + + for _, entry := range e.files { + entry.id.File.EventUnregister(&entry.waiter) + } +} diff --git a/pkg/sentry/kernel/epoll/epoll_state.go b/pkg/sentry/kernel/epoll/epoll_state.go new file mode 100644 index 000000000..dabb32f49 --- /dev/null +++ b/pkg/sentry/kernel/epoll/epoll_state.go @@ -0,0 +1,51 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package epoll + +import ( + "gvisor.googlesource.com/gvisor/pkg/ilist" + "gvisor.googlesource.com/gvisor/pkg/refs" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// afterLoad is invoked by stateify. +func (p *pollEntry) afterLoad() { + p.waiter = waiter.Entry{Callback: &readyCallback{}} + p.waiter.Context = p + p.file = refs.NewWeakRef(p.id.File, p) + p.id.File.EventRegister(&p.waiter, p.mask) +} + +// afterLoad is invoked by stateify. +func (e *EventPoll) afterLoad() { + e.listsMu.Lock() + defer e.listsMu.Unlock() + + for _, ls := range []*ilist.List{&e.waitingList, &e.readyList, &e.disabledList} { + for it := ls.Front(); it != nil; it = it.Next() { + it.(*pollEntry).curList = ls + } + } + + for it := e.waitingList.Front(); it != nil; it = it.Next() { + p := it.(*pollEntry) + if p.id.File.Readiness(p.mask) != 0 { + e.waitingList.Remove(p) + e.readyList.PushBack(p) + p.curList = &e.readyList + e.Notify(waiter.EventIn) + } + } +} diff --git a/pkg/sentry/kernel/epoll/epoll_test.go b/pkg/sentry/kernel/epoll/epoll_test.go new file mode 100644 index 000000000..bc869fc13 --- /dev/null +++ b/pkg/sentry/kernel/epoll/epoll_test.go @@ -0,0 +1,54 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package epoll + +import ( + "testing" + + "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/filetest" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +func TestFileDestroyed(t *testing.T) { + f := filetest.NewTestFile(t) + id := FileIdentifier{f, 12} + + efile := NewEventPoll(contexttest.Context(t)) + e := efile.FileOperations.(*EventPoll) + if err := e.AddEntry(id, 0, waiter.EventIn, [2]int32{}); err != nil { + t.Fatalf("addEntry failed: %v", err) + } + + // Check that we get an event reported twice in a row. + evt := e.ReadEvents(1) + if len(evt) != 1 { + t.Fatalf("Unexpected number of ready events: want %v, got %v", 1, len(evt)) + } + + evt = e.ReadEvents(1) + if len(evt) != 1 { + t.Fatalf("Unexpected number of ready events: want %v, got %v", 1, len(evt)) + } + + // Destroy the file. Check that we get no more events. + f.DecRef() + + evt = e.ReadEvents(1) + if len(evt) != 0 { + t.Fatalf("Unexpected number of ready events: want %v, got %v", 0, len(evt)) + } + +} diff --git a/pkg/sentry/kernel/eventfd/BUILD b/pkg/sentry/kernel/eventfd/BUILD new file mode 100644 index 000000000..2d5a3c693 --- /dev/null +++ b/pkg/sentry/kernel/eventfd/BUILD @@ -0,0 +1,46 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_stateify( + name = "eventfd_state", + srcs = [ + "eventfd.go", + ], + out = "eventfd_state.go", + package = "eventfd", +) + +go_library( + name = "eventfd", + srcs = [ + "eventfd.go", + "eventfd_state.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/eventfd", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/refs", + "//pkg/sentry/context", + "//pkg/sentry/fs", + "//pkg/sentry/fs/anon", + "//pkg/sentry/fs/fsutil", + "//pkg/sentry/usermem", + "//pkg/state", + "//pkg/syserror", + "//pkg/waiter", + ], +) + +go_test( + name = "eventfd_test", + size = "small", + srcs = ["eventfd_test.go"], + embed = [":eventfd"], + deps = [ + "//pkg/sentry/context/contexttest", + "//pkg/sentry/usermem", + "//pkg/waiter", + ], +) diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go new file mode 100644 index 000000000..c9333719e --- /dev/null +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -0,0 +1,172 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package eventfd provides an implementation of Linux's file-based event +// notification. +package eventfd + +import ( + "math" + "sync" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/anon" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// EventOperations represents an event with the semantics of Linux's file-based event +// notification (eventfd). +type EventOperations struct { + fsutil.NoopRelease `state:"nosave"` + fsutil.PipeSeek `state:"nosave"` + fsutil.NotDirReaddir `state:"nosave"` + fsutil.NoFsync `state:"nosave"` + fsutil.NoopFlush `state:"nosave"` + fsutil.NoMMap `state:"nosave"` + fsutil.NoIoctl `state:"nosave"` + + // Mutex that protects accesses to the fields of this event. + mu sync.Mutex `state:"nosave"` + + // Queue is used to notify interested parties when the event object + // becomes readable or writable. + waiter.Queue `state:"nosave"` + + // val is the current value of the event counter. + val uint64 + + // semMode specifies whether the event is in "semaphore" mode. + semMode bool +} + +// New creates a new event object with the supplied initial value and mode. +func New(ctx context.Context, initVal uint64, semMode bool) *fs.File { + // name matches fs/eventfd.c:eventfd_file_create. + dirent := fs.NewDirent(anon.NewInode(ctx), "anon_inode:[eventfd]") + return fs.NewFile(ctx, dirent, fs.FileFlags{Read: true, Write: true}, &EventOperations{ + val: initVal, + semMode: semMode, + }) +} + +// Read implements fs.FileOperations.Read. +func (e *EventOperations) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, _ int64) (int64, error) { + if dst.NumBytes() < 8 { + return 0, syscall.EINVAL + } + if err := e.read(ctx, dst); err != nil { + return 0, err + } + return 8, nil +} + +// Write implements fs.FileOperations.Write. +func (e *EventOperations) Write(ctx context.Context, _ *fs.File, src usermem.IOSequence, _ int64) (int64, error) { + if src.NumBytes() < 8 { + return 0, syscall.EINVAL + } + if err := e.write(ctx, src); err != nil { + return 0, err + } + return 8, nil +} + +func (e *EventOperations) read(ctx context.Context, dst usermem.IOSequence) error { + e.mu.Lock() + + // We can't complete the read if the value is currently zero. + if e.val == 0 { + e.mu.Unlock() + return syserror.ErrWouldBlock + } + + // Update the value based on the mode the event is operating in. + var val uint64 + if e.semMode { + val = 1 + // Consistent with Linux, this is done even if writing to memory fails. + e.val-- + } else { + val = e.val + e.val = 0 + } + + e.mu.Unlock() + + // Notify writers. We do this even if we were already writable because + // it is possible that a writer is waiting to write the maximum value + // to the event. + e.Notify(waiter.EventOut) + + var buf [8]byte + usermem.ByteOrder.PutUint64(buf[:], val) + _, err := dst.CopyOut(ctx, buf[:]) + return err +} + +func (e *EventOperations) write(ctx context.Context, src usermem.IOSequence) error { + var buf [8]byte + if _, err := src.CopyIn(ctx, buf[:]); err != nil { + return err + } + val := usermem.ByteOrder.Uint64(buf[:]) + + return e.Signal(val) +} + +// Signal is an internal function to signal the event fd. +func (e *EventOperations) Signal(val uint64) error { + if val == math.MaxUint64 { + return syscall.EINVAL + } + + e.mu.Lock() + + // We only allow writes that won't cause the value to go over the max + // uint64 minus 1. + if val > math.MaxUint64-1-e.val { + e.mu.Unlock() + return syserror.ErrWouldBlock + } + + e.val += val + e.mu.Unlock() + + // Always trigger a notification. + e.Notify(waiter.EventIn) + + return nil +} + +// Readiness returns the ready events for the event fd. +func (e *EventOperations) Readiness(mask waiter.EventMask) waiter.EventMask { + ready := waiter.EventMask(0) + + e.mu.Lock() + if e.val > 0 { + ready |= waiter.EventIn + } + + if e.val < math.MaxUint64-1 { + ready |= waiter.EventOut + } + e.mu.Unlock() + + return mask & ready +} diff --git a/pkg/sentry/kernel/eventfd/eventfd_test.go b/pkg/sentry/kernel/eventfd/eventfd_test.go new file mode 100644 index 000000000..71326b62f --- /dev/null +++ b/pkg/sentry/kernel/eventfd/eventfd_test.go @@ -0,0 +1,78 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package eventfd + +import ( + "testing" + + "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +func TestEventfd(t *testing.T) { + initVals := []uint64{ + 0, + // Using a non-zero initial value verifies that writing to an + // eventfd signals when the eventfd's counter was already + // non-zero. + 343, + } + + for _, initVal := range initVals { + ctx := contexttest.Context(t) + + // Make a new event that is writable. + event := New(ctx, initVal, false) + + // Register a callback for a write event. + w, ch := waiter.NewChannelEntry(nil) + event.EventRegister(&w, waiter.EventIn) + defer event.EventUnregister(&w) + + data := []byte("00000124") + // Create and submit a write request. + n, err := event.Writev(ctx, usermem.BytesIOSequence(data)) + if err != nil { + t.Fatal(err) + } + if n != 8 { + t.Errorf("eventfd.write wrote %d bytes, not full int64", n) + } + + // Check if the callback fired due to the write event. + select { + case <-ch: + default: + t.Errorf("Didn't get notified of EventIn after write") + } + } +} + +func TestEventfdStat(t *testing.T) { + ctx := contexttest.Context(t) + + // Make a new event that is writable. + event := New(ctx, 0, false) + + // Create and submit an stat request. + uattr, err := event.Dirent.Inode.UnstableAttr(ctx) + if err != nil { + t.Fatalf("eventfd stat request failed: %v", err) + } + if uattr.Size != 0 { + t.Fatal("EventFD size should be 0") + } +} diff --git a/pkg/sentry/kernel/fd_map.go b/pkg/sentry/kernel/fd_map.go new file mode 100644 index 000000000..ef73125fd --- /dev/null +++ b/pkg/sentry/kernel/fd_map.go @@ -0,0 +1,340 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "bytes" + "fmt" + "sort" + "sync" + "sync/atomic" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/refs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/lock" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.googlesource.com/gvisor/pkg/sentry/limits" +) + +// FDs is an ordering of FD's that can be made stable. +type FDs []kdefs.FD + +func (f FDs) Len() int { + return len(f) +} + +func (f FDs) Swap(i, j int) { + f[i], f[j] = f[j], f[i] +} + +func (f FDs) Less(i, j int) bool { + return f[i] < f[j] +} + +// FDFlags define flags for an individual descriptor. +type FDFlags struct { + // CloseOnExec indicates the descriptor should be closed on exec. + CloseOnExec bool +} + +// descriptor holds the details about a file descriptor, namely a pointer the +// file itself and the descriptor flags. +type descriptor struct { + file *fs.File + flags FDFlags +} + +// FDMap is used to manage File references and flags. +type FDMap struct { + refs.AtomicRefCount + k *Kernel + files map[kdefs.FD]descriptor + mu sync.RWMutex `state:"nosave"` + uid uint64 +} + +// ID returns a unique identifier for this FDMap. +func (f *FDMap) ID() uint64 { + return f.uid +} + +// NewFDMap allocates a new FDMap that may be used by tasks in k. +func (k *Kernel) NewFDMap() *FDMap { + return &FDMap{ + k: k, + files: make(map[kdefs.FD]descriptor), + uid: atomic.AddUint64(&k.fdMapUids, 1), + } +} + +// destroy removes all of the file descriptors from the map. +func (f *FDMap) destroy() { + f.RemoveIf(func(*fs.File, FDFlags) bool { + return true + }) +} + +// DecRef implements RefCounter.DecRef with destructor f.destroy. +func (f *FDMap) DecRef() { + f.DecRefWithDestructor(f.destroy) +} + +// Size returns the number of file descriptor slots currently allocated. +func (f *FDMap) Size() int { + f.mu.RLock() + defer f.mu.RUnlock() + + return len(f.files) +} + +// String is a stringer for FDMap. +func (f *FDMap) String() string { + f.mu.RLock() + defer f.mu.RUnlock() + + var b bytes.Buffer + for k, v := range f.files { + n, _ := v.file.Dirent.FullName(nil /* root */) + b.WriteString(fmt.Sprintf("\tfd:%d => name %s\n", k, n)) + } + return b.String() +} + +// NewFDFrom allocates a new FD guaranteed to be the lowest number available +// greater than or equal to from. This property is important as Unix programs +// tend to count on this allocation order. +func (f *FDMap) NewFDFrom(fd kdefs.FD, file *fs.File, flags FDFlags, limitSet *limits.LimitSet) (kdefs.FD, error) { + if fd < 0 { + // Don't accept negative FDs. + return 0, syscall.EINVAL + } + + f.mu.Lock() + defer f.mu.Unlock() + + // Finds the lowest fd not in the handles map. + lim := limitSet.Get(limits.NumberOfFiles) + for i := fd; lim.Cur == limits.Infinity || i < kdefs.FD(lim.Cur); i++ { + if _, ok := f.files[i]; !ok { + file.IncRef() + f.files[i] = descriptor{file, flags} + return i, nil + } + } + + return -1, syscall.EMFILE +} + +// NewFDAt sets the file reference for the given FD. If there is an +// active reference for that FD, the ref count for that existing reference +// is decremented. +func (f *FDMap) NewFDAt(fd kdefs.FD, file *fs.File, flags FDFlags, limitSet *limits.LimitSet) error { + if fd < 0 { + // Don't accept negative FDs. + return syscall.EBADF + } + + // In this one case we do not do a defer of the Unlock. The + // reason is that we must have done all the work needed for + // discarding any old open file before we return to the + // caller. In other words, the DecRef(), below, must have + // completed by the time we return to the caller to ensure + // side effects are, in fact, effected. A classic example is + // dup2(fd1, fd2); if fd2 was already open, it must be closed, + // and we don't want to resume the caller until it is; we have + // to block on the DecRef(). Hence we can not just do a 'go + // oldfile.DecRef()', since there would be no guarantee that + // it would be done before we the caller resumed. Since we + // must wait for the DecRef() to finish, and that could take + // time, it's best to first call f.muUnlock beore so we are + // not blocking other uses of this FDMap on the DecRef() call. + f.mu.Lock() + oldDesc, oldExists := f.files[fd] + lim := limitSet.Get(limits.NumberOfFiles).Cur + // if we're closing one then the effective limit is one + // more than the actual limit. + if oldExists && lim != limits.Infinity { + lim++ + } + if lim != limits.Infinity && fd >= kdefs.FD(lim) { + f.mu.Unlock() + return syscall.EMFILE + } + + file.IncRef() + f.files[fd] = descriptor{file, flags} + f.mu.Unlock() + + if oldExists { + oldDesc.file.DecRef() + } + return nil +} + +// SetFlags sets the flags for the given file descriptor, if it is valid. +func (f *FDMap) SetFlags(fd kdefs.FD, flags FDFlags) { + f.mu.Lock() + defer f.mu.Unlock() + + desc, ok := f.files[fd] + if !ok { + return + } + + f.files[fd] = descriptor{desc.file, flags} +} + +// GetDescriptor returns a reference to the file and the flags for the FD. It +// bumps its reference count as well. It returns nil if there is no File +// for the FD, i.e. if the FD is invalid. The caller must use DecRef +// when they are done. +func (f *FDMap) GetDescriptor(fd kdefs.FD) (*fs.File, FDFlags) { + f.mu.RLock() + defer f.mu.RUnlock() + + if desc, ok := f.files[fd]; ok { + desc.file.IncRef() + return desc.file, desc.flags + } + return nil, FDFlags{} +} + +// GetFile returns a reference to the File for the FD and bumps +// its reference count as well. It returns nil if there is no File +// for the FD, i.e. if the FD is invalid. The caller must use DecRef +// when they are done. +func (f *FDMap) GetFile(fd kdefs.FD) *fs.File { + f.mu.RLock() + if desc, ok := f.files[fd]; ok { + desc.file.IncRef() + f.mu.RUnlock() + return desc.file + } + f.mu.RUnlock() + return nil +} + +// fds returns an ordering of FDs. +func (f *FDMap) fds() FDs { + fds := make(FDs, 0, len(f.files)) + for fd := range f.files { + fds = append(fds, fd) + } + sort.Sort(fds) + return fds +} + +// GetFDs returns a list of valid fds. +func (f *FDMap) GetFDs() FDs { + f.mu.RLock() + defer f.mu.RUnlock() + return f.fds() +} + +// GetRefs returns a stable slice of references to all files and bumps the +// reference count on each. The caller must use DecRef on each reference when +// they're done using the slice. +func (f *FDMap) GetRefs() []*fs.File { + f.mu.RLock() + defer f.mu.RUnlock() + + fds := f.fds() + fs := make([]*fs.File, 0, len(fds)) + for _, fd := range fds { + desc := f.files[fd] + desc.file.IncRef() + fs = append(fs, desc.file) + } + return fs +} + +// Fork returns an independent FDMap pointing to the same descriptors. +func (f *FDMap) Fork() *FDMap { + f.mu.RLock() + defer f.mu.RUnlock() + + clone := f.k.NewFDMap() + + // Grab a extra reference for every file. + for fd, desc := range f.files { + desc.file.IncRef() + clone.files[fd] = desc + } + + // That's it! + return clone +} + +// unlock releases all file locks held by this FDMap's uid. Must only be +// called on a non-nil *fs.File. +func (f *FDMap) unlock(file *fs.File) { + id := lock.UniqueID(f.ID()) + file.Dirent.Inode.LockCtx.Posix.UnlockRegion(id, lock.LockRange{0, lock.LockEOF}) +} + +// inotifyFileClose generates the appropriate inotify events for f being closed. +func inotifyFileClose(f *fs.File) { + var ev uint32 + d := f.Dirent + + if fs.IsDir(d.Inode.StableAttr) { + ev |= linux.IN_ISDIR + } + + if f.Flags().Write { + ev |= linux.IN_CLOSE_WRITE + } else { + ev |= linux.IN_CLOSE_NOWRITE + } + + d.InotifyEvent(ev, 0) +} + +// Remove removes an FD from the FDMap, and returns (File, true) if a File +// one was found. Callers are expected to decrement the reference count on +// the File. Otherwise returns (nil, false). +func (f *FDMap) Remove(fd kdefs.FD) (*fs.File, bool) { + f.mu.Lock() + desc := f.files[fd] + delete(f.files, fd) + f.mu.Unlock() + if desc.file != nil { + f.unlock(desc.file) + inotifyFileClose(desc.file) + return desc.file, true + } + return nil, false +} + +// RemoveIf removes all FDs where cond is true. +func (f *FDMap) RemoveIf(cond func(*fs.File, FDFlags) bool) { + var removed []*fs.File + f.mu.Lock() + for fd, desc := range f.files { + if desc.file != nil && cond(desc.file, desc.flags) { + delete(f.files, fd) + removed = append(removed, desc.file) + } + } + f.mu.Unlock() + + for _, file := range removed { + f.unlock(file) + inotifyFileClose(file) + file.DecRef() + } +} diff --git a/pkg/sentry/kernel/fd_map_test.go b/pkg/sentry/kernel/fd_map_test.go new file mode 100644 index 000000000..e1ac900e8 --- /dev/null +++ b/pkg/sentry/kernel/fd_map_test.go @@ -0,0 +1,134 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "testing" + + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/filetest" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.googlesource.com/gvisor/pkg/sentry/limits" +) + +const ( + // maxFD is the maximum FD to try to create in the map. + // This number of open files has been seen in the wild. + maxFD = 2 * 1024 +) + +func newTestFDMap() *FDMap { + return &FDMap{ + files: make(map[kdefs.FD]descriptor), + } +} + +// TestFDMapMany allocates maxFD FDs, i.e. maxes out the FDMap, +// until there is no room, then makes sure that NewFDAt works +// and also that if we remove one and add one that works too. +func TestFDMapMany(t *testing.T) { + file := filetest.NewTestFile(t) + limitSet := limits.NewLimitSet() + limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}) + + f := newTestFDMap() + for i := 0; i < maxFD; i++ { + if _, err := f.NewFDFrom(0, file, FDFlags{}, limitSet); err != nil { + t.Fatalf("Allocated %v FDs but wanted to allocate %v", i, maxFD) + } + } + + if _, err := f.NewFDFrom(0, file, FDFlags{}, limitSet); err == nil { + t.Fatalf("f.NewFDFrom(0, r) in full map: got nil, wanted error") + } + + if err := f.NewFDAt(1, file, FDFlags{}, limitSet); err != nil { + t.Fatalf("f.NewFDAt(1, r, FDFlags{}): got %v, wanted nil", err) + } +} + +// TestFDMap does a set of simple tests to make sure simple adds, +// removes, GetRefs, and DecRefs work. The ordering is just weird +// enough that a table-driven approach seemed clumsy. +func TestFDMap(t *testing.T) { + file := filetest.NewTestFile(t) + limitSet := limits.NewLimitSet() + limitSet.Set(limits.NumberOfFiles, limits.Limit{1, maxFD}) + + f := newTestFDMap() + if _, err := f.NewFDFrom(0, file, FDFlags{}, limitSet); err != nil { + t.Fatalf("Adding an FD to an empty 1-size map: got %v, want nil", err) + } + + if _, err := f.NewFDFrom(0, file, FDFlags{}, limitSet); err == nil { + t.Fatalf("Adding an FD to a filled 1-size map: got nil, wanted an error") + } + + largeLimit := limits.Limit{maxFD, maxFD} + limitSet.Set(limits.NumberOfFiles, largeLimit) + + if fd, err := f.NewFDFrom(0, file, FDFlags{}, limitSet); err != nil { + t.Fatalf("Adding an FD to a resized map: got %v, want nil", err) + } else if fd != kdefs.FD(1) { + t.Fatalf("Added an FD to a resized map: got %v, want 1", fd) + } + + if err := f.NewFDAt(1, file, FDFlags{}, limitSet); err != nil { + t.Fatalf("Replacing FD 1 via f.NewFDAt(1, r, FDFlags{}): got %v, wanted nil", err) + } + + if err := f.NewFDAt(maxFD+1, file, FDFlags{}, limitSet); err == nil { + t.Fatalf("Using an FD that was too large via f.NewFDAt(%v, r, FDFlags{}): got nil, wanted an error", maxFD+1) + } + + if ref := f.GetFile(1); ref == nil { + t.Fatalf("f.GetFile(1): got nil, wanted %v", file) + } + + if ref := f.GetFile(2); ref != nil { + t.Fatalf("f.GetFile(2): got a %v, wanted nil", ref) + } + + ref, ok := f.Remove(1) + if !ok { + t.Fatalf("f.Remove(1) for an existing FD: failed, want success") + } + ref.DecRef() + + if ref, ok := f.Remove(1); ok { + ref.DecRef() + t.Fatalf("r.Remove(1) for a removed FD: got success, want failure") + } + +} + +func TestDescriptorFlags(t *testing.T) { + file := filetest.NewTestFile(t) + f := newTestFDMap() + limitSet := limits.NewLimitSet() + limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}) + + if err := f.NewFDAt(2, file, FDFlags{CloseOnExec: true}, limitSet); err != nil { + t.Fatalf("f.NewFDAt(2, r, FDFlags{}): got %v, wanted nil", err) + } + + newFile, flags := f.GetDescriptor(2) + if newFile == nil { + t.Fatalf("f.GetFile(2): got a %v, wanted nil", newFile) + } + + if !flags.CloseOnExec { + t.Fatalf("new File flags %d don't match original %d\n", flags, 0) + } +} diff --git a/pkg/sentry/kernel/fs_context.go b/pkg/sentry/kernel/fs_context.go new file mode 100644 index 000000000..9aa6fa951 --- /dev/null +++ b/pkg/sentry/kernel/fs_context.go @@ -0,0 +1,172 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + "sync" + + "gvisor.googlesource.com/gvisor/pkg/refs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" +) + +// FSContext contains filesystem context. +// +// This includes umask and working directory. +type FSContext struct { + refs.AtomicRefCount + + // mu protects below. + mu sync.Mutex `state:"nosave"` + + // root is the filesystem root. Will be nil iff the FSContext has been + // destroyed. + root *fs.Dirent + + // cwd is the current working directory. Will be nil iff the FSContext + // has been destroyed. + cwd *fs.Dirent + + // umask is the current file mode creation mask. When a thread using this + // context invokes a syscall that creates a file, bits set in umask are + // removed from the permissions that the file is created with. + umask uint +} + +// newFSContext returns a new filesystem context. +func newFSContext(root, cwd *fs.Dirent, umask uint) *FSContext { + root.IncRef() + cwd.IncRef() + return &FSContext{ + root: root, + cwd: cwd, + umask: umask, + } +} + +// destroy is the destructor for an FSContext. +// +// This will call DecRef on both root and cwd Dirents. If either call to +// DecRef returns an error, then it will be propigated. If both calls to +// DecRef return an error, then the one from root.DecRef will be propigated. +// +// Note that there may still be calls to WorkingDirectory() or RootDirectory() +// (that return nil). This is because valid references may still be held via +// proc files or other mechanisms. +func (f *FSContext) destroy() { + f.root.DecRef() + f.root = nil + + f.cwd.DecRef() + f.cwd = nil +} + +// DecRef implements RefCounter.DecRef with destructor f.destroy. +func (f *FSContext) DecRef() { + f.DecRefWithDestructor(f.destroy) +} + +// Fork forks this FSContext. +// +// This is not a valid call after destroy. +func (f *FSContext) Fork() *FSContext { + f.mu.Lock() + defer f.mu.Unlock() + f.cwd.IncRef() + f.root.IncRef() + return &FSContext{ + cwd: f.cwd, + root: f.root, + umask: f.umask, + } +} + +// WorkingDirectory returns the current working directory. +// You should call DecRef on the returned Dirent when finished. +// +// This will return nil if called after destroy(). +func (f *FSContext) WorkingDirectory() *fs.Dirent { + f.mu.Lock() + defer f.mu.Unlock() + if f.cwd != nil { + f.cwd.IncRef() + } + return f.cwd +} + +// SetWorkingDirectory sets the current working directory. +// This will take an extra reference on the Dirent. +// +// This is not a valid call after destroy. +func (f *FSContext) SetWorkingDirectory(d *fs.Dirent) { + if d == nil { + panic("FSContext.SetWorkingDirectory called with nil dirent") + } + if f.cwd == nil { + panic(fmt.Sprintf("FSContext.SetWorkingDirectory(%v)) called after destroy", d)) + } + f.mu.Lock() + defer f.mu.Unlock() + old := f.cwd + f.cwd = d + d.IncRef() + old.DecRef() +} + +// RootDirectory returns the current filesystem root. +// You should call DecRef on the returned Dirent when finished. +// +// This will return nil if called after destroy(). +func (f *FSContext) RootDirectory() *fs.Dirent { + f.mu.Lock() + defer f.mu.Unlock() + f.root.IncRef() + return f.root +} + +// SetRootDirectory sets the root directory. +// This will take an extra reference on the Dirent. +// +// This is not a valid call after free. +func (f *FSContext) SetRootDirectory(d *fs.Dirent) { + if d == nil { + panic("FSContext.SetRootDirectory called with nil dirent") + } + if f.root == nil { + panic(fmt.Sprintf("FSContext.SetRootDirectory(%v)) called after destroy", d)) + } + f.mu.Lock() + defer f.mu.Unlock() + old := f.root + f.root = d + d.IncRef() + old.DecRef() +} + +// Umask returns the current umask. +func (f *FSContext) Umask() uint { + f.mu.Lock() + defer f.mu.Unlock() + return f.umask +} + +// SwapUmask atomically sets the current umask and returns the old umask. +func (f *FSContext) SwapUmask(mask uint) uint { + f.mu.Lock() + defer f.mu.Unlock() + old := f.umask + f.umask = mask + return old +} diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD new file mode 100644 index 000000000..de9897c58 --- /dev/null +++ b/pkg/sentry/kernel/futex/BUILD @@ -0,0 +1,48 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_generics:defs.bzl", "go_template_instance") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_template_instance( + name = "waiter_list", + out = "waiter_list.go", + package = "futex", + prefix = "waiter", + template = "//pkg/ilist:generic_list", + types = { + "Linker": "*Waiter", + }, +) + +go_stateify( + name = "futex_state", + srcs = [ + "futex.go", + "waiter_list.go", + ], + out = "futex_state.go", + package = "futex", +) + +go_library( + name = "futex", + srcs = [ + "futex.go", + "futex_state.go", + "waiter_list.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/state", + "//pkg/syserror", + ], +) + +go_test( + name = "futex_test", + size = "small", + srcs = ["futex_test.go"], + embed = [":futex"], +) diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go new file mode 100644 index 000000000..b3ba57a2c --- /dev/null +++ b/pkg/sentry/kernel/futex/futex.go @@ -0,0 +1,405 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package futex provides an implementation of the futex interface as found in +// the Linux kernel. It allows one to easily transform Wait() calls into waits +// on a channel, which is useful in a Go-based kernel, for example. +package futex + +import ( + "sync" + "sync/atomic" + + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// Checker abstracts memory accesses. This is useful because the "addresses" +// used in this package may not be real addresses (they could be indices of an +// array, for example), or they could be mapped via some special mechanism. +// +// TODO: Replace this with usermem.IO. +type Checker interface { + // Check should validate that given address contains the given value. + // If it does not contain the value, syserror.EAGAIN must be returned. + // Any other error may be returned, which will be propagated. + Check(addr uintptr, val uint32) error + + // Op should atomically perform the operation encoded in op on the data + // pointed to by addr, then apply the comparison encoded in op to the + // original value at addr, returning the result. + // Note that op is an opaque operation whose behaviour is defined + // outside of the futex manager. + Op(addr uintptr, op uint32) (bool, error) +} + +// Waiter is the struct which gets enqueued into buckets for wake up routines +// and requeue routines to scan and notify. Once a Waiter has been enqueued by +// WaitPrepare(), callers may listen on C for wake up events. +type Waiter struct { + // Synchronization: + // + // - A Waiter that is not enqueued in a bucket is exclusively owned (no + // synchronization applies). + // + // - A Waiter is enqueued in a bucket by calling WaitPrepare(). After this, + // waiterEntry, complete, and addr are protected by the bucket.mu ("bucket + // lock") of the containing bucket, and bitmask is immutable. complete and + // addr are additionally mutated using atomic memory operations, ensuring + // that they can be read using atomic memory operations without holding the + // bucket lock. + // + // - A Waiter is only guaranteed to be no longer queued after calling + // WaitComplete(). + + // waiterEntry links Waiter into bucket.waiters. + waiterEntry + + // complete is 1 if the Waiter was removed from its bucket by a wakeup and + // 0 otherwise. + complete int32 + + // C is sent to when the Waiter is woken. + C chan struct{} + + // addr is the address being waited on. + addr uintptr + + // The bitmask we're waiting on. + // This is used the case of a FUTEX_WAKE_BITSET. + bitmask uint32 +} + +// NewWaiter returns a new unqueued Waiter. +func NewWaiter() *Waiter { + return &Waiter{ + C: make(chan struct{}, 1), + } +} + +// bucket holds a list of waiters for a given address hash. +type bucket struct { + // mu protects waiters and contained Waiter state. See comment in Waiter. + mu sync.Mutex `state:"nosave"` + + waiters waiterList `state:"zerovalue"` +} + +// wakeLocked wakes up to n waiters matching the bitmask at the addr for this +// bucket and returns the number of waiters woken. +// +// Preconditions: b.mu must be locked. +func (b *bucket) wakeLocked(addr uintptr, bitmask uint32, n int) int { + done := 0 + for w := b.waiters.Front(); done < n && w != nil; { + if w.addr != addr || w.bitmask&bitmask == 0 { + // Not matching. + w = w.Next() + continue + } + + // Remove from the bucket and wake the waiter. + woke := w + w = w.Next() // Next iteration. + b.waiters.Remove(woke) + woke.C <- struct{}{} + + // NOTE: The above channel write establishes a write barrier + // according to the memory model, so nothing may be ordered + // around it. Since we've dequeued w and will never touch it + // again, we can safely store 1 to w.complete here and allow + // the WaitComplete() to short-circuit grabbing the bucket + // lock. If they somehow miss the w.complete, we are still + // holding the lock, so we can know that they won't dequeue w, + // assume it's free and have the below operation afterwards. + atomic.StoreInt32(&woke.complete, 1) + done++ + } + return done +} + +// requeueLocked takes n waiters from the bucket and moves them to naddr on the +// bucket "to". +// +// Preconditions: b and to must be locked. +func (b *bucket) requeueLocked(to *bucket, addr, naddr uintptr, n int) int { + done := 0 + for w := b.waiters.Front(); done < n && w != nil; { + if w.addr != addr { + // Not matching. + w = w.Next() + continue + } + + requeued := w + w = w.Next() // Next iteration. + b.waiters.Remove(requeued) + atomic.StoreUintptr(&requeued.addr, naddr) + to.waiters.PushBack(requeued) + done++ + } + return done +} + +const ( + // bucketCount is the number of buckets per Manager. By having many of + // these we reduce contention when concurrent yet unrelated calls are made. + bucketCount = 1 << bucketCountBits + bucketCountBits = 10 +) + +func checkAddr(addr uintptr) error { + // Ensure the address is aligned. + // It must be a DWORD boundary. + if addr&0x3 != 0 { + return syserror.EINVAL + } + + return nil +} + +// bucketIndexForAddr returns the index into Manager.buckets for addr. +func bucketIndexForAddr(addr uintptr) uintptr { + // - The bottom 2 bits of addr must be 0, per checkAddr. + // + // - On amd64, the top 16 bits of addr (bits 48-63) must be equal to bit 47 + // for a canonical address, and (on all existing platforms) bit 47 must be + // 0 for an application address. + // + // Thus 19 bits of addr are "useless" for hashing, leaving only 45 "useful" + // bits. We choose one of the simplest possible hash functions that at + // least uses all 45 useful bits in the output, given that bucketCountBits + // == 10. This hash function also has the property that it will usually map + // adjacent addresses to adjacent buckets, slightly improving memory + // locality when an application synchronization structure uses multiple + // nearby futexes. + // + // Note that despite the large number of arithmetic operations in the + // function, many components can be computed in parallel, such that the + // critical path is 1 bit shift + 3 additions (2 in h1, then h1 + h2). This + // is also why h1 and h2 are grouped separately; for "(addr >> 2) + ... + + // (addr >> 42)" without any additional grouping, the compiler puts all 4 + // additions in the critical path. + h1 := (addr >> 2) + (addr >> 12) + (addr >> 22) + h2 := (addr >> 32) + (addr >> 42) + return (h1 + h2) % bucketCount +} + +// Manager holds futex state for a single virtual address space. +type Manager struct { + buckets [bucketCount]bucket +} + +// NewManager returns an initialized futex manager. +// N.B. we use virtual address to tag futexes, so it only works for private +// (within a single process) futex. +func NewManager() *Manager { + return &Manager{} +} + +// lockBucket returns a locked bucket for the given addr. +// +// Preconditions: checkAddr(addr) == nil. +func (m *Manager) lockBucket(addr uintptr) *bucket { + b := &m.buckets[bucketIndexForAddr(addr)] + b.mu.Lock() + return b +} + +// lockBuckets returns locked buckets for the given addrs. +// +// Preconditions: checkAddr(addr1) == checkAddr(addr2) == nil. +func (m *Manager) lockBuckets(addr1 uintptr, addr2 uintptr) (*bucket, *bucket) { + i1 := bucketIndexForAddr(addr1) + i2 := bucketIndexForAddr(addr2) + b1 := &m.buckets[i1] + b2 := &m.buckets[i2] + + // Ensure that buckets are locked in a consistent order (lowest index + // first) to avoid circular locking. + switch { + case i1 < i2: + b1.mu.Lock() + b2.mu.Lock() + case i2 < i1: + b2.mu.Lock() + b1.mu.Lock() + default: + b1.mu.Lock() + } + + return b1, b2 +} + +// Wake wakes up to n waiters matching the bitmask on the given addr. +// The number of waiters woken is returned. +func (m *Manager) Wake(addr uintptr, bitmask uint32, n int) (int, error) { + if err := checkAddr(addr); err != nil { + return 0, err + } + + b := m.lockBucket(addr) + // This function is very hot; avoid defer. + r := b.wakeLocked(addr, bitmask, n) + b.mu.Unlock() + return r, nil +} + +func (m *Manager) doRequeue(c Checker, addr uintptr, val uint32, naddr uintptr, nwake int, nreq int) (int, error) { + if err := checkAddr(addr); err != nil { + return 0, err + } + if err := checkAddr(naddr); err != nil { + return 0, err + } + + b1, b2 := m.lockBuckets(addr, naddr) + defer b1.mu.Unlock() + if b2 != b1 { + defer b2.mu.Unlock() + } + + // Check our value. + // This only applied for RequeueCmp(). + if c != nil { + if err := c.Check(addr, val); err != nil { + return 0, err + } + } + + // Wake the number required. + done := b1.wakeLocked(addr, ^uint32(0), nwake) + + // Requeue the number required. + b1.requeueLocked(b2, addr, naddr, nreq) + + return done, nil +} + +// Requeue wakes up to nwake waiters on the given addr, and unconditionally +// requeues up to nreq waiters on naddr. +func (m *Manager) Requeue(addr uintptr, naddr uintptr, nwake int, nreq int) (int, error) { + return m.doRequeue(nil, addr, 0, naddr, nwake, nreq) +} + +// RequeueCmp atomically checks that the addr contains val (via the Checker), +// wakes up to nwake waiters on addr and then unconditionally requeues nreq +// waiters on naddr. +func (m *Manager) RequeueCmp(c Checker, addr uintptr, val uint32, naddr uintptr, nwake int, nreq int) (int, error) { + return m.doRequeue(c, addr, val, naddr, nwake, nreq) +} + +// WakeOp atomically applies op to the memory address addr2, wakes up to nwake1 +// waiters unconditionally from addr1, and, based on the original value at addr2 +// and a comparison encoded in op, wakes up to nwake2 waiters from addr2. +// It returns the total number of waiters woken. +func (m *Manager) WakeOp(c Checker, addr1 uintptr, addr2 uintptr, nwake1 int, nwake2 int, op uint32) (int, error) { + if err := checkAddr(addr1); err != nil { + return 0, err + } + if err := checkAddr(addr2); err != nil { + return 0, err + } + + b1, b2 := m.lockBuckets(addr1, addr2) + + done := 0 + cond, err := c.Op(addr2, op) + if err == nil { + // Wake up up to nwake1 entries from the first bucket. + done = b1.wakeLocked(addr1, ^uint32(0), nwake1) + + // Wake up up to nwake2 entries from the second bucket if the + // operation yielded true. + if cond { + done += b2.wakeLocked(addr2, ^uint32(0), nwake2) + } + } + + b1.mu.Unlock() + if b2 != b1 { + b2.mu.Unlock() + } + return done, err +} + +// WaitPrepare atomically checks that addr contains val (via the Checker), then +// enqueues w to be woken by a send to w.C. If WaitPrepare returns nil, the +// Waiter must be subsequently removed by calling WaitComplete, whether or not +// a wakeup is received on w.C. +func (m *Manager) WaitPrepare(w *Waiter, c Checker, addr uintptr, val uint32, bitmask uint32) error { + if err := checkAddr(addr); err != nil { + return err + } + + // Prepare the Waiter before taking the bucket lock. + w.complete = 0 + select { + case <-w.C: + default: + } + w.addr = addr + w.bitmask = bitmask + + b := m.lockBucket(addr) + // This function is very hot; avoid defer. + + // Perform our atomic check. + if err := c.Check(addr, val); err != nil { + b.mu.Unlock() + return err + } + + // Add the waiter to the bucket. + b.waiters.PushBack(w) + + b.mu.Unlock() + return nil +} + +// WaitComplete must be called when a Waiter previously added by WaitPrepare is +// no longer eligible to be woken. +func (m *Manager) WaitComplete(w *Waiter) { + // Can we short-circuit acquiring the lock? + // This is the happy path where a notification + // was received and we don't need to dequeue this + // waiter from any list (or take any locks). + if atomic.LoadInt32(&w.complete) != 0 { + return + } + + // Take the bucket lock. Note that without holding the bucket lock, the + // waiter is not guaranteed to stay in that bucket, so after we take the + // bucket lock, we must ensure that the bucket hasn't changed: if it + // happens to have changed, we release the old bucket lock and try again + // with the new bucket; if it hasn't changed, we know it won't change now + // because we hold the lock. + var b *bucket + for { + addr := atomic.LoadUintptr(&w.addr) + b = m.lockBucket(addr) + // We still have to use an atomic load here, because if w was racily + // requeued then w.addr is not protected by b.mu. + if addr == atomic.LoadUintptr(&w.addr) { + break + } + b.mu.Unlock() + } + + // Remove waiter from the bucket. w.complete can only be stored with b.mu + // locked, so this load doesn't need to use sync/atomic. + if w.complete == 0 { + b.waiters.Remove(w) + } + b.mu.Unlock() +} diff --git a/pkg/sentry/kernel/futex/futex_test.go b/pkg/sentry/kernel/futex/futex_test.go new file mode 100644 index 000000000..7b81358ec --- /dev/null +++ b/pkg/sentry/kernel/futex/futex_test.go @@ -0,0 +1,500 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package futex + +import ( + "math" + "runtime" + "sync" + "sync/atomic" + "syscall" + "testing" + "unsafe" +) + +const ( + testMutexSize = 4 + testMutexLocked uint32 = 1 + testMutexUnlocked uint32 = 0 +) + +// testData implements the Checker interface, and allows us to +// treat the address passed for futex operations as an index in +// a byte slice for testing simplicity. +type testData []byte + +func newTestData(size uint) testData { + return make([]byte, size) +} + +func (t testData) Check(addr uintptr, val uint32) error { + if val != atomic.LoadUint32((*uint32)(unsafe.Pointer(&t[addr]))) { + return syscall.EAGAIN + } + return nil +} + +func (t testData) Op(addr uintptr, val uint32) (bool, error) { + return val == 0, nil +} + +// testMutex ties together a testData slice, an address, and a +// futex manager in order to implement the sync.Locker interface. +// Beyond being used as a Locker, this is a simple mechanism for +// changing the underlying values for simpler tests. +type testMutex struct { + a uintptr + d testData + m *Manager +} + +func newTestMutex(addr uintptr, d testData, m *Manager) *testMutex { + return &testMutex{a: addr, d: d, m: m} +} + +// Lock acquires the testMutex. +// This may wait for it to be available via the futex manager. +func (t *testMutex) Lock() { + for { + // Attempt to grab the lock. + if atomic.CompareAndSwapUint32( + ((*uint32)(unsafe.Pointer(&t.d[t.a]))), + testMutexUnlocked, + testMutexLocked) { + // Lock held. + return + } + + // Wait for it to be "not locked". + w := NewWaiter() + err := t.m.WaitPrepare(w, t.d, t.a, testMutexLocked, ^uint32(0)) + if err == syscall.EAGAIN { + continue + } + if err != nil { + // Should never happen. + panic("WaitPrepare returned unexpected error: " + err.Error()) + } + <-w.C + t.m.WaitComplete(w) + } +} + +// Unlock releases the testMutex. +// This will notify any waiters via the futex manager. +func (t *testMutex) Unlock() { + // Unlock. + atomic.StoreUint32(((*uint32)(unsafe.Pointer(&t.d[t.a]))), testMutexUnlocked) + + // Notify all waiters. + t.m.Wake(t.a, ^uint32(0), math.MaxInt32) +} + +func TestFutexWake(t *testing.T) { + m := NewManager() + d := newTestData(testMutexSize) + + // Wait for it to be locked. + // (This won't trigger the wake in testMutex) + w := NewWaiter() + m.WaitPrepare(w, d, 0, testMutexUnlocked, ^uint32(0)) + + // Wake the single thread. + if _, err := m.Wake(0, ^uint32(0), 1); err != nil { + t.Error("wake error:", err) + } + + <-w.C + m.WaitComplete(w) +} + +func TestFutexWakeBitmask(t *testing.T) { + m := NewManager() + d := newTestData(testMutexSize) + + // Wait for it to be locked. + // (This won't trigger the wake in testMutex) + w := NewWaiter() + m.WaitPrepare(w, d, 0, testMutexUnlocked, 0x0000ffff) + + // Wake the single thread, not using the bitmask. + if _, err := m.Wake(0, 0xffff0000, 1); err != nil { + t.Error("wake non-matching bitmask error:", err) + } + + select { + case <-w.C: + t.Error("w is alive?") + default: + } + + // Now use a matching bitmask. + if _, err := m.Wake(0, 0x00000001, 1); err != nil { + t.Error("wake matching bitmask error:", err) + } + + <-w.C + m.WaitComplete(w) +} + +func TestFutexWakeTwo(t *testing.T) { + m := NewManager() + d := newTestData(testMutexSize) + + // Wait for it to be locked. + // (This won't trigger the wake in testMutex) + w1 := NewWaiter() + w2 := NewWaiter() + w3 := NewWaiter() + m.WaitPrepare(w1, d, 0, testMutexUnlocked, ^uint32(0)) + m.WaitPrepare(w2, d, 0, testMutexUnlocked, ^uint32(0)) + m.WaitPrepare(w3, d, 0, testMutexUnlocked, ^uint32(0)) + + // Wake exactly two threads. + if _, err := m.Wake(0, ^uint32(0), 2); err != nil { + t.Error("wake error:", err) + } + + // Ensure exactly two are alive. + // We don't get guarantees about exactly which two, + // (although we expect them to be w1 and w2). + awake := 0 + for { + select { + case <-w1.C: + awake++ + case <-w2.C: + awake++ + case <-w3.C: + awake++ + default: + if awake != 2 { + t.Error("awake != 2?") + } + + // Success. + return + } + } +} + +func TestFutexWakeUnrelated(t *testing.T) { + m := NewManager() + d := newTestData(2 * testMutexSize) + + // Wait for it to be locked. + w1 := NewWaiter() + w2 := NewWaiter() + m.WaitPrepare(w1, d, 0*testMutexSize, testMutexUnlocked, ^uint32(0)) + m.WaitPrepare(w2, d, 1*testMutexSize, testMutexUnlocked, ^uint32(0)) + + // Wake only the second one. + if _, err := m.Wake(1*testMutexSize, ^uint32(0), 2); err != nil { + t.Error("wake error:", err) + } + + // Ensure only r2 is alive. + select { + case <-w1.C: + t.Error("w1 is alive?") + default: + } + <-w2.C +} + +// This function was shamelessly stolen from mutex_test.go. +func HammerMutex(l sync.Locker, loops int, cdone chan bool) { + for i := 0; i < loops; i++ { + l.Lock() + runtime.Gosched() + l.Unlock() + } + cdone <- true +} + +func TestFutexStress(t *testing.T) { + m := NewManager() + d := newTestData(testMutexSize) + tm := newTestMutex(0*testMutexSize, d, m) + c := make(chan bool) + + for i := 0; i < 10; i++ { + go HammerMutex(tm, 1000, c) + } + + for i := 0; i < 10; i++ { + <-c + } +} + +func TestWakeOpEmpty(t *testing.T) { + m := NewManager() + d := newTestData(8) + + n, err := m.WakeOp(d, 0, 4, 10, 10, 0) + if err != nil { + t.Fatalf("WakeOp failed: %v", err) + } + + if n != 0 { + t.Fatalf("Invalid number of wakes: want 0, got %d", n) + } +} + +func TestWakeOpFirstNonEmpty(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address 0. + w1 := NewWaiter() + if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w1) + + w2 := NewWaiter() + if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w2) + + // Wake up all waiters on address 0. + n, err := m.WakeOp(d, 0, 4, 10, 10, 0) + if err != nil { + t.Fatalf("WakeOp failed: %v", err) + } + + if n != 2 { + t.Fatalf("Invalid number of wakes: want 2, got %d", n) + } +} + +func TestWakeOpSecondNonEmpty(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address 4. + w1 := NewWaiter() + if err := m.WaitPrepare(w1, d, 4, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w1) + + w2 := NewWaiter() + if err := m.WaitPrepare(w2, d, 4, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w2) + + // Wake up all waiters on address 4. + n, err := m.WakeOp(d, 0, 4, 10, 10, 0) + if err != nil { + t.Fatalf("WakeOp failed: %v", err) + } + + if n != 2 { + t.Fatalf("Invalid number of wakes: want 2, got %d", n) + } +} + +func TestWakeOpSecondNonEmptyFailingOp(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address 4. + w1 := NewWaiter() + if err := m.WaitPrepare(w1, d, 4, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w1) + + w2 := NewWaiter() + if err := m.WaitPrepare(w2, d, 4, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w2) + + // Wake up all waiters on address 4. + n, err := m.WakeOp(d, 0, 4, 10, 10, 1) + if err != nil { + t.Fatalf("WakeOp failed: %v", err) + } + + if n != 0 { + t.Fatalf("Invalid number of wakes: want 0, got %d", n) + } +} + +func TestWakeOpAllNonEmpty(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address 0. + w1 := NewWaiter() + if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w1) + + w2 := NewWaiter() + if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w2) + + // Add two waiters on address 4. + w3 := NewWaiter() + if err := m.WaitPrepare(w3, d, 4, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w3) + + w4 := NewWaiter() + if err := m.WaitPrepare(w4, d, 4, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w4) + + // Wake up all waiters on both addresses. + n, err := m.WakeOp(d, 0, 4, 10, 10, 0) + if err != nil { + t.Fatalf("WakeOp failed: %v", err) + } + + if n != 4 { + t.Fatalf("Invalid number of wakes: want 4, got %d", n) + } +} + +func TestWakeOpAllNonEmptyFailingOp(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address 0. + w1 := NewWaiter() + if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w1) + + w2 := NewWaiter() + if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w2) + + // Add two waiters on address 4. + w3 := NewWaiter() + if err := m.WaitPrepare(w3, d, 4, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w3) + + w4 := NewWaiter() + if err := m.WaitPrepare(w4, d, 4, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w4) + + // Wake up all waiters on both addresses. + n, err := m.WakeOp(d, 0, 4, 10, 10, 1) + if err != nil { + t.Fatalf("WakeOp failed: %v", err) + } + + if n != 2 { + t.Fatalf("Invalid number of wakes: want 2, got %d", n) + } +} + +func TestWakeOpSameAddress(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add four waiters on address 0. + w1 := NewWaiter() + if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w1) + + w2 := NewWaiter() + if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w2) + + w3 := NewWaiter() + if err := m.WaitPrepare(w3, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w3) + + w4 := NewWaiter() + if err := m.WaitPrepare(w4, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w4) + + // Use the same address, with one at most one waiter from each. + n, err := m.WakeOp(d, 0, 0, 1, 1, 0) + if err != nil { + t.Fatalf("WakeOp failed: %v", err) + } + + if n != 2 { + t.Fatalf("Invalid number of wakes: want 2, got %d", n) + } +} + +func TestWakeOpSameAddressFailingOp(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add four waiters on address 0. + w1 := NewWaiter() + if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w1) + + w2 := NewWaiter() + if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w2) + + w3 := NewWaiter() + if err := m.WaitPrepare(w3, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w3) + + w4 := NewWaiter() + if err := m.WaitPrepare(w4, d, 0, 0, ^uint32(0)); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + defer m.WaitComplete(w4) + + // Use the same address, with one at most one waiter from each. + n, err := m.WakeOp(d, 0, 0, 1, 1, 1) + if err != nil { + t.Fatalf("WakeOp failed: %v", err) + } + + if n != 1 { + t.Fatalf("Invalid number of wakes: want 1, got %d", n) + } +} diff --git a/pkg/sentry/kernel/g3doc/run_states.dot b/pkg/sentry/kernel/g3doc/run_states.dot new file mode 100644 index 000000000..7861fe1f5 --- /dev/null +++ b/pkg/sentry/kernel/g3doc/run_states.dot @@ -0,0 +1,99 @@ +digraph { + subgraph { + App; + } + subgraph { + Interrupt; + InterruptAfterSignalDeliveryStop; + } + subgraph { + Syscall; + SyscallAfterPtraceEventSeccomp; + SyscallEnter; + SyscallAfterSyscallEnterStop; + SyscallAfterSysemuStop; + SyscallInvoke; + SyscallAfterPtraceEventClone; + SyscallAfterExecStop; + SyscallAfterVforkStop; + SyscallReinvoke; + SyscallExit; + } + subgraph { + Vsyscall; + VsyscallAfterPtraceEventSeccomp; + VsyscallInvoke; + } + subgraph { + Exit; + ExitMain; // leave thread group, release resources, reparent children, kill PID namespace and wait if TGID 1 + ExitNotify; // signal parent/tracer, become waitable + ExitDone; // represented by t.runState == nil + } + + // Task exit + Exit -> ExitMain; + ExitMain -> ExitNotify; + ExitNotify -> ExitDone; + + // Execution of untrusted application code + App -> App; + + // Interrupts (usually signal delivery) + App -> Interrupt; + Interrupt -> Interrupt; // if other interrupt conditions may still apply + Interrupt -> Exit; // if killed + + // Syscalls + App -> Syscall; + Syscall -> SyscallEnter; + SyscallEnter -> SyscallInvoke; + SyscallInvoke -> SyscallExit; + SyscallExit -> App; + + // exit, exit_group + SyscallInvoke -> Exit; + + // execve + SyscallInvoke -> SyscallAfterExecStop; + SyscallAfterExecStop -> SyscallExit; + SyscallAfterExecStop -> App; // fatal signal pending + + // vfork + SyscallInvoke -> SyscallAfterVforkStop; + SyscallAfterVforkStop -> SyscallExit; + + // Vsyscalls + App -> Vsyscall; + Vsyscall -> VsyscallInvoke; + Vsyscall -> App; // fault while reading return address from stack + VsyscallInvoke -> App; + + // ptrace-specific branches + Interrupt -> InterruptAfterSignalDeliveryStop; + InterruptAfterSignalDeliveryStop -> Interrupt; + SyscallEnter -> SyscallAfterSyscallEnterStop; + SyscallAfterSyscallEnterStop -> SyscallInvoke; + SyscallAfterSyscallEnterStop -> SyscallExit; // skipped by tracer + SyscallAfterSyscallEnterStop -> App; // fatal signal pending + SyscallEnter -> SyscallAfterSysemuStop; + SyscallAfterSysemuStop -> SyscallExit; + SyscallAfterSysemuStop -> App; // fatal signal pending + SyscallInvoke -> SyscallAfterPtraceEventClone; + SyscallAfterPtraceEventClone -> SyscallExit; + SyscallAfterPtraceEventClone -> SyscallAfterVforkStop; + + // seccomp + Syscall -> App; // SECCOMP_RET_TRAP, SECCOMP_RET_ERRNO, SECCOMP_RET_KILL, SECCOMP_RET_TRACE without tracer + Syscall -> SyscallAfterPtraceEventSeccomp; // SECCOMP_RET_TRACE + SyscallAfterPtraceEventSeccomp -> SyscallEnter; + SyscallAfterPtraceEventSeccomp -> SyscallExit; // skipped by tracer + SyscallAfterPtraceEventSeccomp -> App; // fatal signal pending + Vsyscall -> VsyscallAfterPtraceEventSeccomp; + VsyscallAfterPtraceEventSeccomp -> VsyscallInvoke; + VsyscallAfterPtraceEventSeccomp -> App; + + // Autosave + SyscallInvoke -> SyscallReinvoke; + SyscallReinvoke -> SyscallInvoke; +} diff --git a/pkg/sentry/kernel/ipc_namespace.go b/pkg/sentry/kernel/ipc_namespace.go new file mode 100644 index 000000000..78737f58f --- /dev/null +++ b/pkg/sentry/kernel/ipc_namespace.go @@ -0,0 +1,43 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/semaphore" +) + +// IPCNamespace represents an IPC namespace. +type IPCNamespace struct { + semaphores *semaphore.Registry +} + +// NewIPCNamespace creates a new IPC namespace. +func NewIPCNamespace() *IPCNamespace { + return &IPCNamespace{ + semaphores: semaphore.NewRegistry(), + } +} + +// SemaphoreRegistry returns the semanphore set registry for this namespace. +func (i *IPCNamespace) SemaphoreRegistry() *semaphore.Registry { + return i.semaphores +} + +// IPCNamespace returns the task's IPC namespace. +func (t *Task) IPCNamespace() *IPCNamespace { + t.mu.Lock() + defer t.mu.Unlock() + return t.ipcns +} diff --git a/pkg/sentry/kernel/kdefs/BUILD b/pkg/sentry/kernel/kdefs/BUILD new file mode 100644 index 000000000..b6c00042a --- /dev/null +++ b/pkg/sentry/kernel/kdefs/BUILD @@ -0,0 +1,10 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library") + +go_library( + name = "kdefs", + srcs = ["kdefs.go"], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", + visibility = ["//:sandbox"], +) diff --git a/pkg/sentry/kernel/kdefs/kdefs.go b/pkg/sentry/kernel/kdefs/kdefs.go new file mode 100644 index 000000000..bbb476544 --- /dev/null +++ b/pkg/sentry/kernel/kdefs/kdefs.go @@ -0,0 +1,20 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package kdefs defines common kernel definitions. +// +package kdefs + +// FD is a File Descriptor. +type FD int32 diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go new file mode 100644 index 000000000..0932965e0 --- /dev/null +++ b/pkg/sentry/kernel/kernel.go @@ -0,0 +1,957 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package kernel provides an emulation of the Linux kernel. +// +// See README.md for a detailed overview. +// +// Lock order (outermost locks must be taken first): +// +// Kernel.extMu +// TaskSet.mu +// SignalHandlers.mu +// Task.mu +// +// Locking SignalHandlers.mu in multiple SignalHandlers requires locking +// TaskSet.mu exclusively first. Locking Task.mu in multiple Tasks at the same +// time requires locking all of their signal mutexes first. +package kernel + +import ( + "fmt" + "io" + "path/filepath" + "sync" + "sync/atomic" + "time" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/cpuid" + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/timerfd" + "gvisor.googlesource.com/gvisor/pkg/sentry/hostcpu" + "gvisor.googlesource.com/gvisor/pkg/sentry/inet" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/limits" + "gvisor.googlesource.com/gvisor/pkg/sentry/loader" + "gvisor.googlesource.com/gvisor/pkg/sentry/mm" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/port" + sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" + "gvisor.googlesource.com/gvisor/pkg/state" +) + +// Kernel represents an emulated Linux kernel. It must be initialized by calling +// Init() or LoadFrom(). +type Kernel struct { + // extMu serializes external changes to the Kernel with calls to + // Kernel.SaveTo. (Kernel.SaveTo requires that the state of the Kernel + // remains frozen for the duration of the call; it requires that the Kernel + // is paused as a precondition, which ensures that none of the tasks + // running within the Kernel can affect its state, but extMu is required to + // ensure that concurrent users of the Kernel *outside* the Kernel's + // control cannot affect its state by calling e.g. + // Kernel.SendExternalSignal.) + extMu sync.Mutex `state:"nosave"` + + // started is true if Start has been called. Unless otherwise specified, + // all Kernel fields become immutable once started becomes true. + started bool `state:"nosave"` + + // All of the following fields are immutable unless otherwise specified. + + // Platform is the platform that is used to execute tasks in the + // created Kernel. It is embedded so that Kernel can directly serve as + // Platform in mm logic and also serve as platform.MemoryProvider in + // filemem S/R logic. + platform.Platform `state:"nosave"` + + // See InitKernelArgs for the meaning of these fields. + featureSet *cpuid.FeatureSet + timekeeper *Timekeeper + tasks *TaskSet + rootUserNamespace *auth.UserNamespace + networkStack inet.Stack `state:"nosave"` + applicationCores uint + useHostCores bool + extraAuxv []arch.AuxEntry + vdso *loader.VDSO + rootUTSNamespace *UTSNamespace + rootIPCNamespace *IPCNamespace + + // mounts holds the state of the virtual filesystem. mounts is initially + // nil, and must be set by calling Kernel.SetRootMountNamespace before + // Kernel.CreateProcess can succeed. + mounts *fs.MountNamespace + + // globalInit is the thread group whose leader has ID 1 in the root PID + // namespace. globalInit is stored separately so that it is accessible even + // after all tasks in the thread group have exited, such that ID 1 is no + // longer mapped. + // + // globalInit is mutable until it is assigned by the first successful call + // to CreateProcess, and is protected by extMu. + globalInit *ThreadGroup + + // realtimeClock is a ktime.Clock based on timekeeper's Realtime. + realtimeClock *timekeeperClock + + // monotonicClock is a ktime.Clock based on timekeeper's Monotonic. + monotonicClock *timekeeperClock + + // syslog is the kernel log. + syslog syslog + + // cpuClock is incremented every linux.ClockTick. cpuClock is used to + // measure task CPU usage, since sampling monotonicClock twice on every + // syscall turns out to be unreasonably expensive. This is similar to how + // Linux does task CPU accounting on x86 (CONFIG_IRQ_TIME_ACCOUNTING), + // although Linux also uses scheduler timing information to improve + // resolution (kernel/sched/cputime.c:cputime_adjust()), which we can't do + // since "preeemptive" scheduling is managed by the Go runtime, which + // doesn't provide this information. + // + // cpuClock is mutable, and is accessed using atomic memory operations. + cpuClock uint64 + + // cpuClockTicker increments cpuClock. + cpuClockTicker *ktime.Timer `state:"nosave"` + + // fdMapUids is an ever-increasing counter for generating FDMap uids. + // + // fdMapUids is mutable, and is accessed using atomic memory operations. + fdMapUids uint64 + + // uniqueID is used to generate unique identifiers. + // + // uniqueID is mutable, and is accessed using atomic memory operations. + uniqueID uint64 + + // nextInotifyCookie is a monotonically increasing counter used for + // generating unique inotify event cookies. + // + // nextInotifyCookie is mutable, and is accesed using atomic memory + // operations. + nextInotifyCookie uint32 + + // netlinkPorts manages allocation of netlink socket port IDs. + netlinkPorts *port.Manager + + // exitErr is the error causing the sandbox to exit, if any. It is + // protected by extMu. + exitErr error +} + +// InitKernelArgs holds arguments to Init. +type InitKernelArgs struct { + // FeatureSet is the emulated CPU feature set. + FeatureSet *cpuid.FeatureSet + + // Timekeeper manages time for all tasks in the system. + Timekeeper *Timekeeper + + // RootUserNamespace is the root user namespace. + RootUserNamespace *auth.UserNamespace + + // NetworkStack is the TCP/IP network stack. NetworkStack may be nil. + NetworkStack inet.Stack + + // ApplicationCores is the number of logical CPUs visible to sandboxed + // applications. The set of logical CPU IDs is [0, ApplicationCores); thus + // ApplicationCores is analogous to Linux's nr_cpu_ids, the index of the + // most significant bit in cpu_possible_mask + 1. + ApplicationCores uint + + // If UseHostCores is true, Task.CPU() returns the task goroutine's CPU + // instead of a virtualized CPU number, and Task.CopyToCPUMask() is a + // no-op. If ApplicationCores is less than hostcpu.MaxPossibleCPU(), it + // will be overridden. + UseHostCores bool + + // ExtraAuxv contains additional auxiliary vector entries that are added to + // each process by the ELF loader. + ExtraAuxv []arch.AuxEntry + + // Vdso holds the VDSO and its parameter page. + Vdso *loader.VDSO + + // RootUTSNamespace is the root UTS namepsace. + RootUTSNamespace *UTSNamespace + + // RootIPCNamespace is the root IPC namepsace. + RootIPCNamespace *IPCNamespace +} + +// Init initialize the Kernel with no tasks. +// +// Callers must manually set Kernel.Platform before caling Init. +func (k *Kernel) Init(args InitKernelArgs) error { + if args.FeatureSet == nil { + return fmt.Errorf("FeatureSet is nil") + } + if args.Timekeeper == nil { + return fmt.Errorf("Timekeeper is nil") + } + if args.RootUserNamespace == nil { + return fmt.Errorf("RootUserNamespace is nil") + } + if args.ApplicationCores == 0 { + return fmt.Errorf("ApplicationCores is 0") + } + + k.featureSet = args.FeatureSet + k.timekeeper = args.Timekeeper + k.tasks = newTaskSet() + k.rootUserNamespace = args.RootUserNamespace + k.rootUTSNamespace = args.RootUTSNamespace + k.rootIPCNamespace = args.RootIPCNamespace + k.networkStack = args.NetworkStack + k.applicationCores = args.ApplicationCores + if args.UseHostCores { + k.useHostCores = true + maxCPU, err := hostcpu.MaxPossibleCPU() + if err != nil { + return fmt.Errorf("Failed to get maximum CPU number: %v", err) + } + minAppCores := uint(maxCPU) + 1 + if k.applicationCores < minAppCores { + log.Infof("UseHostCores enabled: increasing ApplicationCores from %d to %d", k.applicationCores, minAppCores) + k.applicationCores = minAppCores + } + } + k.extraAuxv = args.ExtraAuxv + k.vdso = args.Vdso + k.realtimeClock = &timekeeperClock{tk: args.Timekeeper, c: sentrytime.Realtime} + k.monotonicClock = &timekeeperClock{tk: args.Timekeeper, c: sentrytime.Monotonic} + k.netlinkPorts = port.New() + + return nil +} + +// SaveTo saves the state of k to w. +// +// Preconditions: The kernel must be paused throughout the call to SaveTo. +func (k *Kernel) SaveTo(w io.Writer) error { + saveStart := time.Now() + ctx := k.SupervisorContext() + + // Do not allow other Kernel methods to affect it while it's being saved. + k.extMu.Lock() + defer k.extMu.Unlock() + + // Stop time. + k.pauseTimeLocked() + defer k.resumeTimeLocked() + + // Flush write operations on open files so data reaches backing storage. + if err := k.tasks.flushWritesToFiles(ctx); err != nil { + return err + } + + // Remove all epoll waiter objects from underlying wait queues. + // NOTE: for programs to resume execution in future snapshot scenarios, + // we will need to re-establish these waiter objects after saving. + k.tasks.unregisterEpollWaiters() + + // Clear the dirent cache before saving because Dirents must be Loaded in a + // particular order (parents before children), and Loading dirents from a cache + // breaks that order. + k.mounts.FlushMountSourceRefs() + + // Ensure that all pending asynchronous work is complete: + // - inode and mount release + // - asynchronuous IO + fs.AsyncBarrier() + + // Once all fs work has completed (flushed references have all been released), + // reset mount mappings. This allows individual mounts to save how inodes map + // to filesystem resources. Without this, fs.Inodes cannot be restored. + fs.SaveInodeMappings() + + // Discard unsavable mappings, such as those for host file descriptors. + // This must be done after waiting for "asynchronous fs work", which + // includes async I/O that may touch application memory. + if err := k.invalidateUnsavableMappings(ctx); err != nil { + return fmt.Errorf("failed to invalidate unsavable mappings: %v", err) + } + + // Save the kernel state. + kernelStart := time.Now() + var stats state.Stats + if err := state.Save(w, k, &stats); err != nil { + return err + } + log.Infof("Kernel save stats: %s", &stats) + log.Infof("Kernel save took [%s].", time.Since(kernelStart)) + + // Save the memory state. + // + // FIXME: In the future, this should not be dispatched via + // an abstract memory type. This should be dispatched to a single + // memory implementation that belongs to the kernel. (There is + // currently a single implementation anyways, it just needs to be + // "unabstracted" and reparented appropriately.) + memoryStart := time.Now() + if err := k.Platform.Memory().SaveTo(w); err != nil { + return err + } + log.Infof("Memory save took [%s].", time.Since(memoryStart)) + + log.Infof("Overall save took [%s].", time.Since(saveStart)) + + return nil +} + +func (ts *TaskSet) flushWritesToFiles(ctx context.Context) error { + ts.mu.RLock() + defer ts.mu.RUnlock() + for t := range ts.Root.tids { + if fdmap := t.FDMap(); fdmap != nil { + for _, desc := range fdmap.files { + if flags := desc.file.Flags(); !flags.Write { + continue + } + if sattr := desc.file.Dirent.Inode.StableAttr; !fs.IsFile(sattr) && !fs.IsDir(sattr) { + continue + } + // Here we need all metadata synced. + syncErr := desc.file.Fsync(ctx, 0, fs.FileMaxOffset, fs.SyncAll) + if err := fs.SaveFileFsyncError(syncErr); err != nil { + name, _ := desc.file.Dirent.FullName(nil /* root */) + return fmt.Errorf("%q was not sufficiently synced: %v", name, err) + } + } + } + } + return nil +} + +// Preconditions: The kernel must be paused. +func (k *Kernel) invalidateUnsavableMappings(ctx context.Context) error { + invalidated := make(map[*mm.MemoryManager]struct{}) + k.tasks.mu.RLock() + defer k.tasks.mu.RUnlock() + for t := range k.tasks.Root.tids { + // We can skip locking Task.mu here since the kernel is paused. + if mm := t.tc.MemoryManager; mm != nil { + if _, ok := invalidated[mm]; !ok { + if err := mm.InvalidateUnsavable(ctx); err != nil { + return err + } + invalidated[mm] = struct{}{} + } + } + // I really wish we just had a sync.Map of all MMs... + if r, ok := t.runState.(*runSyscallAfterExecStop); ok { + if err := r.tc.MemoryManager.InvalidateUnsavable(ctx); err != nil { + return err + } + } + } + return nil +} + +func (ts *TaskSet) unregisterEpollWaiters() { + ts.mu.RLock() + defer ts.mu.RUnlock() + for t := range ts.Root.tids { + if fdmap := t.FDMap(); fdmap != nil { + for _, desc := range fdmap.files { + if desc.file != nil { + if e, ok := desc.file.FileOperations.(*epoll.EventPoll); ok { + e.UnregisterEpollWaiters() + } + } + } + } + } +} + +// LoadFrom returns a new Kernel loaded from args. +func (k *Kernel) LoadFrom(r io.Reader, p platform.Platform, net inet.Stack) error { + loadStart := time.Now() + if p == nil { + return fmt.Errorf("Platform is nil") + } + + k.Platform = p + k.networkStack = net + + initAppCores := k.applicationCores + + // Load the kernel state. + kernelStart := time.Now() + var stats state.Stats + if err := state.Load(r, k, &stats); err != nil { + return err + } + log.Infof("Kernel load stats: %s", &stats) + log.Infof("Kernel load took [%s].", time.Since(kernelStart)) + + // Load the memory state. + // + // See the note in SaveTo. + memoryStart := time.Now() + if err := k.Platform.Memory().LoadFrom(r); err != nil { + return err + } + log.Infof("Memory load took [%s].", time.Since(memoryStart)) + + // Ensure that all pending asynchronous work is complete: + // - namedpipe opening + // - inode file opening + fs.AsyncBarrier() + + log.Infof("Overall load took [%s]", time.Since(loadStart)) + + // Applications may size per-cpu structures based on k.applicationCores, so + // it can't change across save/restore. When we are virtualizing CPU + // numbers, this isn't a problem. However, when we are exposing host CPU + // assignments, we can't tolerate an increase in the number of host CPUs, + // which could result in getcpu(2) returning CPUs that applications expect + // not to exist. + if k.useHostCores && initAppCores > k.applicationCores { + return fmt.Errorf("UseHostCores enabled: can't increase ApplicationCores from %d to %d after restore", k.applicationCores, initAppCores) + } + + return nil +} + +// Destroy releases resources owned by k. +// +// Preconditions: There must be no task goroutines running in k. +func (k *Kernel) Destroy() { + if k.mounts != nil { + k.mounts.DecRef() + k.mounts = nil + } +} + +// UniqueID returns a unique identifier. +func (k *Kernel) UniqueID() uint64 { + id := atomic.AddUint64(&k.uniqueID, 1) + if id == 0 { + panic("unique identifier generator wrapped around") + } + return id +} + +// CreateProcessArgs holds arguments to kernel.CreateProcess. +type CreateProcessArgs struct { + // Filename is the filename to load. + // + // If this is provided as "", then the file will be guessed via Argv[0]. + Filename string + + // Argvv is a list of arguments. + Argv []string + + // Envv is a list of environment variables. + Envv []string + + // WorkingDirectory is the initial working directory. + // + // This defaults to the root if empty. + WorkingDirectory string + + // Credentials is the initial credentials. + Credentials *auth.Credentials + + // FDMap is the initial set of file descriptors. If CreateProcess succeeds, + // it takes a reference on FDMap. + FDMap *FDMap + + // Umask is the initial umask. + Umask uint + + // Limits is the initial resource limits. + Limits *limits.LimitSet + + // MaxSymlinkTraversals is the maximum number of symlinks to follow + // during resolution. + MaxSymlinkTraversals uint + + // UTSNamespace is the initial UTS namespace. + UTSNamespace *UTSNamespace + + // IPCNamespace is the initial IPC namespace. + IPCNamespace *IPCNamespace +} + +// NewContext returns a context.Context that represents the task that will be +// created by args.NewContext(k). +func (args *CreateProcessArgs) NewContext(k *Kernel) *createProcessContext { + return &createProcessContext{ + Logger: log.Log(), + k: k, + args: args, + } +} + +// createProcessContext is a context.Context that represents the context +// associated with a task that is being created. +type createProcessContext struct { + context.NoopSleeper + log.Logger + k *Kernel + args *CreateProcessArgs +} + +// Value implements context.Context.Value. +func (ctx *createProcessContext) Value(key interface{}) interface{} { + switch key { + case CtxKernel: + return ctx.k + case CtxPIDNamespace: + // "The new task ... is in the root PID namespace." - + // Kernel.CreateProcess + return ctx.k.tasks.Root + case CtxUTSNamespace: + return ctx.args.UTSNamespace + case CtxIPCNamespace: + return ctx.args.IPCNamespace + case auth.CtxCredentials: + return ctx.args.Credentials + case fs.CtxRoot: + if ctx.k.mounts == nil { + return nil + } + return ctx.k.mounts.Root() + case ktime.CtxRealtimeClock: + return ctx.k.RealtimeClock() + case limits.CtxLimits: + return ctx.args.Limits + case platform.CtxPlatform: + return ctx.k + case uniqueid.CtxGlobalUniqueID: + return ctx.k.UniqueID() + case uniqueid.CtxInotifyCookie: + return ctx.k.GenerateInotifyCookie() + default: + return nil + } +} + +// CreateProcess creates a new task in a new thread group with the given +// options. The new task has no parent and is in the root PID namespace. +// +// If k.Start() has already been called, the created task will begin running +// immediately. Otherwise, it will be started when k.Start() is called. +// +// CreateProcess has no analogue in Linux; it is used to create the initial +// application task, as well as processes started by the control server. +func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { + k.extMu.Lock() + defer k.extMu.Unlock() + log.Infof("EXEC: %v", args.Argv) + + if k.mounts == nil { + return nil, fmt.Errorf("no kernel MountNamespace") + } + + tg := NewThreadGroup(k.tasks.Root, NewSignalHandlers(), linux.SIGCHLD, args.Limits, k.monotonicClock) + ctx := args.NewContext(k) + + // Grab the root directory. + root := fs.RootFromContext(ctx) + defer root.DecRef() + + // Grab the working directory. + wd := root // Default. + if args.WorkingDirectory != "" { + var err error + wd, err = k.mounts.FindInode(ctx, root, nil, args.WorkingDirectory, args.MaxSymlinkTraversals) + if err != nil { + return nil, fmt.Errorf("failed to find initial working directory %q: %v", args.WorkingDirectory, err) + } + defer wd.DecRef() + } + + if args.Filename == "" { + // Was anything provided? + if len(args.Argv) == 0 { + return nil, fmt.Errorf("no filename or command provided") + } + if !filepath.IsAbs(args.Argv[0]) { + return nil, fmt.Errorf("'%s' is not an absolute path", args.Argv[0]) + } + args.Filename = args.Argv[0] + } + + // Create a fresh task context. + tc, err := k.LoadTaskImage(ctx, k.mounts, root, wd, args.MaxSymlinkTraversals, args.Filename, args.Argv, args.Envv, k.featureSet) + if err != nil { + return nil, err + } + tr := newTaskResources(args.FDMap, newFSContext(root, wd, args.Umask)) + // NewTask unconditionally takes ownership of tr, so we never have to call + // tr.release. + + // Create the task. + config := &TaskConfig{ + Kernel: k, + ThreadGroup: tg, + TaskContext: tc, + TaskResources: tr, + Credentials: args.Credentials, + UTSNamespace: args.UTSNamespace, + IPCNamespace: args.IPCNamespace, + AllowedCPUMask: sched.NewFullCPUSet(k.applicationCores), + } + t, err := k.tasks.NewTask(config) + if err != nil { + return nil, err + } + + // Success. + if k.started { + tid := k.tasks.Root.IDOfTask(t) + t.Start(tid) + } else if k.globalInit == nil { + k.globalInit = tg + } + return tg, nil +} + +// Start starts execution of all tasks in k. +// +// Preconditions: Start may be called exactly once. +func (k *Kernel) Start() error { + k.extMu.Lock() + defer k.extMu.Unlock() + + if k.globalInit == nil { + return fmt.Errorf("kernel contains no tasks") + } + if k.started { + return fmt.Errorf("kernel already started") + } + + k.started = true + k.cpuClockTicker = ktime.NewTimer(k.monotonicClock, kernelCPUClockListener{k}) + k.cpuClockTicker.Swap(ktime.Setting{ + Enabled: true, + Period: linux.ClockTick, + }) + // If k was created by LoadKernelFrom, timers were stopped during + // Kernel.SaveTo and need to be resumed. If k was created by NewKernel, + // this is a no-op. + k.resumeTimeLocked() + // Start task goroutines. + k.tasks.mu.RLock() + defer k.tasks.mu.RUnlock() + for t, tid := range k.tasks.Root.tids { + t.Start(tid) + } + return nil +} + +// pauseTimeLocked pauses all Timers and Timekeeper updates. +// +// Preconditions: Any task goroutines running in k must be stopped. k.extMu +// must be locked. +func (k *Kernel) pauseTimeLocked() { + // k.cpuClockTicker may be nil since Kernel.SaveTo() may be called before + // Kernel.Start(). + if k.cpuClockTicker != nil { + k.cpuClockTicker.Pause() + } + + // By precondition, nothing else can be interacting with PIDNamespace.tids + // or FDMap.files, so we can iterate them without synchronization. (We + // can't hold the TaskSet mutex when pausing thread group timers because + // thread group timers call ThreadGroup.SendSignal, which takes the TaskSet + // mutex, while holding the Timer mutex.) + for t := range k.tasks.Root.tids { + if t == t.tg.leader { + t.tg.tm.pause() + } + // This means we'll iterate FDMaps shared by multiple tasks repeatedly, + // but ktime.Timer.Pause is idempotent so this is harmless. + if fdm := t.tr.FDMap; fdm != nil { + for _, desc := range fdm.files { + if tfd, ok := desc.file.FileOperations.(*timerfd.TimerOperations); ok { + tfd.PauseTimer() + } + } + } + } + k.timekeeper.PauseUpdates() +} + +// resumeTimeLocked resumes all Timers and Timekeeper updates. If +// pauseTimeLocked has not been previously called, resumeTimeLocked has no +// effect. +// +// Preconditions: Any task goroutines running in k must be stopped. k.extMu +// must be locked. +func (k *Kernel) resumeTimeLocked() { + if k.cpuClockTicker != nil { + k.cpuClockTicker.Resume() + } + + k.timekeeper.ResumeUpdates() + for t := range k.tasks.Root.tids { + if t == t.tg.leader { + t.tg.tm.resume() + } + if fdm := t.tr.FDMap; fdm != nil { + for _, desc := range fdm.files { + if tfd, ok := desc.file.FileOperations.(*timerfd.TimerOperations); ok { + tfd.ResumeTimer() + } + } + } + } +} + +// WaitExited blocks until all tasks in k have exited. +func (k *Kernel) WaitExited() { + k.tasks.liveGoroutines.Wait() +} + +// Kill requests that all tasks in k immediately exit as if group exiting with +// status es. Kill does not wait for tasks to exit. +func (k *Kernel) Kill(es ExitStatus) { + k.extMu.Lock() + defer k.extMu.Unlock() + k.tasks.Kill(es) +} + +// Pause requests that all tasks in k temporarily stop executing, and blocks +// until all tasks in k have stopped. Multiple calls to Pause nest and require +// an equal number of calls to Unpause to resume execution. +func (k *Kernel) Pause() { + k.extMu.Lock() + k.tasks.BeginExternalStop() + k.extMu.Unlock() + k.tasks.runningGoroutines.Wait() +} + +// Unpause ends the effect of a previous call to Pause. If Unpause is called +// without a matching preceding call to Pause, Unpause may panic. +func (k *Kernel) Unpause() { + k.extMu.Lock() + defer k.extMu.Unlock() + k.tasks.EndExternalStop() +} + +// SendExternalSignal injects a signal into the kernel. +// +// context is used only for debugging to describe how the signal was received. +// +// Returns false if signal could not be sent because the Kernel is not fully +// initialized yet. +func (k *Kernel) SendExternalSignal(info *arch.SignalInfo, context string) bool { + k.extMu.Lock() + defer k.extMu.Unlock() + return k.sendExternalSignal(info, context) +} + +// FeatureSet returns the FeatureSet. +func (k *Kernel) FeatureSet() *cpuid.FeatureSet { + return k.featureSet +} + +// Timekeeper returns the Timekeeper. +func (k *Kernel) Timekeeper() *Timekeeper { + return k.timekeeper +} + +// TaskSet returns the TaskSet. +func (k *Kernel) TaskSet() *TaskSet { + return k.tasks +} + +// RootUserNamespace returns the root UserNamespace. +func (k *Kernel) RootUserNamespace() *auth.UserNamespace { + return k.rootUserNamespace +} + +// RootUTSNamespace returns the root UTSNamespace. +func (k *Kernel) RootUTSNamespace() *UTSNamespace { + return k.rootUTSNamespace +} + +// RootIPCNamespace returns the root IPCNamespace. +func (k *Kernel) RootIPCNamespace() *IPCNamespace { + return k.rootIPCNamespace +} + +// RootMountNamespace returns the MountNamespace. +func (k *Kernel) RootMountNamespace() *fs.MountNamespace { + k.extMu.Lock() + defer k.extMu.Unlock() + return k.mounts +} + +// SetRootMountNamespace sets the MountNamespace. +func (k *Kernel) SetRootMountNamespace(mounts *fs.MountNamespace) { + k.extMu.Lock() + defer k.extMu.Unlock() + k.mounts = mounts +} + +// NetworkStack returns the network stack. NetworkStack may return nil if no +// network stack is available. +func (k *Kernel) NetworkStack() inet.Stack { + return k.networkStack +} + +// GlobalInit returns the thread group with ID 1 in the root PID namespace, or +// nil if no such thread group exists. GlobalInit may return a thread group +// containing no tasks if the thread group has already exited. +func (k *Kernel) GlobalInit() *ThreadGroup { + k.extMu.Lock() + defer k.extMu.Unlock() + return k.globalInit +} + +// ApplicationCores returns the number of CPUs visible to sandboxed +// applications. +func (k *Kernel) ApplicationCores() uint { + return k.applicationCores +} + +// RealtimeClock returns the application CLOCK_REALTIME clock. +func (k *Kernel) RealtimeClock() ktime.Clock { + return k.realtimeClock +} + +// MonotonicClock returns the application CLOCK_MONOTONIC clock. +func (k *Kernel) MonotonicClock() ktime.Clock { + return k.monotonicClock +} + +// CPUClockNow returns the current value of k.cpuClock. +func (k *Kernel) CPUClockNow() uint64 { + return atomic.LoadUint64(&k.cpuClock) +} + +// Syslog returns the syslog. +func (k *Kernel) Syslog() *syslog { + return &k.syslog +} + +// GenerateInotifyCookie generates a unique inotify event cookie. +// +// Returned values may overlap with previously returned values if the value +// space is exhausted. 0 is not a valid cookie value, all other values +// representable in a uint32 are allowed. +func (k *Kernel) GenerateInotifyCookie() uint32 { + id := atomic.AddUint32(&k.nextInotifyCookie, 1) + // Wrap-around is explicitly allowed for inotify event cookies. + if id == 0 { + id = atomic.AddUint32(&k.nextInotifyCookie, 1) + } + return id +} + +// NetlinkPorts returns the netlink port manager. +func (k *Kernel) NetlinkPorts() *port.Manager { + return k.netlinkPorts +} + +// ExitError returns the sandbox error that caused the kernel to exit. +func (k *Kernel) ExitError() error { + k.extMu.Lock() + defer k.extMu.Unlock() + return k.exitErr +} + +// SetExitError sets the sandbox error that caused the kernel to exit, if one is +// not already set. +func (k *Kernel) SetExitError(err error) { + k.extMu.Lock() + defer k.extMu.Unlock() + if k.exitErr == nil { + k.exitErr = err + } +} + +// SupervisorContext returns a Context with maximum privileges in k. It should +// only be used by goroutines outside the control of the emulated kernel +// defined by e. +// +// Callers are responsible for ensuring that the returned Context is not used +// concurrently with changes to the Kernel. +func (k *Kernel) SupervisorContext() context.Context { + return supervisorContext{ + Logger: log.Log(), + k: k, + } +} + +type supervisorContext struct { + context.NoopSleeper + log.Logger + k *Kernel +} + +// Value implements context.Context. +func (ctx supervisorContext) Value(key interface{}) interface{} { + switch key { + case CtxCanTrace: + // The supervisor context can trace anything. (None of + // supervisorContext's users are expected to invoke ptrace, but ptrace + // permissions are required for certain file accesses.) + return func(*Task, bool) bool { return true } + case CtxKernel: + return ctx.k + case CtxPIDNamespace: + return ctx.k.tasks.Root + case CtxUTSNamespace: + return ctx.k.rootUTSNamespace + case CtxIPCNamespace: + return ctx.k.rootIPCNamespace + case auth.CtxCredentials: + // The supervisor context is global root. + return auth.NewRootCredentials(ctx.k.rootUserNamespace) + case fs.CtxRoot: + return ctx.k.mounts.Root() + case ktime.CtxRealtimeClock: + return ctx.k.RealtimeClock() + case limits.CtxLimits: + // No limits apply. + return limits.NewLimitSet() + case platform.CtxPlatform: + return ctx.k + case uniqueid.CtxGlobalUniqueID: + return ctx.k.UniqueID() + case uniqueid.CtxInotifyCookie: + return ctx.k.GenerateInotifyCookie() + default: + return nil + } +} + +type kernelCPUClockListener struct { + k *Kernel +} + +// Notify implements ktime.TimerListener.Notify. +func (l kernelCPUClockListener) Notify(exp uint64) { + atomic.AddUint64(&l.k.cpuClock, exp) +} + +// Destroy implements ktime.TimerListener.Destroy. +func (l kernelCPUClockListener) Destroy() { +} diff --git a/pkg/sentry/kernel/memevent/BUILD b/pkg/sentry/kernel/memevent/BUILD new file mode 100644 index 000000000..c7779e1d5 --- /dev/null +++ b/pkg/sentry/kernel/memevent/BUILD @@ -0,0 +1,31 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") +load("@io_bazel_rules_go//go:def.bzl", "go_library") + +go_library( + name = "memevent", + srcs = ["memory_events.go"], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/memevent", + visibility = ["//:sandbox"], + deps = [ + ":memory_events_go_proto", + "//pkg/eventchannel", + "//pkg/log", + "//pkg/sentry/kernel", + "//pkg/sentry/usage", + ], +) + +proto_library( + name = "memory_events_proto", + srcs = ["memory_events.proto"], + visibility = ["//visibility:public"], +) + +go_proto_library( + name = "memory_events_go_proto", + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/memevent/memory_events_go_proto", + proto = ":memory_events_proto", + visibility = ["//visibility:public"], +) diff --git a/pkg/sentry/kernel/memevent/memory_events.go b/pkg/sentry/kernel/memevent/memory_events.go new file mode 100644 index 000000000..ecc9151de --- /dev/null +++ b/pkg/sentry/kernel/memevent/memory_events.go @@ -0,0 +1,98 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package memevent implements the memory usage events controller, which +// periodically emits events via the eventchannel. +package memevent + +import ( + "sync" + "time" + + "gvisor.googlesource.com/gvisor/pkg/eventchannel" + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + pb "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/memevent/memory_events_go_proto" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" +) + +// MemoryEvents describes the configuration for the global memory event emitter. +type MemoryEvents struct { + k *kernel.Kernel + + // The period is how often to emit an event. The memory events goroutine + // will ensure a minimum of one event is emitted per this period, regardless + // how of much memory usage has changed. + period time.Duration + + // Writing to this channel indicates the memory goroutine should stop. + stop chan struct{} + + // done is used to signal when the memory event goroutine has exited. + done sync.WaitGroup +} + +// New creates a new MemoryEvents. +func New(k *kernel.Kernel, period time.Duration) *MemoryEvents { + return &MemoryEvents{ + k: k, + period: period, + stop: make(chan struct{}), + } +} + +// Stop stops the memory usage events emitter goroutine. Stop must not be called +// concurrently with Start and may only be called once. +func (m *MemoryEvents) Stop() { + close(m.stop) + m.done.Wait() +} + +// Start starts the memory usage events emitter goroutine. Start must not be +// called concurrently with Stop and may only be called once. +func (m *MemoryEvents) Start() { + if m.period == 0 { + return + } + go m.run() // S/R-SAFE: doesn't interact with saved state. +} + +func (m *MemoryEvents) run() { + m.done.Add(1) + + ticker := time.NewTicker(m.period) + defer ticker.Stop() + + for { + select { + case <-m.stop: + m.done.Done() + return + case <-ticker.C: + m.emit() + } + } +} + +func (m *MemoryEvents) emit() { + totalPlatform, err := m.k.Platform.Memory().TotalUsage() + if err != nil { + log.Warningf("Failed to fetch memory usage for memory events: %v", err) + return + } + snapshot, _ := usage.MemoryAccounting.Copy() + total := totalPlatform + snapshot.Mapped + + eventchannel.Emit(&pb.MemoryUsageEvent{Total: total}) +} diff --git a/pkg/sentry/kernel/memevent/memory_events.proto b/pkg/sentry/kernel/memevent/memory_events.proto new file mode 100644 index 000000000..e6e0bd628 --- /dev/null +++ b/pkg/sentry/kernel/memevent/memory_events.proto @@ -0,0 +1,25 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +syntax = "proto3"; + +package gvisor; + +// MemoryUsageEvent describes the memory usage of the sandbox at a single +// instant in time. These messages are emitted periodically on the eventchannel. +message MemoryUsageEvent { + // The total memory usage of the sandboxed application in bytes, calculated + // using the 'fast' method. + uint64 total = 1; +} diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go new file mode 100644 index 000000000..d8701f47a --- /dev/null +++ b/pkg/sentry/kernel/pending_signals.go @@ -0,0 +1,126 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/bits" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" +) + +const ( + // stdSignalCap is the maximum number of instances of a given standard + // signal that may be pending. ("[If] multiple instances of a standard + // signal are delivered while that signal is currently blocked, then only + // one instance is queued.") - signal(7) + stdSignalCap = 1 + + // rtSignalCap is the maximum number of instances of a given realtime + // signal that may be pending. + // + // TODO: In Linux, the minimum signal queue size is + // RLIMIT_SIGPENDING, which is by default max_threads/2. + rtSignalCap = 32 +) + +// pendingSignals holds a collection of pending signals. The zero value of +// pendingSignals is a valid empty collection. pendingSignals is thread-unsafe; +// users must provide synchronization. +type pendingSignals struct { + // signals contains all pending signals. + // + // Note that signals is zero-indexed, but signal 1 is the first valid + // signal, so signals[0] contains signals with signo 1 etc. This offset is + // usually handled by using Signal.index(). + signals [linux.SignalMaximum]pendingSignalQueue + + // Bit i of pendingSet is set iff there is at least one signal with signo + // i+1 pending. + pendingSet linux.SignalSet +} + +// pendingSignalQueue holds a pendingSignalList for a single signal number. +type pendingSignalQueue struct { + pendingSignalList + length int +} + +type pendingSignal struct { + // pendingSignalEntry links into a pendingSignalList. + pendingSignalEntry + *arch.SignalInfo +} + +// enqueue enqueues the given signal. enqueue returns true on success and false +// on failure (if the given signal's queue is full). +// +// Preconditions: info represents a valid signal. +func (p *pendingSignals) enqueue(info *arch.SignalInfo) bool { + sig := linux.Signal(info.Signo) + q := &p.signals[sig.Index()] + if sig.IsStandard() { + if q.length >= stdSignalCap { + return false + } + } else if q.length >= rtSignalCap { + return false + } + q.pendingSignalList.PushBack(&pendingSignal{SignalInfo: info}) + q.length++ + p.pendingSet |= linux.SignalSetOf(sig) + return true +} + +// dequeue dequeues and returns any pending signal not masked by mask. If no +// unmasked signals are pending, dequeue returns nil. +func (p *pendingSignals) dequeue(mask linux.SignalSet) *arch.SignalInfo { + // "Real-time signals are delivered in a guaranteed order. Multiple + // real-time signals of the same type are delivered in the order they were + // sent. If different real-time signals are sent to a process, they are + // delivered starting with the lowest-numbered signal. (I.e., low-numbered + // signals have highest priority.) By contrast, if multiple standard + // signals are pending for a process, the order in which they are delivered + // is unspecified. If both standard and real-time signals are pending for a + // process, POSIX leaves it unspecified which is delivered first. Linux, + // like many other implementations, gives priority to standard signals in + // this case." - signal(7) + lowestPendingUnblockedBit := bits.TrailingZeros64(uint64(p.pendingSet &^ mask)) + if lowestPendingUnblockedBit >= linux.SignalMaximum { + return nil + } + return p.dequeueSpecific(linux.Signal(lowestPendingUnblockedBit + 1)) +} + +func (p *pendingSignals) dequeueSpecific(sig linux.Signal) *arch.SignalInfo { + q := &p.signals[sig.Index()] + ps := q.pendingSignalList.Front() + if ps == nil { + return nil + } + q.pendingSignalList.Remove(ps) + q.length-- + if q.length == 0 { + p.pendingSet &^= linux.SignalSetOf(sig) + } + return ps.SignalInfo +} + +// discardSpecific causes all pending signals with number sig to be discarded. +func (p *pendingSignals) discardSpecific(sig linux.Signal) { + q := &p.signals[sig.Index()] + q.pendingSignalList.Reset() + q.length = 0 + p.pendingSet &^= linux.SignalSetOf(sig) +} diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD new file mode 100644 index 000000000..ca9825f9d --- /dev/null +++ b/pkg/sentry/kernel/pipe/BUILD @@ -0,0 +1,68 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_stateify( + name = "pipe_state", + srcs = [ + "buffers.go", + "node.go", + "pipe.go", + "reader.go", + "reader_writer.go", + "writer.go", + ], + out = "pipe_state.go", + package = "pipe", +) + +go_library( + name = "pipe", + srcs = [ + "buffers.go", + "device.go", + "node.go", + "pipe.go", + "pipe_state.go", + "reader.go", + "reader_writer.go", + "writer.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/pipe", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/abi/linux", + "//pkg/amutex", + "//pkg/ilist", + "//pkg/log", + "//pkg/refs", + "//pkg/sentry/arch", + "//pkg/sentry/context", + "//pkg/sentry/device", + "//pkg/sentry/fs", + "//pkg/sentry/fs/fsutil", + "//pkg/sentry/usermem", + "//pkg/state", + "//pkg/syserror", + "//pkg/waiter", + ], +) + +go_test( + name = "pipe_test", + size = "small", + srcs = [ + "node_test.go", + "pipe_test.go", + ], + embed = [":pipe"], + deps = [ + "//pkg/sentry/context", + "//pkg/sentry/context/contexttest", + "//pkg/sentry/fs", + "//pkg/sentry/usermem", + "//pkg/syserror", + "//pkg/waiter", + ], +) diff --git a/pkg/sentry/kernel/pipe/buffers.go b/pkg/sentry/kernel/pipe/buffers.go new file mode 100644 index 000000000..f300537c5 --- /dev/null +++ b/pkg/sentry/kernel/pipe/buffers.go @@ -0,0 +1,50 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "gvisor.googlesource.com/gvisor/pkg/ilist" +) + +// Buffer encapsulates a queueable byte buffer that can +// easily be truncated. It is designed only for use with pipes. +type Buffer struct { + ilist.Entry + data []byte +} + +// newBuffer initializes a Buffer. +func newBuffer(buf []byte) *Buffer { + return &Buffer{data: buf} +} + +// bytes returns the bytes contained in the buffer. +func (b *Buffer) bytes() []byte { + return b.data +} + +// size returns the number of bytes contained in the buffer. +func (b *Buffer) size() int { + return len(b.data) +} + +// truncate removes the first n bytes from the buffer. +func (b *Buffer) truncate(n int) int { + if n > len(b.data) { + panic("Trying to truncate past end of array.") + } + b.data = b.data[n:] + return len(b.data) +} diff --git a/pkg/sentry/kernel/pipe/device.go b/pkg/sentry/kernel/pipe/device.go new file mode 100644 index 000000000..8d383577a --- /dev/null +++ b/pkg/sentry/kernel/pipe/device.go @@ -0,0 +1,20 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import "gvisor.googlesource.com/gvisor/pkg/sentry/device" + +// pipeDevice is used for all pipe files. +var pipeDevice = device.NewAnonDevice() diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go new file mode 100644 index 000000000..5b47427ef --- /dev/null +++ b/pkg/sentry/kernel/pipe/node.go @@ -0,0 +1,175 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/amutex" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// inodeOperations wraps fs.InodeOperations operations with common pipe opening semantics. +type inodeOperations struct { + fs.InodeOperations + + // mu protects the fields below. + mu sync.Mutex `state:"nosave"` + + // p is the underlying Pipe object representing this fifo. + p *Pipe + + // Channels for synchronizing the creation of new readers and writers of + // this fifo. See waitFor and newHandleLocked. + // + // These are not saved/restored because all waiters are unblocked on save, + // and either automatically restart (via ERESTARTSYS) or return EINTR on + // resume. On restarts via ERESTARTSYS, the appropriate channel will be + // recreated. + rWakeup chan struct{} `state:"nosave"` + wWakeup chan struct{} `state:"nosave"` +} + +// NewInodeOperations creates a new pipe fs.InodeOperations. +func NewInodeOperations(base fs.InodeOperations, p *Pipe) fs.InodeOperations { + return &inodeOperations{ + InodeOperations: base, + p: p, + } +} + +// GetFile implements fs.InodeOperations.GetFile. Named pipes have special blocking +// semantics during open: +// +// "Normally, opening the FIFO blocks until the other end is opened also. A +// process can open a FIFO in nonblocking mode. In this case, opening for +// read-only will succeed even if no-one has opened on the write side yet, +// opening for write-only will fail with ENXIO (no such device or address) +// unless the other end has already been opened. Under Linux, opening a FIFO +// for read and write will succeed both in blocking and nonblocking mode. POSIX +// leaves this behavior undefined. This can be used to open a FIFO for writing +// while there are no readers available." - fifo(7) +func (i *inodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + i.mu.Lock() + defer i.mu.Unlock() + + switch { + case flags.Read && !flags.Write: // O_RDONLY. + r := i.p.ROpen(ctx) + i.newHandleLocked(&i.rWakeup) + + if i.p.isNamed && !flags.NonBlocking && !i.p.HasWriters() { + if !i.waitFor(&i.wWakeup, ctx) { + r.DecRef() + return nil, syserror.ErrInterrupted + } + } + + // By now, either we're doing a nonblocking open or we have a writer. On + // a nonblocking read-only open, the open succeeds even if no-one has + // opened the write side yet. + return r, nil + + case flags.Write && !flags.Read: // O_WRONLY. + w := i.p.WOpen(ctx) + i.newHandleLocked(&i.wWakeup) + + if i.p.isNamed && !i.p.HasReaders() { + // On a nonblocking, write-only open, the open fails with ENXIO if the + // read side isn't open yet. + if flags.NonBlocking { + w.DecRef() + return nil, syserror.ENXIO + } + + if !i.waitFor(&i.rWakeup, ctx) { + w.DecRef() + return nil, syserror.ErrInterrupted + } + } + return w, nil + + case flags.Read && flags.Write: // O_RDWR. + // Pipes opened for read-write always succeeds without blocking. + rw := i.p.RWOpen(ctx) + i.newHandleLocked(&i.rWakeup) + i.newHandleLocked(&i.wWakeup) + return rw, nil + + default: + return nil, syserror.EINVAL + } +} + +// waitFor blocks until the underlying pipe has at least one reader/writer is +// announced via 'wakeupChan', or until 'sleeper' is cancelled. Any call to this +// function will block for either readers or writers, depending on where +// 'wakeupChan' points. +// +// f.mu must be held by the caller. waitFor returns with f.mu held, but it will +// drop f.mu before blocking for any reader/writers. +func (i *inodeOperations) waitFor(wakeupChan *chan struct{}, sleeper amutex.Sleeper) bool { + // Ideally this function would simply use a condition variable. However, the + // wait needs to be interruptible via 'sleeper', so we must sychronize via a + // channel. The synchronization below relies on the fact that closing a + // channel unblocks all receives on the channel. + + // Does an appropriate wakeup channel already exist? If not, create a new + // one. This is all done under f.mu to avoid races. + if *wakeupChan == nil { + *wakeupChan = make(chan struct{}) + } + + // Grab a local reference to the wakeup channel since it may disappear as + // soon as we drop f.mu. + wakeup := *wakeupChan + + // Drop the lock and prepare to sleep. + i.mu.Unlock() + cancel := sleeper.SleepStart() + + // Wait for either a new reader/write to be signalled via 'wakeup', or + // for the sleep to be cancelled. + select { + case <-wakeup: + sleeper.SleepFinish(true) + case <-cancel: + sleeper.SleepFinish(false) + } + + // Take the lock and check if we were woken. If we were woken and + // interrupted, the former takes priority. + i.mu.Lock() + select { + case <-wakeup: + return true + default: + return false + } +} + +// newHandleLocked signals a new pipe reader or writer depending on where +// 'wakeupChan' points. This unblocks any corresponding reader or writer +// waiting for the other end of the channel to be opened, see Fifo.waitFor. +// +// i.mu must be held. +func (*inodeOperations) newHandleLocked(wakeupChan *chan struct{}) { + if *wakeupChan != nil { + close(*wakeupChan) + *wakeupChan = nil + } +} diff --git a/pkg/sentry/kernel/pipe/node_test.go b/pkg/sentry/kernel/pipe/node_test.go new file mode 100644 index 000000000..cc1ebf4f6 --- /dev/null +++ b/pkg/sentry/kernel/pipe/node_test.go @@ -0,0 +1,308 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "testing" + "time" + + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +type sleeper struct { + context.Context + ch chan struct{} +} + +func newSleeperContext(t *testing.T) context.Context { + return &sleeper{ + Context: contexttest.Context(t), + ch: make(chan struct{}), + } +} + +func (s *sleeper) SleepStart() <-chan struct{} { + return s.ch +} + +func (s *sleeper) SleepFinish(bool) { +} + +func (s *sleeper) Cancel() { + s.ch <- struct{}{} +} + +type openResult struct { + *fs.File + error +} + +func testOpenOrDie(ctx context.Context, t *testing.T, n fs.InodeOperations, flags fs.FileFlags, doneChan chan<- struct{}) (*fs.File, error) { + file, err := n.GetFile(ctx, nil, flags) + if err != nil { + t.Fatalf("open with flags %+v failed: %v", flags, err) + } + if doneChan != nil { + doneChan <- struct{}{} + } + return file, err +} + +func testOpen(ctx context.Context, t *testing.T, n fs.InodeOperations, flags fs.FileFlags, resChan chan<- openResult) (*fs.File, error) { + file, err := n.GetFile(ctx, nil, flags) + if resChan != nil { + resChan <- openResult{file, err} + } + return file, err +} + +func newNamedPipe(t *testing.T) *Pipe { + return NewPipe(contexttest.Context(t), true, DefaultPipeSize, usermem.PageSize) +} + +func newAnonPipe(t *testing.T) *Pipe { + return NewPipe(contexttest.Context(t), false, DefaultPipeSize, usermem.PageSize) +} + +// assertRecvBlocks ensures that a recv attempt on c blocks for at least +// blockDuration. This is useful for checking that a goroutine that is supposed +// to be executing a blocking operation is actually blocking. +func assertRecvBlocks(t *testing.T, c <-chan struct{}, blockDuration time.Duration, failMsg string) { + select { + case <-c: + t.Fatalf(failMsg) + case <-time.After(blockDuration): + // Ok, blocked for the required duration. + } +} + +func TestReadOpenBlocksForWriteOpen(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + rDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone) + + // Verify that the open for read is blocking. + assertRecvBlocks(t, rDone, time.Millisecond*100, + "open for read not blocking with no writers") + + wDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) + + <-wDone + <-rDone +} + +func TestWriteOpenBlocksForReadOpen(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + wDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) + + // Verify that the open for write is blocking + assertRecvBlocks(t, wDone, time.Millisecond*100, + "open for write not blocking with no readers") + + rDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone) + + <-rDone + <-wDone +} + +func TestMultipleWriteOpenDoesntCountAsReadOpen(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + rDone1 := make(chan struct{}) + rDone2 := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone1) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone2) + + assertRecvBlocks(t, rDone1, time.Millisecond*100, + "open for read didn't block with no writers") + assertRecvBlocks(t, rDone2, time.Millisecond*100, + "open for read didn't block with no writers") + + wDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) + + <-wDone + <-rDone2 + <-rDone1 +} + +func TestClosedReaderBlocksWriteOpen(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + rFile, _ := testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true, NonBlocking: true}, nil) + rFile.DecRef() + + wDone := make(chan struct{}) + // This open for write should block because the reader is now gone. + go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) + assertRecvBlocks(t, wDone, time.Millisecond*100, + "open for write didn't block with no concurrent readers") + + // Open for read again. This should unblock the open for write. + rDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone) + + <-rDone + <-wDone +} + +func TestReadWriteOpenNeverBlocks(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + rwDone := make(chan struct{}) + // Open for read-write never wait for a reader or writer, even if the + // nonblocking flag is not set. + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true, Write: true, NonBlocking: false}, rwDone) + <-rwDone +} + +func TestReadWriteOpenUnblocksReadOpen(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + rDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone) + + rwDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true, Write: true}, rwDone) + + <-rwDone + <-rDone +} + +func TestReadWriteOpenUnblocksWriteOpen(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + wDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) + + rwDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true, Write: true}, rwDone) + + <-rwDone + <-wDone +} + +func TestBlockedOpenIsCancellable(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + done := make(chan openResult) + go testOpen(ctx, t, f, fs.FileFlags{Read: true}, done) + select { + case <-done: + t.Fatalf("open for read didn't block with no writers") + case <-time.After(time.Millisecond * 100): + // Ok. + } + + ctx.(*sleeper).Cancel() + // If the cancel on the sleeper didn't work, the open for read would never + // return. + res := <-done + if res.error != syserror.ErrInterrupted { + t.Fatalf("Cancellation didn't cause GetFile to return fs.ErrInterrupted, got %v.", + res.error) + } +} + +func TestNonblockingReadOpenNoWriters(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + if _, err := testOpen(ctx, t, f, fs.FileFlags{Read: true, NonBlocking: true}, nil); err != nil { + t.Fatalf("Nonblocking open for read failed with error %v.", err) + } +} + +func TestNonblockingWriteOpenNoReaders(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + if _, err := testOpen(ctx, t, f, fs.FileFlags{Write: true, NonBlocking: true}, nil); err != syserror.ENXIO { + t.Fatalf("Nonblocking open for write failed unexpected error %v.", err) + } +} + +func TestNonBlockingReadOpenWithWriter(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + wDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) + + // Open for write blocks since there are no readers yet. + assertRecvBlocks(t, wDone, time.Millisecond*100, + "Open for write didn't block with no reader.") + + if _, err := testOpen(ctx, t, f, fs.FileFlags{Read: true, NonBlocking: true}, nil); err != nil { + t.Fatalf("Nonblocking open for read failed with error %v.", err) + } + + // Open for write should now be unblocked. + <-wDone +} + +func TestNonBlockingWriteOpenWithReader(t *testing.T) { + f := NewInodeOperations(nil, newNamedPipe(t)) + ctx := newSleeperContext(t) + + rDone := make(chan struct{}) + go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone) + + // Open for write blocked, since no reader yet. + assertRecvBlocks(t, rDone, time.Millisecond*100, + "Open for reader didn't block with no writer.") + + if _, err := testOpen(ctx, t, f, fs.FileFlags{Write: true, NonBlocking: true}, nil); err != nil { + t.Fatalf("Nonblocking open for write failed with error %v.", err) + } + + // Open for write should now be unblocked. + <-rDone +} + +func TestAnonReadOpen(t *testing.T) { + f := NewInodeOperations(nil, newAnonPipe(t)) + ctx := newSleeperContext(t) + + if _, err := testOpen(ctx, t, f, fs.FileFlags{Read: true}, nil); err != nil { + t.Fatalf("open anon pipe for read failed: %v", err) + } +} + +func TestAnonWriteOpen(t *testing.T) { + f := NewInodeOperations(nil, newAnonPipe(t)) + ctx := newSleeperContext(t) + + if _, err := testOpen(ctx, t, f, fs.FileFlags{Write: true}, nil); err != nil { + t.Fatalf("open anon pipe for write failed: %v", err) + } +} diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go new file mode 100644 index 000000000..1656c6ff3 --- /dev/null +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -0,0 +1,335 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package pipe provides an in-memory implementation of a unidirectional +// pipe. +// +// The goal of this pipe is to emulate the pipe syscall in all of its +// edge cases and guarantees of atomic IO. +package pipe + +import ( + "fmt" + "sync" + "sync/atomic" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/ilist" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// DefaultPipeSize is the system-wide default size of a pipe in bytes. +const DefaultPipeSize = 65536 + +// Pipe is an encapsulation of a platform-independent pipe. +// It manages a buffered byte queue shared between a reader/writer +// pair. +type Pipe struct { + waiter.Queue `state:"nosave"` + + // Whether this is a named or anonymous pipe. + isNamed bool + + // The dirent backing this pipe. Shared by all readers and writers. + dirent *fs.Dirent + + // The buffered byte queue. + data ilist.List + + // Max size of the pipe in bytes. When this max has been reached, + // writers will get EWOULDBLOCK. + max int + + // Current size of the pipe in bytes. + size int + + // Max number of bytes the pipe can guarantee to read or write + // atomically. + atomicIOBytes int + + // The number of active readers for this pipe. Load/store atomically. + readers int32 + + // The number of active writes for this pipe. Load/store atomically. + writers int32 + + // This flag indicates if this pipe ever had a writer. Note that this does + // not necessarily indicate there is *currently* a writer, just that there + // has been a writer at some point since the pipe was created. + // + // Protected by mu. + hadWriter bool + + // Lock protecting all pipe internal state. + mu sync.Mutex `state:"nosave"` +} + +// NewPipe initializes and returns a pipe. A pipe created by this function is +// persistent, and will remain valid even without any open fds to it. Named +// pipes for mknod(2) are created via this function. Note that the +// implementation of blocking semantics for opening the read and write ends of a +// named pipe are left to filesystems. +func NewPipe(ctx context.Context, isNamed bool, sizeBytes, atomicIOBytes int) *Pipe { + p := &Pipe{ + isNamed: isNamed, + max: sizeBytes, + atomicIOBytes: atomicIOBytes, + } + + // Build the fs.Dirent of this pipe, shared by all fs.Files associated + // with this pipe. + ino := pipeDevice.NextIno() + base := fsutil.NewSimpleInodeOperations(fsutil.InodeSimpleAttributes{ + FSType: linux.PIPEFS_MAGIC, + UAttr: fs.WithCurrentTime(ctx, fs.UnstableAttr{ + Owner: fs.FileOwnerFromContext(ctx), + Perms: fs.FilePermissions{ + User: fs.PermMask{Read: true, Write: true}, + }, + Links: 1, + }), + }) + sattr := fs.StableAttr{ + Type: fs.Pipe, + DeviceID: pipeDevice.DeviceID(), + InodeID: ino, + BlockSize: int64(atomicIOBytes), + } + // There is no real filesystem backing this pipe, so we pass in a nil + // Filesystem. + sb := fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}) + p.dirent = fs.NewDirent(fs.NewInode(NewInodeOperations(base, p), sb, sattr), fmt.Sprintf("pipe:[%d]", ino)) + + return p +} + +// NewConnectedPipe initializes a pipe and returns a pair of objects (which +// implement kio.File) representing the read and write ends of the pipe. A pipe +// created by this function becomes invalid as soon as either the read or write +// end is closed, and errors on subsequent operations on either end. Pipes +// for pipe(2) and pipe2(2) are generally created this way. +func NewConnectedPipe(ctx context.Context, sizeBytes int, atomicIOBytes int) (*fs.File, *fs.File) { + p := NewPipe(ctx, false /* isNamed */, sizeBytes, atomicIOBytes) + return p.ROpen(ctx), p.WOpen(ctx) +} + +// ROpen opens the pipe for reading. +func (p *Pipe) ROpen(ctx context.Context) *fs.File { + p.rOpen() + return fs.NewFile(ctx, p.dirent, fs.FileFlags{Read: true}, &Reader{ + ReaderWriter: ReaderWriter{Pipe: p}, + }) +} + +// WOpen opens the pipe for writing. +func (p *Pipe) WOpen(ctx context.Context) *fs.File { + p.wOpen() + return fs.NewFile(ctx, p.dirent, fs.FileFlags{Write: true}, &Writer{ + ReaderWriter: ReaderWriter{Pipe: p}, + }) +} + +// RWOpen opens the pipe for both reading and writing. +func (p *Pipe) RWOpen(ctx context.Context) *fs.File { + p.rOpen() + p.wOpen() + return fs.NewFile(ctx, p.dirent, fs.FileFlags{Read: true, Write: true}, &ReaderWriter{ + Pipe: p, + }) +} + +// read reads data from the pipe into dst and returns the number of bytes +// read, or returns ErrWouldBlock if the pipe is empty. +func (p *Pipe) read(ctx context.Context, dst usermem.IOSequence) (int64, error) { + if !p.HasReaders() { + return 0, syscall.EBADF + } + + // Don't block for a zero-length read even if the pipe is empty. + if dst.NumBytes() == 0 { + return 0, nil + } + + p.mu.Lock() + defer p.mu.Unlock() + // If there is nothing to read at the moment but there is a writer, tell the + // caller to block. + if p.size == 0 { + if !p.HasWriters() { + // There are no writers, return EOF. + return 0, nil + } + return 0, syserror.ErrWouldBlock + } + var n int64 + for b := p.data.Front(); b != nil; b = p.data.Front() { + buffer := b.(*Buffer) + n0, err := dst.CopyOut(ctx, buffer.bytes()) + n += int64(n0) + p.size -= n0 + if buffer.truncate(n0) == 0 { + p.data.Remove(b) + } + dst = dst.DropFirst(n0) + if dst.NumBytes() == 0 || err != nil { + return n, err + } + } + return n, nil +} + +// write writes data from sv into the pipe and returns the number of bytes +// written. If no bytes are written because the pipe is full (or has less than +// atomicIOBytes free capacity), write returns ErrWouldBlock. +func (p *Pipe) write(ctx context.Context, src usermem.IOSequence) (int64, error) { + p.mu.Lock() + defer p.mu.Unlock() + + if !p.HasWriters() { + return 0, syscall.EBADF + } + if !p.HasReaders() { + return 0, syscall.EPIPE + } + + // POSIX requires that a write smaller than atomicIOBytes (PIPE_BUF) be + // atomic, but requires no atomicity for writes larger than this. However, + // Linux appears to provide stronger semantics than this in practice: + // unmerged writes are done one PAGE_SIZE buffer at a time, so for larger + // writes, the writing of each PIPE_BUF-sized chunk is atomic. We implement + // this by writing at most atomicIOBytes at a time if we can't service the + // write in its entirety. + canWrite := src.NumBytes() + if canWrite > int64(p.max-p.size) { + if p.max-p.size >= p.atomicIOBytes { + canWrite = int64(p.atomicIOBytes) + } else { + return 0, syserror.ErrWouldBlock + } + } + + // Copy data from user memory into a pipe-owned buffer. + buf := make([]byte, canWrite) + n, err := src.CopyIn(ctx, buf) + if n > 0 { + p.data.PushBack(newBuffer(buf[:n])) + p.size += n + } + if int64(n) < src.NumBytes() && err == nil { + // Partial write due to full pipe. + err = syserror.ErrWouldBlock + } + return int64(n), err +} + +// rOpen signals a new reader of the pipe. +func (p *Pipe) rOpen() { + atomic.AddInt32(&p.readers, 1) +} + +// wOpen signals a new writer of the pipe. +func (p *Pipe) wOpen() { + p.mu.Lock() + defer p.mu.Unlock() + p.hadWriter = true + atomic.AddInt32(&p.writers, 1) +} + +// rClose signals that a reader has closed their end of the pipe. +func (p *Pipe) rClose() { + newReaders := atomic.AddInt32(&p.readers, -1) + if newReaders < 0 { + panic(fmt.Sprintf("Refcounting bug, pipe has negative readers: %v", newReaders)) + } +} + +// wClose signals that a writer has closed their end of the pipe. +func (p *Pipe) wClose() { + newWriters := atomic.AddInt32(&p.writers, -1) + if newWriters < 0 { + panic(fmt.Sprintf("Refcounting bug, pipe has negative writers: %v.", newWriters)) + } +} + +// HasReaders returns whether the pipe has any active readers. +func (p *Pipe) HasReaders() bool { + return atomic.LoadInt32(&p.readers) > 0 +} + +// HasWriters returns whether the pipe has any active writers. +func (p *Pipe) HasWriters() bool { + return atomic.LoadInt32(&p.writers) > 0 +} + +func (p *Pipe) rReadinessLocked() waiter.EventMask { + ready := waiter.EventMask(0) + if p.HasReaders() && p.data.Front() != nil { + ready |= waiter.EventIn + } + if !p.HasWriters() && p.hadWriter { + // POLLHUP must be supressed until the pipe has had at least one writer + // at some point. Otherwise a reader thread may poll and immediately get + // a POLLHUP before the writer ever opens the pipe, which the reader may + // interpret as the writer opening then closing the pipe. + ready |= waiter.EventHUp + } + return ready +} + +// rReadiness returns a mask that states whether the read end of the pipe is +// ready for reading. +func (p *Pipe) rReadiness() waiter.EventMask { + p.mu.Lock() + defer p.mu.Unlock() + return p.rReadinessLocked() +} + +func (p *Pipe) wReadinessLocked() waiter.EventMask { + ready := waiter.EventMask(0) + if p.HasWriters() && p.size < p.max { + ready |= waiter.EventOut + } + if !p.HasReaders() { + ready |= waiter.EventErr + } + return ready +} + +// wReadiness returns a mask that states whether the write end of the pipe +// is ready for writing. +func (p *Pipe) wReadiness() waiter.EventMask { + p.mu.Lock() + defer p.mu.Unlock() + return p.wReadinessLocked() +} + +// rwReadiness returns a mask that states whether a read-write handle to the +// pipe is ready for IO. +func (p *Pipe) rwReadiness() waiter.EventMask { + p.mu.Lock() + defer p.mu.Unlock() + return p.rReadinessLocked() | p.wReadinessLocked() +} + +func (p *Pipe) queuedSize() int { + p.mu.Lock() + defer p.mu.Unlock() + return p.size +} diff --git a/pkg/sentry/kernel/pipe/pipe_test.go b/pkg/sentry/kernel/pipe/pipe_test.go new file mode 100644 index 000000000..49ef8c8ac --- /dev/null +++ b/pkg/sentry/kernel/pipe/pipe_test.go @@ -0,0 +1,138 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "bytes" + "testing" + + "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +func TestPipeRW(t *testing.T) { + ctx := contexttest.Context(t) + r, w := NewConnectedPipe(ctx, 65536, 4096) + defer r.DecRef() + defer w.DecRef() + + msg := []byte("here's some bytes") + wantN := int64(len(msg)) + n, err := w.Writev(ctx, usermem.BytesIOSequence(msg)) + if n != wantN || err != nil { + t.Fatalf("Writev: got (%d, %v), wanted (%d, nil)", n, err, wantN) + } + + buf := make([]byte, len(msg)) + n, err = r.Readv(ctx, usermem.BytesIOSequence(buf)) + if n != wantN || err != nil || !bytes.Equal(buf, msg) { + t.Fatalf("Readv: got (%d, %v) %q, wanted (%d, nil) %q", n, err, buf, wantN, msg) + } +} + +func TestPipeReadBlock(t *testing.T) { + ctx := contexttest.Context(t) + r, w := NewConnectedPipe(ctx, 65536, 4096) + defer r.DecRef() + defer w.DecRef() + + n, err := r.Readv(ctx, usermem.BytesIOSequence(make([]byte, 1))) + if n != 0 || err != syserror.ErrWouldBlock { + t.Fatalf("Readv: got (%d, %v), wanted (0, %v)", n, err, syserror.ErrWouldBlock) + } +} + +func TestPipeWriteBlock(t *testing.T) { + const atomicIOBytes = 2 + + ctx := contexttest.Context(t) + r, w := NewConnectedPipe(ctx, 10, atomicIOBytes) + defer r.DecRef() + defer w.DecRef() + + msg := []byte("here's some bytes") + n, err := w.Writev(ctx, usermem.BytesIOSequence(msg)) + if wantN, wantErr := int64(atomicIOBytes), syserror.ErrWouldBlock; n != wantN || err != wantErr { + t.Fatalf("Writev: got (%d, %v), wanted (%d, %v)", n, err, wantN, wantErr) + } +} + +func TestPipeWriteUntilEnd(t *testing.T) { + const atomicIOBytes = 2 + + ctx := contexttest.Context(t) + r, w := NewConnectedPipe(ctx, atomicIOBytes, atomicIOBytes) + defer r.DecRef() + defer w.DecRef() + + msg := []byte("here's some bytes") + + wDone := make(chan struct{}, 0) + rDone := make(chan struct{}, 0) + defer func() { + // Signal the reader to stop and wait until it does so. + close(wDone) + <-rDone + }() + + go func() { + defer close(rDone) + // Read from r until done is closed. + ctx := contexttest.Context(t) + buf := make([]byte, len(msg)+1) + dst := usermem.BytesIOSequence(buf) + e, ch := waiter.NewChannelEntry(nil) + r.EventRegister(&e, waiter.EventIn) + defer r.EventUnregister(&e) + for { + n, err := r.Readv(ctx, dst) + dst = dst.DropFirst64(n) + if err == syserror.ErrWouldBlock { + select { + case <-ch: + continue + case <-wDone: + // We expect to have 1 byte left in dst since len(buf) == + // len(msg)+1. + if dst.NumBytes() != 1 || !bytes.Equal(buf[:len(msg)], msg) { + t.Errorf("Reader: got %q (%d bytes remaining), wanted %q", buf, dst.NumBytes(), msg) + } + return + } + } + if err != nil { + t.Fatalf("Readv: got unexpected error %v", err) + } + } + }() + + src := usermem.BytesIOSequence(msg) + e, ch := waiter.NewChannelEntry(nil) + w.EventRegister(&e, waiter.EventOut) + defer w.EventUnregister(&e) + for src.NumBytes() != 0 { + n, err := w.Writev(ctx, src) + src = src.DropFirst64(n) + if err == syserror.ErrWouldBlock { + <-ch + continue + } + if err != nil { + t.Fatalf("Writev: got (%d, %v)", n, err) + } + } +} diff --git a/pkg/sentry/kernel/pipe/reader.go b/pkg/sentry/kernel/pipe/reader.go new file mode 100644 index 000000000..40d5e4943 --- /dev/null +++ b/pkg/sentry/kernel/pipe/reader.go @@ -0,0 +1,37 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// Reader satisfies the fs.FileOperations interface for read-only pipes. +// Reader should be used with !fs.FileFlags.Write to reject writes. +type Reader struct { + ReaderWriter +} + +// Release implements fs.FileOperations.Release. +func (r *Reader) Release() { + r.Pipe.rClose() + // Wake up writers. + r.Pipe.Notify(waiter.EventOut) +} + +// Readiness returns the ready events in the underlying pipe. +func (r *Reader) Readiness(mask waiter.EventMask) waiter.EventMask { + return r.Pipe.rReadiness() & mask +} diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go new file mode 100644 index 000000000..dc642a3a6 --- /dev/null +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -0,0 +1,91 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "fmt" + "math" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// ReaderWriter satisfies the FileOperations interface and services both +// read and write requests. This should only be used directly for named pipes. +// pipe(2) and pipe2(2) only support unidirectional pipes and should use +// either pipe.Reader or pipe.Writer. +type ReaderWriter struct { + fsutil.PipeSeek `state:"nosave"` + fsutil.NotDirReaddir `state:"nosave"` + fsutil.NoFsync `state:"nosave"` + fsutil.NoopFlush `state:"nosave"` + fsutil.NoMMap `state:"nosave"` + *Pipe +} + +// Release implements fs.FileOperations.Release. +func (rw *ReaderWriter) Release() { + rw.Pipe.rClose() + rw.Pipe.wClose() + // Wake up readers and writers. + rw.Pipe.Notify(waiter.EventIn | waiter.EventOut) +} + +// Read implements fs.FileOperations.Read. +func (rw *ReaderWriter) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, _ int64) (int64, error) { + n, err := rw.Pipe.read(ctx, dst) + if n > 0 { + rw.Pipe.Notify(waiter.EventOut) + } + return n, err +} + +// Write implements fs.FileOperations.Write. +func (rw *ReaderWriter) Write(ctx context.Context, _ *fs.File, src usermem.IOSequence, _ int64) (int64, error) { + n, err := rw.Pipe.write(ctx, src) + if n > 0 { + rw.Pipe.Notify(waiter.EventIn) + } + return n, err +} + +// Readiness returns the ready events in the underlying pipe. +func (rw *ReaderWriter) Readiness(mask waiter.EventMask) waiter.EventMask { + return rw.Pipe.rwReadiness() & mask +} + +// Ioctl implements fs.FileOperations.Ioctl. +func (rw *ReaderWriter) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { + // Switch on ioctl request. + switch int(args[1].Int()) { + case syscall.TIOCINQ: + v := rw.queuedSize() + if v > math.MaxInt32 { + panic(fmt.Sprintf("Impossibly large pipe queued size: %d", v)) + } + // Copy result to user-space. + _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), int32(v), usermem.IOOpts{ + AddressSpaceActive: true, + }) + return 0, err + default: + return 0, syscall.ENOTTY + } +} diff --git a/pkg/sentry/kernel/pipe/writer.go b/pkg/sentry/kernel/pipe/writer.go new file mode 100644 index 000000000..fd13008ac --- /dev/null +++ b/pkg/sentry/kernel/pipe/writer.go @@ -0,0 +1,37 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// Writer satisfies the fs.FileOperations interface for write-only pipes. +// Writer should be used with !fs.FileFlags.Read to reject reads. +type Writer struct { + ReaderWriter +} + +// Release implements fs.FileOperations.Release. +func (w *Writer) Release() { + w.Pipe.wClose() + // Wake up readers. + w.Pipe.Notify(waiter.EventHUp) +} + +// Readiness returns the ready events in the underlying pipe. +func (w *Writer) Readiness(mask waiter.EventMask) waiter.EventMask { + return w.Pipe.wReadiness() & mask +} diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go new file mode 100644 index 000000000..20b1c4cd4 --- /dev/null +++ b/pkg/sentry/kernel/ptrace.go @@ -0,0 +1,1054 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// ptrace constants from Linux's include/uapi/linux/ptrace.h. +const ( + _PTRACE_EVENT_SECCOMP = 7 + PTRACE_SEIZE = 0x4206 + PTRACE_INTERRUPT = 0x4207 + PTRACE_LISTEN = 0x4208 + PTRACE_PEEKSIGINFO = 0x4209 + PTRACE_GETSIGMASK = 0x420a + PTRACE_SETSIGMASK = 0x420b + _PTRACE_O_EXITKILL = 1 << 20 + _PTRACE_O_TRACESECCOMP = 1 << _PTRACE_EVENT_SECCOMP +) + +// ptraceOptions are the subset of options controlling a task's ptrace behavior +// that are set by ptrace(PTRACE_SETOPTIONS). +type ptraceOptions struct { + // ExitKill is true if the tracee should be sent SIGKILL when the tracer + // exits. + ExitKill bool + + // If SysGood is true, set bit 7 in the signal number for + // syscall-entry-stop and syscall-exit-stop traps delivered to this task's + // tracer. + SysGood bool + + // TraceClone is true if the tracer wants to receive PTRACE_EVENT_CLONE + // events. + TraceClone bool + + // TraceExec is true if the tracer wants to receive PTRACE_EVENT_EXEC + // events. + TraceExec bool + + // TraceExit is true if the tracer wants to receive PTRACE_EVENT_EXIT + // events. + TraceExit bool + + // TraceFork is true if the tracer wants to receive PTRACE_EVENT_FORK + // events. + TraceFork bool + + // TraceSeccomp is true if the tracer wants to receive PTRACE_EVENT_SECCOMP + // events. + TraceSeccomp bool + + // TraceVfork is true if the tracer wants to receive PTRACE_EVENT_VFORK + // events. + TraceVfork bool + + // TraceVforkDone is true if the tracer wants to receive + // PTRACE_EVENT_VFORK_DONE events. + TraceVforkDone bool +} + +// ptraceSyscallMode controls the behavior of a ptraced task at syscall entry +// and exit. +type ptraceSyscallMode int + +const ( + // ptraceSyscallNone indicates that the task has never ptrace-stopped, or + // that it was resumed from its last ptrace-stop by PTRACE_CONT or + // PTRACE_DETACH. The task's syscalls will not be intercepted. + ptraceSyscallNone ptraceSyscallMode = iota + + // ptraceSyscallIntercept indicates that the task was resumed from its last + // ptrace-stop by PTRACE_SYSCALL. The next time the task enters or exits a + // syscall, a ptrace-stop will occur. + ptraceSyscallIntercept + + // ptraceSyscallEmu indicates that the task was resumed from its last + // ptrace-stop by PTRACE_SYSEMU or PTRACE_SYSEMU_SINGLESTEP. The next time + // the task enters a syscall, the syscall will be skipped, and a + // ptrace-stop will occur. + ptraceSyscallEmu +) + +// CanTrace checks that t is permitted to access target's state, as defined by +// ptrace(2), subsection "Ptrace access mode checking". If attach is true, it +// checks for access mode PTRACE_MODE_ATTACH; otherwise, it checks for access +// mode PTRACE_MODE_READ. +func (t *Task) CanTrace(target *Task, attach bool) bool { + // "1. If the calling thread and the target thread are in the same thread + // group, access is always allowed." - ptrace(2) + // + // Note: Strictly speaking, prior to 73af963f9f30 ("__ptrace_may_access() + // should not deny sub-threads", first released in Linux 3.12), the rule + // only applies if t and target are the same task. But, as that commit + // message puts it, "[any] security check is pointless when the tasks share + // the same ->mm." + if t.tg == target.tg { + return true + } + + // """ + // 2. If the access mode specifies PTRACE_MODE_FSCREDS (ED: snipped, + // doesn't exist until Linux 4.5). + // + // Otherwise, the access mode specifies PTRACE_MODE_REALCREDS, so use the + // caller's real UID and GID for the checks in the next step. (Most APIs + // that check the caller's UID and GID use the effective IDs. For + // historical reasons, the PTRACE_MODE_REALCREDS check uses the real IDs + // instead.) + // + // 3. Deny access if neither of the following is true: + // + // - The real, effective, and saved-set user IDs of the target match the + // caller's user ID, *and* the real, effective, and saved-set group IDs of + // the target match the caller's group ID. + // + // - The caller has the CAP_SYS_PTRACE capability in the user namespace of + // the target. + // + // 4. Deny access if the target process "dumpable" attribute has a value + // other than 1 (SUID_DUMP_USER; see the discussion of PR_SET_DUMPABLE in + // prctl(2)), and the caller does not have the CAP_SYS_PTRACE capability in + // the user namespace of the target process. + // + // 5. The kernel LSM security_ptrace_access_check() interface is invoked to + // see if ptrace access is permitted. The results depend on the LSM(s). The + // implementation of this interface in the commoncap LSM performs the + // following steps: + // + // a) If the access mode includes PTRACE_MODE_FSCREDS, then use the + // caller's effective capability set; otherwise (the access mode specifies + // PTRACE_MODE_REALCREDS, so) use the caller's permitted capability set. + // + // b) Deny access if neither of the following is true: + // + // - The caller and the target process are in the same user namespace, and + // the caller's capabilities are a proper superset of the target process's + // permitted capabilities. + // + // - The caller has the CAP_SYS_PTRACE capability in the target process's + // user namespace. + // + // Note that the commoncap LSM does not distinguish between + // PTRACE_MODE_READ and PTRACE_MODE_ATTACH. (ED: From earlier in this + // section: "the commoncap LSM ... is always invoked".) + // """ + callerCreds := t.Credentials() + targetCreds := target.Credentials() + if callerCreds.HasCapabilityIn(linux.CAP_SYS_PTRACE, targetCreds.UserNamespace) { + return true + } + if cuid := callerCreds.RealKUID; cuid != targetCreds.RealKUID || cuid != targetCreds.EffectiveKUID || cuid != targetCreds.SavedKUID { + return false + } + if cgid := callerCreds.RealKGID; cgid != targetCreds.RealKGID || cgid != targetCreds.EffectiveKGID || cgid != targetCreds.SavedKGID { + return false + } + // TODO: dumpability check + if callerCreds.UserNamespace != targetCreds.UserNamespace { + return false + } + if targetCreds.PermittedCaps&^callerCreds.PermittedCaps != 0 { + return false + } + // TODO: Yama LSM + return true +} + +// Tracer returns t's ptrace Tracer. +func (t *Task) Tracer() *Task { + return t.ptraceTracer.Load().(*Task) +} + +// hasTracer returns true if t has a ptrace tracer attached. +func (t *Task) hasTracer() bool { + // This isn't just inlined into callers so that if Task.Tracer() turns out + // to be too expensive because of e.g. interface conversion, we can switch + // to having a separate atomic flag more easily. + return t.Tracer() != nil +} + +// ptraceStop is a TaskStop placed on tasks in a ptrace-stop. +type ptraceStop struct { + // If frozen is true, the stopped task's tracer is currently operating on + // it, so Task.Kill should not remove the stop. + frozen bool +} + +// Killable implements TaskStop.Killable. +func (s *ptraceStop) Killable() bool { + return !s.frozen +} + +// beginPtraceStopLocked initiates an unfrozen ptrace-stop on t. If t has been +// killed, the stop is skipped, and beginPtraceStopLocked returns false. +// +// beginPtraceStopLocked does not signal t's tracer or wake it if it is +// waiting. +// +// Preconditions: The TaskSet mutex must be locked. The caller must be running +// on the task goroutine. +func (t *Task) beginPtraceStopLocked() bool { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + // This is analogous to Linux's kernel/signal.c:ptrace_stop() => ... => + // kernel/sched/core.c:__schedule() => signal_pending_state() check, which + // is what prevents tasks from entering ptrace-stops after being killed. + // Note that if t was SIGKILLed and beingPtraceStopLocked is being called + // for PTRACE_EVENT_EXIT, the task will have dequeued the signal before + // entering the exit path, so t.killable() will no longer return true. This + // is consistent with Linux: "Bugs: ... A SIGKILL signal may still cause a + // PTRACE_EVENT_EXIT stop before actual signal death. This may be changed + // in the future; SIGKILL is meant to always immediately kill tasks even + // under ptrace. Last confirmed on Linux 3.13." - ptrace(2) + if t.killedLocked() { + return false + } + t.beginInternalStopLocked(&ptraceStop{}) + return true +} + +// Preconditions: The TaskSet mutex must be locked. +func (t *Task) ptraceTrapLocked(code int32) { + t.ptraceCode = code + t.ptraceSiginfo = &arch.SignalInfo{ + Signo: int32(linux.SIGTRAP), + Code: code, + } + t.ptraceSiginfo.SetPid(int32(t.tg.pidns.tids[t])) + t.ptraceSiginfo.SetUid(int32(t.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow())) + if t.beginPtraceStopLocked() { + tracer := t.Tracer() + tracer.signalStop(t, arch.CLD_TRAPPED, int32(linux.SIGTRAP)) + tracer.tg.eventQueue.Notify(EventTraceeStop) + } +} + +// ptraceFreeze checks if t is in a ptraceStop. If so, it freezes the +// ptraceStop, temporarily preventing it from being removed by a concurrent +// Task.Kill, and returns true. Otherwise it returns false. +// +// Preconditions: The TaskSet mutex must be locked. The caller must be running +// on the task goroutine of t's tracer. +func (t *Task) ptraceFreeze() bool { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + if t.stop == nil { + return false + } + s, ok := t.stop.(*ptraceStop) + if !ok { + return false + } + s.frozen = true + return true +} + +// ptraceUnfreeze ends the effect of a previous successful call to +// ptraceFreeze. +// +// Preconditions: t must be in a frozen ptraceStop. +func (t *Task) ptraceUnfreeze() { + // t.tg.signalHandlers is stable because t is in a frozen ptrace-stop, + // preventing its thread group from completing execve. + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + // Do this even if the task has been killed to ensure a panic if t.stop is + // nil or not a ptraceStop. + t.stop.(*ptraceStop).frozen = false + if t.killedLocked() { + t.endInternalStopLocked() + } +} + +// ptraceUnstop implements ptrace request PTRACE_CONT, PTRACE_SYSCALL, +// PTRACE_SINGLESTEP, PTRACE_SYSEMU, or PTRACE_SYSEMU_SINGLESTEP depending on +// mode and singlestep. +// +// Preconditions: t must be in a frozen ptrace stop. +// +// Postconditions: If ptraceUnstop returns nil, t will no longer be in a ptrace +// stop. +func (t *Task) ptraceUnstop(mode ptraceSyscallMode, singlestep bool, sig linux.Signal) error { + if sig != 0 && !sig.IsValid() { + return syserror.EIO + } + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + t.ptraceCode = int32(sig) + t.ptraceSyscallMode = mode + t.ptraceSinglestep = singlestep + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + t.endInternalStopLocked() + return nil +} + +func (t *Task) ptraceTraceme() error { + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + if t.hasTracer() { + return syserror.EPERM + } + if t.parent == nil { + // In Linux, only init can not have a parent, and init is assumed never + // to invoke PTRACE_TRACEME. In the sentry, TGID 1 is an arbitrary user + // application that may invoke PTRACE_TRACEME; having no parent can + // also occur if all tasks in the parent thread group have exited, and + // failed to find a living thread group to reparent to. The former case + // is treated as if TGID 1 has an exited parent in an invisible + // ancestor PID namespace that is an owner of the root user namespace + // (and consequently has CAP_SYS_PTRACE), and the latter case is a + // special form of the exited parent case below. In either case, + // returning nil here is correct. + return nil + } + if !t.parent.CanTrace(t, true) { + return syserror.EPERM + } + if t.parent.exitState != TaskExitNone { + // Fail silently, as if we were successfully attached but then + // immediately detached. This is consistent with Linux. + return nil + } + t.ptraceTracer.Store(t.parent) + t.parent.ptraceTracees[t] = struct{}{} + return nil +} + +// ptraceAttach implements ptrace(PTRACE_ATTACH, target). t is the caller. +func (t *Task) ptraceAttach(target *Task) error { + if t.tg == target.tg { + return syserror.EPERM + } + if !t.CanTrace(target, true) { + return syserror.EPERM + } + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + if target.hasTracer() { + return syserror.EPERM + } + // Attaching to zombies and dead tasks is not permitted; the exit + // notification logic relies on this. Linux allows attaching to PF_EXITING + // tasks, though. + if target.exitState >= TaskExitZombie { + return syserror.EPERM + } + target.ptraceTracer.Store(t) + t.ptraceTracees[target] = struct{}{} + target.tg.signalHandlers.mu.Lock() + target.sendSignalLocked(&arch.SignalInfo{ + Signo: int32(linux.SIGSTOP), + Code: arch.SignalInfoUser, + }, false /* group */) + // Undocumented Linux feature: If the tracee is already group-stopped (and + // consequently will not report the SIGSTOP just sent), force it to leave + // and re-enter the stop so that it will switch to a ptrace-stop. + if target.stop == (*groupStop)(nil) { + target.groupStopRequired = true + target.endInternalStopLocked() + } + target.tg.signalHandlers.mu.Unlock() + return nil +} + +// ptraceDetach implements ptrace(PTRACE_DETACH, target, 0, sig). t is the +// caller. +// +// Preconditions: target must be a tracee of t in a frozen ptrace stop. +// +// Postconditions: If ptraceDetach returns nil, target will no longer be in a +// ptrace stop. +func (t *Task) ptraceDetach(target *Task, sig linux.Signal) error { + if sig != 0 && !sig.IsValid() { + return syserror.EIO + } + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + target.ptraceCode = int32(sig) + target.forgetTracerLocked() + delete(t.ptraceTracees, target) + return nil +} + +// exitPtrace is called in the exit path to detach all of t's tracees. +func (t *Task) exitPtrace() { + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + for target := range t.ptraceTracees { + if target.ptraceOpts.ExitKill { + target.tg.signalHandlers.mu.Lock() + target.sendSignalLocked(&arch.SignalInfo{ + Signo: int32(linux.SIGKILL), + }, false /* group */) + target.tg.signalHandlers.mu.Unlock() + } + // Leave ptraceCode unchanged so that if the task is ptrace-stopped, it + // observes the ptraceCode it set before it entered the stop. I believe + // this is consistent with Linux. + target.forgetTracerLocked() + } + // "nil maps cannot be saved" + t.ptraceTracees = make(map[*Task]struct{}) +} + +// forgetTracerLocked detaches t's tracer and ensures that t is no longer +// ptrace-stopped. +// +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) forgetTracerLocked() { + t.ptraceOpts = ptraceOptions{} + t.ptraceSyscallMode = ptraceSyscallNone + t.ptraceSinglestep = false + t.ptraceTracer.Store((*Task)(nil)) + if t.exitTracerNotified && !t.exitTracerAcked { + t.exitTracerAcked = true + t.exitNotifyLocked(true) + } + // If t is ptrace-stopped, but its thread group is in a group stop and t is + // eligible to participate, make it do so. This is essentially the reverse + // of the special case in ptraceAttach, which converts a group stop to a + // ptrace stop. ("Handling of restart from group-stop is currently buggy, + // but the "as planned" behavior is to leave tracee stopped and waiting for + // SIGCONT." - ptrace(2)) + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + if t.stop == nil { + return + } + if _, ok := t.stop.(*ptraceStop); ok { + if t.exitState < TaskExitInitiated && t.tg.groupStopPhase >= groupStopInitiated { + t.groupStopRequired = true + } + t.endInternalStopLocked() + } +} + +// ptraceSignalLocked is called after signal dequeueing to check if t should +// enter ptrace signal-delivery-stop. +// +// Preconditions: The signal mutex must be locked. The caller must be running +// on the task goroutine. +func (t *Task) ptraceSignalLocked(info *arch.SignalInfo) bool { + if linux.Signal(info.Signo) == linux.SIGKILL { + return false + } + if !t.hasTracer() { + return false + } + // The tracer might change this signal into a stop signal, in which case + // any SIGCONT received after the signal was originally dequeued should + // cancel it. This is consistent with Linux. + if t.tg.groupStopPhase == groupStopNone { + t.tg.groupStopPhase = groupStopDequeued + } + // Can't lock the TaskSet mutex while holding a signal mutex. + t.tg.signalHandlers.mu.Unlock() + defer t.tg.signalHandlers.mu.Lock() + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + tracer := t.Tracer() + if tracer == nil { + return false + } + t.ptraceCode = info.Signo + t.ptraceSiginfo = info + t.Debugf("Entering signal-delivery-stop for signal %d", info.Signo) + if t.beginPtraceStopLocked() { + tracer.signalStop(t, arch.CLD_TRAPPED, info.Signo) + tracer.tg.eventQueue.Notify(EventTraceeStop) + } + return true +} + +// ptraceSeccomp is called when a seccomp-bpf filter returns action +// SECCOMP_RET_TRACE to check if t should enter PTRACE_EVENT_SECCOMP stop. data +// is the lower 16 bits of the filter's return value. +func (t *Task) ptraceSeccomp(data uint16) bool { + if !t.hasTracer() { + return false + } + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + if !t.ptraceOpts.TraceSeccomp { + return false + } + t.Debugf("Entering PTRACE_EVENT_SECCOMP stop") + t.ptraceEventLocked(_PTRACE_EVENT_SECCOMP, uint64(data)) + return true +} + +// ptraceSyscallEnter is called immediately before entering a syscall to check +// if t should enter ptrace syscall-enter-stop. +func (t *Task) ptraceSyscallEnter() (taskRunState, bool) { + if !t.hasTracer() { + return nil, false + } + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + switch t.ptraceSyscallMode { + case ptraceSyscallNone: + return nil, false + case ptraceSyscallIntercept: + t.Debugf("Entering syscall-enter-stop from PTRACE_SYSCALL") + t.ptraceSyscallStopLocked() + return (*runSyscallAfterSyscallEnterStop)(nil), true + case ptraceSyscallEmu: + t.Debugf("Entering syscall-enter-stop from PTRACE_SYSEMU") + t.ptraceSyscallStopLocked() + return (*runSyscallAfterSysemuStop)(nil), true + } + panic(fmt.Sprintf("Unknown ptraceSyscallMode: %v", t.ptraceSyscallMode)) +} + +// ptraceSyscallExit is called immediately after leaving a syscall to check if +// t should enter ptrace syscall-exit-stop. +func (t *Task) ptraceSyscallExit() { + if !t.hasTracer() { + return + } + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + if t.ptraceSyscallMode != ptraceSyscallIntercept { + return + } + t.Debugf("Entering syscall-exit-stop") + t.ptraceSyscallStopLocked() +} + +// Preconditions: The TaskSet mutex must be locked. +func (t *Task) ptraceSyscallStopLocked() { + code := int32(linux.SIGTRAP) + if t.ptraceOpts.SysGood { + code |= 0x80 + } + t.ptraceTrapLocked(code) +} + +type ptraceCloneKind int32 + +const ( + // ptraceCloneKindClone represents a call to Task.Clone where + // TerminationSignal is not SIGCHLD and Vfork is false. + ptraceCloneKindClone ptraceCloneKind = iota + + // ptraceCloneKindFork represents a call to Task.Clone where + // TerminationSignal is SIGCHLD and Vfork is false. + ptraceCloneKindFork + + // ptraceCloneKindVfork represents a call to Task.Clone where Vfork is + // true. + ptraceCloneKindVfork +) + +// ptraceClone is called at the end of a clone or fork syscall to check if t +// should enter PTRACE_EVENT_CLONE, PTRACE_EVENT_FORK, or PTRACE_EVENT_VFORK +// stop. child is the new task. +func (t *Task) ptraceClone(kind ptraceCloneKind, child *Task, opts *CloneOptions) bool { + if !t.hasTracer() { + return false + } + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + event := false + if !opts.Untraced { + switch kind { + case ptraceCloneKindClone: + if t.ptraceOpts.TraceClone { + t.Debugf("Entering PTRACE_EVENT_CLONE stop") + t.ptraceEventLocked(syscall.PTRACE_EVENT_CLONE, uint64(t.tg.pidns.tids[child])) + event = true + } + case ptraceCloneKindFork: + if t.ptraceOpts.TraceFork { + t.Debugf("Entering PTRACE_EVENT_FORK stop") + t.ptraceEventLocked(syscall.PTRACE_EVENT_FORK, uint64(t.tg.pidns.tids[child])) + event = true + } + case ptraceCloneKindVfork: + if t.ptraceOpts.TraceVfork { + t.Debugf("Entering PTRACE_EVENT_VFORK stop") + t.ptraceEventLocked(syscall.PTRACE_EVENT_VFORK, uint64(t.tg.pidns.tids[child])) + event = true + } + default: + panic(fmt.Sprintf("Unknown ptraceCloneKind: %v", kind)) + } + } + // "If the PTRACE_O_TRACEFORK, PTRACE_O_TRACEVFORK, or PTRACE_O_TRACECLONE + // options are in effect, then children created by, respectively, vfork(2) + // or clone(2) with the CLONE_VFORK flag, fork(2) or clone(2) with the exit + // signal set to SIGCHLD, and other kinds of clone(2), are automatically + // attached to the same tracer which traced their parent. SIGSTOP is + // delivered to the children, causing them to enter signal-delivery-stop + // after they exit the system call which created them." - ptrace(2) + // + // clone(2)'s documentation of CLONE_UNTRACED and CLONE_PTRACE is + // confusingly wrong; see kernel/fork.c:_do_fork() => copy_process() => + // include/linux/ptrace.h:ptrace_init_task(). + if event || opts.InheritTracer { + tracer := t.Tracer() + if tracer != nil { + child.ptraceTracer.Store(tracer) + tracer.ptraceTracees[child] = struct{}{} + // "Flags are inherited by new tracees created and "auto-attached" + // via active PTRACE_O_TRACEFORK, PTRACE_O_TRACEVFORK, or + // PTRACE_O_TRACECLONE options." + child.ptraceOpts = t.ptraceOpts + child.tg.signalHandlers.mu.Lock() + // If the child is PT_SEIZED (currently not possible in the sentry + // because PTRACE_SEIZE is unimplemented, but for future + // reference), Linux just sets JOBCTL_TRAP_STOP instead, so the + // child skips signal-delivery-stop and goes directly to + // group-stop. + // + // The child will self-t.interrupt() when its task goroutine starts + // running, so we don't have to. + child.pendingSignals.enqueue(&arch.SignalInfo{ + Signo: int32(linux.SIGSTOP), + }) + child.tg.signalHandlers.mu.Unlock() + } + } + return event +} + +// ptraceVforkDone is called after the end of a vfork stop to check if t should +// enter PTRACE_EVENT_VFORK_DONE stop. child is the new task's thread ID in t's +// PID namespace. +func (t *Task) ptraceVforkDone(child ThreadID) bool { + if !t.hasTracer() { + return false + } + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + if !t.ptraceOpts.TraceVforkDone { + return false + } + t.Debugf("Entering PTRACE_EVENT_VFORK_DONE stop") + t.ptraceEventLocked(syscall.PTRACE_EVENT_VFORK_DONE, uint64(child)) + return true +} + +// ptraceExec is called at the end of an execve syscall to check if t should +// enter PTRACE_EVENT_EXEC stop. oldTID is t's thread ID, in its *tracer's* PID +// namespace, prior to the execve. (If t did not have a tracer at the time +// oldTID was read, oldTID may be 0. This is consistent with Linux.) +func (t *Task) ptraceExec(oldTID ThreadID) { + if !t.hasTracer() { + return + } + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + // Recheck with the TaskSet mutex locked. Most ptrace points don't need to + // do this because detaching resets ptrace options, but PTRACE_EVENT_EXEC + // is special because both TraceExec and !TraceExec do something if a + // tracer is attached. + if !t.hasTracer() { + return + } + if t.ptraceOpts.TraceExec { + t.Debugf("Entering PTRACE_EVENT_EXEC stop") + t.ptraceEventLocked(syscall.PTRACE_EVENT_EXEC, uint64(oldTID)) + return + } + // "If the PTRACE_O_TRACEEXEC option is not in effect for the execing + // tracee, and if the tracee was PTRACE_ATTACHed rather that [sic] + // PTRACE_SEIZEd, the kernel delivers an extra SIGTRAP to the tracee after + // execve(2) returns. This is an ordinary signal (similar to one which can + // be generated by `kill -TRAP`, not a special kind of ptrace-stop. + // Employing PTRACE_GETSIGINFO for this signal returns si_code set to 0 + // (SI_USER). This signal may be blocked by signal mask, and thus may be + // delivered (much) later." - ptrace(2) + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + t.sendSignalLocked(&arch.SignalInfo{ + Signo: int32(linux.SIGTRAP), + Code: arch.SignalInfoUser, + }, false /* group */) +} + +// ptraceExit is called early in the task exit path to check if t should enter +// PTRACE_EVENT_EXIT stop. +func (t *Task) ptraceExit() { + if !t.hasTracer() { + return + } + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + if !t.ptraceOpts.TraceExit { + return + } + t.tg.signalHandlers.mu.Lock() + status := t.exitStatus.Status() + t.tg.signalHandlers.mu.Unlock() + t.Debugf("Entering PTRACE_EVENT_EXIT stop") + t.ptraceEventLocked(syscall.PTRACE_EVENT_EXIT, uint64(status)) +} + +// Preconditions: The TaskSet mutex must be locked. +func (t *Task) ptraceEventLocked(event int32, msg uint64) { + t.ptraceEventMsg = msg + // """ + // PTRACE_EVENT stops are observed by the tracer as waitpid(2) returning + // with WIFSTOPPED(status), and WSTOPSIG(status) returns SIGTRAP. An + // additional bit is set in the higher byte of the status word: the value + // status>>8 will be + // + // (SIGTRAP | PTRACE_EVENT_foo << 8). + // + // ... + // + // """ - ptrace(2) + t.ptraceTrapLocked(int32(linux.SIGTRAP) | (event << 8)) +} + +// ptraceKill implements ptrace(PTRACE_KILL, target). t is the caller. +func (t *Task) ptraceKill(target *Task) error { + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + if target.Tracer() != t { + return syserror.ESRCH + } + target.tg.signalHandlers.mu.Lock() + defer target.tg.signalHandlers.mu.Unlock() + // "This operation is deprecated; do not use it! Instead, send a SIGKILL + // directly using kill(2) or tgkill(2). The problem with PTRACE_KILL is + // that it requires the tracee to be in signal-delivery-stop, otherwise it + // may not work (i.e., may complete successfully but won't kill the + // tracee)." - ptrace(2) + if target.stop == nil { + return nil + } + if _, ok := target.stop.(*ptraceStop); !ok { + return nil + } + target.ptraceCode = int32(linux.SIGKILL) + target.endInternalStopLocked() + return nil +} + +// Ptrace implements the ptrace system call. +func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { + // PTRACE_TRACEME ignores all other arguments. + if req == syscall.PTRACE_TRACEME { + return t.ptraceTraceme() + } + // All other ptrace requests operate on a current or future tracee + // specified by pid. + target := t.tg.pidns.TaskWithID(pid) + if target == nil { + return syserror.ESRCH + } + + // PTRACE_ATTACH (and PTRACE_SEIZE, which is unimplemented) do not require + // that target is not already a tracee. + if req == syscall.PTRACE_ATTACH { + return t.ptraceAttach(target) + } + // PTRACE_KILL (and PTRACE_INTERRUPT, which is unimplemented) require that + // the target is a tracee, but does not require that it is ptrace-stopped. + if req == syscall.PTRACE_KILL { + return t.ptraceKill(target) + } + // All other ptrace requests require that the target is a ptrace-stopped + // tracee, and freeze the ptrace-stop so the tracee can be operated on. + t.tg.pidns.owner.mu.RLock() + if target.Tracer() != t { + t.tg.pidns.owner.mu.RUnlock() + return syserror.ESRCH + } + if !target.ptraceFreeze() { + t.tg.pidns.owner.mu.RUnlock() + // "Most ptrace commands (all except PTRACE_ATTACH, PTRACE_SEIZE, + // PTRACE_TRACEME, PTRACE_INTERRUPT, and PTRACE_KILL) require the + // tracee to be in a ptrace-stop, otherwise they fail with ESRCH." - + // ptrace(2) + return syserror.ESRCH + } + t.tg.pidns.owner.mu.RUnlock() + // Even if the target has a ptrace-stop active, the tracee's task goroutine + // may not yet have reached Task.doStop; wait for it to do so. This is safe + // because there's no way for target to initiate a ptrace-stop and then + // block (by calling Task.block) before entering it. + // + // Caveat: If tasks were just restored, the tracee's first call to + // Task.Activate (in Task.run) occurs before its first call to Task.doStop, + // which may block if the tracer's address space is active. + t.UninterruptibleSleepStart(true) + target.waitGoroutineStoppedOrExited() + t.UninterruptibleSleepFinish(true) + + // Resuming commands end the ptrace stop, but only if successful. + switch req { + case syscall.PTRACE_DETACH: + if err := t.ptraceDetach(target, linux.Signal(data)); err != nil { + target.ptraceUnfreeze() + return err + } + return nil + case syscall.PTRACE_CONT: + if err := target.ptraceUnstop(ptraceSyscallNone, false, linux.Signal(data)); err != nil { + target.ptraceUnfreeze() + return err + } + return nil + case syscall.PTRACE_SYSCALL: + if err := target.ptraceUnstop(ptraceSyscallIntercept, false, linux.Signal(data)); err != nil { + target.ptraceUnfreeze() + return err + } + return nil + case syscall.PTRACE_SINGLESTEP: + if err := target.ptraceUnstop(ptraceSyscallNone, true, linux.Signal(data)); err != nil { + target.ptraceUnfreeze() + return err + } + return nil + case syscall.PTRACE_SYSEMU: + if err := target.ptraceUnstop(ptraceSyscallEmu, false, linux.Signal(data)); err != nil { + target.ptraceUnfreeze() + return err + } + return nil + case syscall.PTRACE_SYSEMU_SINGLESTEP: + if err := target.ptraceUnstop(ptraceSyscallEmu, true, linux.Signal(data)); err != nil { + target.ptraceUnfreeze() + return err + } + return nil + } + // All other ptrace requests expect us to unfreeze the stop. + defer target.ptraceUnfreeze() + + switch req { + case syscall.PTRACE_PEEKTEXT, syscall.PTRACE_PEEKDATA: + // "At the system call level, the PTRACE_PEEKTEXT, PTRACE_PEEKDATA, and + // PTRACE_PEEKUSER requests have a different API: they store the result + // at the address specified by the data parameter, and the return value + // is the error flag." - ptrace(2) + word := t.Arch().Native(0) + if _, err := usermem.CopyObjectIn(t, target.MemoryManager(), addr, word, usermem.IOOpts{ + IgnorePermissions: true, + }); err != nil { + return err + } + _, err := t.CopyOut(data, word) + return err + + case syscall.PTRACE_POKETEXT, syscall.PTRACE_POKEDATA: + _, err := usermem.CopyObjectOut(t, target.MemoryManager(), addr, t.Arch().Native(uintptr(data)), usermem.IOOpts{ + IgnorePermissions: true, + }) + return err + + case syscall.PTRACE_PEEKUSR: // aka PTRACE_PEEKUSER + n, err := target.Arch().PtracePeekUser(uintptr(addr)) + if err != nil { + return err + } + _, err = t.CopyOut(data, n) + return err + + case syscall.PTRACE_POKEUSR: // aka PTRACE_POKEUSER + return target.Arch().PtracePokeUser(uintptr(addr), uintptr(data)) + + case syscall.PTRACE_GETREGS: + // "Copy the tracee's general-purpose ... registers ... to the address + // data in the tracer. ... (addr is ignored.) Note that SPARC systems + // have the meaning of data and addr reversed ..." + _, err := target.Arch().PtraceGetRegs(&usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: data, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }) + return err + + case syscall.PTRACE_GETFPREGS: + _, err := target.Arch().PtraceGetFPRegs(&usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: data, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }) + return err + + case syscall.PTRACE_GETREGSET: + // "Read the tracee's registers. addr specifies, in an + // architecture-dependent way, the type of registers to be read. ... + // data points to a struct iovec, which describes the destination + // buffer's location and length. On return, the kernel modifies iov.len + // to indicate the actual number of bytes returned." - ptrace(2) + ars, err := t.CopyInIovecs(data, 1) + if err != nil { + return err + } + ar := ars.Head() + n, err := target.Arch().PtraceGetRegSet(uintptr(addr), &usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: ar.Start, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }, int(ar.Length())) + if err != nil { + return err + } + ar.End -= usermem.Addr(n) + return t.CopyOutIovecs(data, usermem.AddrRangeSeqOf(ar)) + + case syscall.PTRACE_SETREGS: + _, err := target.Arch().PtraceSetRegs(&usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: data, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }) + return err + + case syscall.PTRACE_SETFPREGS: + _, err := target.Arch().PtraceSetFPRegs(&usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: data, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }) + return err + + case syscall.PTRACE_SETREGSET: + ars, err := t.CopyInIovecs(data, 1) + if err != nil { + return err + } + ar := ars.Head() + n, err := target.Arch().PtraceSetRegSet(uintptr(addr), &usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: ar.Start, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }, int(ar.Length())) + if err != nil { + return err + } + ar.End -= usermem.Addr(n) + return t.CopyOutIovecs(data, usermem.AddrRangeSeqOf(ar)) + + case syscall.PTRACE_GETSIGINFO: + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + if target.ptraceSiginfo == nil { + return syserror.EINVAL + } + _, err := t.CopyOut(data, target.ptraceSiginfo) + return err + + case syscall.PTRACE_SETSIGINFO: + var info arch.SignalInfo + if _, err := t.CopyIn(data, &info); err != nil { + return err + } + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + if target.ptraceSiginfo == nil { + return syserror.EINVAL + } + target.ptraceSiginfo = &info + return nil + + case PTRACE_GETSIGMASK: + if addr != linux.SignalSetSize { + return syserror.EINVAL + } + target.mu.Lock() + defer target.mu.Unlock() + _, err := t.CopyOut(data, target.tr.SignalMask) + return err + + case PTRACE_SETSIGMASK: + if addr != linux.SignalSetSize { + return syserror.EINVAL + } + var mask linux.SignalSet + if _, err := t.CopyIn(data, &mask); err != nil { + return err + } + // The target's task goroutine is stopped, so this is safe: + target.SetSignalMask(mask &^ UnblockableSignals) + return nil + + case syscall.PTRACE_SETOPTIONS: + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + validOpts := uintptr(_PTRACE_O_EXITKILL | syscall.PTRACE_O_TRACESYSGOOD | syscall.PTRACE_O_TRACECLONE | + syscall.PTRACE_O_TRACEEXEC | syscall.PTRACE_O_TRACEEXIT | syscall.PTRACE_O_TRACEFORK | + _PTRACE_O_TRACESECCOMP | syscall.PTRACE_O_TRACEVFORK | syscall.PTRACE_O_TRACEVFORKDONE) + if uintptr(data)&^validOpts != 0 { + return syserror.EINVAL + } + target.ptraceOpts = ptraceOptions{ + ExitKill: data&_PTRACE_O_EXITKILL != 0, + SysGood: data&syscall.PTRACE_O_TRACESYSGOOD != 0, + TraceClone: data&syscall.PTRACE_O_TRACECLONE != 0, + TraceExec: data&syscall.PTRACE_O_TRACEEXEC != 0, + TraceExit: data&syscall.PTRACE_O_TRACEEXIT != 0, + TraceFork: data&syscall.PTRACE_O_TRACEFORK != 0, + TraceSeccomp: data&_PTRACE_O_TRACESECCOMP != 0, + TraceVfork: data&syscall.PTRACE_O_TRACEVFORK != 0, + TraceVforkDone: data&syscall.PTRACE_O_TRACEVFORKDONE != 0, + } + return nil + + case syscall.PTRACE_GETEVENTMSG: + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + _, err := t.CopyOut(usermem.Addr(data), target.ptraceEventMsg) + return err + + default: + // PEEKSIGINFO is unimplemented but seems to have no users anywhere. + return syserror.EIO + } +} diff --git a/pkg/sentry/kernel/rseq.go b/pkg/sentry/kernel/rseq.go new file mode 100644 index 000000000..635372993 --- /dev/null +++ b/pkg/sentry/kernel/rseq.go @@ -0,0 +1,118 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/hostcpu" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// Restartable sequences, as described in https://lwn.net/Articles/650333/. + +// RSEQCriticalRegion describes a restartable sequence critical region. +type RSEQCriticalRegion struct { + // When a task in this thread group has its CPU preempted (as defined by + // platform.ErrContextCPUPreempted) or has a signal delivered to an + // application handler while its instruction pointer is in CriticalSection, + // set the instruction pointer to Restart and application register r10 (on + // amd64) to the former instruction pointer. + CriticalSection usermem.AddrRange + Restart usermem.Addr +} + +// RSEQAvailable returns true if t supports restartable sequences. +func (t *Task) RSEQAvailable() bool { + return t.k.useHostCores && t.k.Platform.DetectsCPUPreemption() +} + +// RSEQCriticalRegion returns a copy of t's thread group's current restartable +// sequence. +func (t *Task) RSEQCriticalRegion() RSEQCriticalRegion { + return *t.tg.rscr.Load().(*RSEQCriticalRegion) +} + +// SetRSEQCriticalRegion replaces t's thread group's restartable sequence. +// +// Preconditions: t.RSEQAvailable() == true. +func (t *Task) SetRSEQCriticalRegion(rscr RSEQCriticalRegion) error { + // These checks are somewhat more lenient than in Linux, which (bizarrely) + // requires rscr.CriticalSection to be non-empty and rscr.Restart to be + // outside of rscr.CriticalSection, even if rscr.CriticalSection.Start == 0 + // (which disables the critical region). + if rscr.CriticalSection.Start == 0 { + rscr.CriticalSection.End = 0 + rscr.Restart = 0 + t.tg.rscr.Store(&rscr) + return nil + } + if rscr.CriticalSection.Start >= rscr.CriticalSection.End { + return syserror.EINVAL + } + if rscr.CriticalSection.Contains(rscr.Restart) { + return syserror.EINVAL + } + // TODO: check that rscr.CriticalSection and rscr.Restart are in + // the application address range, for consistency with Linux + t.tg.rscr.Store(&rscr) + return nil +} + +// RSEQCPUAddr returns the address that RSEQ will keep updated with t's CPU +// number. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) RSEQCPUAddr() usermem.Addr { + return t.rseqCPUAddr +} + +// SetRSEQCPUAddr replaces the address that RSEQ will keep updated with t's CPU +// number. +// +// Preconditions: t.RSEQAvailable() == true. The caller must be running on the +// task goroutine. t's AddressSpace must be active. +func (t *Task) SetRSEQCPUAddr(addr usermem.Addr) error { + t.rseqCPUAddr = addr + if addr != 0 { + if err := t.rseqCopyOutCPU(); err != nil { + t.rseqCPUAddr = 0 + t.rseqCPU = -1 + return syserror.EINVAL // yes, EINVAL, not err or EFAULT + } + } else { + t.rseqCPU = -1 + } + return nil +} + +// Preconditions: The caller must be running on the task goroutine. t's +// AddressSpace must be active. +func (t *Task) rseqCopyOutCPU() error { + t.rseqCPU = int32(hostcpu.GetCPU()) + buf := t.CopyScratchBuffer(4) + usermem.ByteOrder.PutUint32(buf, uint32(t.rseqCPU)) + _, err := t.CopyOutBytes(t.rseqCPUAddr, buf) + return err +} + +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) rseqInterrupt() { + rscr := t.tg.rscr.Load().(*RSEQCriticalRegion) + if ip := t.Arch().IP(); rscr.CriticalSection.Contains(usermem.Addr(ip)) { + t.Debugf("Interrupted RSEQ critical section at %#x; restarting at %#x", ip, rscr.Restart) + t.Arch().SetIP(uintptr(rscr.Restart)) + t.Arch().SetRSEQInterruptedIP(ip) + } +} diff --git a/pkg/sentry/kernel/sched/BUILD b/pkg/sentry/kernel/sched/BUILD new file mode 100644 index 000000000..b533c51c4 --- /dev/null +++ b/pkg/sentry/kernel/sched/BUILD @@ -0,0 +1,20 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") + +go_library( + name = "sched", + srcs = [ + "cpuset.go", + "sched.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched", + visibility = ["//pkg/sentry:internal"], +) + +go_test( + name = "sched_test", + size = "small", + srcs = ["cpuset_test.go"], + embed = [":sched"], +) diff --git a/pkg/sentry/kernel/sched/cpuset.go b/pkg/sentry/kernel/sched/cpuset.go new file mode 100644 index 000000000..0a97603f0 --- /dev/null +++ b/pkg/sentry/kernel/sched/cpuset.go @@ -0,0 +1,105 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package sched + +import "math/bits" + +const ( + bitsPerByte = 8 + bytesPerLong = 8 // only for 64-bit architectures +) + +// CPUSet contains a bitmap to record CPU information. +// +// Note that this definition is only correct for little-endian architectures, +// since Linux's cpumask_t uses unsigned long. +type CPUSet []byte + +// CPUSetSize returns the size in bytes of a CPUSet that can contain num cpus. +func CPUSetSize(num uint) uint { + // NOTE: Applications may expect that the size of a CPUSet in + // bytes is always a multiple of sizeof(unsigned long), since this is true + // in Linux. Thus we always round up. + bytes := (num + bitsPerByte - 1) / bitsPerByte + longs := (bytes + bytesPerLong - 1) / bytesPerLong + return longs * bytesPerLong +} + +// NewCPUSet returns a CPUSet for the given number of CPUs which initially +// contains no CPUs. +func NewCPUSet(num uint) CPUSet { + return CPUSet(make([]byte, CPUSetSize(num))) +} + +// NewFullCPUSet returns a CPUSet for the given number of CPUs, all of which +// are present in the set. +func NewFullCPUSet(num uint) CPUSet { + c := NewCPUSet(num) + var i uint + for ; i < num/bitsPerByte; i++ { + c[i] = 0xff + } + if rem := num % bitsPerByte; rem != 0 { + c[i] = (1 << rem) - 1 + } + return c +} + +// Size returns the size of 'c' in bytes. +func (c CPUSet) Size() uint { + return uint(len(c)) +} + +// NumCPUs returns how many cpus are set in the CPUSet. +func (c CPUSet) NumCPUs() uint { + var n int + for _, b := range c { + n += bits.OnesCount8(b) + } + return uint(n) +} + +// Copy returns a copy of the CPUSet. +func (c CPUSet) Copy() CPUSet { + return append(CPUSet(nil), c...) +} + +// Set sets the bit corresponding to cpu. +func (c *CPUSet) Set(cpu uint) { + (*c)[cpu/bitsPerByte] |= 1 << (cpu % bitsPerByte) +} + +// ClearAbove clears bits corresponding to cpu and all higher cpus. +func (c *CPUSet) ClearAbove(cpu uint) { + i := cpu / bitsPerByte + if i >= c.Size() { + return + } + (*c)[i] &^= 0xff << (cpu % bitsPerByte) + for i++; i < c.Size(); i++ { + (*c)[i] = 0 + } +} + +// ForEachCPU iterates over the CPUSet and calls fn with the cpu index if +// it's set. +func (c CPUSet) ForEachCPU(fn func(uint)) { + for i := uint(0); i < c.Size()*bitsPerByte; i++ { + bit := uint(1) << (i & (bitsPerByte - 1)) + if uint(c[i/bitsPerByte])&bit == bit { + fn(i) + } + } +} diff --git a/pkg/sentry/kernel/sched/cpuset_test.go b/pkg/sentry/kernel/sched/cpuset_test.go new file mode 100644 index 000000000..8a6e12958 --- /dev/null +++ b/pkg/sentry/kernel/sched/cpuset_test.go @@ -0,0 +1,44 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package sched + +import ( + "testing" +) + +func TestNumCPUs(t *testing.T) { + for i := uint(0); i < 1024; i++ { + c := NewCPUSet(i) + for j := uint(0); j < i; j++ { + c.Set(j) + } + n := c.NumCPUs() + if n != i { + t.Errorf("got wrong number of cpus %d, want %d", n, i) + } + } +} + +func TestClearAbove(t *testing.T) { + const n = 1024 + c := NewFullCPUSet(n) + for i := uint(0); i < n; i++ { + cpu := n - i + c.ClearAbove(cpu) + if got := c.NumCPUs(); got != cpu { + t.Errorf("iteration %d: got %d cpus, wanted %d", i, got, cpu) + } + } +} diff --git a/pkg/sentry/kernel/sched/sched.go b/pkg/sentry/kernel/sched/sched.go new file mode 100644 index 000000000..f1de1da60 --- /dev/null +++ b/pkg/sentry/kernel/sched/sched.go @@ -0,0 +1,16 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package sched implements scheduler related features. +package sched diff --git a/pkg/sentry/kernel/seccomp.go b/pkg/sentry/kernel/seccomp.go new file mode 100644 index 000000000..b7c4a507f --- /dev/null +++ b/pkg/sentry/kernel/seccomp.go @@ -0,0 +1,205 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/binary" + "gvisor.googlesource.com/gvisor/pkg/bpf" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +const maxSyscallFilterInstructions = 1 << 15 + +type seccompResult int + +const ( + // seccompResultDeny indicates that a syscall should not be executed. + seccompResultDeny seccompResult = iota + + // seccompResultAllow indicates that a syscall should be executed. + seccompResultAllow + + // seccompResultKill indicates that the task should be killed immediately, + // with the exit status indicating that the task was killed by SIGSYS. + seccompResultKill + + // seccompResultTrace indicates that a ptracer was successfully notified as + // a result of a SECCOMP_RET_TRACE. + seccompResultTrace +) + +// seccompData is equivalent to struct seccomp_data, which contains the data +// passed to seccomp-bpf filters. +type seccompData struct { + // nr is the system call number. + nr int32 + + // arch is an AUDIT_ARCH_* value indicating the system call convention. + arch uint32 + + // instructionPointer is the value of the instruction pointer at the time + // of the system call. + instructionPointer uint64 + + // args contains the first 6 system call arguments. + args [6]uint64 +} + +func (d *seccompData) asBPFInput() bpf.Input { + return bpf.InputBytes{binary.Marshal(nil, usermem.ByteOrder, d), usermem.ByteOrder} +} + +func seccompSiginfo(t *Task, errno, sysno int32, ip usermem.Addr) *arch.SignalInfo { + si := &arch.SignalInfo{ + Signo: int32(linux.SIGSYS), + Errno: errno, + Code: arch.SYS_SECCOMP, + } + si.SetCallAddr(uint64(ip)) + si.SetSyscall(sysno) + si.SetArch(t.SyscallTable().AuditNumber) + return si +} + +// checkSeccompSyscall applies the task's seccomp filters before the execution +// of syscall sysno at instruction pointer ip. (These parameters must be passed +// in because vsyscalls do not use the values in t.Arch().) +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) checkSeccompSyscall(sysno int32, args arch.SyscallArguments, ip usermem.Addr) seccompResult { + result := t.evaluateSyscallFilters(sysno, args, ip) + switch result & linux.SECCOMP_RET_ACTION { + case linux.SECCOMP_RET_TRAP: + // "Results in the kernel sending a SIGSYS signal to the triggering + // task without executing the system call. ... The SECCOMP_RET_DATA + // portion of the return value will be passed as si_errno." - + // Documentation/prctl/seccomp_filter.txt + t.SendSignal(seccompSiginfo(t, int32(result&linux.SECCOMP_RET_DATA), sysno, ip)) + return seccompResultDeny + + case linux.SECCOMP_RET_ERRNO: + // "Results in the lower 16-bits of the return value being passed to + // userland as the errno without executing the system call." + t.Arch().SetReturn(-uintptr(result & linux.SECCOMP_RET_DATA)) + return seccompResultDeny + + case linux.SECCOMP_RET_TRACE: + // "When returned, this value will cause the kernel to attempt to + // notify a ptrace()-based tracer prior to executing the system call. + // If there is no tracer present, -ENOSYS is returned to userland and + // the system call is not executed." + if t.ptraceSeccomp(uint16(result & linux.SECCOMP_RET_DATA)) { + return seccompResultTrace + } + // This useless-looking temporary is needed because Go. + tmp := uintptr(syscall.ENOSYS) + t.Arch().SetReturn(-tmp) + return seccompResultDeny + + case linux.SECCOMP_RET_ALLOW: + // "Results in the system call being executed." + return seccompResultAllow + + case linux.SECCOMP_RET_KILL: + // "Results in the task exiting immediately without executing the + // system call. The exit status of the task will be SIGSYS, not + // SIGKILL." + fallthrough + default: // consistent with Linux + return seccompResultKill + } +} + +func (t *Task) evaluateSyscallFilters(sysno int32, args arch.SyscallArguments, ip usermem.Addr) uint32 { + data := seccompData{ + nr: sysno, + arch: t.tc.st.AuditNumber, + instructionPointer: uint64(ip), + } + // data.args is []uint64 and args is []arch.SyscallArgument (uintptr), so + // we can't do any slicing tricks or even use copy/append here. + for i, arg := range args { + if i >= len(data.args) { + break + } + data.args[i] = arg.Uint64() + } + input := data.asBPFInput() + + ret := uint32(linux.SECCOMP_RET_ALLOW) + // "Every filter successfully installed will be evaluated (in reverse + // order) for each system call the task makes." - kernel/seccomp.c + for i := len(t.syscallFilters) - 1; i >= 0; i-- { + thisRet, err := bpf.Exec(t.syscallFilters[i], input) + if err != nil { + t.Debugf("seccomp-bpf filter %d returned error: %v", i, err) + thisRet = linux.SECCOMP_RET_KILL + } + // "If multiple filters exist, the return value for the evaluation of a + // given system call will always use the highest precedent value." - + // Documentation/prctl/seccomp_filter.txt + // + // (Note that this contradicts prctl(2): "If the filters permit prctl() + // calls, then additional filters can be added; they are run in order + // until the first non-allow result is seen." prctl(2) is incorrect.) + // + // "The ordering ensures that a min_t() over composed return values + // always selects the least permissive choice." - + // include/uapi/linux/seccomp.h + if (thisRet & linux.SECCOMP_RET_ACTION) < (ret & linux.SECCOMP_RET_ACTION) { + ret = thisRet + } + } + + return ret +} + +// AppendSyscallFilter adds BPF program p as a system call filter. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) AppendSyscallFilter(p bpf.Program) error { + // Cap the combined length of all syscall filters (plus a penalty of 4 + // instructions per filter beyond the first) to + // maxSyscallFilterInstructions. (This restriction is inherited from + // Linux.) + totalLength := p.Length() + for _, f := range t.syscallFilters { + totalLength += f.Length() + 4 + } + if totalLength > maxSyscallFilterInstructions { + return syserror.ENOMEM + } + t.mu.Lock() + defer t.mu.Unlock() + t.syscallFilters = append(t.syscallFilters, p) + return nil +} + +// SeccompMode returns a SECCOMP_MODE_* constant indicating the task's current +// seccomp syscall filtering mode, appropriate for both prctl(PR_GET_SECCOMP) +// and /proc/[pid]/status. +func (t *Task) SeccompMode() int { + t.mu.Lock() + defer t.mu.Unlock() + if len(t.syscallFilters) > 0 { + return linux.SECCOMP_MODE_FILTER + } + return linux.SECCOMP_MODE_NONE +} diff --git a/pkg/sentry/kernel/semaphore/BUILD b/pkg/sentry/kernel/semaphore/BUILD new file mode 100644 index 000000000..1656ad126 --- /dev/null +++ b/pkg/sentry/kernel/semaphore/BUILD @@ -0,0 +1,62 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_generics:defs.bzl", "go_template_instance") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_template_instance( + name = "waiter_list", + out = "waiter_list.go", + package = "semaphore", + prefix = "waiter", + template = "//pkg/ilist:generic_list", + types = { + "Linker": "*waiter", + }, +) + +go_stateify( + name = "semaphore_state", + srcs = [ + "semaphore.go", + "waiter_list.go", + ], + out = "semaphore_autogen_state.go", + package = "semaphore", +) + +go_library( + name = "semaphore", + srcs = [ + "semaphore.go", + "semaphore_autogen_state.go", + "waiter_list.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/semaphore", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/abi/linux", + "//pkg/log", + "//pkg/sentry/context", + "//pkg/sentry/fs", + "//pkg/sentry/kernel/auth", + "//pkg/sentry/kernel/time", + "//pkg/state", + "//pkg/state/statefile", + "//pkg/syserror", + ], +) + +go_test( + name = "semaphore_test", + size = "small", + srcs = ["semaphore_test.go"], + embed = [":semaphore"], + deps = [ + "//pkg/abi/linux", + "//pkg/sentry/context", + "//pkg/sentry/context/contexttest", + "//pkg/sentry/kernel/auth", + "//pkg/syserror", + ], +) diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go new file mode 100644 index 000000000..19ad5d537 --- /dev/null +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -0,0 +1,473 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package semaphore implements System V semaphores. +package semaphore + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +const ( + valueMax = 32767 // SEMVMX + + // semaphoresMax is "maximum number of semaphores per semaphore ID" (SEMMSL). + semaphoresMax = 32000 + + // setMax is "system-wide limit on the number of semaphore sets" (SEMMNI). + setsMax = 32000 + + // semaphoresTotalMax is "system-wide limit on the number of semaphores" + // (SEMMNS = SEMMNI*SEMMSL). + semaphoresTotalMax = 1024000000 +) + +// Registry maintains a set of semaphores that can be found by key or ID. +type Registry struct { + // mu protects all fields below. + mu sync.Mutex `state:"nosave"` + semaphores map[int32]*Set + lastIDUsed int32 +} + +// Set represents a set of semaphores that can be operated atomically. +type Set struct { + // Id is a handle that identifies the set. + ID int32 + + // key is an user provided key that can be shared between processes. + key int32 + + // creator is the user that created the set. Immutable. + creator fs.FileOwner + + // mu protects all fields below. + mu sync.Mutex `state:"nosave"` + owner fs.FileOwner + perms fs.FilePermissions + opTime ktime.Time + changeTime ktime.Time + sems []sem + + // dead is set to true when the set is removed and can't be reached anymore. + // All waiters must wake up and fail when set is dead. + dead bool +} + +// sem represents a single semanphore from a set. +type sem struct { + value int16 + waiters waiterList `state:"zerovalue"` +} + +// waiter represents a caller that is waiting for the semaphore value to +// become positive or zero. +type waiter struct { + waiterEntry + + // value represents how much resource the waiter needs to wake up. + value int16 + ch chan struct{} +} + +// NewRegistry creates a new semaphore set registry. +func NewRegistry() *Registry { + return &Registry{semaphores: make(map[int32]*Set)} +} + +// FindOrCreate searches for a semaphore set that matches 'key'. If not found, +// it may create a new one if requested. If private is true, key is ignored and +// a new set is always created. If create is false, it fails if a set cannot +// be found. If exclusive is true, it fails if a set with the same key already +// exists. +func (r *Registry) FindOrCreate(ctx context.Context, key, nsems int32, mode linux.FileMode, private, create, exclusive bool) (*Set, error) { + if nsems < 0 || nsems > semaphoresMax { + return nil, syserror.EINVAL + } + + r.mu.Lock() + defer r.mu.Unlock() + + if !private { + // Look up an existing semaphore. + if set := r.findByKey(key); set != nil { + // Check that caller can access semaphore set. + creds := auth.CredentialsFromContext(ctx) + if !set.checkPerms(creds, fs.PermsFromMode(mode)) { + return nil, syserror.EACCES + } + + // Validate parameters. + if nsems > int32(set.size()) { + return nil, syserror.EINVAL + } + if create && exclusive { + return nil, syserror.EEXIST + } + return set, nil + } + + if !create { + // Semaphore not found and should not be created. + return nil, syserror.ENOENT + } + } + + // Zero is only valid if an existing set is found. + if nsems == 0 { + return nil, syserror.EINVAL + } + + // Apply system limits. + if len(r.semaphores) >= setsMax { + return nil, syserror.EINVAL + } + if r.totalSems() > int(semaphoresTotalMax-nsems) { + return nil, syserror.EINVAL + } + + // Finally create a new set. + owner := fs.FileOwnerFromContext(ctx) + perms := fs.FilePermsFromMode(mode) + return r.newSet(ctx, key, owner, owner, perms, nsems) +} + +// RemoveID removes set with give 'id' from the registry and marks the set as +// dead. All waiters will be awakened and fail. +func (r *Registry) RemoveID(id int32, creds *auth.Credentials) error { + r.mu.Lock() + defer r.mu.Unlock() + + set := r.semaphores[id] + if set == nil { + return syserror.EINVAL + } + + // "The effective user ID of the calling process must match the creator or + // owner of the semaphore set, or the caller must be privileged." + if !set.checkCredentials(creds) && !set.checkCapability(creds) { + return syserror.EACCES + } + + delete(r.semaphores, set.ID) + set.destroy() + return nil +} + +func (r *Registry) newSet(ctx context.Context, key int32, owner, creator fs.FileOwner, perms fs.FilePermissions, nsems int32) (*Set, error) { + set := &Set{ + key: key, + owner: owner, + creator: owner, + perms: perms, + changeTime: ktime.NowFromContext(ctx), + sems: make([]sem, nsems), + } + + // Find the next available ID. + for id := r.lastIDUsed + 1; id != r.lastIDUsed; id++ { + // Handle wrap around. + if id < 0 { + id = 0 + continue + } + if r.semaphores[id] == nil { + r.lastIDUsed = id + r.semaphores[id] = set + set.ID = id + return set, nil + } + } + + log.Warningf("Semaphore map is full, they must be leaking") + return nil, syserror.ENOMEM +} + +// FindByID looks up a set given an ID. +func (r *Registry) FindByID(id int32) *Set { + r.mu.Lock() + defer r.mu.Unlock() + return r.semaphores[id] +} + +func (r *Registry) findByKey(key int32) *Set { + for _, v := range r.semaphores { + if v.key == key { + return v + } + } + return nil +} + +func (r *Registry) totalSems() int { + totalSems := 0 + for _, v := range r.semaphores { + totalSems += v.size() + } + return totalSems +} + +func (s *Set) findSem(num int32) *sem { + if num < 0 || int(num) >= s.size() { + return nil + } + return &s.sems[num] +} + +func (s *Set) size() int { + return len(s.sems) +} + +// Change changes some fields from the set atomically. +func (s *Set) Change(ctx context.Context, creds *auth.Credentials, owner fs.FileOwner, perms fs.FilePermissions) error { + s.mu.Lock() + defer s.mu.Unlock() + + // "The effective UID of the calling process must match the owner or creator + // of the semaphore set, or the caller must be privileged." + if !s.checkCredentials(creds) && !s.checkCapability(creds) { + return syserror.EACCES + } + + s.owner = owner + s.perms = perms + s.changeTime = ktime.NowFromContext(ctx) + return nil +} + +// SetVal overrides a semaphore value, waking up waiters as needed. +func (s *Set) SetVal(ctx context.Context, num int32, val int16, creds *auth.Credentials) error { + if val < 0 || val > valueMax { + return syserror.ERANGE + } + + s.mu.Lock() + defer s.mu.Unlock() + + // "The calling process must have alter permission on the semaphore set." + if !s.checkPerms(creds, fs.PermMask{Write: true}) { + return syserror.EACCES + } + + sem := s.findSem(num) + if sem == nil { + return syserror.ERANGE + } + + // TODO: Clear undo entries in all processes + sem.value = val + s.changeTime = ktime.NowFromContext(ctx) + sem.wakeWaiters() + return nil +} + +// GetVal returns a semaphore value. +func (s *Set) GetVal(num int32, creds *auth.Credentials) (int16, error) { + s.mu.Lock() + defer s.mu.Unlock() + + // "The calling process must have read permission on the semaphore set." + if !s.checkPerms(creds, fs.PermMask{Read: true}) { + return 0, syserror.EACCES + } + + sem := s.findSem(num) + if sem == nil { + return 0, syserror.ERANGE + } + return sem.value, nil +} + +// ExecuteOps attempts to execute a list of operations to the set. It only +// suceeds when all operations can be applied. No changes are made if it fails. +// +// On failure, it may return an error (retries are hopeless) or it may return +// a channel that can be waited on before attempting again. +func (s *Set) ExecuteOps(ctx context.Context, ops []linux.Sembuf, creds *auth.Credentials) (chan struct{}, int32, error) { + s.mu.Lock() + defer s.mu.Unlock() + + // Did it race with a removal operation? + if s.dead { + return nil, 0, syserror.EIDRM + } + + // Validate the operations. + readOnly := true + for _, op := range ops { + if s.findSem(int32(op.SemNum)) == nil { + return nil, 0, syserror.EFBIG + } + if op.SemOp != 0 { + readOnly = false + } + } + + if !s.checkPerms(creds, fs.PermMask{Read: readOnly, Write: !readOnly}) { + return nil, 0, syserror.EACCES + } + + ch, num, err := s.executeOps(ctx, ops) + if err != nil { + return nil, 0, err + } + return ch, num, nil +} + +func (s *Set) executeOps(ctx context.Context, ops []linux.Sembuf) (chan struct{}, int32, error) { + // Changes to semaphores go to this slice temporarily until they all succeed. + tmpVals := make([]int16, len(s.sems)) + for i := range s.sems { + tmpVals[i] = s.sems[i].value + } + + for _, op := range ops { + sem := &s.sems[op.SemNum] + if op.SemOp == 0 { + // Handle 'wait for zero' operation. + if tmpVals[op.SemNum] != 0 { + // Semaphore isn't 0, must wait. + if op.SemFlg&linux.IPC_NOWAIT != 0 { + return nil, 0, syserror.ErrWouldBlock + } + + w := newWaiter(op.SemOp) + sem.waiters.PushBack(w) + return w.ch, int32(op.SemNum), nil + } + } else { + if op.SemOp < 0 { + // Handle 'wait' operation. + if -op.SemOp > valueMax { + return nil, 0, syserror.ERANGE + } + if -op.SemOp > tmpVals[op.SemNum] { + // Not enough resources, must wait. + if op.SemFlg&linux.IPC_NOWAIT != 0 { + return nil, 0, syserror.ErrWouldBlock + } + + w := newWaiter(op.SemOp) + sem.waiters.PushBack(w) + return w.ch, int32(op.SemNum), nil + } + } else { + // op.SemOp > 0: Handle 'signal' operation. + if tmpVals[op.SemNum] > valueMax-op.SemOp { + return nil, 0, syserror.ERANGE + } + } + + tmpVals[op.SemNum] += op.SemOp + } + } + + // All operations succeeded, apply them. + // TODO: handle undo operations. + for i, v := range tmpVals { + s.sems[i].value = v + s.sems[i].wakeWaiters() + } + s.opTime = ktime.NowFromContext(ctx) + return nil, 0, nil +} + +// AbortWait notifies that a waiter is giving up and will not wait on the +// channel anymore. +func (s *Set) AbortWait(num int32, ch chan struct{}) { + s.mu.Lock() + defer s.mu.Unlock() + + sem := &s.sems[num] + for w := sem.waiters.Front(); w != nil; w = w.Next() { + if w.ch == ch { + sem.waiters.Remove(w) + return + } + } + // Waiter may not be found in case it raced with wakeWaiters(). +} + +func (s *Set) checkCredentials(creds *auth.Credentials) bool { + return s.owner.UID == creds.EffectiveKUID || + s.owner.GID == creds.EffectiveKGID || + s.creator.UID == creds.EffectiveKUID || + s.creator.GID == creds.EffectiveKGID +} + +func (s *Set) checkCapability(creds *auth.Credentials) bool { + return creds.HasCapability(linux.CAP_IPC_OWNER) && creds.UserNamespace.MapFromKUID(s.owner.UID).Ok() +} + +func (s *Set) checkPerms(creds *auth.Credentials, reqPerms fs.PermMask) bool { + // Are we owner, or in group, or other? + p := s.perms.Other + if s.owner.UID == creds.EffectiveKUID { + p = s.perms.User + } else if creds.InGroup(s.owner.GID) { + p = s.perms.Group + } + + // Are permissions satisfied without capability checks? + if p.SupersetOf(reqPerms) { + return true + } + + return s.checkCapability(creds) +} + +func (s *Set) destroy() { + s.mu.Lock() + defer s.mu.Unlock() + + // Notify all waiters. Tney will fail on the next attempt to execute + // operations and return error. + s.dead = true + for _, s := range s.sems { + for w := s.waiters.Front(); w != nil; w = w.Next() { + w.ch <- struct{}{} + } + s.waiters.Reset() + } +} + +// wakeWaiters goes over all waiters and checks which of them can be notified. +func (s *sem) wakeWaiters() { + // Note that this will release all waiters waiting for 0 too. + for w := s.waiters.Front(); w != nil; { + if s.value < w.value { + // Still blocked, skip it. + continue + } + w.ch <- struct{}{} + old := w + w = w.Next() + s.waiters.Remove(old) + } +} + +func newWaiter(val int16) *waiter { + return &waiter{ + value: val, + ch: make(chan struct{}, 1), + } +} diff --git a/pkg/sentry/kernel/semaphore/semaphore_test.go b/pkg/sentry/kernel/semaphore/semaphore_test.go new file mode 100644 index 000000000..0386586ab --- /dev/null +++ b/pkg/sentry/kernel/semaphore/semaphore_test.go @@ -0,0 +1,172 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package semaphore + +import ( + "testing" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +func executeOps(ctx context.Context, t *testing.T, set *Set, ops []linux.Sembuf, block bool) chan struct{} { + ch, _, err := set.executeOps(ctx, ops) + if err != nil { + t.Fatalf("ExecuteOps(ops) failed, err: %v, ops: %+v", err, ops) + } + if block { + if ch == nil { + t.Fatalf("ExecuteOps(ops) got: nil, expected: !nil, ops: %+v", ops) + } + if signalled(ch) { + t.Fatalf("ExecuteOps(ops) channel should not have been signalled, ops: %+v", ops) + } + } else { + if ch != nil { + t.Fatalf("ExecuteOps(ops) got: %v, expected: nil, ops: %+v", ch, ops) + } + } + return ch +} + +func signalled(ch chan struct{}) bool { + select { + case <-ch: + return true + default: + return false + } +} + +func TestBasic(t *testing.T) { + ctx := contexttest.Context(t) + set := &Set{ID: 123, sems: make([]sem, 1)} + ops := []linux.Sembuf{ + linux.Sembuf{SemOp: 1}, + } + executeOps(ctx, t, set, ops, false) + + ops[0].SemOp = -1 + executeOps(ctx, t, set, ops, false) + + ops[0].SemOp = -1 + ch1 := executeOps(ctx, t, set, ops, true) + + ops[0].SemOp = 1 + executeOps(ctx, t, set, ops, false) + if !signalled(ch1) { + t.Fatalf("ExecuteOps(ops) channel should not have been signalled, ops: %+v", ops) + } +} + +func TestWaitForZero(t *testing.T) { + ctx := contexttest.Context(t) + set := &Set{ID: 123, sems: make([]sem, 1)} + ops := []linux.Sembuf{ + linux.Sembuf{SemOp: 0}, + } + executeOps(ctx, t, set, ops, false) + + ops[0].SemOp = -2 + ch1 := executeOps(ctx, t, set, ops, true) + + ops[0].SemOp = 0 + executeOps(ctx, t, set, ops, false) + + ops[0].SemOp = 1 + executeOps(ctx, t, set, ops, false) + + ops[0].SemOp = 0 + chZero1 := executeOps(ctx, t, set, ops, true) + + ops[0].SemOp = 0 + chZero2 := executeOps(ctx, t, set, ops, true) + + ops[0].SemOp = 1 + executeOps(ctx, t, set, ops, false) + if !signalled(ch1) { + t.Fatalf("ExecuteOps(ops) channel should have been signalled, ops: %+v, set: %+v", ops, set) + } + + ops[0].SemOp = -2 + executeOps(ctx, t, set, ops, false) + if !signalled(chZero1) { + t.Fatalf("ExecuteOps(ops) channel zero 1 should have been signalled, ops: %+v, set: %+v", ops, set) + } + if !signalled(chZero2) { + t.Fatalf("ExecuteOps(ops) channel zero 2 should have been signalled, ops: %+v, set: %+v", ops, set) + } +} + +func TestNoWait(t *testing.T) { + ctx := contexttest.Context(t) + set := &Set{ID: 123, sems: make([]sem, 1)} + ops := []linux.Sembuf{ + linux.Sembuf{SemOp: 1}, + } + executeOps(ctx, t, set, ops, false) + + ops[0].SemOp = -2 + ops[0].SemFlg = linux.IPC_NOWAIT + if _, _, err := set.executeOps(ctx, ops); err != syserror.ErrWouldBlock { + t.Fatalf("ExecuteOps(ops) wrong result, got: %v, expected: %v", err, syserror.ErrWouldBlock) + } + + ops[0].SemOp = 0 + ops[0].SemFlg = linux.IPC_NOWAIT + if _, _, err := set.executeOps(ctx, ops); err != syserror.ErrWouldBlock { + t.Fatalf("ExecuteOps(ops) wrong result, got: %v, expected: %v", err, syserror.ErrWouldBlock) + } +} + +func TestUnregister(t *testing.T) { + ctx := contexttest.Context(t) + r := NewRegistry() + set, err := r.FindOrCreate(ctx, 123, 2, linux.FileMode(0x600), true, true, true) + if err != nil { + t.Fatalf("FindOrCreate() failed, err: %v", err) + } + if got := r.FindByID(set.ID); got.ID != set.ID { + t.Fatalf("FindById(%d) failed, got: %+v, expected: %+v", set.ID, got, set) + } + + ops := []linux.Sembuf{ + linux.Sembuf{SemOp: -1}, + } + chs := make([]chan struct{}, 0, 5) + for i := 0; i < 5; i++ { + ch := executeOps(ctx, t, set, ops, true) + chs = append(chs, ch) + } + + creds := auth.CredentialsFromContext(ctx) + if err := r.RemoveID(set.ID, creds); err != nil { + t.Fatalf("RemoveID(%d) failed, err: %v", set.ID, err) + } + if !set.dead { + t.Fatalf("set is not dead: %+v", set) + } + if got := r.FindByID(set.ID); got != nil { + t.Fatalf("FindById(%d) failed, got: %+v, expected: nil", set.ID, got) + } + for i, ch := range chs { + if !signalled(ch) { + t.Fatalf("channel %d should have been signalled", i) + } + } +} diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go new file mode 100644 index 000000000..53d8fb844 --- /dev/null +++ b/pkg/sentry/kernel/sessions.go @@ -0,0 +1,462 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/refs" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// SessionID is the public identifier. +type SessionID ThreadID + +// ProcessGroupID is the public identifier. +type ProcessGroupID ThreadID + +// Session contains a leader threadgroup and a list of ProcessGroups. +type Session struct { + refs refs.AtomicRefCount + + // leader is the originator of the Session. + // + // Note that this may no longer be running (and may be reaped), so the + // ID is cached upon initial creation. The leader is still required + // however, since its PIDNamespace defines the scope of the Session. + // + // The leader is immutable. + leader *ThreadGroup + + // id is the cached identifier in the leader's namespace. + // + // The id is immutable. + id SessionID + + // ProcessGroups is a list of process groups in this Session. This is + // protected by TaskSet.mu. + processGroups processGroupList + + // sessionEntry is the embed for TaskSet.sessions. This is protected by + // TaskSet.mu. + sessionEntry +} + +// incRef grabs a reference. +func (s *Session) incRef() { + s.refs.IncRef() +} + +// decRef drops a reference. +// +// Precondition: callers must hold TaskSet.mu for writing. +func (s *Session) decRef() { + s.refs.DecRefWithDestructor(func() { + // Remove translations from the leader. + for ns := s.leader.pidns; ns != nil; ns = ns.parent { + id := ns.sids[s] + delete(ns.sids, s) + delete(ns.sessions, id) + } + + // Remove from the list of global Sessions. + s.leader.pidns.owner.sessions.Remove(s) + }) +} + +// ProcessGroup contains an originator threadgroup and a parent Session. +type ProcessGroup struct { + refs refs.AtomicRefCount // not exported. + + // originator is the originator of the group. + // + // See note re: leader in Session. The same applies here. + // + // The originator is immutable. + originator *ThreadGroup + + // id is the cached identifier in the originator's namespace. + // + // The id is immutable. + id ProcessGroupID + + // Session is the parent Session. + // + // The session is immutable. + session *Session + + // ancestors is the number of thread groups in this process group whose + // parent is in a different process group in the same session. + // + // The name is derived from the fact that process groups where + // ancestors is zero are considered "orphans". + // + // ancestors is protected by TaskSet.mu. + ancestors uint32 + + // processGroupEntry is the embedded entry for Sessions.groups. This is + // protected by TaskSet.mu. + processGroupEntry +} + +// incRefWithParent grabs a reference. +// +// This function is called when this ProcessGroup is being associated with some +// new ThreadGroup, tg. parentPG is the ProcessGroup of tg's parent +// ThreadGroup. If tg is init, then parentPG may be nil. +// +// Precondition: callers must hold TaskSet.mu for writing. +func (pg *ProcessGroup) incRefWithParent(parentPG *ProcessGroup) { + // We acquire an "ancestor" reference in the case of a nil parent. + // This is because the process being associated is init, and init can + // never be orphaned (we count it as always having an ancestor). + if pg != parentPG && (parentPG == nil || pg.session == parentPG.session) { + pg.ancestors++ + } + + pg.refs.IncRef() +} + +// decRefWithParent drops a reference. +// +// parentPG is per incRefWithParent. +// +// Precondition: callers must hold TaskSet.mu for writing. +func (pg *ProcessGroup) decRefWithParent(parentPG *ProcessGroup) { + // See incRefWithParent regarding parent == nil. + if pg != parentPG && (parentPG == nil || pg.session == parentPG.session) { + pg.ancestors-- + } + + alive := true + pg.refs.DecRefWithDestructor(func() { + alive = false // don't bother with handleOrphan. + + // Remove translations from the originator. + for ns := pg.originator.pidns; ns != nil; ns = ns.parent { + id := ns.pgids[pg] + delete(ns.pgids, pg) + delete(ns.processGroups, id) + } + + // Remove the list of process groups. + pg.session.processGroups.Remove(pg) + pg.session.decRef() + }) + if alive { + pg.handleOrphan() + } +} + +// parentPG returns the parent process group. +// +// Precondition: callers must hold TaskSet.mu. +func (tg *ThreadGroup) parentPG() *ProcessGroup { + if tg.leader.parent != nil { + return tg.leader.parent.tg.processGroup + } + return nil +} + +// handleOrphan checks whether the process group is an orphan and has any +// stopped jobs. If yes, then appropriate signals are delivered to each thread +// group within the process group. +// +// Precondition: callers must hold TaskSet.mu for writing. +func (pg *ProcessGroup) handleOrphan() { + // Check if this process is an orphan. + if pg.ancestors != 0 { + return + } + + // See if there are any stopped jobs. + hasStopped := false + pg.originator.pidns.owner.forEachThreadGroupLocked(func(tg *ThreadGroup) { + if tg.processGroup != pg { + return + } + tg.signalHandlers.mu.Lock() + if tg.groupStopPhase == groupStopComplete { + hasStopped = true + } + tg.signalHandlers.mu.Unlock() + }) + if !hasStopped { + return + } + + // Deliver appropriate signals to all thread groups. + pg.originator.pidns.owner.forEachThreadGroupLocked(func(tg *ThreadGroup) { + if tg.processGroup != pg { + return + } + tg.signalHandlers.mu.Lock() + tg.leader.sendSignalLocked(sigPriv(linux.SIGHUP), true /* group */) + tg.leader.sendSignalLocked(sigPriv(linux.SIGCONT), true /* group */) + tg.signalHandlers.mu.Unlock() + }) + + return +} + +// CreateSession creates a new Session, with the ThreadGroup as the leader. +// +// EPERM may be returned if either the given ThreadGroup is already a Session +// leader, or a ProcessGroup already exists for the ThreadGroup's ID. +func (tg *ThreadGroup) CreateSession() error { + tg.pidns.owner.mu.Lock() + defer tg.pidns.owner.mu.Unlock() + return tg.createSession() +} + +// createSession creates a new session for a threadgroup. +// +// Precondition: callers must hold TaskSet.mu for writing. +func (tg *ThreadGroup) createSession() error { + // Get the ID for this thread in the current namespace. + id := tg.pidns.tids[tg.leader] + + // Check if this ThreadGroup already leads a Session, or + // if the proposed group is already taken. + for s := tg.pidns.owner.sessions.Front(); s != nil; s = s.Next() { + if s.leader.pidns != tg.pidns { + continue + } + if s.leader == tg { + return syserror.EPERM + } + if s.id == SessionID(id) { + return syserror.EPERM + } + for pg := s.processGroups.Front(); pg != nil; pg = pg.Next() { + if pg.id == ProcessGroupID(id) { + return syserror.EPERM + } + } + } + + // Create a new Session, with a single reference. + s := &Session{ + id: SessionID(id), + leader: tg, + } + + // Create a new ProcessGroup, belonging to that Session. + // This also has a single reference (assigned below). + // + // Note that since this is a new session and a new process group, there + // will be zero ancestors for this process group. (It is an orphan at + // this point.) + pg := &ProcessGroup{ + id: ProcessGroupID(id), + originator: tg, + session: s, + ancestors: 0, + } + + // Tie them and return the result. + s.processGroups.PushBack(pg) + tg.pidns.owner.sessions.PushBack(s) + + // Leave the current group, and assign the new one. + if tg.processGroup != nil { + oldParentPG := tg.parentPG() + tg.forEachChildThreadGroupLocked(func(childTG *ThreadGroup) { + childTG.processGroup.incRefWithParent(pg) + childTG.processGroup.decRefWithParent(oldParentPG) + }) + tg.processGroup.decRefWithParent(oldParentPG) + tg.processGroup = pg + } else { + // The current process group may be nil only in the case of an + // unparented thread group (i.e. the init process). This would + // not normally occur, but we allow it for the convenience of + // CreateSession working from that point. There will be no + // child processes. We always say that the very first group + // created has ancestors (avoids checks elsewhere). + // + // Note that this mirrors the parent == nil logic in + // incRef/decRef/reparent, which counts nil as an ancestor. + tg.processGroup = pg + tg.processGroup.ancestors++ + } + + // Ensure a translation is added to all namespaces. + for ns := tg.pidns; ns != nil; ns = ns.parent { + local := ns.tids[tg.leader] + ns.sids[s] = SessionID(local) + ns.sessions[SessionID(local)] = s + ns.pgids[pg] = ProcessGroupID(local) + ns.processGroups[ProcessGroupID(local)] = pg + } + + return nil +} + +// CreateProcessGroup creates a new process group. +// +// An EPERM error will be returned if the ThreadGroup belongs to a different +// Session, is a Session leader or the group already exists. +func (tg *ThreadGroup) CreateProcessGroup() error { + tg.pidns.owner.mu.Lock() + defer tg.pidns.owner.mu.Unlock() + + // Get the ID for this thread in the current namespace. + id := tg.pidns.tids[tg.leader] + + // Per above, check for a Session leader or existing group. + for s := tg.pidns.owner.sessions.Front(); s != nil; s = s.Next() { + if s.leader.pidns != tg.pidns { + continue + } + if s.leader == tg { + return syserror.EPERM + } + for pg := s.processGroups.Front(); pg != nil; pg = pg.Next() { + if pg.id == ProcessGroupID(id) { + return syserror.EPERM + } + } + } + + // Create a new ProcessGroup, belonging to the current Session. + // + // We manually adjust the ancestors if the parent is in the same + // session. + tg.processGroup.session.incRef() + pg := &ProcessGroup{ + id: ProcessGroupID(id), + originator: tg, + session: tg.processGroup.session, + } + if tg.leader.parent != nil && tg.leader.parent.tg.processGroup.session == pg.session { + pg.ancestors++ + } + + // Assign the new process group; adjust children. + oldParentPG := tg.parentPG() + tg.forEachChildThreadGroupLocked(func(childTG *ThreadGroup) { + childTG.processGroup.incRefWithParent(pg) + childTG.processGroup.decRefWithParent(oldParentPG) + }) + tg.processGroup.decRefWithParent(oldParentPG) + tg.processGroup = pg + + // Ensure this translation is added to all namespaces. + for ns := tg.pidns; ns != nil; ns = ns.parent { + local := ns.tids[tg.leader] + ns.pgids[pg] = ProcessGroupID(local) + ns.processGroups[ProcessGroupID(local)] = pg + } + + return nil +} + +// JoinProcessGroup joins an existing process group. +// +// This function will return EACCES if an exec has been performed since fork +// by the given ThreadGroup, and EPERM if the Sessions are not the same or the +// group does not exist. +// +// If checkExec is set, then the join is not permitted after the process has +// executed exec at least once. +func (tg *ThreadGroup) JoinProcessGroup(pidns *PIDNamespace, pgid ProcessGroupID, checkExec bool) error { + pidns.owner.mu.Lock() + defer pidns.owner.mu.Unlock() + + // Lookup the ProcessGroup. + pg := pidns.processGroups[pgid] + if pg == nil { + return syserror.EPERM + } + + // Disallow the join if an execve has performed, per POSIX. + if checkExec && tg.execed { + return syserror.EACCES + } + + // See if it's in the same session as ours. + if pg.session != tg.processGroup.session { + return syserror.EPERM + } + + // Join the group; adjust children. + parentPG := tg.parentPG() + pg.incRefWithParent(parentPG) + tg.forEachChildThreadGroupLocked(func(childTG *ThreadGroup) { + childTG.processGroup.incRefWithParent(pg) + childTG.processGroup.decRefWithParent(tg.processGroup) + }) + tg.processGroup.decRefWithParent(parentPG) + tg.processGroup = pg + + return nil +} + +// Session returns the ThreadGroup's Session. +// +// A reference is not taken on the session. +func (tg *ThreadGroup) Session() *Session { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + return tg.processGroup.session +} + +// IDOfSession returns the Session assigned to s in PID namespace ns. +// +// If this group isn't visible in this namespace, zero will be returned. It is +// the callers responsibility to check that before using this function. +func (pidns *PIDNamespace) IDOfSession(s *Session) SessionID { + pidns.owner.mu.RLock() + defer pidns.owner.mu.RUnlock() + return pidns.sids[s] +} + +// SessionWithID returns the Session with the given ID in the PID namespace ns, +// or nil if that given ID is not defined in this namespace. +// +// A reference is not taken on the session. +func (pidns *PIDNamespace) SessionWithID(id SessionID) *Session { + pidns.owner.mu.RLock() + defer pidns.owner.mu.RUnlock() + return pidns.sessions[id] +} + +// ProcessGroup returns the ThreadGroup's ProcessGroup. +// +// A reference is not taken on the process group. +func (tg *ThreadGroup) ProcessGroup() *ProcessGroup { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + return tg.processGroup +} + +// IDOfProcessGroup returns the process group assigned to pg in PID namespace ns. +// +// The same constraints apply as IDOfSession. +func (pidns *PIDNamespace) IDOfProcessGroup(pg *ProcessGroup) ProcessGroupID { + pidns.owner.mu.RLock() + defer pidns.owner.mu.RUnlock() + return pidns.pgids[pg] +} + +// ProcessGroupWithID returns the ProcessGroup with the given ID in the PID +// namespace ns, or nil if that given ID is not defined in this namespace. +// +// A reference is not taken on the process group. +func (pidns *PIDNamespace) ProcessGroupWithID(id ProcessGroupID) *ProcessGroup { + pidns.owner.mu.RLock() + defer pidns.owner.mu.RUnlock() + return pidns.processGroups[id] +} diff --git a/pkg/sentry/kernel/signal.go b/pkg/sentry/kernel/signal.go new file mode 100644 index 000000000..8edd05cdf --- /dev/null +++ b/pkg/sentry/kernel/signal.go @@ -0,0 +1,69 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" +) + +// SignalPanic is used to panic the running threads. It is a signal which +// cannot be used by the application: it must be caught and ignored by the +// runtime (in order to catch possible races). +const SignalPanic = linux.SIGUSR2 + +// sendExternalSignal is called when an asynchronous signal is sent to the +// sentry ("in sentry context"). On some platforms, it may also be called when +// an asynchronous signal is sent to sandboxed application threads ("in +// application context"). +// +// context is used only for debugging to differentiate these cases. +// +// Returns false if signal could not be sent because the Kernel is not fully +// initialized yet. +func (k *Kernel) sendExternalSignal(info *arch.SignalInfo, context string) bool { + switch linux.Signal(info.Signo) { + case platform.SignalInterrupt: + // Assume that a call to platform.Context.Interrupt() misfired. + return true + + case SignalPanic: + // SignalPanic is also specially handled in sentry setup to ensure that + // it causes a panic even after tasks exit, but SignalPanic may also + // be sent here if it is received while in app context. + panic("Signal-induced panic") + + default: + log.Infof("Received external signal %d in %s context", info.Signo, context) + if k.globalInit == nil { + log.Warningf("Received external signal %d before init created", info.Signo) + return false + } + k.globalInit.SendSignal(info) + } + + return true +} + +// sigPriv returns a SignalInfo representing a signal sent by the sentry. (The +// name reflects its equivalence to Linux's SEND_SIG_PRIV.) +func sigPriv(sig linux.Signal) *arch.SignalInfo { + return &arch.SignalInfo{ + Signo: int32(sig), + Code: arch.SignalInfoKernel, + } +} diff --git a/pkg/sentry/kernel/signal_handlers.go b/pkg/sentry/kernel/signal_handlers.go new file mode 100644 index 000000000..21ba4ee70 --- /dev/null +++ b/pkg/sentry/kernel/signal_handlers.go @@ -0,0 +1,79 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" +) + +// SignalHandlers holds information about signal actions. +type SignalHandlers struct { + // mu protects actions, as well as the signal state of all tasks and thread + // groups using this SignalHandlers object. (See comment on + // ThreadGroup.signalHandlers.) + mu sync.Mutex `state:"nosave"` + + // actions is the action to be taken upon receiving each signal. + actions map[linux.Signal]arch.SignalAct +} + +// NewSignalHandlers returns a new SignalHandlers specifying all default +// actions. +func NewSignalHandlers() *SignalHandlers { + return &SignalHandlers{ + actions: make(map[linux.Signal]arch.SignalAct), + } +} + +// Fork returns a copy of sh for a new thread group. +func (sh *SignalHandlers) Fork() *SignalHandlers { + sh2 := NewSignalHandlers() + sh.mu.Lock() + defer sh.mu.Unlock() + for sig, act := range sh.actions { + sh2.actions[sig] = act + } + return sh2 +} + +// CopyForExec returns a copy of sh for a thread group that is undergoing an +// execve. (See comments in Task.finishExec.) +func (sh *SignalHandlers) CopyForExec() *SignalHandlers { + sh2 := NewSignalHandlers() + sh.mu.Lock() + defer sh.mu.Unlock() + for sig, act := range sh.actions { + if act.Handler == arch.SignalActIgnore { + sh2.actions[sig] = arch.SignalAct{ + Handler: arch.SignalActIgnore, + } + } + } + return sh2 +} + +// dequeueActionLocked returns the SignalAct that should be used to handle sig. +// +// Preconditions: sh.mu must be locked. +func (sh *SignalHandlers) dequeueAction(sig linux.Signal) arch.SignalAct { + act := sh.actions[sig] + if act.IsResetHandler() { + delete(sh.actions, sig) + } + return act +} diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go new file mode 100644 index 000000000..e20fa3eb6 --- /dev/null +++ b/pkg/sentry/kernel/syscalls.go @@ -0,0 +1,305 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + "sync" + "sync/atomic" + + "gvisor.googlesource.com/gvisor/pkg/abi" + "gvisor.googlesource.com/gvisor/pkg/bits" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" +) + +// maxSyscallNum is the highest supported syscall number. +// +// The types below create fast lookup slices for all syscalls. This maximum +// serves as a sanity check that we don't allocate huge slices for a very large +// syscall. +const maxSyscallNum = 2000 + +// SyscallFn is a syscall implementation. +type SyscallFn func(t *Task, args arch.SyscallArguments) (uintptr, *SyscallControl, error) + +// MissingFn is a syscall to be called when an implementation is missing. +type MissingFn func(t *Task, sysno uintptr, args arch.SyscallArguments) (uintptr, error) + +// Possible flags for SyscallFlagsTable.enable. +const ( + // syscallPresent indicates that this is not a missing syscall. + // + // This flag is used internally in SyscallFlagsTable. + syscallPresent = 1 << iota + + // StraceEnableLog enables syscall log tracing. + StraceEnableLog + + // StraceEnableEvent enables syscall event tracing. + StraceEnableEvent + + // ExternalBeforeEnable enables the external hook before syscall execution. + ExternalBeforeEnable + + // ExternalAfterEnable enables the external hook after syscall execution. + ExternalAfterEnable +) + +// StraceEnableBits combines both strace log and event flags. +const StraceEnableBits = StraceEnableLog | StraceEnableEvent + +// SyscallFlagsTable manages a set of enable/disable bit fields on a per-syscall +// basis. +type SyscallFlagsTable struct { + // mu protects writes to the fields below. + // + // Atomic loads are always allowed. Atomic stores are allowed only + // while mu is held. + mu sync.Mutex + + // enable contains the enable bits for each syscall. + // + // missing syscalls have the same value in enable as missingEnable to + // avoid an extra branch in Word. + enable []uint32 + + // missingEnable contains the enable bits for missing syscalls. + missingEnable uint32 +} + +// Init initializes the struct, with all syscalls in table set to enable. +// +// max is the largest syscall number in table. +func (e *SyscallFlagsTable) init(table map[uintptr]SyscallFn, max uintptr) { + e.enable = make([]uint32, max+1) + for num := range table { + e.enable[num] = syscallPresent + } +} + +// Word returns the enable bitfield for sysno. +func (e *SyscallFlagsTable) Word(sysno uintptr) uint32 { + if sysno < uintptr(len(e.enable)) { + return atomic.LoadUint32(&e.enable[sysno]) + } + + return atomic.LoadUint32(&e.missingEnable) +} + +// Enable sets enable bit bit for all syscalls based on s. +// +// Syscalls missing from s are disabled. +// +// Syscalls missing from the initial table passed to Init cannot be added as +// individual syscalls. If present in s they will be ignored. +// +// Callers to Word may see either the old or new value while this function +// is executing. +func (e *SyscallFlagsTable) Enable(bit uint32, s map[uintptr]bool, missingEnable bool) { + e.mu.Lock() + defer e.mu.Unlock() + + missingVal := atomic.LoadUint32(&e.missingEnable) + if missingEnable { + missingVal |= bit + } else { + missingVal &^= bit + } + atomic.StoreUint32(&e.missingEnable, missingVal) + + for num := range e.enable { + val := atomic.LoadUint32(&e.enable[num]) + if !bits.IsOn32(val, syscallPresent) { + // Missing. + atomic.StoreUint32(&e.enable[num], missingVal) + continue + } + + if s[uintptr(num)] { + val |= bit + } else { + val &^= bit + } + atomic.StoreUint32(&e.enable[num], val) + } +} + +// EnableAll sets enable bit bit for all syscalls, present and missing. +func (e *SyscallFlagsTable) EnableAll(bit uint32) { + e.mu.Lock() + defer e.mu.Unlock() + + missingVal := atomic.LoadUint32(&e.missingEnable) + missingVal |= bit + atomic.StoreUint32(&e.missingEnable, missingVal) + + for num := range e.enable { + val := atomic.LoadUint32(&e.enable[num]) + if !bits.IsOn32(val, syscallPresent) { + // Missing. + atomic.StoreUint32(&e.enable[num], missingVal) + continue + } + + val |= bit + atomic.StoreUint32(&e.enable[num], val) + } +} + +// Stracer traces syscall execution. +type Stracer interface { + // SyscallEnter is called on syscall entry. + // + // The returned private data is passed to SyscallExit. + // + // TODO: remove kernel imports from the strace package so + // that the type can be used directly. + SyscallEnter(t *Task, sysno uintptr, args arch.SyscallArguments, flags uint32) interface{} + + // SyscallExit is called on syscall exit. + SyscallExit(context interface{}, t *Task, sysno, rval uintptr, err error) +} + +// SyscallTable is a lookup table of system calls. Critically, a SyscallTable +// is *immutable*. In order to make supporting suspend and resume sane, they +// must be uniquely registered and may not change during operation. +type SyscallTable struct { + // OS is the operating system that this syscall table implements. + OS abi.OS `state:"wait"` + + // Arch is the architecture that this syscall table targets. + Arch arch.Arch `state:"wait"` + + // The OS version that this syscall table implements. + Version Version `state:"manual"` + + // AuditNumber is a numeric constant that represents the syscall table. If + // non-zero, auditNumber must be one of the AUDIT_ARCH_* values defined by + // linux/audit.h. + AuditNumber uint32 `state:"manual"` + + // Table is the collection of functions. + Table map[uintptr]SyscallFn `state:"manual"` + + // lookup is a fixed-size array that holds the syscalls (indexed by + // their numbers). It is used for fast look ups. + lookup []SyscallFn `state:"manual"` + + // Emulate is a collection of instruction addresses to emulate. The + // keys are addresses, and the values are system call numbers. + Emulate map[usermem.Addr]uintptr `state:"manual"` + + // The function to call in case of a missing system call. + Missing MissingFn `state:"manual"` + + // Stracer traces this syscall table. + Stracer Stracer `state:"manual"` + + // External is used to handle an external callback. + External func(*Kernel) `state:"manual"` + + // ExternalFilterBefore is called before External is called before the syscall is executed. + // External is not called if it returns false. + ExternalFilterBefore func(*Task, uintptr, arch.SyscallArguments) bool `state:"manual"` + + // ExternalFilterAfter is called before External is called after the syscall is executed. + // External is not called if it returns false. + ExternalFilterAfter func(*Task, uintptr, arch.SyscallArguments) bool `state:"manual"` + + // FeatureEnable stores the strace and one-shot enable bits. + FeatureEnable SyscallFlagsTable `state:"manual"` +} + +// allSyscallTables contains all known tables. +var allSyscallTables []*SyscallTable + +// SyscallTables returns a read-only slice of registered SyscallTables. +func SyscallTables() []*SyscallTable { + return allSyscallTables +} + +// LookupSyscallTable returns the SyscallCall table for the OS/Arch combination. +func LookupSyscallTable(os abi.OS, a arch.Arch) (*SyscallTable, bool) { + for _, s := range allSyscallTables { + if s.OS == os && s.Arch == a { + return s, true + } + } + return nil, false +} + +// RegisterSyscallTable registers a new syscall table for use by a Kernel. +func RegisterSyscallTable(s *SyscallTable) { + if s.Table == nil { + // Ensure non-nil lookup table. + s.Table = make(map[uintptr]SyscallFn) + } + if s.Emulate == nil { + // Ensure non-nil emulate table. + s.Emulate = make(map[usermem.Addr]uintptr) + } + + var max uintptr + for num := range s.Table { + if num > max { + max = num + } + } + + if max > maxSyscallNum { + panic(fmt.Sprintf("SyscallTable %+v contains too large syscall number %d", s, max)) + } + + s.lookup = make([]SyscallFn, max+1) + + // Initialize the fast-lookup table. + for num, fn := range s.Table { + s.lookup[num] = fn + } + + s.FeatureEnable.init(s.Table, max) + + if _, ok := LookupSyscallTable(s.OS, s.Arch); ok { + panic(fmt.Sprintf("Duplicate SyscallTable registered for OS %v Arch %v", s.OS, s.Arch)) + } + + // Save a reference to this table. + // + // This is required for a Kernel to find the table and for save/restore + // operations below. + allSyscallTables = append(allSyscallTables, s) +} + +// Lookup returns the syscall implementation, if one exists. +func (s *SyscallTable) Lookup(sysno uintptr) SyscallFn { + if sysno < uintptr(len(s.lookup)) { + return s.lookup[sysno] + } + + return nil +} + +// LookupEmulate looks up an emulation syscall number. +func (s *SyscallTable) LookupEmulate(addr usermem.Addr) (uintptr, bool) { + sysno, ok := s.Emulate[addr] + return sysno, ok +} + +// mapLookup is similar to Lookup, except that it only uses the syscall table, +// that is, it skips the fast look array. This is available for benchmarking. +func (s *SyscallTable) mapLookup(sysno uintptr) SyscallFn { + return s.Table[sysno] +} diff --git a/pkg/sentry/kernel/syscalls_state.go b/pkg/sentry/kernel/syscalls_state.go new file mode 100644 index 000000000..826809a70 --- /dev/null +++ b/pkg/sentry/kernel/syscalls_state.go @@ -0,0 +1,29 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import "fmt" + +// afterLoad is invoked by stateify. +func (s *SyscallTable) afterLoad() { + otherTable, ok := LookupSyscallTable(s.OS, s.Arch) + if !ok { + // Couldn't find a reference? + panic(fmt.Sprintf("syscall table not found for OS %v Arch %v", s.OS, s.Arch)) + } + + // Copy the table. + *s = *otherTable +} diff --git a/pkg/sentry/kernel/syslog.go b/pkg/sentry/kernel/syslog.go new file mode 100644 index 000000000..31541749e --- /dev/null +++ b/pkg/sentry/kernel/syslog.go @@ -0,0 +1,100 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + "math/rand" + "sync" +) + +// syslog represents a sentry-global kernel log. +// +// Currently, it contains only fun messages for a dmesg easter egg. +type syslog struct { + // mu protects the below. + mu sync.Mutex `state:"nosave"` + + // msg is the syslog message buffer. It is lazily initialized. + msg []byte +} + +// Log returns a copy of the syslog. +func (s *syslog) Log() []byte { + s.mu.Lock() + defer s.mu.Unlock() + + if s.msg != nil { + // Already initialized, just return a copy. + o := make([]byte, len(s.msg)) + copy(o, s.msg) + return o + } + + // Not initialized, create message. + allMessages := []string{ + "Synthesizing system calls...", + "Mounting deweydecimalfs...", + "Moving files to filing cabinet...", + "Digging up root...", + "Constructing home...", + "Segmenting fault lines...", + "Creating bureaucratic processes...", + "Searching for needles in stacks...", + "Preparing for the zombie uprising...", + "Feeding the init monster...", + "Creating cloned children...", + "Daemonizing children...", + "Waiting for children...", + "Gathering forks...", + "Committing treasure map to memory...", + "Reading process obituaries...", + "Searching for socket adapter...", + "Creating process schedule...", + "Generating random numbers by fair dice roll...", + "Rewriting operating system in Javascript...", + "Consulting tar man page...", + "Forking spaghetti code...", + "Checking naughty and nice process list...", + "Checking naughty and nice process list...", // Check it up to twice. + "Granting licence to kill(2)...", // British spelling for British movie. + "Letting the watchdogs out...", + } + + selectMessage := func() string { + i := rand.Intn(len(allMessages)) + m := allMessages[i] + + // Delete the selected message. + allMessages[i] = allMessages[len(allMessages)-1] + allMessages = allMessages[:len(allMessages)-1] + + return m + } + + time := 0.0 + for i := 0; i < 10; i++ { + time += rand.Float64() / 2 + s.msg = append(s.msg, []byte(fmt.Sprintf("<6>[%11.6f] %s\n", time, selectMessage()))...) + } + + time += rand.Float64() / 2 + s.msg = append(s.msg, []byte(fmt.Sprintf("<6>[%11.6f] Ready!\n", time))...) + + // Return a copy. + o := make([]byte, len(s.msg)) + copy(o, s.msg) + return o +} diff --git a/pkg/sentry/kernel/table_test.go b/pkg/sentry/kernel/table_test.go new file mode 100644 index 000000000..71ca75555 --- /dev/null +++ b/pkg/sentry/kernel/table_test.go @@ -0,0 +1,108 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "testing" + + "gvisor.googlesource.com/gvisor/pkg/abi" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" +) + +const ( + maxTestSyscall = 1000 +) + +func createSyscallTable() *SyscallTable { + m := make(map[uintptr]SyscallFn) + for i := uintptr(0); i <= maxTestSyscall; i++ { + j := i + m[i] = func(*Task, arch.SyscallArguments) (uintptr, *SyscallControl, error) { + return j, nil, nil + } + } + + s := &SyscallTable{ + OS: abi.Linux, + Arch: arch.AMD64, + Table: m, + } + + RegisterSyscallTable(s) + return s +} + +func TestTable(t *testing.T) { + table := createSyscallTable() + defer func() { + // Cleanup registered tables to keep tests separate. + allSyscallTables = []*SyscallTable{} + }() + + // Go through all functions and check that they return the right value. + for i := uintptr(0); i < maxTestSyscall; i++ { + fn := table.Lookup(i) + if fn == nil { + t.Errorf("Syscall %v is set to nil", i) + continue + } + + v, _, _ := fn(nil, arch.SyscallArguments{}) + if v != i { + t.Errorf("Wrong return value for syscall %v: expected %v, got %v", i, i, v) + } + } + + // Check that values outside the range return nil. + for i := uintptr(maxTestSyscall + 1); i < maxTestSyscall+100; i++ { + fn := table.Lookup(i) + if fn != nil { + t.Errorf("Syscall %v is not nil: %v", i, fn) + continue + } + } +} + +func BenchmarkTableLookup(b *testing.B) { + table := createSyscallTable() + + b.ResetTimer() + + j := uintptr(0) + for i := 0; i < b.N; i++ { + table.Lookup(j) + j = (j + 1) % 310 + } + + b.StopTimer() + // Cleanup registered tables to keep tests separate. + allSyscallTables = []*SyscallTable{} +} + +func BenchmarkTableMapLookup(b *testing.B) { + table := createSyscallTable() + + b.ResetTimer() + + j := uintptr(0) + for i := 0; i < b.N; i++ { + table.mapLookup(j) + j = (j + 1) % 310 + } + + b.StopTimer() + // Cleanup registered tables to keep tests separate. + allSyscallTables = []*SyscallTable{} +} diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go new file mode 100644 index 000000000..3d2e035e9 --- /dev/null +++ b/pkg/sentry/kernel/task.go @@ -0,0 +1,606 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "sync" + "sync/atomic" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/bpf" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/limits" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + ssync "gvisor.googlesource.com/gvisor/pkg/sync" +) + +// Task represents a thread of execution in the untrusted app. It +// includes registers and any thread-specific state that you would +// normally expect. +// +// Each task is associated with a goroutine, called the task goroutine, that +// executes code (application code, system calls, etc.) on behalf of that task. +// See Task.run (task_run.go). +// +// All fields that are "owned by the task goroutine" can only be mutated by the +// task goroutine while it is running. The task goroutine does not require +// synchronization to read these fields, although it still requires +// synchronization as described for those fields to mutate them. +// +// All fields that are "exclusive to the task goroutine" can only be accessed +// by the task goroutine while it is running. The task goroutine does not +// require synchronization to read or write these fields. +type Task struct { + taskNode + + // runState is what the task goroutine is executing if it is not stopped. + // If runState is nil, the task goroutine should exit or has exited. + // runState is exclusive to the task goroutine. + runState taskRunState + + // haveSyscallReturn is true if tc.Arch().Return() represents a value + // returned by a syscall (or set by ptrace after a syscall). + // + // haveSyscallReturn is exclusive to the task goroutine. + haveSyscallReturn bool + + // interruptChan is notified whenever the task goroutine is interrupted + // (usually by a pending signal). interruptChan is effectively a condition + // variable that can be used in select statements. + // + // interruptChan is not saved; because saving interrupts all tasks, + // interruptChan is always notified after restore (see Task.run). + interruptChan chan struct{} `state:"nosave"` + + // gosched contains the current scheduling state of the task goroutine. + // + // gosched is protected by goschedSeq. gosched is owned by the task + // goroutine. + goschedSeq ssync.SeqCount `state:"nosave"` + gosched TaskGoroutineSchedInfo + + // yieldCount is the number of times the task goroutine has called + // Task.InterruptibleSleepStart, Task.UninterruptibleSleepStart, or + // Task.Yield(), voluntarily ceasing execution. + // + // yieldCount is accessed using atomic memory operations. yieldCount is + // owned by the task goroutine. + yieldCount uint64 + + // pendingSignals is the set of pending signals that may be handled only by + // this task. + // + // pendingSignals is protected by (taskNode.)tg.signalHandlers.mu + // (hereafter "the signal mutex"); see comment on + // ThreadGroup.signalHandlers. + pendingSignals pendingSignals + + // If haveSavedSignalMask is true, savedSignalMask is the signal mask that + // should be applied after the task has either delivered one signal to a + // user handler or is about to resume execution in the untrusted + // application. + // + // Both haveSavedSignalMask and savedSignalMask are exclusive to the task + // goroutine. + haveSavedSignalMask bool + savedSignalMask linux.SignalSet + + // signalStack is the alternate signal stack used by signal handlers for + // which the SA_ONSTACK flag is set. + // + // signalStack is exclusive to the task goroutine. + signalStack arch.SignalStack + + // If groupStopRequired is true, the task should enter a group stop in the + // interrupt path. groupStopRequired is not redundant with + // tg.groupStopPhase != groupStopNone, because ptrace allows tracers to + // resume individual tasks from a group stop without ending the group stop + // as a whole. + // + // groupStopRequired is analogous to JOBCTL_TRAP_STOP in Linux, except that + // Linux only uses that flag for ptraced tasks. + // + // groupStopRequired is protected by the signal mutex. + groupStopRequired bool + + // If groupStopAcknowledged is true, the task has already acknowledged that + // it is entering the most recent group stop that has been initiated on its + // thread group. groupStopAcknowledged is only meaningful if + // tg.groupStopPhase == groupStopInitiated. + // + // groupStopAcknowledged is analogous to !JOBCTL_STOP_CONSUME in Linux. + // + // groupStopAcknowledged is protected by the signal mutex. + groupStopAcknowledged bool + + // If stop is not nil, it is the internally-initiated condition that + // currently prevents the task goroutine from running. + // + // stop is protected by the signal mutex. + stop TaskStop + + // stopCount is the number of active external stops (calls to + // Task.BeginExternalStop that have not been paired with a call to + // Task.EndExternalStop), plus 1 if stop is not nil. Hence stopCount is + // non-zero if the task goroutine should stop. + // + // Mutating stopCount requires both locking the signal mutex and using + // atomic memory operations. Reading stopCount requires either locking the + // signal mutex or using atomic memory operations. This allows Task.doStop + // to require only a single atomic read in the common case where stopCount + // is 0. + // + // stopCount is not saved, because external stops cannot be retained across + // a save/restore cycle. (Suppose a sentryctl command issues an external + // stop; after a save/restore cycle, the restored sentry has no knowledge + // of the pre-save sentryctl command, and the stopped task would remain + // stopped forever.) + stopCount int32 `state:"nosave"` + + // endStopCond is signaled when stopCount transitions to 0. The combination + // of stopCount and endStopCond effectively form a sync.WaitGroup, but + // WaitGroup provides no way to read its counter value. + // + // Invariant: endStopCond.L is the signal mutex. (This is not racy because + // sync.Cond.Wait is the only user of sync.Cond.L; only the task goroutine + // calls sync.Cond.Wait; and only the task goroutine can change the + // identity of the signal mutex, in Task.finishExec.) + endStopCond sync.Cond `state:"nosave"` + + // exitStatus is the task's exit status. + // + // exitStatus is protected by the signal mutex. + exitStatus ExitStatus + + // syscallRestartBlock represents a custom restart function to run in + // restart_syscall(2) to resume an interrupted syscall. + // + // syscallRestartBlock is exclusive to the task goroutine. + syscallRestartBlock SyscallRestartBlock + + // mu protects some of the following fields. + mu sync.Mutex `state:"nosave"` + + // tc and tr form the majority of the task's data. + // + // tc and tr are protected by mu. tc and tr are owned by the task + // goroutine. tr.signalMask is protected by the signal mutex and must be + // written using atomic memory operations (such that reading tr.signalMask + // is safe if the signal mutex is locked or if atomic memory operations are + // used), but is also owned by the task goroutine. + tc TaskContext + tr TaskResources + + // p provides the mechanism by which the task runs code in userspace. The p + // interface object is immutable. + p platform.Context `state:"nosave"` + + // k is the Kernel that this task belongs to. The k pointer is immutable. + k *Kernel + + // If vforkParent is not nil, it is the task that created this task with + // vfork() or clone(CLONE_VFORK), and should have its vforkStop ended when + // this TaskContext is released. + // + // vforkParent is protected by the TaskSet mutex. + vforkParent *Task + + // exitState is the task's progress through the exit path. + // + // exitState is protected by the TaskSet mutex. exitState is owned by the + // task goroutine. + exitState TaskExitState + + // exitTracerNotified is true if the exit path has either signaled the + // task's tracer to indicate the exit, or determined that no such signal is + // needed. exitTracerNotified can only be true if exitState is + // TaskExitZombie or TaskExitDead. + // + // exitTracerNotified is protected by the TaskSet mutex. + exitTracerNotified bool + + // exitTracerAcked is true if exitTracerNotified is true and either the + // task's tracer has acknowledged the exit notification, or the exit path + // has determined that no such notification is needed. + // + // exitTracerAcked is protected by the TaskSet mutex. + exitTracerAcked bool + + // exitParentNotified is true if the exit path has either signaled the + // task's parent to indicate the exit, or determined that no such signal is + // needed. exitParentNotified can only be true if exitState is + // TaskExitZombie or TaskExitDead. + // + // exitParentNotified is protected by the TaskSet mutex. + exitParentNotified bool + + // exitParentAcked is true if exitParentNotified is true and either the + // task's parent has acknowledged the exit notification, or the exit path + // has determined that no such acknowledgment is needed. + // + // exitParentAcked is protected by the TaskSet mutex. + exitParentAcked bool + + // goroutineStopped is a WaitGroup whose counter value is 1 when the task + // goroutine is running and 0 when the task goroutine is stopped or has + // exited. + goroutineStopped sync.WaitGroup `state:"nosave"` + + // ptraceTracer is the task that is ptrace-attached to this one. If + // ptraceTracer is nil, this task is not being traced. Note that due to + // atomic.Value limitations (atomic.Value.Store(nil) panics), a nil + // ptraceTracer is always represented as a typed nil (i.e. (*Task)(nil)). + // + // ptraceTracer is protected by the TaskSet mutex, and accessed with atomic + // operations. This allows paths that wouldn't otherwise lock the TaskSet + // mutex, notably the syscall path, to check if ptraceTracer is nil without + // additional synchronization. + ptraceTracer atomic.Value `state:".(*Task)"` + + // ptraceTracees is the set of tasks that this task is ptrace-attached to. + // + // ptraceTracees is protected by the TaskSet mutex. + ptraceTracees map[*Task]struct{} + + // ptraceOpts contains ptrace options explicitly set by the tracer. If + // ptraceTracer is nil, ptraceOpts is expected to be the zero value. + // + // ptraceOpts is protected by the TaskSet mutex. + ptraceOpts ptraceOptions + + // ptraceSyscallMode controls ptrace behavior around syscall entry and + // exit. + // + // ptraceSyscallMode is protected by the TaskSet mutex. + ptraceSyscallMode ptraceSyscallMode + + // If ptraceSinglestep is true, the next time the task executes application + // code, single-stepping should be enabled. ptraceSinglestep is stored + // independently of the architecture-specific trap flag because tracer + // detaching (which can happen concurrently with the tracee's execution if + // the tracer exits) must disable single-stepping, and the task's + // architectural state is implicitly exclusive to the task goroutine (no + // synchronization occurs before passing registers to SwitchToApp). + // + // ptraceSinglestep is analogous to Linux's TIF_SINGLESTEP. + // + // ptraceSinglestep is protected by the TaskSet mutex. + ptraceSinglestep bool + + // If t is ptrace-stopped, ptraceCode is a ptrace-defined value set at the + // time that t entered the ptrace stop, reset to 0 when the tracer + // acknowledges the stop with a wait*() syscall. Otherwise, it is the + // signal number passed to the ptrace operation that ended the last ptrace + // stop on this task. In the latter case, the effect of ptraceCode depends + // on the nature of the ptrace stop; signal-delivery-stop uses it to + // conditionally override ptraceSiginfo, syscall-entry/exit-stops send the + // signal to the task after leaving the stop, and PTRACE_EVENT stops and + // traced group stops ignore it entirely. + // + // Linux contextually stores the equivalent of ptraceCode in + // task_struct::exit_code. + // + // ptraceCode is protected by the TaskSet mutex. + ptraceCode int32 + + // ptraceSiginfo is the value returned to the tracer by + // ptrace(PTRACE_GETSIGINFO) and modified by ptrace(PTRACE_SETSIGINFO). + // (Despite the name, PTRACE_PEEKSIGINFO is completely unrelated.) + // ptraceSiginfo is nil if the task is in a ptraced group-stop (this is + // required for PTRACE_GETSIGINFO to return EINVAL during such stops, which + // is in turn required to distinguish group stops from other ptrace stops, + // per subsection "Group-stop" in ptrace(2)). + // + // ptraceSiginfo is analogous to Linux's task_struct::last_siginfo. + // + // ptraceSiginfo is protected by the TaskSet mutex. + ptraceSiginfo *arch.SignalInfo + + // ptraceEventMsg is the value set by PTRACE_EVENT stops and returned to + // the tracer by ptrace(PTRACE_GETEVENTMSG). + // + // ptraceEventMsg is protected by the TaskSet mutex. + ptraceEventMsg uint64 + + // The struct that holds the IO-related usage. The ioUsage pointer is + // immutable. + ioUsage *usage.IO + + // logPrefix is a string containing the task's thread ID in the root PID + // namespace, and is prepended to log messages emitted by Task.Infof etc. + logPrefix atomic.Value `state:".(string)"` + + // creds is the task's credentials. + // + // creds is protected by mu. + creds *auth.Credentials + + // utsns is the task's UTS namespace. + // + // utsns is protected by mu. + utsns *UTSNamespace + + // ipcns is the task's IPC namespace. + // + // ipcns is protected by mu. + ipcns *IPCNamespace + + // parentDeathSignal is sent to this task's thread group when its parent exits. + // + // parentDeathSignal is protected by mu. + parentDeathSignal linux.Signal + + // syscallFilters is all seccomp-bpf syscall filters applicable to the + // task, in the order in which they were installed. + // + // syscallFilters is protected by mu. syscallFilters is owned by the task + // goroutine. + syscallFilters []bpf.Program + + // If cleartid is non-zero, treat it as a pointer to a ThreadID in the + // task's virtual address space; when the task exits, set the pointed-to + // ThreadID to 0, and wake any futex waiters. + // + // cleartid is exclusive to the task goroutine. + cleartid usermem.Addr + + // This is mostly a fake cpumask just for sched_set/getaffinity as we + // don't really control the affinity. + // + // Invariant: allowedCPUMask.Size() == + // sched.CPUMaskSize(Kernel.applicationCores). + // + // allowedCPUMask is protected by mu. + allowedCPUMask sched.CPUSet + + // cpu is the fake cpu number returned by getcpu(2). cpu is ignored + // entirely if Kernel.useHostCores is true. + // + // cpu is accessed using atomic memory operations. + cpu int32 + + // This is used to keep track of changes made to a process' priority/niceness. + // It is mostly used to provide some reasonable return value from + // getpriority(2) after a call to setpriority(2) has been made. + // We currently do not actually modify a process' scheduling priority. + // NOTE: This represents the userspace view of priority (nice). + // This means that the value should be in the range [-20, 19]. + // + // niceness is protected by mu. + niceness int + + // This is used to track the numa policy for the current thread. This can be + // modified through a set_mempolicy(2) syscall. Since we always report a + // single numa node, all policies are no-ops. We only track this information + // so that we can return reasonable values if the application calls + // get_mempolicy(2) after setting a non-default policy. Note that in the + // real syscall, nodemask can be longer than 4 bytes, but we always report a + // single node so never need to save more than a single bit. + // + // numaPolicy and numaNodeMask are protected by mu. + numaPolicy int32 + numaNodeMask uint32 + + // If netns is true, the task is in a non-root network namespace. Network + // namespaces aren't currently implemented in full; being in a network + // namespace simply prevents the task from observing any network devices + // (including loopback) or using abstract socket addresses (see unix(7)). + // + // netns is protected by mu. netns is owned by the task goroutine. + netns bool + + // If rseqPreempted is true, before the next call to p.Switch(), interrupt + // RSEQ critical regions as defined by tg.rseq and write the task + // goroutine's CPU number to rseqCPUAddr. rseqCPU is the last CPU number + // written to rseqCPUAddr. + // + // If rseqCPUAddr is 0, rseqCPU is -1. + // + // rseqCPUAddr, rseqCPU, and rseqPreempted are exclusive to the task + // goroutine. + rseqPreempted bool `state:"nosave"` + rseqCPUAddr usermem.Addr + rseqCPU int32 + + // copyScratchBuffer is a buffer available to CopyIn/CopyOut + // implementations that require an intermediate buffer to copy data + // into/out of. It prevents these buffers from being allocated/zeroed in + // each syscall and eventually garbage collected. + // + // copyScratchBuffer is exclusive to the task goroutine. + copyScratchBuffer [copyScratchBufferLen]byte `state:"nosave"` + + // blockingTimer is used for blocking timeouts. blockingTimerChan is the + // channel that is sent to when blockingTimer fires. + // + // blockingTimer is exclusive to the task goroutine. + blockingTimer *ktime.Timer `state:"nosave"` + blockingTimerChan <-chan struct{} `state:"nosave"` + + // futexWaiter is used for futex(FUTEX_WAIT) syscalls. + // + // futexWaiter is exclusive to the task goroutine. + futexWaiter *futex.Waiter `state:"nosave"` + + // startTime is the real time at which the task started. It is set when + // a Task is created or invokes execve(2). + // + // startTime is protected by mu. + startTime ktime.Time +} + +func (t *Task) savePtraceTracer() *Task { + return t.ptraceTracer.Load().(*Task) +} + +func (t *Task) loadPtraceTracer(tracer *Task) { + t.ptraceTracer.Store(tracer) +} + +func (t *Task) saveLogPrefix() string { + return t.logPrefix.Load().(string) +} + +func (t *Task) loadLogPrefix(prefix string) { + t.logPrefix.Store(prefix) +} + +// afterLoad is invoked by stateify. +func (t *Task) afterLoad() { + t.interruptChan = make(chan struct{}, 1) + t.gosched.State = TaskGoroutineNonexistent + if t.stop != nil { + t.stopCount = 1 + } + t.endStopCond.L = &t.tg.signalHandlers.mu + t.p = t.k.Platform.NewContext() + t.rseqPreempted = true + t.futexWaiter = futex.NewWaiter() +} + +// copyScratchBufferLen is the length of the copyScratchBuffer field of the Task +// struct. +const copyScratchBufferLen = 52 + +// TaskMaybe is the interface for extracting Tasks out of things which may be +// or contain Task objects. +type TaskMaybe interface { + // ExtractTask returns the Task. + ExtractTask() *Task +} + +// CopyScratchBuffer returns a scratch buffer to be used in CopyIn/CopyOut +// functions. It must only be used within those functions and can only be used +// by the task goroutine; it exists to improve performance and thus +// intentionally lacks any synchronization. +// +// Callers should pass a constant value as an argument, which will allow the +// compiler to inline and optimize out the if statement below. +func (t *Task) CopyScratchBuffer(size int) []byte { + if size > copyScratchBufferLen { + return make([]byte, size) + } + return t.copyScratchBuffer[:size] +} + +// FutexWaiter returns the Task's futex.Waiter. +func (t *Task) FutexWaiter() *futex.Waiter { + return t.futexWaiter +} + +// ExtractTask implements TaskMaybe.ExtractTask. +func (t *Task) ExtractTask() *Task { + return t +} + +// TaskContext returns t's TaskContext. +// +// Precondition: The caller must be running on the task goroutine, or t.mu must +// be locked. +func (t *Task) TaskContext() *TaskContext { + return &t.tc +} + +// TaskResources returns t's TaskResources. +// +// Precondition: The caller must be running on the task goroutine, or t.mu must +// be locked. +func (t *Task) TaskResources() *TaskResources { + return &t.tr +} + +// WithMuLocked executes f with t.mu locked. +func (t *Task) WithMuLocked(f func(*Task)) { + t.mu.Lock() + defer t.mu.Unlock() + f(t) +} + +// Kernel returns the Kernel containing t. +func (t *Task) Kernel() *Kernel { + return t.k +} + +// Value implements context.Context.Value. +func (t *Task) Value(key interface{}) interface{} { + switch key { + case CtxCanTrace: + return t.CanTrace + case CtxKernel: + return t.k + case CtxPIDNamespace: + return t.tg.pidns + case CtxUTSNamespace: + return t.utsns + case CtxIPCNamespace: + return t.ipcns + case CtxTask: + return t + case auth.CtxCredentials: + return t.creds + case fs.CtxRoot: + return t.FSContext().RootDirectory() + case ktime.CtxRealtimeClock: + return t.k.RealtimeClock() + case limits.CtxLimits: + return t.tg.limits + case platform.CtxPlatform: + return t.k + case uniqueid.CtxGlobalUniqueID: + return t.k.UniqueID() + case uniqueid.CtxInotifyCookie: + return t.k.GenerateInotifyCookie() + default: + return nil + } +} + +// SetClearTID sets t's cleartid. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) SetClearTID(addr usermem.Addr) { + t.cleartid = addr +} + +// SetSyscallRestartBlock sets the restart block for use in +// restart_syscall(2). After registering a restart block, a syscall should +// return ERESTART_RESTARTBLOCK to request a restart using the block. +// +// Precondition: The caller must be running on the task goroutine. +func (t *Task) SetSyscallRestartBlock(r SyscallRestartBlock) { + t.syscallRestartBlock = r +} + +// SyscallRestartBlock returns the currently registered restart block for use in +// restart_syscall(2). This function is *not* idempotent and may be called once +// per syscall. This function must not be called if a restart block has not been +// registered for the current syscall. +// +// Precondition: The caller must be running on the task goroutine. +func (t *Task) SyscallRestartBlock() SyscallRestartBlock { + r := t.syscallRestartBlock + // Explicitly set the restart block to nil so that a future syscall can't + // accidentally reuse it. + t.syscallRestartBlock = nil + return r +} diff --git a/pkg/sentry/kernel/task_acct.go b/pkg/sentry/kernel/task_acct.go new file mode 100644 index 000000000..ce12cdb64 --- /dev/null +++ b/pkg/sentry/kernel/task_acct.go @@ -0,0 +1,111 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +// Accounting, limits, timers. + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/limits" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" +) + +// IOUsage returns the io usage of the thread. +func (t *Task) IOUsage() *usage.IO { + return t.ioUsage +} + +// IOUsage returns the total io usage of all dead and live threads in the group. +func (tg *ThreadGroup) IOUsage() *usage.IO { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + + io := *tg.ioUsage + // Account for active tasks. + for t := tg.tasks.Front(); t != nil; t = t.Next() { + io.Accumulate(t.IOUsage()) + } + return &io +} + +// Name returns t's name. +func (t *Task) Name() string { + t.mu.Lock() + defer t.mu.Unlock() + return t.tc.Name +} + +// SetName changes t's name. +func (t *Task) SetName(name string) { + t.mu.Lock() + defer t.mu.Unlock() + t.tc.Name = name + t.Debugf("Set thread name to %q", name) +} + +// SetCPUTimer is used by setrlimit(RLIMIT_CPU) to enforce the hard and soft +// limits on CPU time used by this process. +func (tg *ThreadGroup) SetCPUTimer(l *limits.Limit) { + tg.Timer().applyCPULimits(*l) +} + +// Limits implements context.Context.Limits. +func (t *Task) Limits() *limits.LimitSet { + return t.ThreadGroup().Limits() +} + +// StartTime returns t's start time. +func (t *Task) StartTime() ktime.Time { + t.mu.Lock() + defer t.mu.Unlock() + return t.startTime +} + +// MaxRSS returns the maximum resident set size of the task in bytes. which +// should be one of RUSAGE_SELF, RUSAGE_CHILDREN, RUSAGE_THREAD, or +// RUSAGE_BOTH. See getrusage(2) for documentation on the behavior of these +// flags. +func (t *Task) MaxRSS(which int32) uint64 { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + + switch which { + case linux.RUSAGE_SELF, linux.RUSAGE_THREAD: + // If there's an active mm we can use its value. + if mm := t.MemoryManager(); mm != nil { + if mmMaxRSS := mm.MaxResidentSetSize(); mmMaxRSS > t.tg.maxRSS { + return mmMaxRSS + } + } + return t.tg.maxRSS + case linux.RUSAGE_CHILDREN: + return t.tg.childMaxRSS + case linux.RUSAGE_BOTH: + maxRSS := t.tg.maxRSS + if maxRSS < t.tg.childMaxRSS { + maxRSS = t.tg.childMaxRSS + } + if mm := t.MemoryManager(); mm != nil { + if mmMaxRSS := mm.MaxResidentSetSize(); mmMaxRSS > maxRSS { + return mmMaxRSS + } + } + return maxRSS + default: + // We'll only get here if which is invalid. + return 0 + } +} diff --git a/pkg/sentry/kernel/task_block.go b/pkg/sentry/kernel/task_block.go new file mode 100644 index 000000000..9fd24f134 --- /dev/null +++ b/pkg/sentry/kernel/task_block.go @@ -0,0 +1,207 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "time" + + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// BlockWithTimeout blocks t until an event is received from C, the application +// monotonic clock indicates that timeout has elapsed (only if haveTimeout is true), +// or t is interrupted. It returns: +// +// - The remaining timeout, which is guaranteed to be 0 if the timeout expired, +// and is unspecified if haveTimeout is false. +// +// - An error which is nil if an event is received from C, ETIMEDOUT if the timeout +// expired, and syserror.ErrInterrupted if t is interrupted. +func (t *Task) BlockWithTimeout(C chan struct{}, haveTimeout bool, timeout time.Duration) (time.Duration, error) { + if !haveTimeout { + return timeout, t.block(C, nil) + } + + start := t.Kernel().MonotonicClock().Now() + deadline := start.Add(timeout) + err := t.BlockWithDeadline(C, true, deadline) + + // Timeout, explicitly return a remaining duration of 0. + if err == syserror.ETIMEDOUT { + return 0, err + } + + // Compute the remaining timeout. Note that even if block() above didn't + // return due to a timeout, we may have used up any of the remaining time + // since then. We cap the remaining timeout to 0 to make it easier to + // directly use the returned duration. + end := t.Kernel().MonotonicClock().Now() + remainingTimeout := timeout - end.Sub(start) + if remainingTimeout < 0 { + remainingTimeout = 0 + } + + return remainingTimeout, err +} + +// BlockWithDeadline blocks t until an event is received from C, the +// application monotonic clock indicates a time of deadline (only if +// haveDeadline is true), or t is interrupted. It returns nil if an event is +// received from C, ETIMEDOUT if the deadline expired, and +// syserror.ErrInterrupted if t is interrupted. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) BlockWithDeadline(C chan struct{}, haveDeadline bool, deadline ktime.Time) error { + if !haveDeadline { + return t.block(C, nil) + } + + // Start the timeout timer. + t.blockingTimer.Swap(ktime.Setting{ + Enabled: true, + Next: deadline, + }) + + err := t.block(C, t.blockingTimerChan) + + // Stop the timeout timer and drain the channel. + t.blockingTimer.Swap(ktime.Setting{}) + select { + case <-t.blockingTimerChan: + default: + } + + return err +} + +// BlockWithTimer blocks t until an event is received from C or tchan, or t is +// interrupted. It returns nil if an event is received from C, ETIMEDOUT if an +// event is received from tchan, and syserror.ErrInterrupted if t is +// interrupted. +// +// Most clients should use BlockWithDeadline or BlockWithTimeout instead. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) BlockWithTimer(C chan struct{}, tchan <-chan struct{}) error { + return t.block(C, tchan) +} + +// Block blocks t until an event is received from C or t is interrupted. It +// returns nil if an event is received from C and syserror.ErrInterrupted if t +// is interrupted. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) Block(C chan struct{}) error { + return t.block(C, nil) +} + +// block blocks a task on one of many events. +// N.B. defer is too expensive to be used here. +func (t *Task) block(C chan struct{}, timerChan <-chan struct{}) error { + // Fast path if the request is already done. + select { + case <-C: + return nil + default: + } + + // Deactive our address space, we don't need it. + interrupt := t.SleepStart() + + select { + case <-C: + t.SleepFinish(true) + return nil + + case <-interrupt: + t.SleepFinish(false) + // Return the indicated error on interrupt. + return syserror.ErrInterrupted + + case <-timerChan: + // We've timed out. + t.SleepFinish(true) + return syserror.ETIMEDOUT + } +} + +// SleepStart implements amutex.Sleeper.SleepStart. +func (t *Task) SleepStart() <-chan struct{} { + t.Deactivate() + t.accountTaskGoroutineEnter(TaskGoroutineBlockedInterruptible) + return t.interruptChan +} + +// SleepFinish implements amutex.Sleeper.SleepFinish. +func (t *Task) SleepFinish(success bool) { + if !success { + // The interrupted notification is consumed only at the top-level + // (Run). Therefore we attempt to reset the pending notification. + // This will also elide our next entry back into the task, so we + // will process signals, state changes, etc. + t.interruptSelf() + } + t.accountTaskGoroutineLeave(TaskGoroutineBlockedInterruptible) + t.Activate() +} + +// UninterruptibleSleepStart implements context.Context.UninterruptibleSleepStart. +func (t *Task) UninterruptibleSleepStart(deactivate bool) { + if deactivate { + t.Deactivate() + } + t.accountTaskGoroutineEnter(TaskGoroutineBlockedUninterruptible) +} + +// UninterruptibleSleepFinish implements context.Context.UninterruptibleSleepFinish. +func (t *Task) UninterruptibleSleepFinish(activate bool) { + t.accountTaskGoroutineLeave(TaskGoroutineBlockedUninterruptible) + if activate { + t.Activate() + } +} + +// interrupted returns true if interrupt or interruptSelf has been called at +// least once since the last call to interrupted. +func (t *Task) interrupted() bool { + select { + case <-t.interruptChan: + return true + default: + return false + } +} + +// interrupt unblocks the task and interrupts it if it's currently running in +// userspace. +func (t *Task) interrupt() { + t.interruptSelf() + t.p.Interrupt() +} + +// interruptSelf is like Interrupt, but can only be called by the task +// goroutine. +func (t *Task) interruptSelf() { + select { + case t.interruptChan <- struct{}{}: + t.Debugf("Interrupt queued") + default: + t.Debugf("Dropping duplicate interrupt") + } + // platform.Context.Interrupt() is unnecessary since a task goroutine + // calling interruptSelf() cannot also be blocked in + // platform.Context.Switch(). +} diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go new file mode 100644 index 000000000..3a74abdfb --- /dev/null +++ b/pkg/sentry/kernel/task_clone.go @@ -0,0 +1,475 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/bpf" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// SharingOptions controls what resources are shared by a new task created by +// Task.Clone, or an existing task affected by Task.Unshare. +type SharingOptions struct { + // If NewAddressSpace is true, the task should have an independent virtual + // address space. + NewAddressSpace bool + + // If NewSignalHandlers is true, the task should use an independent set of + // signal handlers. + NewSignalHandlers bool + + // If NewThreadGroup is true, the task should be the leader of its own + // thread group. TerminationSignal is the signal that the thread group + // will send to its parent when it exits. If NewThreadGroup is false, + // TerminationSignal is ignored. + NewThreadGroup bool + TerminationSignal linux.Signal + + // If NewPIDNamespace is true: + // + // - In the context of Task.Clone, the new task should be the init task + // (TID 1) in a new PID namespace. + // + // - In the context of Task.Unshare, the task should create a new PID + // namespace, and all subsequent clones of the task should be members of + // the new PID namespace. + NewPIDNamespace bool + + // If NewUserNamespace is true, the task should have an independent user + // namespace. + NewUserNamespace bool + + // If NewNetworkNamespace is true, the task should have an independent + // network namespace. (Note that network namespaces are not really + // implemented; see comment on Task.netns for details.) + NewNetworkNamespace bool + + // If NewFiles is true, the task should use an independent file descriptor + // table. + NewFiles bool + + // If NewFSContext is true, the task should have an independent FSContext. + NewFSContext bool + + // If NewUTSNamespace is true, the task should have an independent UTS + // namespace. + NewUTSNamespace bool + + // If NewIPCNamespace is true, the task should have an independent IPC + // namespace. + NewIPCNamespace bool +} + +// CloneOptions controls the behavior of Task.Clone. +type CloneOptions struct { + // SharingOptions defines the set of resources that the new task will share + // with its parent. + SharingOptions + + // Stack is the initial stack pointer of the new task. If Stack is 0, the + // new task will start with the same stack pointer as its parent. + Stack usermem.Addr + + // If SetTLS is true, set the new task's TLS (thread-local storage) + // descriptor to TLS. If SetTLS is false, TLS is ignored. + SetTLS bool + TLS usermem.Addr + + // If ChildClearTID is true, when the child exits, 0 is written to the + // address ChildTID in the child's memory, and if the write is successful a + // futex wake on the same address is performed. + // + // If ChildSetTID is true, the child's thread ID (in the child's PID + // namespace) is written to address ChildTID in the child's memory. (As in + // Linux, failed writes are silently ignored.) + ChildClearTID bool + ChildSetTID bool + ChildTID usermem.Addr + + // If ParentSetTID is true, the child's thread ID (in the parent's PID + // namespace) is written to address ParentTID in the parent's memory. (As + // in Linux, failed writes are silently ignored.) + // + // Older versions of the clone(2) man page state that CLONE_PARENT_SETTID + // causes the child's thread ID to be written to ptid in both the parent + // and child's memory, but this is a documentation error fixed by + // 87ab04792ced ("clone.2: Fix description of CLONE_PARENT_SETTID"). + ParentSetTID bool + ParentTID usermem.Addr + + // If Vfork is true, place the parent in vforkStop until the cloned task + // releases its TaskContext. + Vfork bool + + // If Untraced is true, do not report PTRACE_EVENT_CLONE/FORK/VFORK for + // this clone(), and do not ptrace-attach the caller's tracer to the new + // task. (PTRACE_EVENT_VFORK_DONE will still be reported if appropriate). + Untraced bool + + // If InheritTracer is true, ptrace-attach the caller's tracer to the new + // task, even if no PTRACE_EVENT_CLONE/FORK/VFORK event would be reported + // for it. If both Untraced and InheritTracer are true, no event will be + // reported, but tracer inheritance will still occur. + InheritTracer bool +} + +// Clone implements the clone(2) syscall and returns the thread ID of the new +// task in t's PID namespace. Clone may return both a non-zero thread ID and a +// non-nil error. +// +// Preconditions: The caller must be running Task.doSyscallInvoke on the task +// goroutine. +func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { + // Since signal actions may refer to application signal handlers by virtual + // address, any set of signal handlers must refer to the same address + // space. + if !opts.NewSignalHandlers && opts.NewAddressSpace { + return 0, nil, syserror.EINVAL + } + // In order for the behavior of thread-group-directed signals to be sane, + // all tasks in a thread group must share signal handlers. + if !opts.NewThreadGroup && opts.NewSignalHandlers { + return 0, nil, syserror.EINVAL + } + // All tasks in a thread group must be in the same PID namespace. + if !opts.NewThreadGroup && (opts.NewPIDNamespace || t.childPIDNamespace != nil) { + return 0, nil, syserror.EINVAL + } + // The two different ways of specifying a new PID namespace are + // incompatible. + if opts.NewPIDNamespace && t.childPIDNamespace != nil { + return 0, nil, syserror.EINVAL + } + // Thread groups and FS contexts cannot span user namespaces. + if opts.NewUserNamespace && (!opts.NewThreadGroup || !opts.NewFSContext) { + return 0, nil, syserror.EINVAL + } + + // "If CLONE_NEWUSER is specified along with other CLONE_NEW* flags in a + // single clone(2) or unshare(2) call, the user namespace is guaranteed to + // be created first, giving the child (clone(2)) or caller (unshare(2)) + // privileges over the remaining namespaces created by the call." - + // user_namespaces(7) + creds := t.Credentials() + var userns *auth.UserNamespace + if opts.NewUserNamespace { + var err error + // "EPERM (since Linux 3.9): CLONE_NEWUSER was specified in flags and + // the caller is in a chroot environment (i.e., the caller's root + // directory does not match the root directory of the mount namespace + // in which it resides)." - clone(2). Neither chroot(2) nor + // user_namespaces(7) document this. + if t.IsChrooted() { + return 0, nil, syserror.EPERM + } + userns, err = creds.NewChildUserNamespace() + if err != nil { + return 0, nil, err + } + } + if (opts.NewPIDNamespace || opts.NewNetworkNamespace || opts.NewUTSNamespace) && !creds.HasCapability(linux.CAP_SYS_ADMIN) { + return 0, nil, syserror.EPERM + } + + utsns := t.UTSNamespace() + if opts.NewUTSNamespace { + // Note that this must happen after NewUserNamespace so we get + // the new userns if there is one. + utsns = t.UTSNamespace().Clone(userns) + } + + ipcns := t.IPCNamespace() + if opts.NewIPCNamespace { + // Note that "If CLONE_NEWIPC is set, then create the process in a new IPC + // namespace" + ipcns = NewIPCNamespace() + } + + tc, err := t.tc.Fork(t, !opts.NewAddressSpace) + if err != nil { + return 0, nil, err + } + // clone() returns 0 in the child. + tc.Arch.SetReturn(0) + if opts.Stack != 0 { + tc.Arch.SetStack(uintptr(opts.Stack)) + } + if opts.SetTLS { + tc.Arch.StateData().Regs.Fs_base = uint64(opts.TLS) + } + + pidns := t.tg.pidns + if t.childPIDNamespace != nil { + pidns = t.childPIDNamespace + } else if opts.NewPIDNamespace { + pidns = pidns.NewChild(userns) + } + tg := t.tg + parent := t.parent + if opts.NewThreadGroup { + sh := t.tg.signalHandlers + if opts.NewSignalHandlers { + sh = sh.Fork() + } + tg = NewThreadGroup(pidns, sh, opts.TerminationSignal, tg.limits.GetCopy(), t.k.monotonicClock) + parent = t + } + cfg := &TaskConfig{ + Kernel: t.k, + Parent: parent, + ThreadGroup: tg, + TaskContext: tc, + TaskResources: t.tr.Fork(!opts.NewFiles, !opts.NewFSContext), + Niceness: t.Niceness(), + Credentials: creds.Fork(), + NetworkNamespaced: t.netns, + AllowedCPUMask: t.CPUMask(), + UTSNamespace: utsns, + IPCNamespace: ipcns, + } + if opts.NewNetworkNamespace { + cfg.NetworkNamespaced = true + } + nt, err := t.tg.pidns.owner.NewTask(cfg) + if err != nil { + if opts.NewThreadGroup { + tg.release() + } + return 0, nil, err + } + + // "A child process created via fork(2) inherits a copy of its parent's + // alternate signal stack settings" - sigaltstack(2). + // + // However kernel/fork.c:copy_process() adds a limitation to this: + // "sigaltstack should be cleared when sharing the same VM". + if opts.NewAddressSpace || opts.Vfork { + nt.SetSignalStack(t.SignalStack()) + } + + if userns != nil { + if err := nt.SetUserNamespace(userns); err != nil { + // This shouldn't be possible: userns was created from nt.creds, so + // nt should have CAP_SYS_ADMIN in userns. + panic("Task.Clone: SetUserNamespace failed: " + err.Error()) + } + } + + // This has to happen last, because e.g. ptraceClone may send a SIGSTOP to + // nt that it must receive before its task goroutine starts running. + tid := nt.k.tasks.Root.IDOfTask(nt) + defer nt.Start(tid) + + // "If fork/clone and execve are allowed by @prog, any child processes will + // be constrained to the same filters and system call ABI as the parent." - + // Documentation/prctl/seccomp_filter.txt + nt.syscallFilters = append([]bpf.Program(nil), t.syscallFilters...) + if opts.Vfork { + nt.vforkParent = t + } + + if opts.ChildClearTID { + nt.SetClearTID(opts.ChildTID) + } + if opts.ChildSetTID { + // Can't use Task.CopyOut, which assumes AddressSpaceActive. + usermem.CopyObjectOut(t, nt.MemoryManager(), opts.ChildTID, nt.ThreadID(), usermem.IOOpts{}) + } + ntid := t.tg.pidns.IDOfTask(nt) + if opts.ParentSetTID { + t.CopyOut(opts.ParentTID, ntid) + } + + kind := ptraceCloneKindClone + if opts.Vfork { + kind = ptraceCloneKindVfork + } else if opts.TerminationSignal == linux.SIGCHLD { + kind = ptraceCloneKindFork + } + if t.ptraceClone(kind, nt, opts) { + if opts.Vfork { + return ntid, &SyscallControl{next: &runSyscallAfterPtraceEventClone{vforkChild: nt, vforkChildTID: ntid}}, nil + } + return ntid, &SyscallControl{next: &runSyscallAfterPtraceEventClone{}}, nil + } + if opts.Vfork { + t.maybeBeginVforkStop(nt) + return ntid, &SyscallControl{next: &runSyscallAfterVforkStop{childTID: ntid}}, nil + } + return ntid, nil, nil +} + +// maybeBeginVforkStop checks if a previously-started vfork child is still +// running and has not yet released its MM, such that its parent t should enter +// a vforkStop. +// +// Preconditions: The caller must be running on t's task goroutine. +func (t *Task) maybeBeginVforkStop(child *Task) { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + if t.killedLocked() { + child.vforkParent = nil + return + } + if child.vforkParent == t { + t.beginInternalStopLocked((*vforkStop)(nil)) + } +} + +func (t *Task) unstopVforkParent() { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + if p := t.vforkParent; p != nil { + p.tg.signalHandlers.mu.Lock() + defer p.tg.signalHandlers.mu.Unlock() + if _, ok := p.stop.(*vforkStop); ok { + p.endInternalStopLocked() + } + // Parent no longer needs to be unstopped. + t.vforkParent = nil + } +} + +type runSyscallAfterPtraceEventClone struct { + vforkChild *Task + + // If vforkChild is not nil, vforkChildTID is its thread ID in the parent's + // PID namespace. vforkChildTID must be stored since the child may exit and + // release its TID before the PTRACE_EVENT stop ends. + vforkChildTID ThreadID +} + +func (r *runSyscallAfterPtraceEventClone) execute(t *Task) taskRunState { + if r.vforkChild != nil { + t.maybeBeginVforkStop(r.vforkChild) + return &runSyscallAfterVforkStop{r.vforkChildTID} + } + return (*runSyscallExit)(nil) +} + +type runSyscallAfterVforkStop struct { + // childTID has the same meaning as + // runSyscallAfterPtraceEventClone.vforkChildTID. + childTID ThreadID +} + +func (r *runSyscallAfterVforkStop) execute(t *Task) taskRunState { + t.ptraceVforkDone(r.childTID) + return (*runSyscallExit)(nil) +} + +// Unshare changes the set of resources t shares with other tasks, as specified +// by opts. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) Unshare(opts *SharingOptions) error { + // In Linux unshare(2), NewThreadGroup implies NewSignalHandlers and + // NewSignalHandlers implies NewAddressSpace. All three flags are no-ops if + // t is the only task using its MM, which due to clone(2)'s rules imply + // that it is also the only task using its signal handlers / in its thread + // group, and cause EINVAL to be returned otherwise. + // + // Since we don't count the number of tasks using each address space or set + // of signal handlers, we reject NewSignalHandlers and NewAddressSpace + // altogether, and interpret NewThreadGroup as requiring that t be the only + // member of its thread group. This seems to be logically coherent, in the + // sense that clone(2) allows a task to share signal handlers and address + // spaces with tasks in other thread groups. + if opts.NewAddressSpace || opts.NewSignalHandlers { + return syserror.EINVAL + } + if opts.NewThreadGroup { + t.tg.signalHandlers.mu.Lock() + if t.tg.tasksCount != 1 { + t.tg.signalHandlers.mu.Unlock() + return syserror.EINVAL + } + t.tg.signalHandlers.mu.Unlock() + // This isn't racy because we're the only living task, and therefore + // the only task capable of creating new ones, in our thread group. + } + if opts.NewUserNamespace { + if t.IsChrooted() { + return syserror.EPERM + } + // This temporary is needed because Go. + creds := t.Credentials() + newUserNS, err := creds.NewChildUserNamespace() + if err != nil { + return err + } + err = t.SetUserNamespace(newUserNS) + if err != nil { + return err + } + } + haveCapSysAdmin := t.HasCapability(linux.CAP_SYS_ADMIN) + if opts.NewPIDNamespace { + if !haveCapSysAdmin { + return syserror.EPERM + } + t.childPIDNamespace = t.tg.pidns.NewChild(t.UserNamespace()) + } + t.mu.Lock() + defer t.mu.Unlock() + if opts.NewNetworkNamespace { + if !haveCapSysAdmin { + return syserror.EPERM + } + t.netns = true + } + if opts.NewUTSNamespace { + if !haveCapSysAdmin { + return syserror.EPERM + } + // Note that this must happen after NewUserNamespace, so the + // new user namespace is used if there is one. + t.utsns = t.utsns.Clone(t.creds.UserNamespace) + } + if opts.NewIPCNamespace { + if !haveCapSysAdmin { + return syserror.EPERM + } + // Note that "If CLONE_NEWIPC is set, then create the process in a new IPC + // namespace" + t.ipcns = NewIPCNamespace() + } + if opts.NewFiles { + oldFDMap := t.tr.FDMap + t.tr.FDMap = oldFDMap.Fork() + oldFDMap.DecRef() + } + if opts.NewFSContext { + oldFS := t.tr.FSContext + t.tr.FSContext = oldFS.Fork() + oldFS.DecRef() + } + return nil +} + +// vforkStop is a TaskStop imposed on a task that creates a child with +// CLONE_VFORK or vfork(2), that ends when the child task ceases to use its +// current MM. (Normally, CLONE_VFORK is used in conjunction with CLONE_VM, so +// that the child and parent share mappings until the child execve()s into a +// new process image or exits.) +type vforkStop struct{} + +// StopIgnoresKill implements TaskStop.Killable. +func (*vforkStop) Killable() bool { return true } diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go new file mode 100644 index 000000000..5c563ba08 --- /dev/null +++ b/pkg/sentry/kernel/task_context.go @@ -0,0 +1,179 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "errors" + "fmt" + + "gvisor.googlesource.com/gvisor/pkg/cpuid" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" + "gvisor.googlesource.com/gvisor/pkg/sentry/loader" + "gvisor.googlesource.com/gvisor/pkg/sentry/mm" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" +) + +// ErrNoSyscalls is returned if there is no syscall table. +var ErrNoSyscalls = errors.New("no syscall table found") + +// Auxmap contains miscellaneous data for the task. +type Auxmap map[string]interface{} + +// TaskContext is the subset of a task's data that is provided by the loader. +type TaskContext struct { + // Name is the thread name set by the prctl(PR_SET_NAME) system call. + Name string + + // Arch is the architecture-specific context (registers, etc.) + Arch arch.Context + + // MemoryManager is the task's address space. + MemoryManager *mm.MemoryManager + + // fu implements futexes in the address space. + fu *futex.Manager + + // st is the task's syscall table. + st *SyscallTable +} + +// release releases all resources held by the TaskContext. release is called by +// the task when it execs into a new TaskContext or exits. +func (tc *TaskContext) release() { + // Nil out pointers so that if the task is saved after release, it doesn't + // follow the pointers to possibly now-invalid objects. + if tc.MemoryManager != nil { + // TODO + tc.MemoryManager.DecUsers(context.Background()) + tc.MemoryManager = nil + } + tc.fu = nil +} + +// Fork returns a duplicate of tc. The copied TaskContext always has an +// independent arch.Context. If shareAddressSpace is true, the copied +// TaskContext shares an address space with the original; otherwise, the copied +// TaskContext has an independent address space that is initially a duplicate +// of the original's. +func (tc *TaskContext) Fork(ctx context.Context, shareAddressSpace bool) (*TaskContext, error) { + newTC := &TaskContext{ + Arch: tc.Arch.Fork(), + st: tc.st, + } + if shareAddressSpace { + newTC.MemoryManager = tc.MemoryManager + if newTC.MemoryManager != nil { + if !newTC.MemoryManager.IncUsers() { + // Shouldn't be possible since tc.MemoryManager should be a + // counted user. + panic(fmt.Sprintf("TaskContext.Fork called with userless TaskContext.MemoryManager")) + } + } + newTC.fu = tc.fu + } else { + newMM, err := tc.MemoryManager.Fork(ctx) + if err != nil { + return nil, err + } + newTC.MemoryManager = newMM + // TODO: revisit when shmem is supported. + newTC.fu = futex.NewManager() + } + return newTC, nil +} + +// Arch returns t's arch.Context. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) Arch() arch.Context { + return t.tc.Arch +} + +// MemoryManager returns t's MemoryManager. MemoryManager does not take an +// additional reference on the returned MM. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) MemoryManager() *mm.MemoryManager { + return t.tc.MemoryManager +} + +// Futex returns t's futex manager. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) Futex() *futex.Manager { + return t.tc.fu +} + +// SyscallTable returns t's syscall table. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) SyscallTable() *SyscallTable { + return t.tc.st +} + +// Stack returns the userspace stack. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) Stack() *arch.Stack { + return &arch.Stack{t.Arch(), t.MemoryManager(), usermem.Addr(t.Arch().Stack())} +} + +// LoadTaskImage loads filename into a new TaskContext. +// +// It takes several arguments: +// * mounts: MountNamespace to lookup filename in +// * root: Root to lookup filename under +// * wd: Working directory to lookup filename under +// * maxTraversals: maximum number of symlinks to follow +// * filename: path to binary to load +// * argv: Binary argv +// * envv: Binary envv +// * fs: Binary FeatureSet +func (k *Kernel) LoadTaskImage(ctx context.Context, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals uint, filename string, argv, envv []string, fs *cpuid.FeatureSet) (*TaskContext, error) { + // Prepare a new user address space to load into. + m := mm.NewMemoryManager(k) + defer m.DecUsers(ctx) + + os, ac, name, err := loader.Load(ctx, m, mounts, root, wd, maxTraversals, fs, filename, argv, envv, k.extraAuxv, k.vdso) + if err != nil { + return nil, err + } + + // Lookup our new syscall table. + st, ok := LookupSyscallTable(os, ac.Arch()) + if !ok { + // No syscall table found. Yikes. + return nil, ErrNoSyscalls + } + + if !m.IncUsers() { + panic("Failed to increment users count on new MM") + } + return &TaskContext{ + Name: name, + Arch: ac, + MemoryManager: m, + fu: futex.NewManager(), + st: st, + }, nil +} diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go new file mode 100644 index 000000000..2285847a2 --- /dev/null +++ b/pkg/sentry/kernel/task_exec.go @@ -0,0 +1,240 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +// This file implements the machinery behind the execve() syscall. In brief, a +// thread executes an execve() by killing all other threads in its thread +// group, assuming the leader's identity, and then switching process images. +// +// This design is effectively mandated by Linux. From ptrace(2): +// +// """ +// execve(2) under ptrace +// When one thread in a multithreaded process calls execve(2), the +// kernel destroys all other threads in the process, and resets the +// thread ID of the execing thread to the thread group ID (process ID). +// (Or, to put things another way, when a multithreaded process does an +// execve(2), at completion of the call, it appears as though the +// execve(2) occurred in the thread group leader, regardless of which +// thread did the execve(2).) This resetting of the thread ID looks +// very confusing to tracers: +// +// * All other threads stop in PTRACE_EVENT_EXIT stop, if the +// PTRACE_O_TRACEEXIT option was turned on. Then all other threads +// except the thread group leader report death as if they exited via +// _exit(2) with exit code 0. +// +// * The execing tracee changes its thread ID while it is in the +// execve(2). (Remember, under ptrace, the "pid" returned from +// waitpid(2), or fed into ptrace calls, is the tracee's thread ID.) +// That is, the tracee's thread ID is reset to be the same as its +// process ID, which is the same as the thread group leader's thread +// ID. +// +// * Then a PTRACE_EVENT_EXEC stop happens, if the PTRACE_O_TRACEEXEC +// option was turned on. +// +// * If the thread group leader has reported its PTRACE_EVENT_EXIT stop +// by this time, it appears to the tracer that the dead thread leader +// "reappears from nowhere". (Note: the thread group leader does not +// report death via WIFEXITED(status) until there is at least one +// other live thread. This eliminates the possibility that the +// tracer will see it dying and then reappearing.) If the thread +// group leader was still alive, for the tracer this may look as if +// thread group leader returns from a different system call than it +// entered, or even "returned from a system call even though it was +// not in any system call". If the thread group leader was not +// traced (or was traced by a different tracer), then during +// execve(2) it will appear as if it has become a tracee of the +// tracer of the execing tracee. +// +// All of the above effects are the artifacts of the thread ID change in +// the tracee. +// """ + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// execStop is a TaskStop that a task sets on itself when it wants to execve +// and is waiting for the other tasks in its thread group to exit first. +type execStop struct{} + +// Killable implements TaskStop.Killable. +func (*execStop) Killable() bool { return true } + +// Execve implements the execve(2) syscall by killing all other tasks in its +// thread group and switching to newTC. Execve always takes ownership of newTC. +// +// Preconditions: The caller must be running Task.doSyscallInvoke on the task +// goroutine. +func (t *Task) Execve(newTC *TaskContext) (*SyscallControl, error) { + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + + if t.tg.exiting || t.tg.execing != nil { + // We lost to a racing group-exit, kill, or exec from another thread + // and should just exit. + newTC.release() + return nil, syserror.EINTR + } + + // Cancel any racing group stops. + t.tg.endGroupStopLocked(false) + + // If the task has any siblings, they have to exit before the exec can + // continue. + t.tg.execing = t + if t.tg.tasks.Front() != t.tg.tasks.Back() { + // "[All] other threads except the thread group leader report death as + // if they exited via _exit(2) with exit code 0." - ptrace(2) + for sibling := t.tg.tasks.Front(); sibling != nil; sibling = sibling.Next() { + if t != sibling { + sibling.killLocked() + } + } + // The last sibling to exit will wake t. + t.beginInternalStopLocked((*execStop)(nil)) + } + + return &SyscallControl{next: &runSyscallAfterExecStop{newTC}, ignoreReturn: true}, nil +} + +// The runSyscallAfterExecStop state continues execve(2) after all siblings of +// a thread in the execve syscall have exited. +type runSyscallAfterExecStop struct { + tc *TaskContext +} + +func (r *runSyscallAfterExecStop) execute(t *Task) taskRunState { + t.tg.pidns.owner.mu.Lock() + t.tg.execing = nil + if t.killed() { + t.tg.pidns.owner.mu.Unlock() + r.tc.release() + return (*runInterrupt)(nil) + } + // We are the thread group leader now. Save our old thread ID for + // PTRACE_EVENT_EXEC. This is racy in that if a tracer attaches after this + // point it will get a PID of 0, but this is consistent with Linux. + oldTID := ThreadID(0) + if tracer := t.Tracer(); tracer != nil { + oldTID = tracer.tg.pidns.tids[t] + } + t.promoteLocked() + // "During an execve(2), the dispositions of handled signals are reset to + // the default; the dispositions of ignored signals are left unchanged. ... + // [The] signal mask is preserved across execve(2). ... [The] pending + // signal set is preserved across an execve(2)." - signal(7) + // + // Details: + // + // - If the thread group is sharing its signal handlers with another thread + // group via CLONE_SIGHAND, execve forces the signal handlers to be copied + // (see Linux's fs/exec.c:de_thread). We're not reference-counting signal + // handlers, so we always make a copy. + // + // - "Disposition" only means sigaction::sa_handler/sa_sigaction; flags, + // restorer (if present), and mask are always reset. (See Linux's + // fs/exec.c:setup_new_exec => kernel/signal.c:flush_signal_handlers.) + t.tg.signalHandlers = t.tg.signalHandlers.CopyForExec() + t.endStopCond.L = &t.tg.signalHandlers.mu + // "Any alternate signal stack is not preserved (sigaltstack(2))." - execve(2) + t.signalStack = arch.SignalStack{Flags: arch.SignalStackFlagDisable} + // "The termination signal is reset to SIGCHLD (see clone(2))." + t.tg.terminationSignal = linux.SIGCHLD + // execed indicates that the process can no longer join a process group + // in some scenarios (namely, the parent call setpgid(2) on the child). + // See the JoinProcessGroup function in sessions.go for more context. + t.tg.execed = true + // Maximum RSS is preserved across execve(2). + t.updateRSSLocked() + // Restartable sequence state is discarded. + t.rseqPreempted = false + t.rseqCPUAddr = 0 + t.rseqCPU = -1 + t.tg.rscr.Store(&RSEQCriticalRegion{}) + t.tg.pidns.owner.mu.Unlock() + + // Remove FDs with the CloseOnExec flag set. + t.FDMap().RemoveIf(func(file *fs.File, flags FDFlags) bool { + return flags.CloseOnExec + }) + + // Switch to the new process. + t.MemoryManager().Deactivate() + t.mu.Lock() + // Update credentials to reflect the execve. This should precede switching + // MMs to ensure that dumpability has been reset first, if needed. + t.updateCredsForExecLocked() + t.tc.release() + t.tc = *r.tc + t.mu.Unlock() + t.unstopVforkParent() + // NOTE: All locks must be dropped prior to calling Activate. + t.MemoryManager().Activate() + + t.ptraceExec(oldTID) + return (*runSyscallExit)(nil) +} + +// promoteLocked makes t the leader of its thread group. If t is already the +// thread group leader, promoteLocked is a no-op. +// +// Preconditions: All other tasks in t's thread group, including the existing +// leader (if it is not t), have reached TaskExitZombie. The TaskSet mutex must +// be locked for writing. +func (t *Task) promoteLocked() { + oldLeader := t.tg.leader + if t == oldLeader { + return + } + // Swap the leader's TIDs with the execing task's. The latter will be + // released when the old leader is reaped below. + for ns := t.tg.pidns; ns != nil; ns = ns.parent { + oldTID, leaderTID := ns.tids[t], ns.tids[oldLeader] + ns.tids[oldLeader] = oldTID + ns.tids[t] = leaderTID + ns.tasks[oldTID] = oldLeader + ns.tasks[leaderTID] = t + } + + // Inherit the old leader's start time. + oldStartTime := oldLeader.StartTime() + t.mu.Lock() + t.startTime = oldStartTime + t.mu.Unlock() + + t.tg.leader = t + t.Infof("Becoming TID %d (in root PID namespace)", t.tg.pidns.owner.Root.tids[t]) + t.updateLogPrefixLocked() + // Reap the original leader. If it has a tracer, detach it instead of + // waiting for it to acknowledge the original leader's death. + oldLeader.exitParentNotified = true + oldLeader.exitParentAcked = true + if tracer := oldLeader.Tracer(); tracer != nil { + delete(tracer.ptraceTracees, oldLeader) + oldLeader.forgetTracerLocked() + // Notify the tracer that it will no longer be receiving these events + // from the tracee. + tracer.tg.eventQueue.Notify(EventExit | EventTraceeStop | EventGroupContinue) + } + oldLeader.exitNotifyLocked(false) +} diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go new file mode 100644 index 000000000..3d49ae350 --- /dev/null +++ b/pkg/sentry/kernel/task_exit.go @@ -0,0 +1,1139 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +// This file implements the task exit cycle: +// +// - Tasks are asynchronously requested to exit with Task.Kill. +// +// - When able, the task goroutine enters the exit path starting from state +// runExit. +// +// - Other tasks observe completed exits with Task.Wait (which implements the +// wait*() family of syscalls). + +import ( + "errors" + "fmt" + "strconv" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// An ExitStatus is a value communicated from an exiting task or thread group +// to the party that reaps it. +type ExitStatus struct { + // Code is the numeric value passed to the call to exit or exit_group that + // caused the exit. If the exit was not caused by such a call, Code is 0. + Code int + + // Signo is the signal that caused the exit. If the exit was not caused by + // a signal, Signo is 0. + Signo int +} + +// Signaled returns true if the ExitStatus indicates that the exiting task or +// thread group was killed by a signal. +func (es ExitStatus) Signaled() bool { + return es.Signo != 0 +} + +// Status returns the numeric representation of the ExitStatus returned by e.g. +// the wait4() system call. +func (es ExitStatus) Status() uint32 { + return ((uint32(es.Code) & 0xff) << 8) | (uint32(es.Signo) & 0xff) +} + +// ShellExitCode returns the numeric exit code that Bash would return for an +// exit status of es. +func (es ExitStatus) ShellExitCode() int { + if es.Signaled() { + return 128 + es.Signo + } + return es.Code +} + +// TaskExitState represents a step in the task exit path. +// +// "Exiting" and "exited" are often ambiguous; prefer to name specific states. +type TaskExitState int + +const ( + // TaskExitNone indicates that the task has not begun exiting. + TaskExitNone TaskExitState = iota + + // TaskExitInitiated indicates that the task goroutine has entered the exit + // path, and the task is no longer eligible to participate in group stops + // or group signal handling. TaskExitInitiated is analogous to Linux's + // PF_EXITING. + TaskExitInitiated + + // TaskExitZombie indicates that the task has released its resources, and + // the task no longer prevents a sibling thread from completing execve. + TaskExitZombie + + // TaskExitDead indicates that the task's thread IDs have been released, + // and the task no longer prevents its thread group leader from being + // reaped. ("Reaping" refers to the transitioning of a task from + // TaskExitZombie to TaskExitDead.) + TaskExitDead +) + +// String implements fmt.Stringer. +func (t TaskExitState) String() string { + switch t { + case TaskExitNone: + return "TaskExitNone" + case TaskExitInitiated: + return "TaskExitInitiated" + case TaskExitZombie: + return "TaskExitZombie" + case TaskExitDead: + return "TaskExitDead" + default: + return strconv.Itoa(int(t)) + } +} + +// killLocked marks t as killed by enqueueing a SIGKILL, without causing the +// thread-group-affecting side effects SIGKILL usually has. +// +// Preconditions: The signal mutex must be locked. +func (t *Task) killLocked() { + // Clear killable stops. + if t.stop != nil && t.stop.Killable() { + t.endInternalStopLocked() + } + t.groupStopRequired = false + t.pendingSignals.enqueue(&arch.SignalInfo{ + Signo: int32(linux.SIGKILL), + // Linux just sets SIGKILL in the pending signal bitmask without + // enqueueing an actual siginfo, such that + // kernel/signal.c:collect_signal() initalizes si_code to SI_USER. + Code: arch.SignalInfoUser, + }) + t.interrupt() +} + +// killed returns true if t has a SIGKILL pending. killed is analogous to +// Linux's fatal_signal_pending(). +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) killed() bool { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + return t.killedLocked() +} + +func (t *Task) killedLocked() bool { + return t.pendingSignals.pendingSet&linux.SignalSetOf(linux.SIGKILL) != 0 +} + +// PrepareExit indicates an exit with status es. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) PrepareExit(es ExitStatus) { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + t.exitStatus = es +} + +// PrepareGroupExit indicates a group exit with status es to t's thread group. +// +// PrepareGroupExit is analogous to Linux's do_group_exit(), except that it +// does not tail-call do_exit(), except that it *does* set Task.exitStatus. +// (Linux does not do so until within do_exit(), since it reuses exit_code for +// ptrace.) +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) PrepareGroupExit(es ExitStatus) { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + if t.tg.exiting || t.tg.execing != nil { + // Note that if t.tg.exiting is false but t.tg.execing is not nil, i.e. + // this "group exit" is being executed by the killed sibling of an + // execing task, then Task.Execve never set t.tg.exitStatus, so it's + // still the zero value. This is consistent with Linux, both in intent + // ("all other threads ... report death as if they exited via _exit(2) + // with exit code 0" - ptrace(2), "execve under ptrace") and in + // implementation (compare fs/exec.c:de_thread() => + // kernel/signal.c:zap_other_threads() and + // kernel/exit.c:do_group_exit() => + // include/linux/sched.h:signal_group_exit()). + t.exitStatus = t.tg.exitStatus + return + } + t.tg.exiting = true + t.tg.exitStatus = es + t.exitStatus = es + for sibling := t.tg.tasks.Front(); sibling != nil; sibling = sibling.Next() { + if sibling != t { + sibling.killLocked() + } + } +} + +// Kill requests that all tasks in ts exit as if group exiting with status es. +// Kill does not wait for tasks to exit. +// +// Kill has no analogue in Linux; it's provided for save/restore only. +func (ts *TaskSet) Kill(es ExitStatus) { + ts.mu.Lock() + defer ts.mu.Unlock() + ts.Root.exiting = true + for t := range ts.Root.tids { + t.tg.signalHandlers.mu.Lock() + if !t.tg.exiting { + t.tg.exiting = true + t.tg.exitStatus = es + } + t.killLocked() + t.tg.signalHandlers.mu.Unlock() + } +} + +// advanceExitStateLocked checks that t's current exit state is oldExit, then +// sets it to newExit. If t's current exit state is not oldExit, +// advanceExitStateLocked panics. +// +// Preconditions: The TaskSet mutex must be locked. +func (t *Task) advanceExitStateLocked(oldExit, newExit TaskExitState) { + if t.exitState != oldExit { + panic(fmt.Sprintf("Transitioning from exit state %v to %v: unexpected preceding state %v", oldExit, newExit, t.exitState)) + } + t.Debugf("Transitioning from exit state %v to %v", oldExit, newExit) + t.exitState = newExit +} + +// runExit is the entry point into the task exit path. +type runExit struct{} + +func (*runExit) execute(t *Task) taskRunState { + t.ptraceExit() + return (*runExitMain)(nil) +} + +type runExitMain struct{} + +func (*runExitMain) execute(t *Task) taskRunState { + lastExiter := t.exitThreadGroup() + + // If the task has a cleartid, and the thread group wasn't killed by a + // signal, handle that before releasing the MM. + if t.cleartid != 0 { + t.tg.signalHandlers.mu.Lock() + signaled := t.tg.exiting && t.tg.exitStatus.Signaled() + t.tg.signalHandlers.mu.Unlock() + if !signaled { + if _, err := t.CopyOut(t.cleartid, ThreadID(0)); err == nil { + t.Futex().Wake(uintptr(t.cleartid), ^uint32(0), 1) + } + // If the CopyOut fails, there's nothing we can do. + } + } + + // Deactivate the address space before releasing the MM. + t.Deactivate() + + // Update the max resident set size before releasing t.tc.mm. + t.tg.pidns.owner.mu.Lock() + t.updateRSSLocked() + t.tg.pidns.owner.mu.Unlock() + + // Release all of the task's resources. + t.mu.Lock() + t.tc.release() + t.tr.release() + t.mu.Unlock() + t.unstopVforkParent() + + // If this is the last task to exit from the thread group, release the + // thread group's resources. + if lastExiter { + t.tg.release() + } + + // Detach tracees. + t.exitPtrace() + + // Reparent the task's children. + t.exitChildren() + + // Don't tail-call runExitNotify, as exitChildren may have initiated a stop + // to wait for a PID namespace to die. + return (*runExitNotify)(nil) +} + +// exitThreadGroup transitions t to TaskExitInitiated, indicating to t's thread +// group that it is no longer eligible to participate in group activities. It +// returns true if t is the last task in its thread group to call +// exitThreadGroup. +func (t *Task) exitThreadGroup() bool { + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + t.tg.signalHandlers.mu.Lock() + // Can't defer unlock: see below. + + t.advanceExitStateLocked(TaskExitNone, TaskExitInitiated) + t.tg.activeTasks-- + last := t.tg.activeTasks == 0 + + // Ensure that someone will handle the signals we can't. + t.setSignalMaskLocked(^linux.SignalSet(0)) + + // Check if this task's exit interacts with an initiated group stop. + if t.tg.groupStopPhase != groupStopInitiated { + t.tg.signalHandlers.mu.Unlock() + return last + } + if t.groupStopAcknowledged { + // Un-acknowledge the group stop. + t.tg.groupStopCount-- + t.groupStopAcknowledged = false + // If the group stop wasn't complete before, then there is still at + // least one other task that hasn't acknowledged the group stop, so + // it is still not complete now. + t.tg.signalHandlers.mu.Unlock() + return last + } + if t.tg.groupStopCount != t.tg.activeTasks { + t.tg.signalHandlers.mu.Unlock() + return last + } + t.Debugf("Completing group stop") + t.tg.groupStopPhase = groupStopComplete + t.tg.groupStopWaitable = true + sig := t.tg.groupStopSignal + t.tg.groupContNotify = false + t.tg.groupContWaitable = false + // signalStop must be called with t's signal mutex unlocked. + t.tg.signalHandlers.mu.Unlock() + if t.tg.leader.parent != nil { + t.tg.leader.parent.signalStop(t, arch.CLD_STOPPED, int32(sig)) + t.tg.leader.parent.tg.eventQueue.Notify(EventChildGroupStop) + } + return last +} + +func (t *Task) exitChildren() { + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + newParent := t.findReparentTargetLocked() + if newParent == nil { + // "If the init process of a PID namespace terminates, the kernel + // terminates all of the processes in the namespace via a SIGKILL + // signal." - pid_namespaces(7) + t.Debugf("Init process terminating, killing namespace") + t.tg.pidns.exiting = true + for other := range t.tg.pidns.tids { + if other.tg != t.tg { + other.tg.signalHandlers.mu.Lock() + other.sendSignalLocked(&arch.SignalInfo{ + Signo: int32(linux.SIGKILL), + }, false /* group */) + other.tg.signalHandlers.mu.Unlock() + } + } + // TODO: The init process waits for all processes in the + // namespace to exit before completing its own exit + // (kernel/pid_namespace.c:zap_pid_ns_processes()). Stop until all + // other tasks in the namespace are dead, except possibly for this + // thread group's leader (which can't be reaped until this task exits). + } + // This is correct even if newParent is nil (it ensures that children don't + // wait for a parent to reap them.) + for c := range t.children { + if sig := c.ParentDeathSignal(); sig != 0 { + siginfo := &arch.SignalInfo{ + Signo: int32(sig), + Code: arch.SignalInfoUser, + } + siginfo.SetPid(int32(c.tg.pidns.tids[t])) + siginfo.SetUid(int32(t.Credentials().RealKUID.In(c.UserNamespace()).OrOverflow())) + c.tg.signalHandlers.mu.Lock() + c.sendSignalLocked(siginfo, true /* group */) + c.tg.signalHandlers.mu.Unlock() + } + c.reparentLocked(newParent) + if newParent != nil { + newParent.children[c] = struct{}{} + } + } +} + +// findReparentTargetLocked returns the task to which t's children should be +// reparented. If no such task exists, findNewParentLocked returns nil. +// +// Preconditions: The TaskSet mutex must be locked. +func (t *Task) findReparentTargetLocked() *Task { + // Reparent to any sibling in the same thread group that hasn't begun + // exiting. + if t2 := t.tg.anyNonExitingTaskLocked(); t2 != nil { + return t2 + } + // "A child process that is orphaned within the namespace will be + // reparented to [the init process for the namespace] ..." - + // pid_namespaces(7) + if init := t.tg.pidns.tasks[InitTID]; init != nil { + return init.tg.anyNonExitingTaskLocked() + } + return nil +} + +func (tg *ThreadGroup) anyNonExitingTaskLocked() *Task { + for t := tg.tasks.Front(); t != nil; t = t.Next() { + if t.exitState == TaskExitNone { + return t + } + } + return nil +} + +// reparentLocked changes t's parent. The new parent may be nil. +// +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) reparentLocked(parent *Task) { + oldParent := t.parent + t.parent = parent + // If a thread group leader's parent changes, reset the thread group's + // termination signal to SIGCHLD and re-check exit notification. (Compare + // kernel/exit.c:reparent_leader().) + if t != t.tg.leader { + return + } + if oldParent == nil && parent == nil { + return + } + if oldParent != nil && parent != nil && oldParent.tg == parent.tg { + return + } + t.tg.terminationSignal = linux.SIGCHLD + if t.exitParentNotified && !t.exitParentAcked { + t.exitParentNotified = false + t.exitNotifyLocked(false) + } +} + +// When a task exits, other tasks in the system, notably the task's parent and +// ptracer, may want to be notified. The exit notification system ensures that +// interested tasks receive signals and/or are woken from blocking calls to +// wait*() syscalls; these notifications must be resolved before exiting tasks +// can be reaped and disappear from the system. +// +// Each task may have a parent task and/or a tracer task. If both a parent and +// a tracer exist, they may be the same task, different tasks in the same +// thread group, or tasks in different thread groups. (In the last case, Linux +// refers to the task as being ptrace-reparented due to an implementation +// detail; we avoid this terminology to avoid confusion.) +// +// A thread group is *empty* if all non-leader tasks in the thread group are +// dead, and the leader is either a zombie or dead. The exit of a thread group +// leader is never waitable - by either the parent or tracer - until the thread +// group is empty. +// +// There are a few ways for an exit notification to be resolved: +// +// - The exit notification may be acknowledged by a call to Task.Wait with +// WaitOptions.ConsumeEvent set (e.g. due to a wait4() syscall). +// +// - If the notified party is the parent, and the parent thread group is not +// also the tracer thread group, and the notification signal is SIGCHLD, the +// parent may explicitly ignore the notification (see quote in exitNotify). +// Note that it's possible for the notified party to ignore the signal in other +// cases, but the notification is only resolved under the above conditions. +// (Actually, there is one exception; see the last paragraph of the "leader, +// has tracer, tracer thread group is parent thread group" case below.) +// +// - If the notified party is the parent, and the parent does not exist, the +// notification is resolved as if ignored. (This is only possible in the +// sentry. In Linux, the only task / thread group without a parent is global +// init, and killing global init causes a kernel panic.) +// +// - If the notified party is a tracer, the tracer may detach the traced task. +// (Zombie tasks cannot be ptrace-attached, so the reverse is not possible.) +// +// In addition, if the notified party is the parent, the parent may exit and +// cause the notifying task to be reparented to another thread group. This does +// not resolve the notification; instead, the notification must be resent to +// the new parent. +// +// The series of notifications generated for a given task's exit depend on +// whether it is a thread group leader; whether the task is ptraced; and, if +// so, whether the tracer thread group is the same as the parent thread group. +// +// - Non-leader, no tracer: No notification is generated; the task is reaped +// immediately. +// +// - Non-leader, has tracer: SIGCHLD is sent to the tracer. When the tracer +// notification is resolved (by waiting or detaching), the task is reaped. (For +// non-leaders, whether the tracer and parent thread groups are the same is +// irrelevant.) +// +// - Leader, no tracer: The task remains a zombie, with no notification sent, +// until all other tasks in the thread group are dead. (In Linux terms, this +// condition is indicated by include/linux/sched.h:thread_group_empty(); tasks +// are removed from their thread_group list in kernel/exit.c:release_task() => +// __exit_signal() => __unhash_process().) Then the thread group's termination +// signal is sent to the parent. When the parent notification is resolved (by +// waiting or ignoring), the task is reaped. +// +// - Leader, has tracer, tracer thread group is not parent thread group: +// SIGCHLD is sent to the tracer. When the tracer notification is resolved (by +// waiting or detaching), and all other tasks in the thread group are dead, the +// thread group's termination signal is sent to the parent. (Note that the +// tracer cannot resolve the exit notification by waiting until the thread +// group is empty.) When the parent notification is resolved, the task is +// reaped. +// +// - Leader, has tracer, tracer thread group is parent thread group: +// +// If all other tasks in the thread group are dead, the thread group's +// termination signal is sent to the parent. At this point, the notification +// can only be resolved by waiting. If the parent detaches from the task as a +// tracer, the notification is not resolved, but the notification can now be +// resolved by waiting or ignoring. When the parent notification is resolved, +// the task is reaped. +// +// If at least one task in the thread group is not dead, SIGCHLD is sent to the +// parent. At this point, the notification cannot be resolved at all; once the +// thread group becomes empty, it can be resolved only by waiting. If the +// parent detaches from the task as a tracer before all remaining tasks die, +// then exit notification proceeds as in the case where the leader never had a +// tracer. If the parent detaches from the task as a tracer after all remaining +// tasks die, the notification is not resolved, but the notification can now be +// resolved by waiting or ignoring. When the parent notification is resolved, +// the task is reaped. +// +// In both of the above cases, when the parent detaches from the task as a +// tracer while the thread group is empty, whether or not the parent resolves +// the notification by ignoring it is based on the parent's SIGCHLD signal +// action, whether or not the thread group's termination signal is SIGCHLD +// (Linux: kernel/ptrace.c:__ptrace_detach() => ignoring_children()). +// +// There is one final wrinkle: A leader can become a non-leader due to a +// sibling execve. In this case, the execing thread detaches the leader's +// tracer (if one exists) and reaps the leader immediately. In Linux, this is +// in fs/exec.c:de_thread(); in the sentry, this is in Task.promoteLocked(). + +type runExitNotify struct{} + +func (*runExitNotify) execute(t *Task) taskRunState { + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + t.advanceExitStateLocked(TaskExitInitiated, TaskExitZombie) + t.tg.liveTasks-- + // Check if this completes a sibling's execve. + if t.tg.execing != nil && t.tg.liveTasks == 1 { + // execing blocks the addition of new tasks to the thread group, so + // the sole living task must be the execing one. + e := t.tg.execing + e.tg.signalHandlers.mu.Lock() + if _, ok := e.stop.(*execStop); ok { + e.endInternalStopLocked() + } + e.tg.signalHandlers.mu.Unlock() + } + t.exitNotifyLocked(false) + // The task goroutine will now exit. + return nil +} + +// exitNotifyLocked is called after changes to t's state that affect exit +// notification. +// +// If fromPtraceDetach is true, the caller is ptraceDetach or exitPtrace; +// thanks to Linux's haphazard implementation of this functionality, such cases +// determine whether parent notifications are ignored based on the parent's +// handling of SIGCHLD, regardless of what the exited task's thread group's +// termination signal is. +// +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) exitNotifyLocked(fromPtraceDetach bool) { + if t.exitState != TaskExitZombie { + return + } + if !t.exitTracerNotified { + t.exitTracerNotified = true + tracer := t.Tracer() + if tracer == nil { + t.exitTracerAcked = true + } else if t != t.tg.leader || t.parent == nil || tracer.tg != t.parent.tg { + // Don't set exitParentNotified if t is non-leader, even if the + // tracer is in the parent thread group, so that if the parent + // detaches the following call to exitNotifyLocked passes through + // the !exitParentNotified case below and causes t to be reaped + // immediately. + // + // Tracer notification doesn't care about about + // SIG_IGN/SA_NOCLDWAIT. + tracer.tg.signalHandlers.mu.Lock() + tracer.sendSignalLocked(t.exitNotificationSignal(linux.SIGCHLD, tracer), true /* group */) + tracer.tg.signalHandlers.mu.Unlock() + // Wake EventTraceeStop waiters as well since this task will never + // ptrace-stop again. + tracer.tg.eventQueue.Notify(EventExit | EventTraceeStop) + } else { + // t is a leader and the tracer is in the parent thread group. + t.exitParentNotified = true + sig := linux.SIGCHLD + if t.tg.tasksCount == 1 { + sig = t.tg.terminationSignal + } + // This notification doesn't care about SIG_IGN/SA_NOCLDWAIT either + // (in Linux, the check in do_notify_parent() is gated by + // !tsk->ptrace.) + t.parent.tg.signalHandlers.mu.Lock() + t.parent.sendSignalLocked(t.exitNotificationSignal(sig, t.parent), true /* group */) + t.parent.tg.signalHandlers.mu.Unlock() + // See below for rationale for this event mask. + t.parent.tg.eventQueue.Notify(EventExit | EventChildGroupStop | EventGroupContinue) + } + } + if t.exitTracerAcked && !t.exitParentNotified { + if t != t.tg.leader { + t.exitParentNotified = true + t.exitParentAcked = true + } else if t.tg.tasksCount == 1 { + t.exitParentNotified = true + if t.parent == nil { + t.exitParentAcked = true + } else { + // "POSIX.1-2001 specifies that if the disposition of SIGCHLD is + // set to SIG_IGN or the SA_NOCLDWAIT flag is set for SIGCHLD (see + // sigaction(2)), then children that terminate do not become + // zombies and a call to wait() or waitpid() will block until all + // children have terminated, and then fail with errno set to + // ECHILD. (The original POSIX standard left the behavior of + // setting SIGCHLD to SIG_IGN unspecified. Note that even though + // the default disposition of SIGCHLD is "ignore", explicitly + // setting the disposition to SIG_IGN results in different + // treatment of zombie process children.) Linux 2.6 conforms to + // this specification." - wait(2) + // + // Some undocumented Linux-specific details: + // + // - All of the above is ignored if the termination signal isn't + // SIGCHLD. + // + // - SA_NOCLDWAIT causes the leader to be immediately reaped, but + // does not suppress the SIGCHLD. + signalParent := t.tg.terminationSignal.IsValid() + t.parent.tg.signalHandlers.mu.Lock() + if t.tg.terminationSignal == linux.SIGCHLD || fromPtraceDetach { + if act, ok := t.parent.tg.signalHandlers.actions[linux.SIGCHLD]; ok { + if act.Handler == arch.SignalActIgnore { + t.exitParentAcked = true + signalParent = false + } else if act.Flags&arch.SignalFlagNoCldWait != 0 { + t.exitParentAcked = true + } + } + } + if signalParent { + t.parent.tg.leader.sendSignalLocked(t.exitNotificationSignal(t.tg.terminationSignal, t.parent), true /* group */) + } + t.parent.tg.signalHandlers.mu.Unlock() + // If a task in the parent was waiting for a child group stop + // or continue, it needs to be notified of the exit, because + // there may be no remaining eligible tasks (so that wait + // should return ECHILD). + t.parent.tg.eventQueue.Notify(EventExit | EventChildGroupStop | EventGroupContinue) + } + } + } + if t.exitTracerAcked && t.exitParentAcked { + t.advanceExitStateLocked(TaskExitZombie, TaskExitDead) + for ns := t.tg.pidns; ns != nil; ns = ns.parent { + tid := ns.tids[t] + delete(ns.tasks, tid) + delete(ns.tids, t) + } + t.tg.exitedCPUStats.Accumulate(t.CPUStats()) + t.tg.ioUsage.Accumulate(t.ioUsage) + t.tg.signalHandlers.mu.Lock() + t.tg.tasks.Remove(t) + if t.tg.lastTimerSignalTask == t { + t.tg.lastTimerSignalTask = nil + } + t.tg.tasksCount-- + tc := t.tg.tasksCount + t.tg.signalHandlers.mu.Unlock() + if tc == 1 && t != t.tg.leader { + // Our fromPtraceDetach doesn't matter here (in Linux terms, this + // is via a call to release_task()). + t.tg.leader.exitNotifyLocked(false) + } else if tc == 0 { + t.tg.processGroup.decRefWithParent(t.tg.parentPG()) + } + if t.parent != nil { + delete(t.parent.children, t) + t.parent = nil + } + } +} + +// Preconditions: The TaskSet mutex must be locked. +func (t *Task) exitNotificationSignal(sig linux.Signal, receiver *Task) *arch.SignalInfo { + info := &arch.SignalInfo{ + Signo: int32(sig), + } + info.SetPid(int32(receiver.tg.pidns.tids[t])) + info.SetUid(int32(t.Credentials().RealKUID.In(receiver.UserNamespace()).OrOverflow())) + if t.exitStatus.Signaled() { + info.Code = arch.CLD_KILLED + info.SetStatus(int32(t.exitStatus.Signo)) + } else { + info.Code = arch.CLD_EXITED + info.SetStatus(int32(t.exitStatus.Code)) + } + // TODO: Set utime, stime. + return info +} + +// ExitStatus returns t's exit status, which is only guaranteed to be +// meaningful if t.ExitState() != TaskExitNone. +func (t *Task) ExitStatus() ExitStatus { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + return t.exitStatus +} + +// ExitStatus returns the exit status that would be returned by a consuming +// wait*() on tg. +func (tg *ThreadGroup) ExitStatus() ExitStatus { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + tg.signalHandlers.mu.Lock() + defer tg.signalHandlers.mu.Unlock() + if tg.exiting { + return tg.exitStatus + } + return tg.leader.exitStatus +} + +// TerminationSignal returns the thread group's termination signal. +func (tg *ThreadGroup) TerminationSignal() linux.Signal { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + return tg.terminationSignal +} + +// Task events that can be waited for. +const ( + // EventExit represents an exit notification generated for a child thread + // group leader or a tracee under the conditions specified in the comment + // above runExitNotify. + EventExit waiter.EventMask = 1 << iota + + // EventChildGroupStop occurs when a child thread group completes a group + // stop (i.e. all tasks in the child thread group have entered a stopped + // state as a result of a group stop). + EventChildGroupStop + + // EventTraceeStop occurs when a task that is ptraced by a task in the + // notified thread group enters a ptrace stop (see ptrace(2)). + EventTraceeStop + + // EventGroupContinue occurs when a child thread group, or a thread group + // whose leader is ptraced by a task in the notified thread group, that had + // initiated or completed a group stop leaves the group stop, due to the + // child thread group or any task in the child thread group being sent + // SIGCONT. + EventGroupContinue +) + +// WaitOptions controls the behavior of Task.Wait. +type WaitOptions struct { + // If SpecificTID is non-zero, only events from the task with thread ID + // SpecificTID are eligible to be waited for. SpecificTID is resolved in + // the PID namespace of the waiter (the method receiver of Task.Wait). If + // no such task exists, or that task would not otherwise be eligible to be + // waited for by the waiting task, then there are no waitable tasks and + // Wait will return ECHILD. + SpecificTID ThreadID + + // If SpecificPGID is non-zero, only events from ThreadGroups with a + // matching ProcessGroupID are eligible to be waited for. (Same + // constraints as SpecificTID apply.) + SpecificPGID ProcessGroupID + + // Terminology note: Per waitpid(2), "a clone child is one which delivers + // no signal, or a signal other than SIGCHLD to its parent upon + // termination." In Linux, termination signal is technically a per-task + // property rather than a per-thread-group property. However, clone() + // forces no termination signal for tasks created with CLONE_THREAD, and + // execve() resets the termination signal to SIGCHLD, so all + // non-group-leader threads have no termination signal and are therefore + // "clone tasks". + + // If NonCloneTasks is true, events from non-clone tasks are eligible to be + // waited for. + NonCloneTasks bool + + // If CloneTasks is true, events from clone tasks are eligible to be waited + // for. + CloneTasks bool + + // Events is a bitwise combination of the events defined above that specify + // what events are of interest to the call to Wait. + Events waiter.EventMask + + // If ConsumeEvent is true, the Wait should consume the event such that it + // cannot be returned by a future Wait. Note that if a task exit is + // consumed in this way, in most cases the task will be reaped. + ConsumeEvent bool + + // If BlockInterruptErr is not nil, Wait will block until either an event + // is available or there are no tasks that could produce a waitable event; + // if that blocking is interrupted, Wait returns BlockInterruptErr. If + // BlockInterruptErr is nil, Wait will not block. + BlockInterruptErr error +} + +// Preconditions: The TaskSet mutex must be locked (for reading or writing). +func (o *WaitOptions) matchesTask(t *Task, pidns *PIDNamespace) bool { + if o.SpecificTID != 0 && o.SpecificTID != pidns.tids[t] { + return false + } + if o.SpecificPGID != 0 && o.SpecificPGID != pidns.pgids[t.tg.processGroup] { + return false + } + if t == t.tg.leader && t.tg.terminationSignal == linux.SIGCHLD { + return o.NonCloneTasks + } + return o.CloneTasks +} + +// ErrNoWaitableEvent is returned by non-blocking Task.Waits (e.g. +// waitpid(WNOHANG)) that find no waitable events, but determine that waitable +// events may exist in the future. (In contrast, if a non-blocking or blocking +// Wait determines that there are no tasks that can produce a waitable event, +// Task.Wait returns ECHILD.) +var ErrNoWaitableEvent = errors.New("non-blocking Wait found eligible threads but no waitable events") + +// WaitResult contains information about a waited-for event. +type WaitResult struct { + // Task is the task that reported the event. + Task *Task + + // TID is the thread ID of Task in the PID namespace of the task that + // called Wait (that is, the method receiver of the call to Task.Wait). TID + // is provided because consuming exit waits cause the thread ID to be + // deallocated. + TID ThreadID + + // UID is the real UID of Task in the user namespace of the task that + // called Wait. + UID auth.UID + + // Event is exactly one of the events defined above. + Event waiter.EventMask + + // Status is the numeric status associated with the event. + Status uint32 +} + +// Wait waits for an event from a thread group that is a child of t's thread +// group, or a task in such a thread group, or a task that is ptraced by t, +// subject to the options specified in opts. +func (t *Task) Wait(opts *WaitOptions) (*WaitResult, error) { + if opts.BlockInterruptErr == nil { + return t.waitOnce(opts) + } + w, ch := waiter.NewChannelEntry(nil) + t.tg.eventQueue.EventRegister(&w, opts.Events) + defer t.tg.eventQueue.EventUnregister(&w) + for { + wr, err := t.waitOnce(opts) + if err != ErrNoWaitableEvent { + // This includes err == nil. + return wr, err + } + if err := t.Block(ch); err != nil { + return wr, syserror.ConvertIntr(err, opts.BlockInterruptErr) + } + } +} + +func (t *Task) waitOnce(opts *WaitOptions) (*WaitResult, error) { + anyWaitableTasks := false + + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + + // Without the (unimplemented) __WNOTHREAD flag, a task can wait on the + // children and tracees of any task in the same thread group. + for parent := t.tg.tasks.Front(); parent != nil; parent = parent.Next() { + for child := range parent.children { + if !opts.matchesTask(child, parent.tg.pidns) { + continue + } + // Non-leaders don't notify parents on exit and aren't eligible to + // be waited on. + if opts.Events&EventExit != 0 && child == child.tg.leader && !child.exitParentAcked { + anyWaitableTasks = true + if wr := t.waitCollectZombieLocked(child, opts, false); wr != nil { + return wr, nil + } + } + // Check for group stops and continues. Tasks that have passed + // TaskExitInitiated can no longer participate in group stops. + if opts.Events&(EventChildGroupStop|EventGroupContinue) == 0 { + continue + } + if child.exitState >= TaskExitInitiated { + continue + } + // If the waiter is in the same thread group as the task's + // tracer, do not report its group stops; they will be reported + // as ptrace stops instead. This also skips checking for group + // continues, but they'll be checked for when scanning tracees + // below. (Per kernel/exit.c:wait_consider_task(): "If a + // ptracer wants to distinguish the two events for its own + // children, it should create a separate process which takes + // the role of real parent.") + if tracer := child.Tracer(); tracer != nil && tracer.tg == parent.tg { + continue + } + anyWaitableTasks = true + if opts.Events&EventChildGroupStop != 0 { + if wr := t.waitCollectChildGroupStopLocked(child, opts); wr != nil { + return wr, nil + } + } + if opts.Events&EventGroupContinue != 0 { + if wr := t.waitCollectGroupContinueLocked(child, opts); wr != nil { + return wr, nil + } + } + } + for tracee := range parent.ptraceTracees { + if !opts.matchesTask(tracee, parent.tg.pidns) { + continue + } + // Non-leaders do notify tracers on exit. + if opts.Events&EventExit != 0 && !tracee.exitTracerAcked { + anyWaitableTasks = true + if wr := t.waitCollectZombieLocked(tracee, opts, true); wr != nil { + return wr, nil + } + } + if opts.Events&(EventTraceeStop|EventGroupContinue) == 0 { + continue + } + if tracee.exitState >= TaskExitInitiated { + continue + } + anyWaitableTasks = true + if opts.Events&EventTraceeStop != 0 { + if wr := t.waitCollectTraceeStopLocked(tracee, opts); wr != nil { + return wr, nil + } + } + if opts.Events&EventGroupContinue != 0 { + if wr := t.waitCollectGroupContinueLocked(tracee, opts); wr != nil { + return wr, nil + } + } + } + } + + if anyWaitableTasks { + return nil, ErrNoWaitableEvent + } + return nil, syserror.ECHILD +} + +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) waitCollectZombieLocked(target *Task, opts *WaitOptions, asPtracer bool) *WaitResult { + if asPtracer && !target.exitTracerNotified { + return nil + } + if !asPtracer && !target.exitParentNotified { + return nil + } + // Zombied thread group leaders are never waitable until their thread group + // is otherwise empty. Usually this is caught by the + // target.exitParentNotified check above, but if t is both (in the thread + // group of) target's tracer and parent, asPtracer may be true. + if target == target.tg.leader && target.tg.tasksCount != 1 { + return nil + } + pid := t.tg.pidns.tids[target] + uid := target.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow() + status := target.exitStatus.Status() + if !opts.ConsumeEvent { + return &WaitResult{ + Task: target, + TID: pid, + UID: uid, + Event: EventExit, + Status: status, + } + } + // Surprisingly, the exit status reported by a non-consuming wait can + // differ from that reported by a consuming wait; the latter will return + // the group exit code if one is available. + if target.tg.exiting { + status = target.tg.exitStatus.Status() + } + // t may be (in the thread group of) target's parent, tracer, or both. We + // don't need to check for !exitTracerAcked because tracees are detached + // here, and we don't need to check for !exitParentAcked because zombies + // will be reaped here. + if tracer := target.Tracer(); tracer != nil && tracer.tg == t.tg && target.exitTracerNotified { + target.exitTracerAcked = true + target.ptraceTracer.Store((*Task)(nil)) + delete(t.ptraceTracees, target) + } + if target.parent != nil && target.parent.tg == t.tg && target.exitParentNotified { + target.exitParentAcked = true + if target == target.tg.leader { + // target.tg.exitedCPUStats doesn't include target.CPUStats() yet, + // and won't until after target.exitNotifyLocked() (maybe). Include + // target.CPUStats() explicitly. This is consistent with Linux, + // which accounts an exited task's cputime to its thread group in + // kernel/exit.c:release_task() => __exit_signal(), and uses + // thread_group_cputime_adjusted() in wait_task_zombie(). + t.tg.childCPUStats.Accumulate(target.CPUStats()) + t.tg.childCPUStats.Accumulate(target.tg.exitedCPUStats) + t.tg.childCPUStats.Accumulate(target.tg.childCPUStats) + // Update t's child max resident set size. The size will be the maximum + // of this thread's size and all its childrens' sizes. + if t.tg.childMaxRSS < target.tg.maxRSS { + t.tg.childMaxRSS = target.tg.maxRSS + } + if t.tg.childMaxRSS < target.tg.childMaxRSS { + t.tg.childMaxRSS = target.tg.childMaxRSS + } + } + } + target.exitNotifyLocked(false) + return &WaitResult{ + Task: target, + TID: pid, + UID: uid, + Event: EventExit, + Status: status, + } +} + +// updateRSSLocked updates t.tg.maxRSS. +// +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) updateRSSLocked() { + if mmMaxRSS := t.MemoryManager().MaxResidentSetSize(); t.tg.maxRSS < mmMaxRSS { + t.tg.maxRSS = mmMaxRSS + } +} + +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) waitCollectChildGroupStopLocked(target *Task, opts *WaitOptions) *WaitResult { + target.tg.signalHandlers.mu.Lock() + defer target.tg.signalHandlers.mu.Unlock() + if !target.tg.groupStopWaitable { + return nil + } + pid := t.tg.pidns.tids[target] + uid := target.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow() + sig := target.tg.groupStopSignal + if opts.ConsumeEvent { + target.tg.groupStopWaitable = false + } + return &WaitResult{ + Task: target, + TID: pid, + UID: uid, + Event: EventChildGroupStop, + // There is no name for these status constants. + Status: (uint32(sig)&0xff)<<8 | 0x7f, + } +} + +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) waitCollectGroupContinueLocked(target *Task, opts *WaitOptions) *WaitResult { + target.tg.signalHandlers.mu.Lock() + defer target.tg.signalHandlers.mu.Unlock() + if !target.tg.groupContWaitable { + return nil + } + pid := t.tg.pidns.tids[target] + uid := target.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow() + if opts.ConsumeEvent { + target.tg.groupContWaitable = false + } + return &WaitResult{ + Task: target, + TID: pid, + UID: uid, + Event: EventGroupContinue, + Status: 0xffff, + } +} + +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) waitCollectTraceeStopLocked(target *Task, opts *WaitOptions) *WaitResult { + target.tg.signalHandlers.mu.Lock() + defer target.tg.signalHandlers.mu.Unlock() + if target.stop == nil { + return nil + } + if _, ok := target.stop.(*ptraceStop); !ok { + return nil + } + if target.ptraceCode == 0 { + return nil + } + pid := t.tg.pidns.tids[target] + uid := target.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow() + code := target.ptraceCode + if opts.ConsumeEvent { + target.ptraceCode = 0 + } + return &WaitResult{ + Task: target, + TID: pid, + UID: uid, + Event: EventTraceeStop, + Status: uint32(code)<<8 | 0x7f, + } +} + +// ExitState returns t's current progress through the exit path. +func (t *Task) ExitState() TaskExitState { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + return t.exitState +} + +// ParentDeathSignal returns t's parent death signal. +func (t *Task) ParentDeathSignal() linux.Signal { + t.mu.Lock() + defer t.mu.Unlock() + return t.parentDeathSignal +} + +// SetParentDeathSignal sets t's parent death signal. +func (t *Task) SetParentDeathSignal(sig linux.Signal) { + t.mu.Lock() + defer t.mu.Unlock() + t.parentDeathSignal = sig +} diff --git a/pkg/sentry/kernel/task_identity.go b/pkg/sentry/kernel/task_identity.go new file mode 100644 index 000000000..a51fa9d7e --- /dev/null +++ b/pkg/sentry/kernel/task_identity.go @@ -0,0 +1,557 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// Credentials returns t's credentials by value. +func (t *Task) Credentials() auth.Credentials { + t.mu.Lock() + defer t.mu.Unlock() + return *t.creds // Copy out with lock held. +} + +// UserNamespace returns the user namespace associated with the task. +func (t *Task) UserNamespace() *auth.UserNamespace { + t.mu.Lock() + defer t.mu.Unlock() + return t.creds.UserNamespace +} + +// HasCapabilityIn checks if the task has capability cp in user namespace ns. +func (t *Task) HasCapabilityIn(cp linux.Capability, ns *auth.UserNamespace) bool { + t.mu.Lock() + defer t.mu.Unlock() + return t.creds.HasCapabilityIn(cp, ns) +} + +// HasCapability checks if the task has capability cp in its user namespace. +func (t *Task) HasCapability(cp linux.Capability) bool { + t.mu.Lock() + defer t.mu.Unlock() + return t.creds.HasCapability(cp) +} + +// SetUID implements the semantics of setuid(2). +func (t *Task) SetUID(uid auth.UID) error { + // setuid considers -1 to be invalid. + if !uid.Ok() { + return syserror.EINVAL + } + t.mu.Lock() + defer t.mu.Unlock() + kuid := t.creds.UserNamespace.MapToKUID(uid) + if !kuid.Ok() { + return syserror.EINVAL + } + // "setuid() sets the effective user ID of the calling process. If the + // effective UID of the caller is root (more precisely: if the caller has + // the CAP_SETUID capability), the real UID and saved set-user-ID are also + // set." - setuid(2) + if t.creds.HasCapability(linux.CAP_SETUID) { + t.setKUIDsUncheckedLocked(kuid, kuid, kuid) + return nil + } + // "EPERM: The user is not privileged (Linux: does not have the CAP_SETUID + // capability) and uid does not match the real UID or saved set-user-ID of + // the calling process." + if kuid != t.creds.RealKUID && kuid != t.creds.SavedKUID { + return syserror.EPERM + } + t.setKUIDsUncheckedLocked(t.creds.RealKUID, kuid, t.creds.SavedKUID) + return nil +} + +// SetREUID implements the semantics of setreuid(2). +func (t *Task) SetREUID(r, e auth.UID) error { + t.mu.Lock() + defer t.mu.Unlock() + // "Supplying a value of -1 for either the real or effective user ID forces + // the system to leave that ID unchanged." - setreuid(2) + newR := t.creds.RealKUID + if r.Ok() { + newR = t.creds.UserNamespace.MapToKUID(r) + if !newR.Ok() { + return syserror.EINVAL + } + } + newE := t.creds.EffectiveKUID + if e.Ok() { + newE = t.creds.UserNamespace.MapToKUID(e) + if !newE.Ok() { + return syserror.EINVAL + } + } + if !t.creds.HasCapability(linux.CAP_SETUID) { + // "Unprivileged processes may only set the effective user ID to the + // real user ID, the effective user ID, or the saved set-user-ID." + if newE != t.creds.RealKUID && newE != t.creds.EffectiveKUID && newE != t.creds.SavedKUID { + return syserror.EPERM + } + // "Unprivileged users may only set the real user ID to the real user + // ID or the effective user ID." + if newR != t.creds.RealKUID && newR != t.creds.EffectiveKUID { + return syserror.EPERM + } + } + // "If the real user ID is set (i.e., ruid is not -1) or the effective user + // ID is set to a value not equal to the previous real user ID, the saved + // set-user-ID will be set to the new effective user ID." + newS := t.creds.SavedKUID + if r.Ok() || (e.Ok() && newE != t.creds.EffectiveKUID) { + newS = newE + } + t.setKUIDsUncheckedLocked(newR, newE, newS) + return nil +} + +// SetRESUID implements the semantics of the setresuid(2) syscall. +func (t *Task) SetRESUID(r, e, s auth.UID) error { + t.mu.Lock() + defer t.mu.Unlock() + // "Unprivileged user processes may change the real UID, effective UID, and + // saved set-user-ID, each to one of: the current real UID, the current + // effective UID or the current saved set-user-ID. Privileged processes (on + // Linux, those having the CAP_SETUID capability) may set the real UID, + // effective UID, and saved set-user-ID to arbitrary values. If one of the + // arguments equals -1, the corresponding value is not changed." - + // setresuid(2) + var err error + newR := t.creds.RealKUID + if r.Ok() { + newR, err = t.creds.UseUID(r) + if err != nil { + return err + } + } + newE := t.creds.EffectiveKUID + if e.Ok() { + newE, err = t.creds.UseUID(e) + if err != nil { + return err + } + } + newS := t.creds.SavedKUID + if s.Ok() { + newS, err = t.creds.UseUID(s) + if err != nil { + return err + } + } + t.setKUIDsUncheckedLocked(newR, newE, newS) + return nil +} + +// Preconditions: t.mu must be locked. +func (t *Task) setKUIDsUncheckedLocked(newR, newE, newS auth.KUID) { + root := t.creds.UserNamespace.MapToKUID(auth.RootUID) + oldR, oldE, oldS := t.creds.RealKUID, t.creds.EffectiveKUID, t.creds.SavedKUID + t.creds.RealKUID, t.creds.EffectiveKUID, t.creds.SavedKUID = newR, newE, newS + + // "1. If one or more of the real, effective or saved set user IDs was + // previously 0, and as a result of the UID changes all of these IDs have a + // nonzero value, then all capabilities are cleared from the permitted and + // effective capability sets." - capabilities(7) + if (oldR == root || oldE == root || oldS == root) && (newR != root && newE != root && newS != root) { + // prctl(2): "PR_SET_KEEPCAP: Set the state of the calling thread's + // "keep capabilities" flag, which determines whether the thread's permitted + // capability set is cleared when a change is made to the + // thread's user IDs such that the thread's real UID, effective + // UID, and saved set-user-ID all become nonzero when at least + // one of them previously had the value 0. By default, the + // permitted capability set is cleared when such a change is + // made; setting the "keep capabilities" flag prevents it from + // being cleared." (A thread's effective capability set is always + // cleared when such a credential change is made, + // regardless of the setting of the "keep capabilities" flag.) + if !t.creds.KeepCaps { + t.creds.PermittedCaps = 0 + t.creds.EffectiveCaps = 0 + } + } + // """ + // 2. If the effective user ID is changed from 0 to nonzero, then all + // capabilities are cleared from the effective set. + // + // 3. If the effective user ID is changed from nonzero to 0, then the + // permitted set is copied to the effective set. + // """ + if oldE == root && newE != root { + t.creds.EffectiveCaps = 0 + } else if oldE != root && newE == root { + t.creds.EffectiveCaps = t.creds.PermittedCaps + } + // "4. If the filesystem user ID is changed from 0 to nonzero (see + // setfsuid(2)), then the following capabilities are cleared from the + // effective set: ..." + // (filesystem UIDs aren't implemented, nor are any of the capabilities in + // question) + + // Not documented, but compare Linux's kernel/cred.c:commit_creds(). + if oldE != newE { + t.parentDeathSignal = 0 + } +} + +// SetGID implements the semantics of setgid(2). +func (t *Task) SetGID(gid auth.GID) error { + if !gid.Ok() { + return syserror.EINVAL + } + t.mu.Lock() + defer t.mu.Unlock() + kgid := t.creds.UserNamespace.MapToKGID(gid) + if !kgid.Ok() { + return syserror.EINVAL + } + if t.creds.HasCapability(linux.CAP_SETGID) { + t.setKGIDsUncheckedLocked(kgid, kgid, kgid) + return nil + } + if kgid != t.creds.RealKGID && kgid != t.creds.SavedKGID { + return syserror.EPERM + } + t.setKGIDsUncheckedLocked(t.creds.RealKGID, kgid, t.creds.SavedKGID) + return nil +} + +// SetREGID implements the semantics of setregid(2). +func (t *Task) SetREGID(r, e auth.GID) error { + t.mu.Lock() + defer t.mu.Unlock() + newR := t.creds.RealKGID + if r.Ok() { + newR = t.creds.UserNamespace.MapToKGID(r) + if !newR.Ok() { + return syserror.EINVAL + } + } + newE := t.creds.EffectiveKGID + if e.Ok() { + newE = t.creds.UserNamespace.MapToKGID(e) + if !newE.Ok() { + return syserror.EINVAL + } + } + if !t.creds.HasCapability(linux.CAP_SETGID) { + if newE != t.creds.RealKGID && newE != t.creds.EffectiveKGID && newE != t.creds.SavedKGID { + return syserror.EPERM + } + if newR != t.creds.RealKGID && newR != t.creds.EffectiveKGID { + return syserror.EPERM + } + } + newS := t.creds.SavedKGID + if r.Ok() || (e.Ok() && newE != t.creds.EffectiveKGID) { + newS = newE + } + t.setKGIDsUncheckedLocked(newR, newE, newS) + return nil +} + +// SetRESGID implements the semantics of the setresgid(2) syscall. +func (t *Task) SetRESGID(r, e, s auth.GID) error { + t.mu.Lock() + defer t.mu.Unlock() + var err error + newR := t.creds.RealKGID + if r.Ok() { + newR, err = t.creds.UseGID(r) + if err != nil { + return err + } + } + newE := t.creds.EffectiveKGID + if e.Ok() { + newE, err = t.creds.UseGID(e) + if err != nil { + return err + } + } + newS := t.creds.SavedKGID + if s.Ok() { + newS, err = t.creds.UseGID(s) + if err != nil { + return err + } + } + t.setKGIDsUncheckedLocked(newR, newE, newS) + return nil +} + +func (t *Task) setKGIDsUncheckedLocked(newR, newE, newS auth.KGID) { + oldE := t.creds.EffectiveKGID + t.creds.RealKGID, t.creds.EffectiveKGID, t.creds.SavedKGID = newR, newE, newS + + // Not documented, but compare Linux's kernel/cred.c:commit_creds(). + if oldE != newE { + t.parentDeathSignal = 0 + } +} + +// SetExtraGIDs attempts to change t's supplemental groups. All IDs are +// interpreted as being in t's user namespace. +func (t *Task) SetExtraGIDs(gids []auth.GID) error { + t.mu.Lock() + defer t.mu.Unlock() + if !t.creds.HasCapability(linux.CAP_SETGID) { + return syserror.EPERM + } + kgids := make([]auth.KGID, len(gids)) + for i, gid := range gids { + kgid := t.creds.UserNamespace.MapToKGID(gid) + if !kgid.Ok() { + return syserror.EINVAL + } + kgids[i] = kgid + } + t.creds.ExtraKGIDs = kgids + return nil +} + +// SetCapabilitySets attempts to change t's permitted, inheritable, and +// effective capability sets. +func (t *Task) SetCapabilitySets(permitted, inheritable, effective auth.CapabilitySet) error { + t.mu.Lock() + defer t.mu.Unlock() + // "Permitted: This is a limiting superset for the effective capabilities + // that the thread may assume." - capabilities(7) + if effective & ^permitted != 0 { + return syserror.EPERM + } + // "It is also a limiting superset for the capabilities that may be added + // to the inheritable set by a thread that does not have the CAP_SETPCAP + // capability in its effective set." + if !t.creds.HasCapability(linux.CAP_SETPCAP) && (inheritable & ^(t.creds.InheritableCaps|t.creds.PermittedCaps) != 0) { + return syserror.EPERM + } + // "If a thread drops a capability from its permitted set, it can never + // reacquire that capability (unless it execve(2)s ..." + if permitted & ^t.creds.PermittedCaps != 0 { + return syserror.EPERM + } + // "... if a capability is not in the bounding set, then a thread can't add + // this capability to its inheritable set, even if it was in its permitted + // capabilities ..." + if inheritable & ^(t.creds.InheritableCaps|t.creds.BoundingCaps) != 0 { + return syserror.EPERM + } + t.creds.PermittedCaps = permitted + t.creds.InheritableCaps = inheritable + t.creds.EffectiveCaps = effective + return nil +} + +// DropBoundingCapability attempts to drop capability cp from t's capability +// bounding set. +func (t *Task) DropBoundingCapability(cp linux.Capability) error { + t.mu.Lock() + defer t.mu.Unlock() + if !t.creds.HasCapability(linux.CAP_SETPCAP) { + return syserror.EPERM + } + t.creds.BoundingCaps &^= auth.CapabilitySetOf(cp) + return nil +} + +// SetUserNamespace attempts to move c into ns. +func (t *Task) SetUserNamespace(ns *auth.UserNamespace) error { + t.mu.Lock() + defer t.mu.Unlock() + + // "A process reassociating itself with a user namespace must have the + // CAP_SYS_ADMIN capability in the target user namespace." - setns(2) + // + // If t just created ns, then t.creds is guaranteed to have CAP_SYS_ADMIN + // in ns (by rule 3 in auth.Credentials.HasCapability). + if !t.creds.HasCapabilityIn(linux.CAP_SYS_ADMIN, ns) { + return syserror.EPERM + } + + t.creds.UserNamespace = ns + // "The child process created by clone(2) with the CLONE_NEWUSER flag + // starts out with a complete set of capabilities in the new user + // namespace. Likewise, a process that creates a new user namespace using + // unshare(2) or joins an existing user namespace using setns(2) gains a + // full set of capabilities in that namespace." + t.creds.PermittedCaps = auth.AllCapabilities + t.creds.InheritableCaps = 0 + t.creds.EffectiveCaps = auth.AllCapabilities + t.creds.BoundingCaps = auth.AllCapabilities + // "A call to clone(2), unshare(2), or setns(2) using the CLONE_NEWUSER + // flag sets the "securebits" flags (see capabilities(7)) to their default + // values (all flags disabled) in the child (for clone(2)) or caller (for + // unshare(2), or setns(2)." - user_namespaces(7) + t.creds.KeepCaps = false + + return nil +} + +// SetKeepCaps will set the keep capabilities flag PR_SET_KEEPCAPS. +func (t *Task) SetKeepCaps(k bool) { + t.mu.Lock() + defer t.mu.Unlock() + t.creds.KeepCaps = k +} + +// updateCredsForExec updates t.creds to reflect an execve(). +// +// NOTE: We currently do not implement privileged executables +// (set-user/group-ID bits and file capabilities). This allows us to make a lot +// of simplifying assumptions: +// +// - We assume the no_new_privs bit (set by prctl(SET_NO_NEW_PRIVS)), which +// disables the features we don't support anyway, is always set. This +// drastically simplifies this function. +// +// - We don't implement AT_SECURE, because no_new_privs always being set means +// that the conditions that require AT_SECURE never arise. (Compare Linux's +// security/commoncap.c:cap_bprm_set_creds() and cap_bprm_secureexec().) +// +// - We don't check for CAP_SYS_ADMIN in prctl(PR_SET_SECCOMP), since +// seccomp-bpf is also allowed if the task has no_new_privs set. +// +// - Task.ptraceAttach does not serialize with execve as it does in Linux, +// since no_new_privs being set has the same effect as the presence of an +// unprivileged tracer. +// +// Preconditions: t.mu must be locked. +func (t *Task) updateCredsForExecLocked() { + // """ + // During an execve(2), the kernel calculates the new capabilities of + // the process using the following algorithm: + // + // P'(permitted) = (P(inheritable) & F(inheritable)) | + // (F(permitted) & cap_bset) + // + // P'(effective) = F(effective) ? P'(permitted) : 0 + // + // P'(inheritable) = P(inheritable) [i.e., unchanged] + // + // where: + // + // P denotes the value of a thread capability set before the + // execve(2) + // + // P' denotes the value of a thread capability set after the + // execve(2) + // + // F denotes a file capability set + // + // cap_bset is the value of the capability bounding set + // + // ... + // + // In order to provide an all-powerful root using capability sets, during + // an execve(2): + // + // 1. If a set-user-ID-root program is being executed, or the real user ID + // of the process is 0 (root) then the file inheritable and permitted sets + // are defined to be all ones (i.e. all capabilities enabled). + // + // 2. If a set-user-ID-root program is being executed, then the file + // effective bit is defined to be one (enabled). + // + // The upshot of the above rules, combined with the capabilities + // transformations described above, is that when a process execve(2)s a + // set-user-ID-root program, or when a process with an effective UID of 0 + // execve(2)s a program, it gains all capabilities in its permitted and + // effective capability sets, except those masked out by the capability + // bounding set. + // """ - capabilities(7) + // (ambient capability sets omitted) + // + // As the last paragraph implies, the case of "a set-user-ID root program + // is being executed" also includes the case where (namespace) root is + // executing a non-set-user-ID program; the actual check is just based on + // the effective user ID. + var newPermitted auth.CapabilitySet // since F(inheritable) == F(permitted) == 0 + fileEffective := false + root := t.creds.UserNamespace.MapToKUID(auth.RootUID) + if t.creds.EffectiveKUID == root || t.creds.RealKUID == root { + newPermitted = t.creds.InheritableCaps | t.creds.BoundingCaps + if t.creds.EffectiveKUID == root { + fileEffective = true + } + } + + // Now we enter poorly-documented, somewhat confusing territory. (The + // accompanying comment in Linux's security/commoncap.c:cap_bprm_set_creds + // is not very helpful.) My reading of it is: + // + // If at least one of the following is true: + // + // A1. The execing task is ptraced, and the tracer did not have + // CAP_SYS_PTRACE in the execing task's user namespace at the time of + // PTRACE_ATTACH. + // + // A2. The execing task shares its FS context with at least one task in + // another thread group. + // + // A3. The execing task has no_new_privs set. + // + // AND at least one of the following is true: + // + // B1. The new effective user ID (which may come from set-user-ID, or be the + // execing task's existing effective user ID) is not equal to the task's + // real UID. + // + // B2. The new effective group ID (which may come from set-group-ID, or be + // the execing task's existing effective group ID) is not equal to the + // task's real GID. + // + // B3. The new permitted capability set contains capabilities not in the + // task's permitted capability set. + // + // Then: + // + // C1. Limit the new permitted capability set to the task's permitted + // capability set. + // + // C2. If either the task does not have CAP_SETUID in its user namespace, or + // the task has no_new_privs set, force the new effective UID and GID to + // the task's real UID and GID. + // + // But since no_new_privs is always set (A3 is always true), this becomes + // much simpler. If B1 and B2 are false, C2 is a no-op. If B3 is false, C1 + // is a no-op. So we can just do C1 and C2 unconditionally. + if t.creds.EffectiveKUID != t.creds.RealKUID || t.creds.EffectiveKGID != t.creds.RealKGID { + t.creds.EffectiveKUID = t.creds.RealKUID + t.creds.EffectiveKGID = t.creds.RealKGID + t.parentDeathSignal = 0 + } + // (Saved set-user-ID is always set to the new effective user ID, and saved + // set-group-ID is always set to the new effective group ID, regardless of + // the above.) + t.creds.SavedKUID = t.creds.RealKUID + t.creds.SavedKGID = t.creds.RealKGID + t.creds.PermittedCaps &= newPermitted + if fileEffective { + t.creds.EffectiveCaps = t.creds.PermittedCaps + } else { + t.creds.EffectiveCaps = 0 + } + + // prctl(2): The "keep capabilities" value will be reset to 0 on subsequent + // calls to execve(2). + t.creds.KeepCaps = false + + // "The bounding set is inherited at fork(2) from the thread's parent, and + // is preserved across an execve(2)". So we're done. +} diff --git a/pkg/sentry/kernel/task_log.go b/pkg/sentry/kernel/task_log.go new file mode 100644 index 000000000..18efacb19 --- /dev/null +++ b/pkg/sentry/kernel/task_log.go @@ -0,0 +1,137 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + "sort" + + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" +) + +const ( + // maxStackDebugBytes is the maximum number of user stack bytes that may be + // printed by debugDumpStack. + maxStackDebugBytes = 1024 +) + +// Infof logs an formatted info message by calling log.Infof. +func (t *Task) Infof(fmt string, v ...interface{}) { + if log.IsLogging(log.Info) { + log.Infof(t.logPrefix.Load().(string)+fmt, v...) + } +} + +// Warningf logs a warning string by calling log.Warningf. +func (t *Task) Warningf(fmt string, v ...interface{}) { + if log.IsLogging(log.Warning) { + log.Warningf(t.logPrefix.Load().(string)+fmt, v...) + } +} + +// Debugf creates a debug string that includes the task ID. +func (t *Task) Debugf(fmt string, v ...interface{}) { + if log.IsLogging(log.Debug) { + log.Debugf(t.logPrefix.Load().(string)+fmt, v...) + } +} + +// IsLogging returns true iff this level is being logged. +func (t *Task) IsLogging(level log.Level) bool { + return log.IsLogging(level) +} + +// DebugDumpState logs task state at log level debug. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) DebugDumpState() { + t.debugDumpRegisters() + t.debugDumpStack() + if mm := t.MemoryManager(); mm != nil { + t.Debugf("Mappings:\n%s", mm) + } + t.Debugf("FDMap:\n%s", t.FDMap()) +} + +// debugDumpRegisters logs register state at log level debug. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) debugDumpRegisters() { + if !t.IsLogging(log.Debug) { + return + } + regmap, err := t.Arch().RegisterMap() + if err != nil { + t.Debugf("Registers: %v", err) + } else { + t.Debugf("Registers:") + var regs []string + for reg := range regmap { + regs = append(regs, reg) + } + sort.Strings(regs) + for _, reg := range regs { + t.Debugf("%-8s = %016x", reg, regmap[reg]) + } + } +} + +// debugDumpStack logs user stack contents at log level debug. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) debugDumpStack() { + if !t.IsLogging(log.Debug) { + return + } + m := t.MemoryManager() + if m == nil { + t.Debugf("Memory manager for task is gone, skipping application stack dump.") + return + } + t.Debugf("Stack:") + start := usermem.Addr(t.Arch().Stack()) + // Round addr down to a 16-byte boundary. + start &= ^usermem.Addr(15) + // Print 16 bytes per line, one byte at a time. + for offset := uint64(0); offset < maxStackDebugBytes; offset += 16 { + addr, ok := start.AddLength(offset) + if !ok { + break + } + var data [16]byte + n, err := m.CopyIn(t, addr, data[:], usermem.IOOpts{ + IgnorePermissions: true, + }) + // Print as much of the line as we can, even if an error was + // encountered. + if n > 0 { + t.Debugf("%x: % x", addr, data[:n]) + } + if err != nil { + t.Debugf("Error reading stack at address %x: %v", addr+usermem.Addr(n), err) + break + } + } +} + +// updateLogPrefix updates the task's cached log prefix to reflect its +// current thread ID. +// +// Preconditions: The task's owning TaskSet.mu must be locked. +func (t *Task) updateLogPrefixLocked() { + // Use the task's TID in the root PID namespace for logging. + t.logPrefix.Store(fmt.Sprintf("[% 4d] ", t.tg.pidns.owner.Root.tids[t])) +} diff --git a/pkg/sentry/kernel/task_net.go b/pkg/sentry/kernel/task_net.go new file mode 100644 index 000000000..4df2e53d3 --- /dev/null +++ b/pkg/sentry/kernel/task_net.go @@ -0,0 +1,35 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/inet" +) + +// IsNetworkNamespaced returns true if t is in a non-root network namespace. +func (t *Task) IsNetworkNamespaced() bool { + t.mu.Lock() + defer t.mu.Unlock() + return t.netns +} + +// NetworkContext returns the network stack used by the task. NetworkContext +// may return nil if no network stack is available. +func (t *Task) NetworkContext() inet.Stack { + if t.IsNetworkNamespaced() { + return nil + } + return t.k.networkStack +} diff --git a/pkg/sentry/kernel/task_resources.go b/pkg/sentry/kernel/task_resources.go new file mode 100644 index 000000000..e529f0c2d --- /dev/null +++ b/pkg/sentry/kernel/task_resources.go @@ -0,0 +1,126 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" +) + +// TaskResources is the subset of a task's data provided by its creator that is +// not provided by the loader. +type TaskResources struct { + // SignalMask is the set of signals whose delivery is currently blocked. + // + // FIXME: Determine if we also need RealSignalMask + SignalMask linux.SignalSet + + // FSContext is the filesystem context. + *FSContext + + // FDMap provides access to files to the task. + *FDMap + + // Tracks abstract sockets that are in use. + AbstractSockets *AbstractSocketNamespace +} + +// newTaskResources returns a new TaskResources, taking an additional reference +// on fdm. +func newTaskResources(fdm *FDMap, fc *FSContext) *TaskResources { + fdm.IncRef() + return &TaskResources{ + FDMap: fdm, + FSContext: fc, + AbstractSockets: NewAbstractSocketNamespace(), + } +} + +// release releases all resources held by the TaskResources. release is called +// by the task when it exits. +func (tr *TaskResources) release() { + tr.FDMap.DecRef() + tr.FDMap = nil + tr.FSContext.DecRef() + tr.FSContext = nil + tr.AbstractSockets = nil +} + +// Fork returns a duplicate of tr. +// +// FIXME: Preconditions: When tr is owned by a Task, that task's +// signal mutex must be locked, or Fork must be called by the task's goroutine. +func (tr *TaskResources) Fork(shareFiles bool, shareFSContext bool) *TaskResources { + var fdmap *FDMap + if shareFiles { + fdmap = tr.FDMap + fdmap.IncRef() + } else { + fdmap = tr.FDMap.Fork() + } + + var fsc *FSContext + if shareFSContext { + fsc = tr.FSContext + fsc.IncRef() + } else { + fsc = tr.FSContext.Fork() + } + + return &TaskResources{ + SignalMask: tr.SignalMask, + FDMap: fdmap, + FSContext: fsc, + AbstractSockets: tr.AbstractSockets, + } +} + +// FDMap returns t's FDMap. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) FDMap() *FDMap { + return t.tr.FDMap +} + +// FSContext returns t's FSContext. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) FSContext() *FSContext { + return t.tr.FSContext +} + +// MountNamespace returns t's MountNamespace. MountNamespace does not take an additional +// reference on the returned MountNamespace. +func (t *Task) MountNamespace() *fs.MountNamespace { + return t.k.mounts +} + +// AbstractSockets returns t's AbstractSocketNamespace. +func (t *Task) AbstractSockets() *AbstractSocketNamespace { + return t.tr.AbstractSockets +} + +// IsChrooted returns true if the root directory of t's FSContext is not the +// root directory of t's MountNamespace. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) IsChrooted() bool { + realRoot := t.k.mounts.Root() + defer realRoot.DecRef() + return t.tr.FSContext.root != realRoot +} diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go new file mode 100644 index 000000000..94ce5582b --- /dev/null +++ b/pkg/sentry/kernel/task_run.go @@ -0,0 +1,346 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "bytes" + "runtime" + "sync/atomic" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// A taskRunState is a reified state in the task state machine. See README.md +// for details. The canonical list of all run states, as well as transitions +// between them, is given in run_states.dot. +// +// The set of possible states is enumerable and completely defined by the +// kernel package, so taskRunState would ideally be represented by a +// discriminated union. However, Go does not support sum types. +// +// Hence, as with TaskStop, data-free taskRunStates should be represented as +// typecast nils to avoid unnecessary allocation. +type taskRunState interface { + // execute executes the code associated with this state over the given task + // and returns the following state. If execute returns nil, the task + // goroutine should exit. + // + // It is valid to tail-call a following state's execute to avoid the + // overhead of converting the following state to an interface object and + // checking for stops, provided that the tail-call cannot recurse. + execute(*Task) taskRunState +} + +// run runs the task goroutine. +// +// threadID a dummy value set to the task's TID in the root PID namespace to +// make it visible in stack dumps. A goroutine for a given task can be identified +// searching for Task.run()'s argument value. +func (t *Task) run(threadID uintptr) { + // Construct t.blockingTimer here. We do this here because we can't + // reconstruct t.blockingTimer during restore in Task.afterLoad(), because + // kernel.timekeeper.SetClocks() hasn't been called yet. + blockingTimerNotifier, blockingTimerChan := ktime.NewChannelNotifier() + t.blockingTimer = ktime.NewTimer(t.k.MonotonicClock(), blockingTimerNotifier) + defer t.blockingTimer.Destroy() + t.blockingTimerChan = blockingTimerChan + + // Activate our address space. + t.Activate() + // The corresponding t.Deactivate occurs in the exit path + // (runExitMain.execute) so that when + // Platform.CooperativelySharesAddressSpace() == true, we give up the + // AddressSpace before the task goroutine finishes executing. + + // Ensure that thread group timers for execution time reflect that this + // task now exists. + t.tg.tm.kick() + + // If this is a newly-started task, it should check for participation in + // group stops. If this is a task resuming after restore, it was + // interrupted by saving. In either case, the task is initially + // interrupted. + t.interruptSelf() + + for { + // Explanation for this ordering: + // + // - A freshly-started task that is stopped should not do anything + // before it enters the stop. + // + // - If taskRunState.execute returns nil, the task goroutine should + // exit without checking for a stop. + // + // - Task.Start won't start Task.run if t.runState is nil, so this + // ordering is safe. + t.doStop() + t.runState = t.runState.execute(t) + if t.runState == nil { + t.accountTaskGoroutineEnter(TaskGoroutineNonexistent) + t.goroutineStopped.Done() + t.tg.liveGoroutines.Done() + t.tg.pidns.owner.liveGoroutines.Done() + t.tg.pidns.owner.runningGoroutines.Done() + + // Keep argument alive because stack trace for dead variables may not be correct. + runtime.KeepAlive(threadID) + return + } + } +} + +// doStop is called by Task.run to block until the task is not stopped. +func (t *Task) doStop() { + if atomic.LoadInt32(&t.stopCount) == 0 { + return + } + t.Deactivate() + // NOTE: t.Activate() must be called without any locks held, so + // this defer must precede the defer for unlocking the signal mutex. + defer t.Activate() + t.accountTaskGoroutineEnter(TaskGoroutineStopped) + defer t.accountTaskGoroutineLeave(TaskGoroutineStopped) + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + t.tg.pidns.owner.runningGoroutines.Add(-1) + defer t.tg.pidns.owner.runningGoroutines.Add(1) + t.goroutineStopped.Add(-1) + defer t.goroutineStopped.Add(1) + for t.stopCount > 0 { + t.endStopCond.Wait() + } +} + +// The runApp state checks for interrupts before executing untrusted +// application code. +type runApp struct{} + +func (*runApp) execute(t *Task) taskRunState { + if t.interrupted() { + // Checkpointing instructs tasks to stop by sending an interrupt, so we + // must check for stops before entering runInterrupt (instead of + // tail-calling it). + return (*runInterrupt)(nil) + } + + // We're about to switch to the application again. If there's still a + // unhandled SyscallRestartErrno that wasn't translated to an EINTR, + // restart the syscall that was interrupted. If there's a saved signal + // mask, restore it. (Note that restoring the saved signal mask may unblock + // a pending signal, causing another interruption, but that signal should + // not interact with the interrupted syscall.) + if t.haveSyscallReturn { + if sre, ok := SyscallRestartErrnoFromReturn(t.Arch().Return()); ok { + if sre == ERESTART_RESTARTBLOCK { + t.Debugf("Restarting syscall %d with restart block after errno %d: not interrupted by handled signal", t.Arch().SyscallNo(), sre) + t.Arch().RestartSyscallWithRestartBlock() + } else { + t.Debugf("Restarting syscall %d after errno %d: not interrupted by handled signal", t.Arch().SyscallNo(), sre) + t.Arch().RestartSyscall() + } + } + t.haveSyscallReturn = false + } + if t.haveSavedSignalMask { + t.SetSignalMask(t.savedSignalMask) + t.haveSavedSignalMask = false + if t.interrupted() { + return (*runInterrupt)(nil) + } + } + + // Apply restartable sequences. + if t.rseqPreempted { + t.rseqPreempted = false + if t.rseqCPUAddr != 0 { + if err := t.rseqCopyOutCPU(); err != nil { + t.Warningf("Failed to copy CPU to %#x for RSEQ: %v", t.rseqCPUAddr, err) + t.forceSignal(linux.SIGSEGV, false) + t.SendSignal(sigPriv(linux.SIGSEGV)) + // Re-enter the task run loop for signal delivery. + return (*runApp)(nil) + } + } + t.rseqInterrupt() + } + + // Check if we need to enable single-stepping. Tracers expect that the + // kernel preserves the value of the single-step flag set by PTRACE_SETREGS + // whether or not PTRACE_SINGLESTEP/PTRACE_SYSEMU_SINGLESTEP is used (this + // includes our ptrace platform, by the way), so we should only clear the + // single-step flag if we're responsible for setting it. (clearSinglestep + // is therefore analogous to Linux's TIF_FORCED_TF.) + // + // Strictly speaking, we should also not clear the single-step flag if we + // single-step through an instruction that sets the single-step flag + // (arch/x86/kernel/step.c:is_setting_trap_flag()). But nobody sets their + // own TF. (Famous last words, I know.) + clearSinglestep := false + if t.hasTracer() { + t.tg.pidns.owner.mu.RLock() + if t.ptraceSinglestep { + clearSinglestep = !t.Arch().SingleStep() + t.Arch().SetSingleStep() + } + t.tg.pidns.owner.mu.RUnlock() + } + + t.accountTaskGoroutineEnter(TaskGoroutineRunningApp) + info, at, err := t.p.Switch(t.MemoryManager().AddressSpace(), t.Arch(), t.rseqCPU) + t.accountTaskGoroutineLeave(TaskGoroutineRunningApp) + + if clearSinglestep { + t.Arch().ClearSingleStep() + } + + switch err { + case nil: + // Handle application system call. + return t.doSyscall() + + case platform.ErrContextInterrupt: + // Interrupted by platform.Context.Interrupt(). Re-enter the run + // loop to figure out why. + return (*runApp)(nil) + + case platform.ErrContextSignal: + // Looks like a signal has been delivered to us. If it's a synchronous + // signal (SEGV, SIGBUS, etc.), it should be sent to the application + // thread that received it. + sig := linux.Signal(info.Signo) + + // Was it a fault that we should handle internally? If so, this wasn't + // an application-generated signal and we should continue execution + // normally. + if at.Any() { + addr := usermem.Addr(info.Addr()) + err := t.MemoryManager().HandleUserFault(t, addr, at, usermem.Addr(t.Arch().Stack())) + if err == nil { + // The fault was handled appropriately. + // We can resume running the application. + return (*runApp)(nil) + } + + // Is this a vsyscall that we need emulate? + if at.Execute { + if sysno, ok := t.tc.st.LookupEmulate(addr); ok { + return t.doVsyscall(addr, sysno) + } + } + + // The JVM will trigger these errors constantly, so don't + // spam logs with this error. + if err == syserror.EFAULT || err == syserror.EPERM { + t.Debugf("Unhandled user fault: addr=%x ip=%x access=%v err=%v", addr, t.Arch().IP(), at, err) + } else { + t.Warningf("Unhandled user fault: addr=%x ip=%x access=%v err=%v", addr, t.Arch().IP(), at, err) + } + t.DebugDumpState() + + // Continue to signal handling. + // + // Convert a BusError error to a SIGBUS from a SIGSEGV. All + // other info bits stay the same (address, etc.). + if _, ok := err.(*memmap.BusError); ok { + sig = linux.SIGBUS + info.Signo = int32(linux.SIGBUS) + } + } + + switch sig { + case linux.SIGILL: + // N.B. The debug stuff here is arguably + // expensive. Don't fret. This gets called + // about 5 times for a typical application, if + // that. + t.Debugf("SIGILL @ %x", t.Arch().IP()) + + // Is this a CPUID instruction? + expected := arch.CPUIDInstruction[:] + found := make([]byte, len(expected)) + _, err := t.CopyIn(usermem.Addr(t.Arch().IP()), &found) + if err == nil && bytes.Equal(expected, found) { + // Skip the cpuid instruction. + t.Arch().CPUIDEmulate(t) + t.Arch().SetIP(t.Arch().IP() + uintptr(len(expected))) + break + } + + // Treat it like any other synchronous signal. + fallthrough + + case linux.SIGSEGV, linux.SIGBUS, linux.SIGFPE, linux.SIGTRAP: + // Synchronous signal. Send it to ourselves. Assume the signal is + // legitimate and force it (work around the signal being ignored or + // blocked) like Linux does. Conveniently, this is even the correct + // behavior for SIGTRAP from single-stepping. + t.forceSignal(linux.Signal(sig), false /* unconditional */) + t.SendSignal(info) + + case platform.SignalInterrupt: + // Assume that a call to platform.Context.Interrupt() misfired. + + case linux.SIGPROF: + // It's a profiling interrupt: there's not much + // we can do. We've already paid a decent cost + // by intercepting the signal, at this point we + // simply ignore it. + + default: + // Asynchronous signal. Let the system deal with it. + t.k.sendExternalSignal(info, "application") + } + + return (*runApp)(nil) + + case platform.ErrContextCPUPreempted: + // Ensure that RSEQ critical sections are interrupted and per-thread + // CPU values are updated before the next platform.Context.Switch(). + t.rseqPreempted = true + return (*runApp)(nil) + + default: + // What happened? Can't continue. + t.Warningf("Unexpected SwitchToApp error: %v", err) + t.PrepareExit(ExitStatus{Code: t.ExtractErrno(err, -1)}) + return (*runExit)(nil) + } +} + +// waitGoroutineStoppedOrExited blocks until t's task goroutine stops or exits. +func (t *Task) waitGoroutineStoppedOrExited() { + t.goroutineStopped.Wait() +} + +// WaitExited blocks until all task goroutines in tg have exited. +// +// WaitExited does not correspond to anything in Linux; it's provided so that +// external callers of Kernel.CreateProcess can wait for the created thread +// group to terminate. +func (tg *ThreadGroup) WaitExited() { + tg.liveGoroutines.Wait() +} + +// Yield yields the processor for the calling task. +func (t *Task) Yield() { + atomic.AddUint64(&t.yieldCount, 1) + runtime.Gosched() +} diff --git a/pkg/sentry/kernel/task_sched.go b/pkg/sentry/kernel/task_sched.go new file mode 100644 index 000000000..b50139077 --- /dev/null +++ b/pkg/sentry/kernel/task_sched.go @@ -0,0 +1,329 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +// CPU scheduling, real and fake. + +import ( + "fmt" + "sync/atomic" + "time" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/hostcpu" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// TaskGoroutineState is a coarse representation of the current execution +// status of a kernel.Task goroutine. +type TaskGoroutineState int + +const ( + // TaskGoroutineNonexistent indicates that the task goroutine has either + // not yet been created by Task.Start() or has returned from Task.run(). + // This must be the zero value for TaskGoroutineState. + TaskGoroutineNonexistent TaskGoroutineState = iota + + // TaskGoroutineRunningSys indicates that the task goroutine is executing + // sentry code. + TaskGoroutineRunningSys + + // TaskGoroutineRunningApp indicates that the task goroutine is executing + // application code. + TaskGoroutineRunningApp + + // TaskGoroutineBlockedInterruptible indicates that the task goroutine is + // blocked in Task.block(), and hence may be woken by Task.interrupt() + // (e.g. due to signal delivery). + TaskGoroutineBlockedInterruptible + + // TaskGoroutineBlockedUninterruptible indicates that the task goroutine is + // stopped outside of Task.block() and Task.doStop(), and hence cannot be + // woken by Task.interrupt(). + TaskGoroutineBlockedUninterruptible + + // TaskGoroutineStopped indicates that the task goroutine is blocked in + // Task.doStop(). TaskGoroutineStopped is similar to + // TaskGoroutineBlockedUninterruptible, but is a separate state to make it + // possible to determine when Task.stop is meaningful. + TaskGoroutineStopped +) + +// TaskGoroutineSchedInfo contains task goroutine scheduling state which must +// be read and updated atomically. +type TaskGoroutineSchedInfo struct { + // Timestamp was the value of Kernel.cpuClock when this + // TaskGoroutineSchedInfo was last updated. + Timestamp uint64 + + // State is the current state of the task goroutine. + State TaskGoroutineState + + // UserTicks is the amount of time the task goroutine has spent executing + // its associated Task's application code, in units of linux.ClockTick. + UserTicks uint64 + + // SysTicks is the amount of time the task goroutine has spent executing in + // the sentry, in units of linux.ClockTick. + SysTicks uint64 +} + +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) accountTaskGoroutineEnter(state TaskGoroutineState) { + now := t.k.CPUClockNow() + if t.gosched.State != TaskGoroutineRunningSys { + panic(fmt.Sprintf("Task goroutine switching from state %v (expected %v) to %v", t.gosched.State, TaskGoroutineRunningSys, state)) + } + t.goschedSeq.BeginWrite() + // This function is very hot; avoid defer. + t.gosched.SysTicks += now - t.gosched.Timestamp + t.gosched.Timestamp = now + t.gosched.State = state + t.goschedSeq.EndWrite() +} + +// Preconditions: The caller must be running on the task goroutine, and leaving +// a state indicated by a previous call to +// t.accountTaskGoroutineEnter(state). +func (t *Task) accountTaskGoroutineLeave(state TaskGoroutineState) { + now := t.k.CPUClockNow() + if t.gosched.State != state { + panic(fmt.Sprintf("Task goroutine switching from state %v (expected %v) to %v", t.gosched.State, state, TaskGoroutineRunningSys)) + } + t.goschedSeq.BeginWrite() + // This function is very hot; avoid defer. + if state == TaskGoroutineRunningApp { + t.gosched.UserTicks += now - t.gosched.Timestamp + } + t.gosched.Timestamp = now + t.gosched.State = TaskGoroutineRunningSys + t.goschedSeq.EndWrite() +} + +// TaskGoroutineSchedInfo returns a copy of t's task goroutine scheduling info. +// Most clients should use t.CPUStats() instead. +func (t *Task) TaskGoroutineSchedInfo() TaskGoroutineSchedInfo { + return SeqAtomicLoadTaskGoroutineSchedInfo(&t.goschedSeq, &t.gosched) +} + +// CPUStats returns the CPU usage statistics of t. +func (t *Task) CPUStats() usage.CPUStats { + return t.cpuStatsAt(t.k.CPUClockNow()) +} + +// Preconditions: now <= Kernel.CPUClockNow(). (Since Kernel.cpuClock is +// monotonic, this is satisfied if now is the result of a previous call to +// Kernel.CPUClockNow().) This requirement exists because otherwise a racing +// change to t.gosched can cause cpuStatsAt to adjust stats by too much, making +// the returned stats non-monotonic. +func (t *Task) cpuStatsAt(now uint64) usage.CPUStats { + tsched := t.TaskGoroutineSchedInfo() + if tsched.Timestamp < now { + // Update stats to reflect execution since the last update to + // t.gosched. + switch tsched.State { + case TaskGoroutineRunningSys: + tsched.SysTicks += now - tsched.Timestamp + case TaskGoroutineRunningApp: + tsched.UserTicks += now - tsched.Timestamp + } + } + return usage.CPUStats{ + UserTime: time.Duration(tsched.UserTicks * uint64(linux.ClockTick)), + SysTime: time.Duration(tsched.SysTicks * uint64(linux.ClockTick)), + VoluntarySwitches: atomic.LoadUint64(&t.yieldCount), + } +} + +// CPUStats returns the combined CPU usage statistics of all past and present +// threads in tg. +func (tg *ThreadGroup) CPUStats() usage.CPUStats { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + // Hack to get a pointer to the Kernel. + if tg.leader == nil { + // Per comment on tg.leader, this is only possible if nothing in the + // ThreadGroup has ever executed anyway. + return usage.CPUStats{} + } + now := tg.leader.k.CPUClockNow() + stats := tg.exitedCPUStats + // Account for active tasks. + for t := tg.tasks.Front(); t != nil; t = t.Next() { + stats.Accumulate(t.cpuStatsAt(now)) + } + return stats +} + +// JoinedChildCPUStats implements the semantics of RUSAGE_CHILDREN: "Return +// resource usage statistics for all children of [tg] that have terminated and +// been waited for. These statistics will include the resources used by +// grandchildren, and further removed descendants, if all of the intervening +// descendants waited on their terminated children." +func (tg *ThreadGroup) JoinedChildCPUStats() usage.CPUStats { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + return tg.childCPUStats +} + +// StateStatus returns a string representation of the task's current state, +// appropriate for /proc/[pid]/status. +func (t *Task) StateStatus() string { + switch s := t.TaskGoroutineSchedInfo().State; s { + case TaskGoroutineNonexistent: + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + switch t.exitState { + case TaskExitZombie: + return "Z (zombie)" + case TaskExitDead: + return "X (dead)" + default: + // The task goroutine can't exit before passing through + // runExitNotify, so this indicates that the task has been created, + // but the task goroutine hasn't yet started. The Linux equivalent + // is struct task_struct::state == TASK_NEW + // (kernel/fork.c:copy_process() => + // kernel/sched/core.c:sched_fork()), but the TASK_NEW bit is + // masked out by TASK_REPORT for /proc/[pid]/status, leaving only + // TASK_RUNNING. + return "R (running)" + } + case TaskGoroutineRunningSys, TaskGoroutineRunningApp: + return "R (running)" + case TaskGoroutineBlockedInterruptible: + return "S (sleeping)" + case TaskGoroutineStopped: + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + switch t.stop.(type) { + case *groupStop: + return "T (stopped)" + case *ptraceStop: + return "t (tracing stop)" + } + fallthrough + case TaskGoroutineBlockedUninterruptible: + // This is the name Linux uses for TASK_UNINTERRUPTIBLE and + // TASK_KILLABLE (= TASK_UNINTERRUPTIBLE | TASK_WAKEKILL): + // fs/proc/array.c:task_state_array. + return "D (disk sleep)" + default: + panic(fmt.Sprintf("Invalid TaskGoroutineState: %v", s)) + } +} + +// CPUMask returns a copy of t's allowed CPU mask. +func (t *Task) CPUMask() sched.CPUSet { + t.mu.Lock() + defer t.mu.Unlock() + return t.allowedCPUMask.Copy() +} + +// SetCPUMask sets t's allowed CPU mask based on mask. It takes ownership of +// mask. +// +// Preconditions: mask.Size() == +// sched.CPUSetSize(t.Kernel().ApplicationCores()). +func (t *Task) SetCPUMask(mask sched.CPUSet) error { + if want := sched.CPUSetSize(t.k.applicationCores); mask.Size() != want { + panic(fmt.Sprintf("Invalid CPUSet %v (expected %d bytes)", mask, want)) + } + + // Remove CPUs in mask above Kernel.applicationCores. + mask.ClearAbove(t.k.applicationCores) + + // Ensure that at least 1 CPU is still allowed. + if mask.NumCPUs() == 0 { + return syserror.EINVAL + } + + if t.k.useHostCores { + // No-op; pretend the mask was immediately changed back. + return nil + } + + t.tg.pidns.owner.mu.RLock() + rootTID := t.tg.pidns.owner.Root.tids[t] + t.tg.pidns.owner.mu.RUnlock() + + t.mu.Lock() + defer t.mu.Unlock() + t.allowedCPUMask = mask + atomic.StoreInt32(&t.cpu, assignCPU(mask, rootTID)) + return nil +} + +// CPU returns the cpu id for a given task. +func (t *Task) CPU() int32 { + if t.k.useHostCores { + return int32(hostcpu.GetCPU()) + } + + return atomic.LoadInt32(&t.cpu) +} + +// assignCPU returns the virtualized CPU number for the task with global TID +// tid and allowedCPUMask allowed. +func assignCPU(allowed sched.CPUSet, tid ThreadID) (cpu int32) { + // To pretend that threads are evenly distributed to allowed CPUs, choose n + // to be less than the number of CPUs in allowed ... + n := int(tid) % int(allowed.NumCPUs()) + // ... then pick the nth CPU in allowed. + allowed.ForEachCPU(func(c uint) { + if n--; n == 0 { + cpu = int32(c) + } + }) + return cpu +} + +// Niceness returns t's niceness. +func (t *Task) Niceness() int { + t.mu.Lock() + defer t.mu.Unlock() + return t.niceness +} + +// Priority returns t's priority. +func (t *Task) Priority() int { + t.mu.Lock() + defer t.mu.Unlock() + return t.niceness + 20 +} + +// SetNiceness sets t's niceness to n. +func (t *Task) SetNiceness(n int) { + t.mu.Lock() + defer t.mu.Unlock() + t.niceness = n +} + +// NumaPolicy returns t's current numa policy. +func (t *Task) NumaPolicy() (policy int32, nodeMask uint32) { + t.mu.Lock() + defer t.mu.Unlock() + return t.numaPolicy, t.numaNodeMask +} + +// SetNumaPolicy sets t's numa policy. +func (t *Task) SetNumaPolicy(policy int32, nodeMask uint32) { + t.mu.Lock() + defer t.mu.Unlock() + t.numaPolicy = policy + t.numaNodeMask = nodeMask +} diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go new file mode 100644 index 000000000..2340256b0 --- /dev/null +++ b/pkg/sentry/kernel/task_signals.go @@ -0,0 +1,1056 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +// This file defines the behavior of task signal handling. + +import ( + "fmt" + "sync/atomic" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// SignalAction is an internal signal action. +type SignalAction int + +// Available signal actions. +// Note that although we refer the complete set internally, +// the application is only capable of using the Default and +// Ignore actions from the system call interface. +const ( + SignalActionTerm SignalAction = iota + SignalActionCore + SignalActionStop + SignalActionIgnore + SignalActionHandler +) + +// Default signal handler actions. Note that for most signals, +// (except SIGKILL and SIGSTOP) these can be overridden by the app. +var defaultActions = map[linux.Signal]SignalAction{ + // POSIX.1-1990 standard. + linux.SIGHUP: SignalActionTerm, + linux.SIGINT: SignalActionTerm, + linux.SIGQUIT: SignalActionCore, + linux.SIGILL: SignalActionCore, + linux.SIGABRT: SignalActionCore, + linux.SIGFPE: SignalActionCore, + linux.SIGKILL: SignalActionTerm, // but see ThreadGroup.applySignalSideEffects + linux.SIGSEGV: SignalActionCore, + linux.SIGPIPE: SignalActionTerm, + linux.SIGALRM: SignalActionTerm, + linux.SIGTERM: SignalActionTerm, + linux.SIGUSR1: SignalActionTerm, + linux.SIGUSR2: SignalActionTerm, + linux.SIGCHLD: SignalActionIgnore, + linux.SIGCONT: SignalActionIgnore, // but see ThreadGroup.applySignalSideEffects + linux.SIGSTOP: SignalActionStop, + linux.SIGTSTP: SignalActionStop, + linux.SIGTTIN: SignalActionStop, + linux.SIGTTOU: SignalActionStop, + // POSIX.1-2001 standard. + linux.SIGBUS: SignalActionCore, + linux.SIGPROF: SignalActionTerm, + linux.SIGSYS: SignalActionCore, + linux.SIGTRAP: SignalActionCore, + linux.SIGURG: SignalActionIgnore, + linux.SIGVTALRM: SignalActionTerm, + linux.SIGXCPU: SignalActionCore, + linux.SIGXFSZ: SignalActionCore, + // The rest on linux. + linux.SIGSTKFLT: SignalActionTerm, + linux.SIGIO: SignalActionTerm, + linux.SIGPWR: SignalActionTerm, + linux.SIGWINCH: SignalActionIgnore, +} + +// computeAction figures out what to do given a signal number +// and an arch.SignalAct. SIGSTOP always results in a SignalActionStop, +// and SIGKILL always results in a SignalActionTerm. +// Signal 0 is always ignored as many programs use it for various internal functions +// and don't expect it to do anything. +// +// In the event the signal is not one of these, act.Handler determines what +// happens next. +// If act.Handler is: +// 0, the default action is taken; +// 1, the signal is ignored; +// anything else, the function returns SignalActionHandler. +func computeAction(sig linux.Signal, act arch.SignalAct) SignalAction { + switch sig { + case linux.SIGSTOP: + return SignalActionStop + case linux.SIGKILL: + return SignalActionTerm + case linux.Signal(0): + return SignalActionIgnore + } + + switch act.Handler { + case arch.SignalActDefault: + return defaultActions[sig] + case arch.SignalActIgnore: + return SignalActionIgnore + default: + return SignalActionHandler + } +} + +// UnblockableSignals contains the set of signals which cannot be blocked. +var UnblockableSignals = linux.MakeSignalSet(linux.SIGKILL, linux.SIGSTOP) + +// StopSignals is the set of signals whose default action is SignalActionStop. +var StopSignals = linux.MakeSignalSet(linux.SIGSTOP, linux.SIGTSTP, linux.SIGTTIN, linux.SIGTTOU) + +// dequeueSignalLocked returns a pending unmasked signal. If there are no +// pending unmasked signals, dequeueSignalLocked returns nil. +// +// Preconditions: t.tg.signalHandlers.mu must be locked. +func (t *Task) dequeueSignalLocked() *arch.SignalInfo { + if info := t.pendingSignals.dequeue(t.tr.SignalMask); info != nil { + return info + } + if info := t.tg.pendingSignals.dequeue(t.tr.SignalMask); info != nil { + return info + } + return nil +} + +// TakeSignal returns a pending signal not blocked by mask. Signal handlers are +// not affected. If there are no pending signals not blocked by mask, +// TakeSignal returns a nil SignalInfo. +func (t *Task) TakeSignal(mask linux.SignalSet) *arch.SignalInfo { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + if info := t.pendingSignals.dequeue(mask); info != nil { + return info + } + if info := t.tg.pendingSignals.dequeue(mask); info != nil { + return info + } + return nil +} + +// discardSpecificLocked removes all instances of the given signal from all +// signal queues in tg. +// +// Preconditions: The signal mutex must be locked. +func (tg *ThreadGroup) discardSpecificLocked(sig linux.Signal) { + tg.pendingSignals.discardSpecific(sig) + for t := tg.tasks.Front(); t != nil; t = t.Next() { + t.pendingSignals.discardSpecific(sig) + } +} + +// PendingSignals returns the set of pending signals. +func (t *Task) PendingSignals() linux.SignalSet { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + return t.pendingSignals.pendingSet | t.tg.pendingSignals.pendingSet +} + +// deliverSignal delivers the given signal and returns the following run state. +func (t *Task) deliverSignal(info *arch.SignalInfo, act arch.SignalAct) taskRunState { + sigact := computeAction(linux.Signal(info.Signo), act) + + if t.haveSyscallReturn { + if sre, ok := SyscallRestartErrnoFromReturn(t.Arch().Return()); ok { + // Signals that are ignored, cause a thread group stop, or + // terminate the thread group do not interact with interrupted + // syscalls; in Linux terms, they are never returned to the signal + // handling path from get_signal => get_signal_to_deliver. The + // behavior of an interrupted syscall is determined by the first + // signal that is actually handled (by userspace). + if sigact == SignalActionHandler { + switch { + case sre == ERESTARTNOHAND: + fallthrough + case sre == ERESTART_RESTARTBLOCK: + fallthrough + case (sre == ERESTARTSYS && !act.IsRestart()): + t.Debugf("Not restarting syscall %d after errno %d: interrupted by signal %d", t.Arch().SyscallNo(), sre, info.Signo) + t.Arch().SetReturn(uintptr(-t.ExtractErrno(syserror.EINTR, -1))) + default: + t.Debugf("Restarting syscall %d after errno %d: interrupted by signal %d", t.Arch().SyscallNo(), sre, info.Signo) + t.Arch().RestartSyscall() + } + } + } + } + + switch sigact { + case SignalActionTerm, SignalActionCore: + // "Default action is to terminate the process." - signal(7) + t.Debugf("Signal %d: terminating thread group", info.Signo) + t.PrepareGroupExit(ExitStatus{Signo: int(info.Signo)}) + return (*runExit)(nil) + + case SignalActionStop: + // "Default action is to stop the process." + t.initiateGroupStop(info) + + case SignalActionIgnore: + // "Default action is to ignore the signal." + t.Debugf("Signal %d: ignored", info.Signo) + + case SignalActionHandler: + // Try to deliver the signal to the user-configured handler. + t.Debugf("Signal %d: delivering to handler", info.Signo) + if err := t.deliverSignalToHandler(info, act); err != nil { + t.Warningf("Failed to deliver signal %+v to user handler: %v", info, err) + // Send a forced SIGSEGV. If the signal that couldn't be delivered + // was a SIGSEGV, force the handler to SIG_DFL. + t.forceSignal(linux.SIGSEGV, linux.Signal(info.Signo) == linux.SIGSEGV /* unconditional */) + t.SendSignal(sigPriv(linux.SIGSEGV)) + } + + default: + panic(fmt.Sprintf("Unknown signal action %+v, %d?", info, computeAction(linux.Signal(info.Signo), act))) + } + return (*runInterrupt)(nil) +} + +// deliverSignalToHandler changes the task's userspace state to enter the given +// user-configured handler for the given signal. +func (t *Task) deliverSignalToHandler(info *arch.SignalInfo, act arch.SignalAct) error { + // Signal delivery to an application handler interrupts restartable + // sequences. + t.rseqInterrupt() + + // Are executing on the main stack, + // or the provided alternate stack? + sp := usermem.Addr(t.Arch().Stack()) + + // N.B. This is a *copy* of the alternate stack that the user's signal + // handler expects to see in its ucontext (even if it's not in use). + alt := t.signalStack + if act.IsOnStack() && alt.IsEnabled() { + alt.SetOnStack() + if !t.OnSignalStack(alt) { + sp = usermem.Addr(alt.Top()) + } + } + + // Set up the signal handler. If we have a saved signal mask, the signal + // handler should run with the current mask, but sigreturn should restore + // the saved one. + st := &arch.Stack{t.Arch(), t.MemoryManager(), sp} + mask := t.tr.SignalMask + if t.haveSavedSignalMask { + mask = t.savedSignalMask + } + if err := t.Arch().SignalSetup(st, &act, info, &alt, mask); err != nil { + return err + } + t.haveSavedSignalMask = false + + // Add our signal mask. + newMask := t.tr.SignalMask | act.Mask + if !act.IsNoDefer() { + newMask |= linux.SignalSetOf(linux.Signal(info.Signo)) + } + t.SetSignalMask(newMask) + + return nil +} + +var ctrlResume = &SyscallControl{ignoreReturn: true} + +// SignalReturn implements sigreturn(2) (if rt is false) or rt_sigreturn(2) (if +// rt is true). +func (t *Task) SignalReturn(rt bool) (*SyscallControl, error) { + st := t.Stack() + sigset, err := t.Arch().SignalRestore(st, rt) + if err != nil { + return nil, err + } + + // Restore our signal mask. SIGKILL and SIGSTOP should not be blocked. + t.SetSignalMask(sigset &^ UnblockableSignals) + + // TODO: sys_rt_sigreturn also calls restore_altstack from + // uc.stack, allowing the signal handler to implicitly mutate the signal + // stack. + + return ctrlResume, nil +} + +// SendSignal sends the given signal to t. +// +// The following errors may be returned: +// +// syserror.ESRCH - The task has exited. +// syserror.EINVAL - The signal is not valid. +// syserror.EAGAIN - THe signal is realtime, and cannot be queued. +// +func (t *Task) SendSignal(info *arch.SignalInfo) error { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + return t.sendSignalLocked(info, false /* group */) +} + +// SendGroupSignal sends the given signal to t's thread group. +func (t *Task) SendGroupSignal(info *arch.SignalInfo) error { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + return t.sendSignalLocked(info, true /* group */) +} + +// SendSignal sends the given signal to tg, using tg's leader to determine if +// the signal is blocked. +func (tg *ThreadGroup) SendSignal(info *arch.SignalInfo) error { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + tg.signalHandlers.mu.Lock() + defer tg.signalHandlers.mu.Unlock() + return tg.leader.sendSignalLocked(info, true /* group */) +} + +// Preconditions: The TaskSet mutex must be locked. +func (t *Task) onCPULocked(includeSys bool) bool { + // Task is exiting. + if t.exitState != TaskExitNone { + return false + } + + switch t.TaskGoroutineSchedInfo().State { + case TaskGoroutineRunningSys: + return includeSys + case TaskGoroutineRunningApp: + return true + default: + return false + } +} + +// SendTimerSignal mimics the process timer signal delivery behavior in linux: +// signals are delivered to the thread that triggers the timer expiration (see +// kernel/time/posix-cpu-timers.c:check_process_timers(). This +// means +// 1) the thread is running on cpu at the time. +// 2) a thread runs more frequently will get more of those signals. +// +// We approximate this behavior by selecting a running task in a round-robin +// fashion. Statistically, a thread running more often should have a higher +// probability to be selected. +func (tg *ThreadGroup) SendTimerSignal(info *arch.SignalInfo, includeSys bool) error { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + tg.signalHandlers.mu.Lock() + defer tg.signalHandlers.mu.Unlock() + + // Find the next running threads. + var t *Task + if tg.lastTimerSignalTask == nil { + t = tg.tasks.Front() + } else { + t = tg.lastTimerSignalTask.Next() + } + + // Iterate from lastTimerSignalTask.Next() to the last task in the task list. + for t != nil { + if t.onCPULocked(includeSys) { + tg.lastTimerSignalTask = t + return t.sendSignalLocked(info, true /* group */) + } + t = t.Next() + } + + // t is nil when we reach here. If lastTimerSignalTask is not nil, iterate + // from Front to lastTimerSignalTask. + if tg.lastTimerSignalTask != nil { + for t := tg.tasks.Front(); t != tg.lastTimerSignalTask.Next(); t = t.Next() { + if t.onCPULocked(includeSys) { + tg.lastTimerSignalTask = t + return t.sendSignalLocked(info, true /* group */) + } + } + } + + // No running threads? Just try the leader. + tg.lastTimerSignalTask = tg.leader + return tg.leader.sendSignalLocked(info, true /* group */) +} + +func (t *Task) sendSignalLocked(info *arch.SignalInfo, group bool) error { + if t.exitState == TaskExitDead { + return syserror.ESRCH + } + sig := linux.Signal(info.Signo) + if sig == 0 { + return nil + } + if !sig.IsValid() { + return syserror.EINVAL + } + + // Signal side effects apply even if the signal is ultimately discarded. + t.tg.applySignalSideEffectsLocked(sig) + + // TODO: "Only signals for which the "init" process has established a + // signal handler can be sent to the "init" process by other members of the + // PID namespace. This restriction applies even to privileged processes, + // and prevents other members of the PID namespace from accidentally + // killing the "init" process." - pid_namespaces(7). We don't currently do + // this for child namespaces, though we should; we also don't do this for + // the root namespace (the same restriction applies to global init on + // Linux), where whether or not we should is much murkier. In practice, + // most sandboxed applications are not prepared to function as an init + // process. + + // Unmasked, ignored signals are discarded without being queued, unless + // they will be visible to a tracer. Even for group signals, it's the + // originally-targeted task's signal mask and tracer that matter; compare + // Linux's kernel/signal.c:__send_signal() => prepare_signal() => + // sig_ignored(). + ignored := computeAction(sig, t.tg.signalHandlers.actions[sig]) == SignalActionIgnore + if linux.SignalSetOf(sig)&t.tr.SignalMask == 0 && ignored && !t.hasTracer() { + t.Debugf("Discarding ignored signal %d", sig) + return nil + } + + q := &t.pendingSignals + if group { + q = &t.tg.pendingSignals + } + if !q.enqueue(info) { + if sig.IsRealtime() { + return syserror.EAGAIN + } + t.Debugf("Discarding duplicate signal %d", sig) + return nil + } + + // Find a receiver to notify. Note that the task we choose to notify, if + // any, may not be the task that actually dequeues and handles the signal; + // e.g. a racing signal mask change may cause the notified task to become + // ineligible, or a racing sibling task may dequeue the signal first. + if t.canReceiveSignalLocked(sig) { + t.Debugf("Notified of signal %d", sig) + t.interrupt() + return nil + } + if group { + if nt := t.tg.findSignalReceiverLocked(sig); nt != nil { + nt.Debugf("Notified of group signal %d", sig) + nt.interrupt() + return nil + } + } + t.Debugf("No task notified of signal %d", sig) + return nil +} + +func (tg *ThreadGroup) applySignalSideEffectsLocked(sig linux.Signal) { + switch { + case linux.SignalSetOf(sig)&StopSignals != 0: + // Stop signals cause all prior SIGCONT to be discarded. (This is + // despite the fact this has little effect since SIGCONT's most + // important effect is applied when the signal is sent in the branch + // below, not when the signal is delivered.) + tg.discardSpecificLocked(linux.SIGCONT) + case sig == linux.SIGCONT: + // "The SIGCONT signal has a side effect of waking up (all threads of) + // a group-stopped process. This side effect happens before + // signal-delivery-stop. The tracer can't suppress this side effect (it + // can only suppress signal injection, which only causes the SIGCONT + // handler to not be executed in the tracee, if such a handler is + // installed." - ptrace(2) + tg.endGroupStopLocked(true) + case sig == linux.SIGKILL: + // "SIGKILL does not generate signal-delivery-stop and therefore the + // tracer can't suppress it. SIGKILL kills even within system calls + // (syscall-exit-stop is not generated prior to death by SIGKILL)." - + // ptrace(2) + // + // Note that this differs from ThreadGroup.requestExit in that it + // ignores tg.execing. + if !tg.exiting { + tg.exiting = true + tg.exitStatus = ExitStatus{Signo: int(linux.SIGKILL)} + } + for t := tg.tasks.Front(); t != nil; t = t.Next() { + t.killLocked() + } + } +} + +// canReceiveSignalLocked returns true if t should be interrupted to receive +// the given signal. canReceiveSignalLocked is analogous to Linux's +// kernel/signal.c:wants_signal(), but see below for divergences. +// +// Preconditions: The signal mutex must be locked. +func (t *Task) canReceiveSignalLocked(sig linux.Signal) bool { + // - Do not choose tasks that are blocking the signal. + if linux.SignalSetOf(sig)&t.tr.SignalMask != 0 { + return false + } + // - No need to check Task.exitState, as the exit path sets every bit in the + // signal mask when it transitions from TaskExitNone to TaskExitInitiated. + // - No special case for SIGKILL: SIGKILL already interrupted all tasks in the + // task group via applySignalSideEffects => killLocked. + // - Do not choose stopped tasks, which cannot handle signals. + if t.stop != nil { + return false + } + // - TODO: No special case for when t is also the sending task, + // because the identity of the sender is unknown. + // - Do not choose tasks that have already been interrupted, as they may be + // busy handling another signal. + if len(t.interruptChan) != 0 { + return false + } + return true +} + +// findSignalReceiverLocked returns a task in tg that should be interrupted to +// receive the given signal. If no such task exists, findSignalReceiverLocked +// returns nil. +// +// Linux actually records curr_target to balance the group signal targets. +// +// Preconditions: The signal mutex must be locked. +func (tg *ThreadGroup) findSignalReceiverLocked(sig linux.Signal) *Task { + for t := tg.tasks.Front(); t != nil; t = t.Next() { + if t.canReceiveSignalLocked(sig) { + return t + } + } + return nil +} + +// forceSignal ensures that the task is not ignoring or blocking the given +// signal. If unconditional is true, forceSignal takes action even if the +// signal isn't being ignored or blocked. +func (t *Task) forceSignal(sig linux.Signal, unconditional bool) { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + t.forceSignalLocked(sig, unconditional) +} + +func (t *Task) forceSignalLocked(sig linux.Signal, unconditional bool) { + blocked := linux.SignalSetOf(sig)&t.tr.SignalMask != 0 + act := t.tg.signalHandlers.actions[sig] + ignored := act.Handler == arch.SignalActIgnore + if blocked || ignored || unconditional { + act.Handler = arch.SignalActDefault + t.tg.signalHandlers.actions[sig] = act + if blocked { + t.setSignalMaskLocked(t.tr.SignalMask &^ linux.SignalSetOf(sig)) + } + } +} + +// SignalMask returns a copy of t's signal mask. +func (t *Task) SignalMask() linux.SignalSet { + return linux.SignalSet(atomic.LoadUint64((*uint64)(&t.tr.SignalMask))) +} + +// SetSignalMask sets t's signal mask. +// +// Preconditions: SetSignalMask can only be called by the task goroutine. +// t.exitState < TaskExitZombie. +func (t *Task) SetSignalMask(mask linux.SignalSet) { + // By precondition, t prevents t.tg from completing an execve and mutating + // t.tg.signalHandlers, so we can skip the TaskSet mutex. + t.tg.signalHandlers.mu.Lock() + t.setSignalMaskLocked(mask) + t.tg.signalHandlers.mu.Unlock() +} + +// Preconditions: The signal mutex must be locked. +func (t *Task) setSignalMaskLocked(mask linux.SignalSet) { + oldMask := t.tr.SignalMask + atomic.StoreUint64((*uint64)(&t.tr.SignalMask), uint64(mask)) + + // If the new mask blocks any signals that were not blocked by the old + // mask, and at least one such signal is pending in tg.pendingSignals, and + // t has been woken, it could be the case that t was woken to handle that + // signal, but will no longer do so as a result of its new signal mask, so + // we have to pick a replacement. + blocked := mask &^ oldMask + blockedGroupPending := blocked & t.tg.pendingSignals.pendingSet + if blockedGroupPending != 0 && t.interrupted() { + linux.ForEachSignal(blockedGroupPending, func(sig linux.Signal) { + if nt := t.tg.findSignalReceiverLocked(sig); nt != nil { + nt.interrupt() + return + } + }) + // We have to re-issue the interrupt consumed by t.interrupted() since + // it might have been for a different reason. + t.interruptSelf() + } + + // Conversely, if the new mask unblocks any signals that were blocked by + // the old mask, and at least one such signal is pending, we may now need + // to handle that signal. + unblocked := oldMask &^ mask + unblockedPending := unblocked & (t.pendingSignals.pendingSet | t.tg.pendingSignals.pendingSet) + if unblockedPending != 0 { + t.interruptSelf() + } +} + +// SetSavedSignalMask sets the saved signal mask (see Task.savedSignalMask's +// comment). +// +// Preconditions: SetSavedSignalMask can only be called by the task goroutine. +func (t *Task) SetSavedSignalMask(mask linux.SignalSet) { + t.savedSignalMask = mask + t.haveSavedSignalMask = true +} + +// SignalStack returns the task-private signal stack. +func (t *Task) SignalStack() arch.SignalStack { + return t.signalStack +} + +// OnSignalStack returns true if, when the task resumes running, it will run on +// the task-private signal stack. +func (t *Task) OnSignalStack(s arch.SignalStack) bool { + sp := usermem.Addr(t.Arch().Stack()) + return usermem.Addr(s.Addr) <= sp && sp < usermem.Addr(s.Addr+s.Size) +} + +// SetSignalStack sets the task-private signal stack and clears the +// SignalStackFlagDisable, since we have a signal stack. +func (t *Task) SetSignalStack(alt arch.SignalStack) error { + // Mask out irrelevant parts: only disable matters. + alt.Flags &= arch.SignalStackFlagDisable + t.signalStack = alt + return nil +} + +// SetSignalAct atomically sets the thread group's signal action for signal sig +// to *actptr (if actptr is not nil) and returns the old signal action. +func (tg *ThreadGroup) SetSignalAct(sig linux.Signal, actptr *arch.SignalAct) (arch.SignalAct, error) { + if !sig.IsValid() { + return arch.SignalAct{}, syserror.EINVAL + } + + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + sh := tg.signalHandlers + sh.mu.Lock() + defer sh.mu.Unlock() + oldact := sh.actions[sig] + if actptr != nil { + if sig == linux.SIGKILL || sig == linux.SIGSTOP { + return oldact, syserror.EINVAL + } + + act := *actptr + act.Mask &^= UnblockableSignals + sh.actions[sig] = act + // From POSIX, by way of Linux: + // + // "Setting a signal action to SIG_IGN for a signal that is pending + // shall cause the pending signal to be discarded, whether or not it is + // blocked." + // + // "Setting a signal action to SIG_DFL for a signal that is pending and + // whose default action is to ignore the signal (for example, SIGCHLD), + // shall cause the pending signal to be discarded, whether or not it is + // blocked." + if computeAction(sig, act) == SignalActionIgnore { + tg.discardSpecificLocked(sig) + } + } + return oldact, nil +} + +// CopyOutSignalAct converts the given SignalAct into an architecture-specific +// type and then copies it out to task memory. +func (t *Task) CopyOutSignalAct(addr usermem.Addr, s *arch.SignalAct) error { + n := t.Arch().NewSignalAct() + n.SerializeFrom(s) + _, err := t.CopyOut(addr, n) + return err +} + +// CopyInSignalAct copies an architecture-specific sigaction type from task +// memory and then converts it into a SignalAct. +func (t *Task) CopyInSignalAct(addr usermem.Addr) (arch.SignalAct, error) { + n := t.Arch().NewSignalAct() + var s arch.SignalAct + if _, err := t.CopyIn(addr, n); err != nil { + return s, err + } + n.DeserializeTo(&s) + return s, nil +} + +// CopyOutSignalStack converts the given SignalStack into an +// architecture-specific type and then copies it out to task memory. +func (t *Task) CopyOutSignalStack(addr usermem.Addr, s *arch.SignalStack) error { + n := t.Arch().NewSignalStack() + n.SerializeFrom(s) + _, err := t.CopyOut(addr, n) + return err +} + +// CopyInSignalStack copies an architecture-specific stack_t from task memory +// and then converts it into a SignalStack. +func (t *Task) CopyInSignalStack(addr usermem.Addr) (arch.SignalStack, error) { + n := t.Arch().NewSignalStack() + var s arch.SignalStack + if _, err := t.CopyIn(addr, n); err != nil { + return s, err + } + n.DeserializeTo(&s) + return s, nil +} + +// groupStop is a TaskStop placed on tasks that have received a stop signal +// (SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU). (The term "group-stop" originates from +// the ptrace man page.) +type groupStop struct{} + +// Killable implements TaskStop.Killable. +func (*groupStop) Killable() bool { return true } + +type groupStopPhase int + +const ( + // groupStopNone indicates that a thread group is not in, or attempting to + // enter or leave, a group stop. + groupStopNone groupStopPhase = iota + + // groupStopDequeued indicates that at least one task in a thread group has + // dequeued a stop signal (or dequeued any signal and entered a + // signal-delivery-stop as a result, which allows ptrace to change the + // signal into a stop signal), but temporarily dropped the signal mutex + // without initiating the group stop. + // + // groupStopDequeued is analogous to JOBCTL_STOP_DEQUEUED in Linux. + groupStopDequeued + + // groupStopInitiated indicates that a task in a thread group has initiated + // a group stop, but not all tasks in the thread group have acknowledged + // entering the group stop. + // + // groupStopInitiated is represented by JOBCTL_STOP_PENDING && + // !SIGNAL_STOP_STOPPED in Linux. + groupStopInitiated + + // groupStopComplete indicates that all tasks in a thread group have + // acknowledged entering the group stop, and the last one to do so has + // notified the thread group's parent. + // + // groupStopComplete is represented by JOBCTL_STOP_PENDING && + // SIGNAL_STOP_STOPPED in Linux. + groupStopComplete +) + +// initiateGroupStop attempts to initiate a group stop based on a +// previously-dequeued stop signal. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) initiateGroupStop(info *arch.SignalInfo) { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + if t.tg.groupStopPhase != groupStopDequeued { + t.Debugf("Signal %d: not stopping thread group: lost to racing signal", info.Signo) + return + } + if t.tg.exiting { + t.Debugf("Signal %d: not stopping thread group: lost to racing group exit", info.Signo) + return + } + if t.tg.execing != nil { + t.Debugf("Signal %d: not stopping thread group: lost to racing execve", info.Signo) + return + } + t.Debugf("Signal %d: stopping thread group", info.Signo) + t.tg.groupStopPhase = groupStopInitiated + t.tg.groupStopSignal = linux.Signal(info.Signo) + t.tg.groupStopCount = 0 + for t2 := t.tg.tasks.Front(); t2 != nil; t2 = t2.Next() { + t2.groupStopRequired = true + t2.groupStopAcknowledged = false + t2.interrupt() + } +} + +// endGroupStopLocked ensures that all prior stop signals received by tg are +// not stopping tg and will not stop tg in the future. If broadcast is true, +// parent and tracer notification will be scheduled if appropriate. +// +// Preconditions: The signal mutex must be locked. +func (tg *ThreadGroup) endGroupStopLocked(broadcast bool) { + // Discard all previously-queued stop signals. + linux.ForEachSignal(StopSignals, tg.discardSpecificLocked) + + if tg.groupStopPhase != groupStopNone { + tg.leader.Debugf("Ending group stop currently in phase %d", tg.groupStopPhase) + if tg.groupStopPhase == groupStopInitiated || tg.groupStopPhase == groupStopComplete { + tg.groupStopSignal = 0 + for t := tg.tasks.Front(); t != nil; t = t.Next() { + if _, ok := t.stop.(*groupStop); ok { + t.endInternalStopLocked() + } + } + if broadcast { + // Instead of notifying the parent here, set groupContNotify so + // that one of the continuing tasks does so. (Linux does + // something similar.) The reason we do this is to keep locking + // sane. In order to send a signal to the parent, we need to + // lock its signal mutex, but we're already holding tg's signal + // mutex, and the TaskSet mutex must be locked for writing for + // us to hold two signal mutexes. Since we don't want to + // require this for endGroupStopLocked (which is called from + // signal-sending paths), nor do we want to lose atomicity by + // releasing the mutexes we're already holding, just let the + // continuing thread group deal with it. + tg.groupContNotify = true + tg.groupContInterrupted = tg.groupStopPhase == groupStopInitiated + tg.groupContWaitable = true + } + } + // If groupStopPhase was groupStopDequeued, setting it to groupStopNone + // will cause following calls to initiateGroupStop to recognize that + // the group stop has been cancelled. + tg.groupStopPhase = groupStopNone + } +} + +// signalStop sends a signal to t's thread group of a new group stop, group +// continue, or ptrace stop, if appropriate. code and status are set in the +// signal sent to tg, if any. +// +// Preconditions: The TaskSet mutex must be locked (for reading or writing). +func (t *Task) signalStop(target *Task, code int32, status int32) { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + act, ok := t.tg.signalHandlers.actions[linux.SIGCHLD] + if !ok || (act.Handler != arch.SignalActIgnore && act.Flags&arch.SignalFlagNoCldStop == 0) { + sigchld := &arch.SignalInfo{ + Signo: int32(linux.SIGCHLD), + Code: code, + } + sigchld.SetPid(int32(t.tg.pidns.tids[target])) + sigchld.SetUid(int32(target.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow())) + sigchld.SetStatus(status) + // TODO: Set utime, stime. + t.sendSignalLocked(sigchld, true /* group */) + } +} + +// The runInterrupt state handles conditions indicated by interrupts. +type runInterrupt struct{} + +func (*runInterrupt) execute(t *Task) taskRunState { + // Interrupts are de-duplicated (if t is interrupted twice before + // t.interrupted() is called, t.interrupted() will only return true once), + // so early exits from this function must re-enter the runInterrupt state + // to check for more interrupt-signaled conditions. + + t.tg.signalHandlers.mu.Lock() + + // Did we just leave a group stop? + if t.tg.groupContNotify { + t.tg.groupContNotify = false + sig := t.tg.groupStopSignal + intr := t.tg.groupContInterrupted + t.tg.signalHandlers.mu.Unlock() + t.tg.pidns.owner.mu.RLock() + // For consistency with Linux, if the parent and (thread group + // leader's) tracer are in the same thread group, deduplicate + // notifications. + notifyParent := t.tg.leader.parent != nil + if tracer := t.tg.leader.ptraceTracer.Load().(*Task); tracer != nil { + if notifyParent && tracer.tg == t.tg.leader.parent.tg { + notifyParent = false + } + // Sending CLD_STOPPED to the tracer doesn't really make any sense; + // the thread group leader may have already entered the stop and + // notified its tracer accordingly. But it's consistent with + // Linux... + if intr { + tracer.signalStop(t.tg.leader, arch.CLD_STOPPED, int32(sig)) + if !notifyParent { + tracer.tg.eventQueue.Notify(EventGroupContinue | EventTraceeStop | EventChildGroupStop) + } else { + tracer.tg.eventQueue.Notify(EventGroupContinue | EventTraceeStop) + } + } else { + tracer.signalStop(t.tg.leader, arch.CLD_CONTINUED, int32(sig)) + tracer.tg.eventQueue.Notify(EventGroupContinue) + } + } + if notifyParent { + // If groupContInterrupted, do as Linux does and pretend the group + // stop completed just before it ended. The theoretical behavior in + // this case would be to send a SIGCHLD indicating the completed + // stop, followed by a SIGCHLD indicating the continue. However, + // SIGCHLD is a standard signal, so the latter would always be + // dropped. Hence sending only the former is equivalent. + if intr { + t.tg.leader.parent.signalStop(t.tg.leader, arch.CLD_STOPPED, int32(sig)) + t.tg.leader.parent.tg.eventQueue.Notify(EventGroupContinue | EventChildGroupStop) + } else { + t.tg.leader.parent.signalStop(t.tg.leader, arch.CLD_CONTINUED, int32(sig)) + t.tg.leader.parent.tg.eventQueue.Notify(EventGroupContinue) + } + } + t.tg.pidns.owner.mu.RUnlock() + return (*runInterrupt)(nil) + } + + // Do we need to enter a group stop? + if t.groupStopRequired { + t.groupStopRequired = false + sig := t.tg.groupStopSignal + notifyParent := false + if !t.groupStopAcknowledged { + t.groupStopAcknowledged = true + t.tg.groupStopCount++ + if t.tg.groupStopCount == t.tg.activeTasks { + t.Debugf("Completing group stop") + notifyParent = true + t.tg.groupStopPhase = groupStopComplete + t.tg.groupStopWaitable = true + t.tg.groupContNotify = false + t.tg.groupContWaitable = false + } + } + // Drop the signal mutex so we can take the TaskSet mutex. + t.tg.signalHandlers.mu.Unlock() + + t.tg.pidns.owner.mu.RLock() + if t.tg.leader.parent == nil { + notifyParent = false + } + if tracer := t.Tracer(); tracer != nil { + t.ptraceCode = int32(sig) + t.ptraceSiginfo = nil + if t.beginPtraceStopLocked() { + tracer.signalStop(t, arch.CLD_STOPPED, int32(sig)) + // For consistency with Linux, if the parent and tracer are in the + // same thread group, deduplicate notification signals. + if notifyParent && tracer.tg == t.tg.leader.parent.tg { + notifyParent = false + tracer.tg.eventQueue.Notify(EventChildGroupStop | EventTraceeStop) + } else { + tracer.tg.eventQueue.Notify(EventTraceeStop) + } + } + } else { + t.tg.signalHandlers.mu.Lock() + if !t.killedLocked() { + t.beginInternalStopLocked((*groupStop)(nil)) + } + t.tg.signalHandlers.mu.Unlock() + } + if notifyParent { + t.tg.leader.parent.signalStop(t.tg.leader, arch.CLD_STOPPED, int32(sig)) + t.tg.leader.parent.tg.eventQueue.Notify(EventChildGroupStop) + } + t.tg.pidns.owner.mu.RUnlock() + + return (*runInterrupt)(nil) + } + + // Are there signals pending? + if info := t.dequeueSignalLocked(); info != nil { + if linux.SignalSetOf(linux.Signal(info.Signo))&StopSignals != 0 && t.tg.groupStopPhase == groupStopNone { + // Indicate that we've dequeued a stop signal before + // unlocking the signal mutex; initiateGroupStop will check + // that the phase hasn't changed (or is at least another + // "stop signal dequeued" phase) after relocking it. + t.tg.groupStopPhase = groupStopDequeued + } + if t.ptraceSignalLocked(info) { + // Dequeueing the signal action must wait until after the + // signal-delivery-stop ends since the tracer can change or + // suppress the signal. + t.tg.signalHandlers.mu.Unlock() + return (*runInterruptAfterSignalDeliveryStop)(nil) + } + act := t.tg.signalHandlers.dequeueAction(linux.Signal(info.Signo)) + t.tg.signalHandlers.mu.Unlock() + return t.deliverSignal(info, act) + } + + t.tg.signalHandlers.mu.Unlock() + return (*runApp)(nil) +} + +type runInterruptAfterSignalDeliveryStop struct{} + +func (*runInterruptAfterSignalDeliveryStop) execute(t *Task) taskRunState { + t.tg.pidns.owner.mu.Lock() + // Can't defer unlock: deliverSignal must be called without holding TaskSet + // mutex. + sig := linux.Signal(t.ptraceCode) + defer func() { + t.ptraceSiginfo = nil + }() + if !sig.IsValid() { + t.tg.pidns.owner.mu.Unlock() + return (*runInterrupt)(nil) + } + info := t.ptraceSiginfo + if sig != linux.Signal(info.Signo) { + info.Signo = int32(sig) + info.Errno = 0 + info.Code = arch.SignalInfoUser + // pid isn't a valid field for all signal numbers, but Linux + // doesn't care (kernel/signal.c:ptrace_signal()). + // + // Linux uses t->parent for the tid and uid here, which is the tracer + // if it hasn't detached or the real parent otherwise. + parent := t.parent + if tracer := t.Tracer(); tracer != nil { + parent = tracer + } + if parent == nil { + // Tracer has detached and t was created by Kernel.CreateProcess(). + // Pretend the parent is in an ancestor PID + user namespace. + info.SetPid(0) + info.SetUid(int32(auth.OverflowUID)) + } else { + info.SetPid(int32(t.tg.pidns.tids[parent])) + info.SetUid(int32(parent.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow())) + } + } + t.tg.signalHandlers.mu.Lock() + t.tg.pidns.owner.mu.Unlock() + // If the signal is masked, re-queue it. + if linux.SignalSetOf(sig)&t.tr.SignalMask != 0 { + t.sendSignalLocked(info, false /* group */) + t.tg.signalHandlers.mu.Unlock() + return (*runInterrupt)(nil) + } + act := t.tg.signalHandlers.dequeueAction(linux.Signal(info.Signo)) + t.tg.signalHandlers.mu.Unlock() + return t.deliverSignal(info, act) +} diff --git a/pkg/sentry/kernel/task_start.go b/pkg/sentry/kernel/task_start.go new file mode 100644 index 000000000..801cb3395 --- /dev/null +++ b/pkg/sentry/kernel/task_start.go @@ -0,0 +1,252 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// TaskConfig defines the configuration of a new Task (see below). +type TaskConfig struct { + // Kernel is the owning Kernel. + *Kernel + + // Parent is the new task's parent. Parent may be nil. + Parent *Task + + // ThreadGroup is the ThreadGroup the new task belongs to. + *ThreadGroup + + // TaskContext is the TaskContext of the new task. + *TaskContext + + // TaskResources is the TaskResources of the new task. + *TaskResources + + // Credentials is the Credentials of the new task. + Credentials *auth.Credentials + + // Niceness is the niceness of the new task. + Niceness int + + // If NetworkNamespaced is true, the new task should observe a non-root + // network namespace. + NetworkNamespaced bool + + // AllowedCPUMask contains the cpus that this task can run on. + AllowedCPUMask sched.CPUSet + + // UTSNamespace is the UTSNamespace of the new task. + UTSNamespace *UTSNamespace + + // IPCNamespace is the IPCNamespace of the new task. + IPCNamespace *IPCNamespace +} + +// NewTask creates a new task defined by TaskConfig. +// Whether or not NewTask is successful, it takes ownership of both TaskContext +// and TaskResources of the TaskConfig. +// +// NewTask does not start the returned task; the caller must call Task.Start. +func (ts *TaskSet) NewTask(cfg *TaskConfig) (*Task, error) { + t, err := ts.newTask(cfg) + if err != nil { + cfg.TaskContext.release() + cfg.TaskResources.release() + return nil, err + } + return t, nil +} + +// newTask is a helper for TaskSet.NewTask that only takes ownership of TaskContext +// and TaskResources of the TaskConfig if it succeeds. +func (ts *TaskSet) newTask(cfg *TaskConfig) (*Task, error) { + tg := cfg.ThreadGroup + tc := cfg.TaskContext + t := &Task{ + taskNode: taskNode{ + tg: tg, + parent: cfg.Parent, + children: make(map[*Task]struct{}), + }, + runState: (*runApp)(nil), + interruptChan: make(chan struct{}, 1), + signalStack: arch.SignalStack{Flags: arch.SignalStackFlagDisable}, + tc: *tc, + tr: *cfg.TaskResources, + p: cfg.Kernel.Platform.NewContext(), + k: cfg.Kernel, + ptraceTracees: make(map[*Task]struct{}), + allowedCPUMask: cfg.AllowedCPUMask.Copy(), + ioUsage: &usage.IO{}, + creds: cfg.Credentials, + niceness: cfg.Niceness, + netns: cfg.NetworkNamespaced, + utsns: cfg.UTSNamespace, + ipcns: cfg.IPCNamespace, + rseqCPU: -1, + futexWaiter: futex.NewWaiter(), + } + t.endStopCond.L = &t.tg.signalHandlers.mu + t.ptraceTracer.Store((*Task)(nil)) + // We don't construct t.blockingTimer until Task.run(); see that function + // for justification. + + // Make the new task (and possibly thread group) visible to the rest of + // the system atomically. + ts.mu.Lock() + defer ts.mu.Unlock() + tg.signalHandlers.mu.Lock() + defer tg.signalHandlers.mu.Unlock() + if tg.exiting || tg.execing != nil { + // If the caller is in the same thread group, then what we return + // doesn't matter too much since the caller will exit before it returns + // to userspace. If the caller isn't in the same thread group, then + // we're in uncharted territory and can return whatever we want. + return nil, syserror.EINTR + } + if err := ts.assignTIDsLocked(t); err != nil { + return nil, err + } + // Below this point, newTask is expected not to fail (there is no rollback + // of assignTIDsLocked or any of the following). + + // Logging on t's behalf will panic if t.logPrefix hasn't been initialized. + // This is the earliest point at which we can do so (since t now has thread + // IDs). + t.updateLogPrefixLocked() + + if t.parent != nil { + t.parent.children[t] = struct{}{} + } + + if tg.leader == nil { + // New thread group. + tg.leader = t + if parentPG := tg.parentPG(); parentPG == nil { + tg.createSession() + } else { + // Inherit the process group. + parentPG.incRefWithParent(parentPG) + tg.processGroup = parentPG + } + } + tg.tasks.PushBack(t) + tg.tasksCount++ + tg.liveTasks++ + tg.activeTasks++ + + // Propagate external TaskSet stops to the new task. + t.stopCount = ts.stopCount + + t.mu.Lock() + defer t.mu.Unlock() + + t.cpu = assignCPU(t.allowedCPUMask, ts.Root.tids[t]) + + t.startTime = t.k.RealtimeClock().Now() + + return t, nil +} + +// assignTIDsLocked ensures that new task t is visible in all PID namespaces in +// which it should be visible. +// +// Preconditions: ts.mu must be locked for writing. +func (ts *TaskSet) assignTIDsLocked(t *Task) error { + type allocatedTID struct { + ns *PIDNamespace + tid ThreadID + } + var allocatedTIDs []allocatedTID + for ns := t.tg.pidns; ns != nil; ns = ns.parent { + tid, err := ns.allocateTID() + if err != nil { + // Failure. Remove the tids we already allocated in descendant + // namespaces. + for _, a := range allocatedTIDs { + delete(a.ns.tasks, a.tid) + delete(a.ns.tids, t) + } + return err + } + ns.tasks[tid] = t + ns.tids[t] = tid + allocatedTIDs = append(allocatedTIDs, allocatedTID{ns, tid}) + } + return nil +} + +// allocateTID returns an unused ThreadID from ns. +// +// Preconditions: ns.owner.mu must be locked for writing. +func (ns *PIDNamespace) allocateTID() (ThreadID, error) { + if ns.exiting { + // "In this case, a subsequent fork(2) into this PID namespace will + // fail with the error ENOMEM; it is not possible to create a new + // processes [sic] in a PID namespace whose init process has + // terminated." - pid_namespaces(7) + return 0, syserror.ENOMEM + } + tid := ns.last + for { + // Next. + tid++ + if tid > TasksLimit { + tid = InitTID + 1 + } + + // Is it available? + _, ok := ns.tasks[tid] + if !ok { + ns.last = tid + return tid, nil + } + + // Did we do a full cycle? + if tid == ns.last { + // No tid available. + return 0, syserror.EAGAIN + } + } +} + +// Start starts the task goroutine. Start must be called exactly once for each +// task returned by NewTask. +// +// 'tid' must be the task's TID in the root PID namespace and it's used for +// debugging purposes only (set as parameter to Task.run to make it visible +// in stack dumps). +func (t *Task) Start(tid ThreadID) { + // If the task was restored, it may be "starting" after having already exited. + if t.runState == nil { + return + } + t.goroutineStopped.Add(1) + t.tg.liveGoroutines.Add(1) + t.tg.pidns.owner.liveGoroutines.Add(1) + t.tg.pidns.owner.runningGoroutines.Add(1) + + // Task is now running in system mode. + t.accountTaskGoroutineLeave(TaskGoroutineNonexistent) + + // Use the task's TID in the root PID namespace to make it visible in stack dumps. + go t.run(uintptr(tid)) // S/R-SAFE: synchronizes with saving through stops +} diff --git a/pkg/sentry/kernel/task_stop.go b/pkg/sentry/kernel/task_stop.go new file mode 100644 index 000000000..feaf6cae4 --- /dev/null +++ b/pkg/sentry/kernel/task_stop.go @@ -0,0 +1,226 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +// This file implements task stops, which represent the equivalent of Linux's +// uninterruptible sleep states in a way that is compatible with save/restore. +// Task stops comprise both internal stops (which form part of the task's +// "normal" control flow) and external stops (which do not); see README.md for +// details. +// +// There are multiple interfaces for interacting with stops because there are +// multiple cases to consider: +// +// - A task goroutine can begin a stop on its associated task (e.g. a +// vfork() syscall stopping the calling task until the child task releases its +// MM). In this case, calling Task.interrupt is both unnecessary (the task +// goroutine obviously cannot be blocked in Task.block or executing application +// code) and undesirable (as it may spuriously interrupt a in-progress +// syscall). +// +// Beginning internal stops in this case is implemented by +// Task.beginInternalStop / Task.beginInternalStopLocked. As of this writing, +// there are no instances of this case that begin external stops, except for +// autosave; however, autosave terminates the sentry without ending the +// external stop, so the spurious interrupt is moot. +// +// - An arbitrary goroutine can begin a stop on an unrelated task (e.g. all +// tasks being stopped in preparation for state checkpointing). If the task +// goroutine may be in Task.block or executing application code, it must be +// interrupted by Task.interrupt for it to actually enter the stop; since, +// strictly speaking, we have no way of determining this, we call +// Task.interrupt unconditionally. +// +// Beginning external stops in this case is implemented by +// Task.BeginExternalStop. As of this writing, there are no instances of this +// case that begin internal stops. +// +// - An arbitrary goroutine can end a stop on an unrelated task (e.g. an +// exiting task resuming a sibling task that has been blocked in an execve() +// syscall waiting for other tasks to exit). In this case, Task.endStopCond +// must be notified to kick the task goroutine out of Task.doStop. +// +// Ending internal stops in this case is implemented by +// Task.endInternalStopLocked. Ending external stops in this case is +// implemented by Task.EndExternalStop. +// +// - Hypothetically, a task goroutine can end an internal stop on its +// associated task. As of this writing, there are no instances of this case. +// However, any instances of this case could still use the above functions, +// since notifying Task.endStopCond would be unnecessary but harmless. + +import ( + "fmt" + "sync/atomic" +) + +// A TaskStop is a condition visible to the task control flow graph that +// prevents a task goroutine from running or exiting, i.e. an internal stop. +// +// NOTE: Most TaskStops don't contain any data; they're +// distinguished by their type. The obvious way to implement such a TaskStop +// is: +// +// type groupStop struct{} +// func (groupStop) Killable() bool { return true } +// ... +// t.beginInternalStop(groupStop{}) +// +// However, this doesn't work because the state package can't serialize values, +// only pointers. Furthermore, the correctness of save/restore depends on the +// ability to pass a TaskStop to endInternalStop that will compare equal to the +// TaskStop that was passed to beginInternalStop, even if a save/restore cycle +// occurred between the two. As a result, the current idiom is to always use a +// typecast nil for data-free TaskStops: +// +// type groupStop struct{} +// func (*groupStop) Killable() bool { return true } +// ... +// t.beginInternalStop((*groupStop)(nil)) +// +// This is pretty gross, but the alternatives seem grosser. +type TaskStop interface { + // Killable returns true if Task.Kill should end the stop prematurely. + // Killable is analogous to Linux's TASK_WAKEKILL. + Killable() bool +} + +// beginInternalStop indicates the start of an internal stop that applies to t. +// +// Preconditions: The task must not already be in an internal stop (i.e. t.stop +// == nil). The caller must be running on the task goroutine. +func (t *Task) beginInternalStop(s TaskStop) { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + t.beginInternalStopLocked(s) +} + +// Preconditions: The signal mutex must be locked. All preconditions for +// Task.beginInternalStop also apply. +func (t *Task) beginInternalStopLocked(s TaskStop) { + if t.stop != nil { + panic(fmt.Sprintf("Attempting to enter internal stop %#v when already in internal stop %#v", s, t.stop)) + } + t.Debugf("Entering internal stop %#v", s) + t.stop = s + t.beginStopLocked() +} + +// endInternalStopLocked indicates the end of an internal stop that applies to +// t. endInternalStopLocked does not wait for the task to resume. +// +// The caller is responsible for ensuring that the internal stop they expect +// actually applies to t; this requires holding the signal mutex which protects +// t.stop, which is why there is no endInternalStop that locks the signal mutex +// for you. +// +// Preconditions: The signal mutex must be locked. The task must be in an +// internal stop (i.e. t.stop != nil). +func (t *Task) endInternalStopLocked() { + if t.stop == nil { + panic("Attempting to leave non-existent internal stop") + } + t.Debugf("Leaving internal stop %#v", t.stop) + t.stop = nil + t.endStopLocked() +} + +// BeginExternalStop indicates the start of an external stop that applies to t. +// BeginExternalStop does not wait for t's task goroutine to stop. +func (t *Task) BeginExternalStop() { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + t.beginStopLocked() + t.interrupt() +} + +// EndExternalStop indicates the end of an external stop started by a previous +// call to Task.BeginExternalStop. EndExternalStop does not wait for t's task +// goroutine to resume. +func (t *Task) EndExternalStop() { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + t.endStopLocked() +} + +// beginStopLocked increments t.stopCount to indicate that a new internal or +// external stop applies to t. +// +// Preconditions: The signal mutex must be locked. +func (t *Task) beginStopLocked() { + if newval := atomic.AddInt32(&t.stopCount, 1); newval <= 0 { + // Most likely overflow. + panic(fmt.Sprintf("Invalid stopCount: %d", newval)) + } +} + +// endStopLocked decerements t.stopCount to indicate that an existing internal +// or external stop no longer applies to t. +// +// Preconditions: The signal mutex must be locked. +func (t *Task) endStopLocked() { + if newval := atomic.AddInt32(&t.stopCount, -1); newval < 0 { + panic(fmt.Sprintf("Invalid stopCount: %d", newval)) + } else if newval == 0 { + t.endStopCond.Signal() + } +} + +// BeginExternalStop indicates the start of an external stop that applies to +// all current and future tasks in ts. BeginExternalStop does not wait for +// task goroutines to stop. +func (ts *TaskSet) BeginExternalStop() { + ts.mu.Lock() + defer ts.mu.Unlock() + ts.stopCount++ + if ts.stopCount <= 0 { + panic(fmt.Sprintf("Invalid stopCount: %d", ts.stopCount)) + } + if ts.Root == nil { + return + } + for t := range ts.Root.tids { + t.tg.signalHandlers.mu.Lock() + t.beginStopLocked() + t.tg.signalHandlers.mu.Unlock() + t.interrupt() + } +} + +// EndExternalStop indicates the end of an external stop started by a previous +// call to TaskSet.BeginExternalStop. EndExternalStop does not wait for task +// goroutines to resume. +func (ts *TaskSet) EndExternalStop() { + ts.mu.Lock() + defer ts.mu.Unlock() + ts.stopCount-- + if ts.stopCount < 0 { + panic(fmt.Sprintf("Invalid stopCount: %d", ts.stopCount)) + } + if ts.Root == nil { + return + } + for t := range ts.Root.tids { + t.tg.signalHandlers.mu.Lock() + t.endStopLocked() + t.tg.signalHandlers.mu.Unlock() + } +} diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go new file mode 100644 index 000000000..79f4ff60c --- /dev/null +++ b/pkg/sentry/kernel/task_syscall.go @@ -0,0 +1,434 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + "os" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/bits" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// SyscallRestartErrno represents a ERESTART* errno defined in the Linux's kernel +// include/linux/errno.h. These errnos are never returned to userspace +// directly, but are used to communicate the expected behavior of an +// interrupted syscall from the syscall to signal handling. +type SyscallRestartErrno int + +// These numeric values are significant because ptrace syscall exit tracing can +// observe them. +// +// For all of the following errnos, if the syscall is not interrupted by a +// signal delivered to a user handler, the syscall is restarted. +const ( + // ERESTARTSYS is returned by an interrupted syscall to indicate that it + // should be converted to EINTR if interrupted by a signal delivered to a + // user handler without SA_RESTART set, and restarted otherwise. + ERESTARTSYS = SyscallRestartErrno(512) + + // ERESTARTNOINTR is returned by an interrupted syscall to indicate that it + // should always be restarted. + ERESTARTNOINTR = SyscallRestartErrno(513) + + // ERESTARTNOHAND is returned by an interrupted syscall to indicate that it + // should be converted to EINTR if interrupted by a signal delivered to a + // user handler, and restarted otherwise. + ERESTARTNOHAND = SyscallRestartErrno(514) + + // ERESTART_RESTARTBLOCK is returned by an interrupted syscall to indicate + // that it should be restarted using a custom function. The interrupted + // syscall must register a custom restart function by calling + // Task.SetRestartSyscallFn. + ERESTART_RESTARTBLOCK = SyscallRestartErrno(516) +) + +// Error implements error.Error. +func (e SyscallRestartErrno) Error() string { + // Descriptions are borrowed from strace. + switch e { + case ERESTARTSYS: + return "to be restarted if SA_RESTART is set" + case ERESTARTNOINTR: + return "to be restarted" + case ERESTARTNOHAND: + return "to be restarted if no handler" + case ERESTART_RESTARTBLOCK: + return "interrupted by signal" + default: + return "(unknown interrupt error)" + } +} + +// SyscallRestartErrnoFromReturn returns the SyscallRestartErrno represented by +// rv, the value in a syscall return register. +func SyscallRestartErrnoFromReturn(rv uintptr) (SyscallRestartErrno, bool) { + switch int(rv) { + case -int(ERESTARTSYS): + return ERESTARTSYS, true + case -int(ERESTARTNOINTR): + return ERESTARTNOINTR, true + case -int(ERESTARTNOHAND): + return ERESTARTNOHAND, true + case -int(ERESTART_RESTARTBLOCK): + return ERESTART_RESTARTBLOCK, true + default: + return 0, false + } +} + +// SyscallRestartBlock represents the restart block for a syscall restartable +// with a custom function. It encapsulates the state required to restart a +// syscall across a S/R. +type SyscallRestartBlock interface { + Restart(t *Task) (uintptr, error) +} + +// SyscallControl is returned by syscalls to control the behavior of +// Task.doSyscallInvoke. +type SyscallControl struct { + // next is the state that the task goroutine should switch to. If next is + // nil, the task goroutine should continue to syscall exit as usual. + next taskRunState + + // If ignoreReturn is true, Task.doSyscallInvoke should not store any value + // in the task's syscall return value register. + ignoreReturn bool +} + +var ( + // CtrlDoExit is returned by the implementations of the exit and exit_group + // syscalls to enter the task exit path directly, skipping syscall exit + // tracing. + CtrlDoExit = &SyscallControl{next: (*runExit)(nil), ignoreReturn: true} + + // ctrlStopAndReinvokeSyscall is returned by syscalls using the external + // feature before syscall execution. This causes Task.doSyscallInvoke + // to return runSyscallReinvoke, allowing Task.run to check for stops + // before immediately re-invoking the syscall (skipping the re-checking + // of seccomp filters and ptrace which would confuse userspace + // tracing). + ctrlStopAndReinvokeSyscall = &SyscallControl{next: (*runSyscallReinvoke)(nil), ignoreReturn: true} + + // ctrlStopBeforeSyscallExit is returned by syscalls that initiate a stop at + // their end. This causes Task.doSyscallInvoke to return runSyscallExit, rather + // than tail-calling it, allowing stops to be checked before syscall exit. + ctrlStopBeforeSyscallExit = &SyscallControl{next: (*runSyscallExit)(nil)} +) + +func (t *Task) invokeExternal() { + t.BeginExternalStop() + go func() { // S/R-SAFE: External control flow. + defer t.EndExternalStop() + t.SyscallTable().External(t.Kernel()) + }() +} + +func (t *Task) executeSyscall(sysno uintptr, args arch.SyscallArguments) (rval uintptr, ctrl *SyscallControl, err error) { + s := t.SyscallTable() + + fe := s.FeatureEnable.Word(sysno) + + var straceContext interface{} + if bits.IsAnyOn32(fe, StraceEnableBits) { + straceContext = s.Stracer.SyscallEnter(t, sysno, args, fe) + } + + if bits.IsOn32(fe, ExternalBeforeEnable) && (s.ExternalFilterBefore == nil || s.ExternalFilterBefore(t, sysno, args)) { + t.invokeExternal() + // Ensure we check for stops, then invoke the syscall again. + ctrl = ctrlStopAndReinvokeSyscall + } else { + fn := s.Lookup(sysno) + if fn != nil { + // Call our syscall implementation. + rval, ctrl, err = fn(t, args) + } else { + // Use the missing function if not found. + rval, err = t.SyscallTable().Missing(t, sysno, args) + } + } + + if bits.IsOn32(fe, ExternalAfterEnable) && (s.ExternalFilterAfter == nil || s.ExternalFilterAfter(t, sysno, args)) { + t.invokeExternal() + // Don't reinvoke the syscall. + } + + if bits.IsAnyOn32(fe, StraceEnableBits) { + s.Stracer.SyscallExit(straceContext, t, sysno, rval, err) + } + + return +} + +// doSyscall is the entry point for an invocation of a system call specified by +// the current state of t's registers. +// +// The syscall path is very hot; avoid defer. +func (t *Task) doSyscall() taskRunState { + sysno := t.Arch().SyscallNo() + args := t.Arch().SyscallArgs() + + // Tracers expect to see this between when the task traps into the kernel + // to perform a syscall and when the syscall is actually invoked. + // This useless-looking temporary is needed because Go. + tmp := uintptr(syscall.ENOSYS) + t.Arch().SetReturn(-tmp) + + // Check seccomp filters. The nil check is for performance (as seccomp use + // is rare), not needed for correctness. + if t.syscallFilters != nil { + switch r := t.checkSeccompSyscall(int32(sysno), args, usermem.Addr(t.Arch().IP())); r { + case seccompResultDeny: + t.Debugf("Syscall %d: denied by seccomp", sysno) + return (*runSyscallExit)(nil) + case seccompResultAllow: + // ok + case seccompResultKill: + t.Debugf("Syscall %d: killed by seccomp", sysno) + t.PrepareExit(ExitStatus{Signo: int(linux.SIGSYS)}) + return (*runExit)(nil) + case seccompResultTrace: + t.Debugf("Syscall %d: stopping for PTRACE_EVENT_SECCOMP", sysno) + return (*runSyscallAfterPtraceEventSeccomp)(nil) + default: + panic(fmt.Sprintf("Unknown seccomp result %d", r)) + } + } + + return t.doSyscallEnter(sysno, args) +} + +type runSyscallAfterPtraceEventSeccomp struct{} + +func (*runSyscallAfterPtraceEventSeccomp) execute(t *Task) taskRunState { + if t.killed() { + // "[S]yscall-exit-stop is not generated prior to death by SIGKILL." - + // ptrace(2) + return (*runInterrupt)(nil) + } + sysno := t.Arch().SyscallNo() + // "The tracer can skip the system call by changing the syscall number to + // -1." - Documentation/prctl/seccomp_filter.txt + if sysno == ^uintptr(0) { + return (*runSyscallExit)(nil).execute(t) + } + args := t.Arch().SyscallArgs() + return t.doSyscallEnter(sysno, args) +} + +func (t *Task) doSyscallEnter(sysno uintptr, args arch.SyscallArguments) taskRunState { + if next, ok := t.ptraceSyscallEnter(); ok { + return next + } + return t.doSyscallInvoke(sysno, args) +} + +type runSyscallAfterSyscallEnterStop struct{} + +func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState { + if sig := linux.Signal(t.ptraceCode); sig.IsValid() { + t.tg.signalHandlers.mu.Lock() + t.sendSignalLocked(sigPriv(sig), false /* group */) + t.tg.signalHandlers.mu.Unlock() + } + if t.killed() { + return (*runInterrupt)(nil) + } + sysno := t.Arch().SyscallNo() + if sysno == ^uintptr(0) { + return (*runSyscallExit)(nil) + } + args := t.Arch().SyscallArgs() + return t.doSyscallInvoke(sysno, args) +} + +type runSyscallAfterSysemuStop struct{} + +func (*runSyscallAfterSysemuStop) execute(t *Task) taskRunState { + if sig := linux.Signal(t.ptraceCode); sig.IsValid() { + t.tg.signalHandlers.mu.Lock() + t.sendSignalLocked(sigPriv(sig), false /* group */) + t.tg.signalHandlers.mu.Unlock() + } + if t.killed() { + return (*runInterrupt)(nil) + } + return (*runSyscallExit)(nil).execute(t) +} + +func (t *Task) doSyscallInvoke(sysno uintptr, args arch.SyscallArguments) taskRunState { + rval, ctrl, err := t.executeSyscall(sysno, args) + + if ctrl != nil { + if !ctrl.ignoreReturn { + t.Arch().SetReturn(rval) + } + if ctrl.next != nil { + return ctrl.next + } + } else if err != nil { + t.Arch().SetReturn(uintptr(-t.ExtractErrno(err, int(sysno)))) + t.haveSyscallReturn = true + } else { + t.Arch().SetReturn(rval) + } + + return (*runSyscallExit)(nil).execute(t) +} + +type runSyscallReinvoke struct{} + +func (*runSyscallReinvoke) execute(t *Task) taskRunState { + if t.killed() { + // It's possible that since the last execution, the task has + // been forcible killed. Invoking the system call here could + // result in an infinite loop if it is again preempted by an + // external stop and reinvoked. + return (*runInterrupt)(nil) + } + + sysno := t.Arch().SyscallNo() + args := t.Arch().SyscallArgs() + return t.doSyscallInvoke(sysno, args) +} + +type runSyscallExit struct{} + +func (*runSyscallExit) execute(t *Task) taskRunState { + t.ptraceSyscallExit() + return (*runApp)(nil) +} + +// doVsyscall is the entry point for a vsyscall invocation of syscall sysno, as +// indicated by an execution fault at address addr. doVsyscall returns the +// task's next run state. +func (t *Task) doVsyscall(addr usermem.Addr, sysno uintptr) taskRunState { + // Grab the caller up front, to make sure there's a sensible stack. + caller := t.Arch().Native(uintptr(0)) + if _, err := t.CopyIn(usermem.Addr(t.Arch().Stack()), caller); err != nil { + t.Debugf("vsyscall %d: error reading return address from stack: %v", sysno, err) + t.forceSignal(linux.SIGSEGV, false /* unconditional */) + t.SendSignal(sigPriv(linux.SIGSEGV)) + return (*runApp)(nil) + } + + // For _vsyscalls_, there is no need to translate System V calling convention + // to syscall ABI because they both use RDI, RSI, and RDX for the first three + // arguments and none of the vsyscalls uses more than two arguments. + args := t.Arch().SyscallArgs() + if t.syscallFilters != nil { + switch r := t.checkSeccompSyscall(int32(sysno), args, addr); r { + case seccompResultDeny: + t.Debugf("vsyscall %d, caller %x: denied by seccomp", sysno, t.Arch().Value(caller)) + return (*runApp)(nil) + case seccompResultAllow: + // ok + case seccompResultTrace: + t.Debugf("vsyscall %d, caller %x: stopping for PTRACE_EVENT_SECCOMP", sysno, t.Arch().Value(caller)) + return &runVsyscallAfterPtraceEventSeccomp{addr, sysno, caller} + default: + panic(fmt.Sprintf("Unknown seccomp result %d", r)) + } + } + + return t.doVsyscallInvoke(sysno, args, caller) +} + +type runVsyscallAfterPtraceEventSeccomp struct { + addr usermem.Addr + sysno uintptr + caller interface{} +} + +func (r *runVsyscallAfterPtraceEventSeccomp) execute(t *Task) taskRunState { + if t.killed() { + return (*runInterrupt)(nil) + } + sysno := t.Arch().SyscallNo() + // "... the syscall may not be changed to another system call using the + // orig_rax register. It may only be changed to -1 order [sic] to skip the + // currently emulated call. ... The tracer MUST NOT modify rip or rsp." - + // Documentation/prctl/seccomp_filter.txt. On Linux, changing orig_ax or ip + // causes do_exit(SIGSYS), and changing sp is ignored. + if (sysno != ^uintptr(0) && sysno != r.sysno) || usermem.Addr(t.Arch().IP()) != r.addr { + t.PrepareExit(ExitStatus{Signo: int(linux.SIGSYS)}) + return (*runExit)(nil) + } + if sysno == ^uintptr(0) { + return (*runApp)(nil) + } + return t.doVsyscallInvoke(sysno, t.Arch().SyscallArgs(), r.caller) +} + +func (t *Task) doVsyscallInvoke(sysno uintptr, args arch.SyscallArguments, caller interface{}) taskRunState { + rval, ctrl, err := t.executeSyscall(sysno, args) + if ctrl != nil { + t.Debugf("vsyscall %d, caller %x: syscall control: %v", sysno, t.Arch().Value(caller), ctrl) + // Set the return value. The stack has already been adjusted. + t.Arch().SetReturn(0) + } else if err == nil { + t.Debugf("vsyscall %d, caller %x: successfully emulated syscall", sysno, t.Arch().Value(caller)) + // Set the return value. The stack has already been adjusted. + t.Arch().SetReturn(uintptr(rval)) + } else { + t.Debugf("vsyscall %d, caller %x: emulated syscall returned error: %v", sysno, t.Arch().Value(caller), err) + if err == syserror.EFAULT { + t.forceSignal(linux.SIGSEGV, false /* unconditional */) + t.SendSignal(sigPriv(linux.SIGSEGV)) + // A return is not emulated in this case. + return (*runApp)(nil) + } + t.Arch().SetReturn(uintptr(-t.ExtractErrno(err, int(sysno)))) + } + t.Arch().SetIP(t.Arch().Value(caller)) + t.Arch().SetStack(t.Arch().Stack() + uintptr(t.Arch().Width())) + return (*runApp)(nil) +} + +// ExtractErrno extracts an integer error number from the error. +// The syscall number is purely for context in the error case. Use -1 if +// syscall number is unknown. +func (t *Task) ExtractErrno(err error, sysno int) int { + switch err := err.(type) { + case nil: + return 0 + case syscall.Errno: + return int(err) + case SyscallRestartErrno: + return int(err) + case *memmap.BusError: + // Bus errors may generate SIGBUS, but for syscalls they still + // return EFAULT. See case in task_run.go where the fault is + // handled (and the SIGBUS is delivered). + return int(syscall.EFAULT) + case *os.PathError: + return t.ExtractErrno(err.Err, sysno) + case *os.LinkError: + return t.ExtractErrno(err.Err, sysno) + case *os.SyscallError: + return t.ExtractErrno(err.Err, sysno) + default: + if errno, ok := syserror.TranslateError(err); ok { + return int(errno) + } + } + panic(fmt.Sprintf("Unknown syscall %d error: %v", sysno, err)) +} diff --git a/pkg/sentry/kernel/task_test.go b/pkg/sentry/kernel/task_test.go new file mode 100644 index 000000000..82ef858a1 --- /dev/null +++ b/pkg/sentry/kernel/task_test.go @@ -0,0 +1,69 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "testing" + + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" +) + +func TestTaskCPU(t *testing.T) { + for _, test := range []struct { + mask sched.CPUSet + tid ThreadID + cpu int32 + }{ + { + mask: []byte{0xff}, + tid: 1, + cpu: 0, + }, + { + mask: []byte{0xff}, + tid: 10, + cpu: 1, + }, + { + // more than 8 cpus. + mask: []byte{0xff, 0xff}, + tid: 10, + cpu: 9, + }, + { + // missing the first cpu. + mask: []byte{0xfe}, + tid: 1, + cpu: 1, + }, + { + mask: []byte{0xfe}, + tid: 10, + cpu: 3, + }, + { + // missing the fifth cpu. + mask: []byte{0xef}, + tid: 10, + cpu: 2, + }, + } { + assigned := assignCPU(test.mask, test.tid) + if test.cpu != assigned { + t.Errorf("assignCPU(%v, %v) got %v, want %v", test.mask, test.tid, assigned, test.cpu) + } + } + +} diff --git a/pkg/sentry/kernel/task_usermem.go b/pkg/sentry/kernel/task_usermem.go new file mode 100644 index 000000000..7a62ab674 --- /dev/null +++ b/pkg/sentry/kernel/task_usermem.go @@ -0,0 +1,298 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "math" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// _MAX_RW_COUNT is the maximum size in bytes of a single read or write. +// Reads and writes that exceed this size may be silently truncated. +// (Linux: include/linux/fs.h:MAX_RW_COUNT) +var _MAX_RW_COUNT = int(usermem.Addr(math.MaxInt32).RoundDown()) + +// Activate ensures that the task has an active address space. +func (t *Task) Activate() { + if mm := t.MemoryManager(); mm != nil { + if err := mm.Activate(); err != nil { + panic("unable to activate mm: " + err.Error()) + } + } +} + +// Deactivate relinquishes the task's active address space. +func (t *Task) Deactivate() { + if mm := t.MemoryManager(); mm != nil { + if err := mm.Deactivate(); err != nil { + panic("unable to deactivate mm: " + err.Error()) + } + } +} + +// CopyIn copies a fixed-size value or slice of fixed-size values in from the +// task's memory. The copy will fail with syscall.EFAULT if it traverses user +// memory that is unmapped or not readable by the user. +// +// This Task's AddressSpace must be active. +func (t *Task) CopyIn(addr usermem.Addr, dst interface{}) (int, error) { + return usermem.CopyObjectIn(t, t.MemoryManager(), addr, dst, usermem.IOOpts{ + AddressSpaceActive: true, + }) +} + +// CopyInBytes is a fast version of CopyIn if the caller can serialize the +// data without reflection and pass in a byte slice. +// +// This Task's AddressSpace must be active. +func (t *Task) CopyInBytes(addr usermem.Addr, dst []byte) (int, error) { + return t.MemoryManager().CopyIn(t, addr, dst, usermem.IOOpts{ + AddressSpaceActive: true, + }) +} + +// CopyOut copies a fixed-size value or slice of fixed-size values out to the +// task's memory. The copy will fail with syscall.EFAULT if it traverses user +// memory that is unmapped or not writeable by the user. +// +// This Task's AddressSpace must be active. +func (t *Task) CopyOut(addr usermem.Addr, src interface{}) (int, error) { + return usermem.CopyObjectOut(t, t.MemoryManager(), addr, src, usermem.IOOpts{ + AddressSpaceActive: true, + }) +} + +// CopyOutBytes is a fast version of CopyOut if the caller can serialize the +// data without reflection and pass in a byte slice. +// +// This Task's AddressSpace must be active. +func (t *Task) CopyOutBytes(addr usermem.Addr, src []byte) (int, error) { + return t.MemoryManager().CopyOut(t, addr, src, usermem.IOOpts{ + AddressSpaceActive: true, + }) +} + +// CopyInString copies a NUL-terminated string of length at most maxlen in from +// the task's memory. The copy will fail with syscall.EFAULT if it traverses +// user memory that is unmapped or not readable by the user. +// +// This Task's AddressSpace must be active. +func (t *Task) CopyInString(addr usermem.Addr, maxlen int) (string, error) { + return usermem.CopyStringIn(t, t.MemoryManager(), addr, maxlen, usermem.IOOpts{ + AddressSpaceActive: true, + }) +} + +// CopyInVector copies a NULL-terminated vector of strings from the task's +// memory. The copy will fail with syscall.EFAULT if it traverses +// user memory that is unmapped or not readable by the user. +// +// maxElemSize is the maximum size of each individual element. +// +// maxTotalSize is the maximum total length of all elements plus the total +// number of elements. For example, the following strings correspond to +// the following set of sizes: +// +// { "a", "b", "c" } => 6 (3 for lengths, 3 for elements) +// { "abc" } => 4 (3 for length, 1 for elements) +// +// This Task's AddressSpace must be active. +func (t *Task) CopyInVector(addr usermem.Addr, maxElemSize, maxTotalSize int) ([]string, error) { + var v []string + for { + argAddr := t.Arch().Native(0) + if _, err := t.CopyIn(addr, argAddr); err != nil { + return v, err + } + if t.Arch().Value(argAddr) == 0 { + break + } + // Each string has a zero terminating byte counted, so copying out a string + // requires at least one byte of space. Also, see the calculation below. + if maxTotalSize <= 0 { + return nil, syserror.ENOMEM + } + thisMax := maxElemSize + if maxTotalSize < thisMax { + thisMax = maxTotalSize + } + arg, err := t.CopyInString(usermem.Addr(t.Arch().Value(argAddr)), thisMax) + if err != nil { + return v, err + } + v = append(v, arg) + addr += usermem.Addr(t.Arch().Width()) + maxTotalSize -= len(arg) + 1 + } + return v, nil +} + +// CopyOutIovecs converts src to an array of struct iovecs and copies it to the +// memory mapped at addr. +// +// Preconditions: As for usermem.IO.CopyOut. The caller must be running on the +// task goroutine. t's AddressSpace must be active. +func (t *Task) CopyOutIovecs(addr usermem.Addr, src usermem.AddrRangeSeq) error { + switch t.Arch().Width() { + case 8: + const itemLen = 16 + if _, ok := addr.AddLength(uint64(src.NumRanges()) * itemLen); !ok { + return syserror.EFAULT + } + + b := t.CopyScratchBuffer(itemLen) + for ; !src.IsEmpty(); src = src.Tail() { + ar := src.Head() + usermem.ByteOrder.PutUint64(b[0:8], uint64(ar.Start)) + usermem.ByteOrder.PutUint64(b[8:16], uint64(ar.Length())) + if _, err := t.CopyOutBytes(addr, b); err != nil { + return err + } + addr += itemLen + } + + default: + return syserror.ENOSYS + } + + return nil +} + +// CopyInIovecs copies an array of numIovecs struct iovecs from the memory +// mapped at addr, converts them to usermem.AddrRanges, and returns them as a +// usermem.AddrRangeSeq. +// +// CopyInIovecs shares the following properties with Linux's +// lib/iov_iter.c:import_iovec() => fs/read_write.c:rw_copy_check_uvector(): +// +// - If the length of any AddrRange would exceed the range of an ssize_t, +// CopyInIovecs returns EINVAL. +// +// - If the length of any AddrRange would cause its end to overflow, +// CopyInIovecs returns EFAULT. +// +// - The combined length of all AddrRanges is limited to _MAX_RW_COUNT. If the +// combined length of all AddrRanges would otherwise exceed this amount, ranges +// beyond _MAX_RW_COUNT are silently truncated. +// +// Preconditions: As for usermem.IO.CopyIn. The caller must be running on the +// task goroutine. t's AddressSpace must be active. +func (t *Task) CopyInIovecs(addr usermem.Addr, numIovecs int) (usermem.AddrRangeSeq, error) { + if numIovecs == 0 { + return usermem.AddrRangeSeq{}, nil + } + + var dst []usermem.AddrRange + if numIovecs > 1 { + dst = make([]usermem.AddrRange, 0, numIovecs) + } + + switch t.Arch().Width() { + case 8: + const itemLen = 16 + if _, ok := addr.AddLength(uint64(numIovecs) * itemLen); !ok { + return usermem.AddrRangeSeq{}, syserror.EFAULT + } + + b := t.CopyScratchBuffer(itemLen) + for i := 0; i < numIovecs; i++ { + if _, err := t.CopyInBytes(addr, b); err != nil { + return usermem.AddrRangeSeq{}, err + } + + base := usermem.Addr(usermem.ByteOrder.Uint64(b[0:8])) + length := usermem.ByteOrder.Uint64(b[8:16]) + if length > math.MaxInt64 { + return usermem.AddrRangeSeq{}, syserror.EINVAL + } + ar, ok := base.ToRange(length) + if !ok { + return usermem.AddrRangeSeq{}, syserror.EFAULT + } + + if numIovecs == 1 { + // Special case to avoid allocating dst. + return usermem.AddrRangeSeqOf(ar).TakeFirst(_MAX_RW_COUNT), nil + } + dst = append(dst, ar) + + addr += itemLen + } + + default: + return usermem.AddrRangeSeq{}, syserror.ENOSYS + } + + // Truncate to _MAX_RW_COUNT. + var total uint64 + for i := range dst { + dstlen := uint64(dst[i].Length()) + if rem := uint64(_MAX_RW_COUNT) - total; rem < dstlen { + dst[i].End -= usermem.Addr(dstlen - rem) + dstlen = rem + } + total += dstlen + } + + return usermem.AddrRangeSeqFromSlice(dst), nil +} + +// SingleIOSequence returns a usermem.IOSequence representing [addr, +// addr+length) in t's address space. If length exceeds _MAX_RW_COUNT, it is +// silently truncated. +// +// SingleIOSequence is analogous to Linux's +// lib/iov_iter.c:import_single_range(). (Note that the non-vectorized read and +// write syscalls in Linux do not use import_single_range(), but are still +// truncated to _MAX_RW_COUNT by fs/read_write.c:rw_verify_area().) +func (t *Task) SingleIOSequence(addr usermem.Addr, length int, opts usermem.IOOpts) (usermem.IOSequence, error) { + if length > _MAX_RW_COUNT { + length = _MAX_RW_COUNT + } + ar, ok := addr.ToRange(uint64(length)) + if !ok { + return usermem.IOSequence{}, syserror.EFAULT + } + return usermem.IOSequence{ + IO: t.MemoryManager(), + Addrs: usermem.AddrRangeSeqOf(ar), + Opts: opts, + }, nil +} + +// IovecsIOSequence returns a usermem.IOSequence representing the array of +// iovcnt struct iovecs at addr in t's address space. opts applies to the +// returned IOSequence, not the reading of the struct iovec array. +// +// IovecsIOSequence is analogous to Linux's lib/iov_iter.c:import_iovec(). +// +// Preconditions: As for Task.CopyInIovecs. +func (t *Task) IovecsIOSequence(addr usermem.Addr, iovcnt int, opts usermem.IOOpts) (usermem.IOSequence, error) { + if iovcnt < 0 || iovcnt > linux.UIO_MAXIOV { + return usermem.IOSequence{}, syserror.EINVAL + } + ars, err := t.CopyInIovecs(addr, iovcnt) + if err != nil { + return usermem.IOSequence{}, err + } + return usermem.IOSequence{ + IO: t.MemoryManager(), + Addrs: ars, + Opts: opts, + }, nil +} diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go new file mode 100644 index 000000000..8fffd3446 --- /dev/null +++ b/pkg/sentry/kernel/thread_group.go @@ -0,0 +1,269 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "sync" + "sync/atomic" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/limits" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" +) + +// A ThreadGroup is a logical grouping of tasks that has widespread +// significance to other kernel features (e.g. signal handling). ("Thread +// groups" are usually called "processes" in userspace documentation.) +// +// ThreadGroup is a superset of Linux's struct signal_struct. +type ThreadGroup struct { + threadGroupNode + + // signalHandlers is the set of signal handlers used by every task in this + // thread group. (signalHandlers may also be shared with other thread + // groups.) + // + // signalHandlers.mu (hereafter "the signal mutex") protects state related + // to signal handling, as well as state that usually needs to be atomic + // with signal handling, for all ThreadGroups and Tasks using + // signalHandlers. (This is analogous to Linux's use of struct + // sighand_struct::siglock.) + // + // The signalHandlers pointer can only be mutated during an execve + // (Task.finishExec). Consequently, when it's possible for a task in the + // thread group to be completing an execve, signalHandlers is protected by + // the owning TaskSet.mu. Otherwise, it is possible to read the + // signalHandlers pointer without synchronization. In particular, + // completing an execve requires that all other tasks in the thread group + // have exited, so task goroutines do not need the owning TaskSet.mu to + // read the signalHandlers pointer of their thread groups. + signalHandlers *SignalHandlers + + // pendingSignals is the set of pending signals that may be handled by any + // task in this thread group. + // + // pendingSignals is protected by the signal mutex. + pendingSignals pendingSignals + + // lastTimerSignalTask records the last task we deliver a process timer signal to. + // Please see SendTimerSignal for more details. + // + // lastTimerSignalTask is protected by the signal mutex. + lastTimerSignalTask *Task + + // groupStopPhase indicates the state of a group stop in progress on the + // thread group, if any. + // + // groupStopPhase is protected by the signal mutex. + groupStopPhase groupStopPhase + + // groupStopSignal is the signal that caused a group stop to be initiated. + // groupStopSignal is only meaningful if groupStopPhase is + // groupStopInitiated or groupStopComplete. + // + // groupStopSignal is protected by the signal mutex. + groupStopSignal linux.Signal + + // groupStopCount is the number of non-exited tasks in the thread group + // that have acknowledged an initiated group stop. groupStopCount is only + // meaningful if groupStopPhase is groupStopInitiated. + // + // groupStopCount is protected by the signal mutex. + groupStopCount int + + // If groupStopWaitable is true, the thread group is indicating a waitable + // group stop event (as defined by EventChildGroupStop). + // + // Linux represents the analogous state as SIGNAL_STOP_STOPPED being set + // and group_exit_code being non-zero. + // + // groupStopWaitable is protected by the signal mutex. + groupStopWaitable bool + + // If groupContNotify is true, then a SIGCONT has recently ended a group + // stop on this thread group, and the first task to observe it should + // notify its parent. + // + // groupContNotify is protected by the signal mutex. + groupContNotify bool + + // If groupContNotify is true, groupContInterrupted is true iff SIGCONT + // ended a group stop in phase groupStopInitiated. If groupContNotify is + // false, groupContInterrupted is meaningless. + // + // Analogues in Linux: + // + // - groupContNotify && groupContInterrupted is represented by + // SIGNAL_CLD_STOPPED. + // + // - groupContNotify && !groupContInterrupted is represented by + // SIGNAL_CLD_CONTINUED. + // + // - !groupContNotify is represented by neither flag being set. + // + // groupContInterrupted is protected by the signal mutex. + groupContInterrupted bool + + // If groupContWaitable is true, the thread group is indicating a waitable + // continue event (as defined by EventGroupContinue). + // + // groupContWaitable is analogous to Linux's SIGNAL_STOP_CONTINUED. + // + // groupContWaitable is protected by the signal mutex. + groupContWaitable bool + + // exiting is true if all tasks in the ThreadGroup should exit. exiting is + // analogous to Linux's SIGNAL_GROUP_EXIT. + // + // exiting is protected by the signal mutex. exiting can only transition + // from false to true. + exiting bool + + // exitStatus is the thread group's exit status. + // + // While exiting is false, exitStatus is protected by the signal mutex. + // When exiting becomes true, exitStatus becomes immutable. + exitStatus ExitStatus + + // terminationSignal is the signal that this thread group's leader will + // send to its parent when it exits. + // + // terminationSignal is protected by the TaskSet mutex. + terminationSignal linux.Signal + + // liveGoroutines is the number of non-exited task goroutines in the thread + // group. + // + // liveGoroutines is not saved; it is reset as task goroutines are + // restarted by Task.Start. + liveGoroutines sync.WaitGroup `state:"nosave"` + + // tm contains process timers. TimerManager fields are immutable. + tm TimerManager + + // exitedCPUStats is the CPU usage for all exited tasks in the thread + // group. exitedCPUStats is protected by the TaskSet mutex. + exitedCPUStats usage.CPUStats + + // childCPUStats is the CPU usage of all joined descendants of this thread + // group. childCPUStats is protected by the TaskSet mutex. + childCPUStats usage.CPUStats + + // ioUsage is the I/O usage for all exited tasks in the thread group. + // The ioUsage pointer is immutable. + ioUsage *usage.IO + + // maxRSS is the historical maximum resident set size of the thread group, updated when: + // + // - A task in the thread group exits, since after all tasks have + // exited the MemoryManager is no longer reachable. + // + // - The thread group completes an execve, since this changes + // MemoryManagers. + // + // maxRSS is protected by the TaskSet mutex. + maxRSS uint64 + + // childMaxRSS is the maximum resident set size in bytes of all joined + // descendants of this thread group. + // + // childMaxRSS is protected by the TaskSet mutex. + childMaxRSS uint64 + + // Resource limits for this ThreadGroup. The limits pointer is immutable. + limits *limits.LimitSet + + // processGroup is the processGroup for this thread group. + // + // processGroup is protected by the TaskSet mutex. + processGroup *ProcessGroup + + // execed indicates an exec has occurred since creation. This will be + // set by finishExec, and new TheadGroups will have this field cleared. + // When execed is set, the processGroup may no longer be changed. + // + // execed is protected by the TaskSet mutex. + execed bool + + // rscr is the thread group's RSEQ critical region. + rscr atomic.Value `state:".(*RSEQCriticalRegion)"` +} + +// NewThreadGroup returns a new, empty thread group in PID namespace ns. The +// thread group leader will send its parent terminationSignal when it exits. +// The new thread group isn't visible to the system until a task has been +// created inside of it by a successful call to TaskSet.NewTask. +func NewThreadGroup(ns *PIDNamespace, sh *SignalHandlers, terminationSignal linux.Signal, limits *limits.LimitSet, monotonicClock *timekeeperClock) *ThreadGroup { + tg := &ThreadGroup{ + threadGroupNode: threadGroupNode{ + pidns: ns, + }, + signalHandlers: sh, + terminationSignal: terminationSignal, + ioUsage: &usage.IO{}, + limits: limits, + } + tg.tm = newTimerManager(tg, monotonicClock) + tg.rscr.Store(&RSEQCriticalRegion{}) + return tg +} + +// saveRscr is invopked by stateify. +func (tg *ThreadGroup) saveRscr() *RSEQCriticalRegion { + return tg.rscr.Load().(*RSEQCriticalRegion) +} + +// loadRscr is invoked by stateify. +func (tg *ThreadGroup) loadRscr(rscr *RSEQCriticalRegion) { + tg.rscr.Store(rscr) +} + +// SignalHandlers returns the signal handlers used by tg. +// +// Preconditions: The caller must provide the synchronization required to read +// tg.signalHandlers, as described in the field's comment. +func (tg *ThreadGroup) SignalHandlers() *SignalHandlers { + return tg.signalHandlers +} + +// Timer returns tg's timers. +func (tg *ThreadGroup) Timer() *TimerManager { + return &tg.tm +} + +// Limits returns tg's limits. +func (tg *ThreadGroup) Limits() *limits.LimitSet { + return tg.limits +} + +// release releases the thread group's resources. +func (tg *ThreadGroup) release() { + // This must be done without holding the TaskSet mutex since thread group + // timers call SendSignal with Timer.mu locked. + tg.tm.destroy() +} + +// forEachChildThreadGroupLocked indicates over all child ThreadGroups. +// +// Precondition: TaskSet.mu must be held. +func (tg *ThreadGroup) forEachChildThreadGroupLocked(fn func(*ThreadGroup)) { + for t := tg.tasks.Front(); t != nil; t = t.Next() { + for child := range t.children { + if child == child.tg.leader { + fn(child.tg) + } + } + } +} diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go new file mode 100644 index 000000000..440da9dad --- /dev/null +++ b/pkg/sentry/kernel/threads.go @@ -0,0 +1,443 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + "sync" + + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// TasksLimit is the maximum number of threads for untrusted application. +// Linux doesn't really limit this directly, rather it is limited by total +// memory size, stacks allocated and a global maximum. There's no real reason +// for us to limit it either, (esp. since threads are backed by go routines), +// and we would expect to hit resource limits long before hitting this number. +// However, for correctness, we still check that the user doesn't exceed this +// number. +// +// Note that because of the way futexes are implemented, there *are* in fact +// serious restrictions on valid thread IDs. They are limited to 2^30 - 1 +// (kernel/fork.c:MAX_THREADS). +const TasksLimit = (1 << 16) + +// ThreadID is a generic thread identifier. +type ThreadID int32 + +// String returns a decimal representation of the ThreadID. +func (tid ThreadID) String() string { + return fmt.Sprintf("%d", tid) +} + +// InitTID is the TID given to the first task added to each PID namespace. The +// thread group led by InitTID is called the namespace's init process. The +// death of a PID namespace's init process causes all tasks visible in that +// namespace to be killed. +const InitTID ThreadID = 1 + +// A TaskSet comprises all tasks in a system. +type TaskSet struct { + // mu protects all relationships betweens tasks and thread groups in the + // TaskSet. (mu is approximately equivalent to Linux's tasklist_lock.) + mu sync.RWMutex `state:"nosave"` + + // Root is the root PID namespace, in which all tasks in the TaskSet are + // visible. The Root pointer is immutable. + Root *PIDNamespace + + // sessions is the set of all sessions. + sessions sessionList + + // stopCount is the number of active external stops applicable to all tasks + // in the TaskSet (calls to TaskSet.BeginExternalStop that have not been + // paired with a call to TaskSet.EndExternalStop). stopCount is protected + // by mu. + // + // stopCount is not saved for the same reason as Task.stopCount; it is + // always reset to zero after restore. + stopCount int32 `state:"nosave"` + + // liveGoroutines is the number of non-exited task goroutines in the + // TaskSet. + // + // liveGoroutines is not saved; it is reset as task goroutines are + // restarted by Task.Start. + liveGoroutines sync.WaitGroup `state:"nosave"` + + // runningGoroutines is the number of running task goroutines in the + // TaskSet. + // + // runningGoroutines is not saved; its counter value is required to be zero + // at time of save (but note that this is not necessarily the same thing as + // sync.WaitGroup's zero value). + runningGoroutines sync.WaitGroup `state:"nosave"` +} + +// newTaskSet returns a new, empty TaskSet. +func newTaskSet() *TaskSet { + ts := &TaskSet{} + ts.Root = newPIDNamespace(ts, nil /* parent */, auth.NewRootUserNamespace()) + return ts +} + +// forEachThreadGroupLocked applies f to each thread group in ts. +// +// Preconditions: ts.mu must be locked (for reading or writing). +func (ts *TaskSet) forEachThreadGroupLocked(f func(tg *ThreadGroup)) { + for t := range ts.Root.tids { + if t == t.tg.leader { + f(t.tg) + } + } +} + +// A PIDNamespace represents a PID namespace, a bimap between thread IDs and +// tasks. See the pid_namespaces(7) man page for further details. +// +// N.B. A task is said to be visible in a PID namespace if the PID namespace +// contains a thread ID that maps to that task. +type PIDNamespace struct { + // owner is the TaskSet that this PID namespace belongs to. The owner + // pointer is immutable. + owner *TaskSet + + // parent is the PID namespace of the process that created this one. If + // this is the root PID namespace, parent is nil. The parent pointer is + // immutable. + // + // Invariant: All tasks that are visible in this namespace are also visible + // in all ancestor namespaces. + parent *PIDNamespace + + // userns is the user namespace with which this PID namespace is + // associated. Privileged operations on this PID namespace must have + // appropriate capabilities in userns. The userns pointer is immutable. + userns *auth.UserNamespace + + // The following fields are protected by owner.mu. + + // last is the last ThreadID to be allocated in this namespace. + last ThreadID + + // tasks is a mapping from ThreadIDs in this namespace to tasks visible in + // the namespace. + tasks map[ThreadID]*Task + + // tids is a mapping from tasks visible in this namespace to their + // identifiers in this namespace. + tids map[*Task]ThreadID + + // sessions is a mapping from SessionIDs in this namespace to sessions + // visible in the namespace. + sessions map[SessionID]*Session + + // sids is a mapping from sessions visible in this namespace to their + // identifiers in this namespace. + sids map[*Session]SessionID + + // processGroups is a mapping from ProcessGroupIDs in this namespace to + // process groups visible in the namespace. + processGroups map[ProcessGroupID]*ProcessGroup + + // pgids is a mapping from process groups visible in this namespace to + // their identifiers in this namespace. + pgids map[*ProcessGroup]ProcessGroupID + + // exiting indicates that the namespace's init process is exiting or has + // exited. + exiting bool +} + +func newPIDNamespace(ts *TaskSet, parent *PIDNamespace, userns *auth.UserNamespace) *PIDNamespace { + return &PIDNamespace{ + owner: ts, + parent: parent, + userns: userns, + tasks: make(map[ThreadID]*Task), + tids: make(map[*Task]ThreadID), + sessions: make(map[SessionID]*Session), + sids: make(map[*Session]SessionID), + processGroups: make(map[ProcessGroupID]*ProcessGroup), + pgids: make(map[*ProcessGroup]ProcessGroupID), + } +} + +// NewChild returns a new, empty PID namespace that is a child of ns. Authority +// over the new PID namespace is controlled by userns. +func (ns *PIDNamespace) NewChild(userns *auth.UserNamespace) *PIDNamespace { + return newPIDNamespace(ns.owner, ns, userns) +} + +// TaskWithID returns the task with thread ID tid in PID namespace ns. If no +// task has that TID, TaskWithID returns nil. +func (ns *PIDNamespace) TaskWithID(tid ThreadID) *Task { + ns.owner.mu.RLock() + defer ns.owner.mu.RUnlock() + return ns.tasks[tid] +} + +// ThreadGroupWithID returns the thread group lead by the task with thread ID +// tid in PID namespace ns. If no task has that TID, or if the task with that +// TID is not a thread group leader, ThreadGroupWithID returns nil. +func (ns *PIDNamespace) ThreadGroupWithID(tid ThreadID) *ThreadGroup { + ns.owner.mu.RLock() + defer ns.owner.mu.RUnlock() + t := ns.tasks[tid] + if t == nil { + return nil + } + if t != t.tg.leader { + return nil + } + return t.tg +} + +// IDOfTask returns the TID assigned to the given task in PID namespace ns. If +// the task is not visible in that namespace, IDOfTask returns 0. (This return +// value is significant in some cases, e.g. getppid() is documented as +// returning 0 if the caller's parent is in an ancestor namespace and +// consequently not visible to the caller.) If the task is nil, IDOfTask returns +// 0. +func (ns *PIDNamespace) IDOfTask(t *Task) ThreadID { + ns.owner.mu.RLock() + defer ns.owner.mu.RUnlock() + return ns.tids[t] +} + +// IDOfThreadGroup returns the TID assigned to tg's leader in PID namespace ns. +// If the task is not visible in that namespace, IDOfThreadGroup returns 0. +func (ns *PIDNamespace) IDOfThreadGroup(tg *ThreadGroup) ThreadID { + ns.owner.mu.RLock() + defer ns.owner.mu.RUnlock() + return ns.tids[tg.leader] +} + +// Tasks returns a snapshot of the tasks in ns. +func (ns *PIDNamespace) Tasks() []*Task { + ns.owner.mu.RLock() + defer ns.owner.mu.RUnlock() + tasks := make([]*Task, 0, len(ns.tasks)) + for t := range ns.tids { + tasks = append(tasks, t) + } + return tasks +} + +// ThreadGroups returns a snapshot of the thread groups in ns. +func (ns *PIDNamespace) ThreadGroups() []*ThreadGroup { + ns.owner.mu.RLock() + defer ns.owner.mu.RUnlock() + var tgs []*ThreadGroup + for t := range ns.tids { + if t == t.tg.leader { + tgs = append(tgs, t.tg) + } + } + return tgs +} + +// UserNamespace returns the user namespace associated with PID namespace ns. +func (ns *PIDNamespace) UserNamespace() *auth.UserNamespace { + return ns.userns +} + +// A threadGroupNode defines the relationship between a thread group and the +// rest of the system. Conceptually, threadGroupNode is data belonging to the +// owning TaskSet, as if TaskSet contained a field `nodes +// map[*ThreadGroup]*threadGroupNode`. However, for practical reasons, +// threadGroupNode is embedded in the ThreadGroup it represents. +// (threadGroupNode is an anonymous field in ThreadGroup; this is to expose +// threadGroupEntry's methods on ThreadGroup to make it implement +// threadGroupLinker.) +type threadGroupNode struct { + // pidns is the PID namespace containing the thread group and all of its + // member tasks. The pidns pointer is immutable. + pidns *PIDNamespace + + // eventQueue is notified whenever a event of interest to Task.Wait occurs + // in a child of this thread group, or a ptrace tracee of a task in this + // thread group. Events are defined in task_exit.go. + // + // Note that we cannot check and save this wait queue similarly to other + // wait queues, as the queue will not be empty by the time of saving, due + // to the wait sourced from Exec(). + eventQueue waiter.Queue `state:"nosave"` + + // leader is the thread group's leader, which is the oldest task in the + // thread group; usually the last task in the thread group to call + // execve(), or if no such task exists then the first task in the thread + // group, which was created by a call to fork() or clone() without + // CLONE_THREAD. Once a thread group has been made visible to the rest of + // the system by TaskSet.newTask, leader is never nil. + // + // Note that it's possible for the leader to exit without causing the rest + // of the thread group to exit; in such a case, leader will still be valid + // and non-nil, but leader will not be in tasks. + // + // leader is protected by the TaskSet mutex. + leader *Task + + // If execing is not nil, it is a task in the thread group that has killed + // all other tasks so that it can become the thread group leader and + // perform an execve. (execing may already be the thread group leader.) + // + // execing is analogous to Linux's signal_struct::group_exit_task. + // + // execing is protected by the TaskSet mutex. + execing *Task + + // tasks is all tasks in the thread group that have not yet been reaped. + // + // tasks is protected by both the TaskSet mutex and the signal mutex: + // Mutating tasks requires locking the TaskSet mutex for writing *and* + // locking the signal mutex. Reading tasks requires locking the TaskSet + // mutex *or* locking the signal mutex. + tasks taskList + + // tasksCount is the number of tasks in the thread group that have not yet + // been reaped; equivalently, tasksCount is the number of tasks in tasks. + // + // tasksCount is protected by both the TaskSet mutex and the signal mutex, + // as with tasks. + tasksCount int + + // liveTasks is the number of tasks in the thread group that have not yet + // reached TaskExitZombie. + // + // liveTasks is protected by the TaskSet mutex (NOT the signal mutex). + liveTasks int + + // activeTasks is the number of tasks in the thread group that have not yet + // reached TaskExitInitiated. + // + // activeTasks is protected by both the TaskSet mutex and the signal mutex, + // as with tasks. + activeTasks int +} + +// PIDNamespace returns the PID namespace containing tg. +func (tg *ThreadGroup) PIDNamespace() *PIDNamespace { + return tg.pidns +} + +// TaskSet returns the TaskSet containing tg. +func (tg *ThreadGroup) TaskSet() *TaskSet { + return tg.pidns.owner +} + +// Leader returns tg's leader. +func (tg *ThreadGroup) Leader() *Task { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + return tg.leader +} + +// Count returns the number of non-exited threads in the group. +func (tg *ThreadGroup) Count() int { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + var count int + for t := tg.tasks.Front(); t != nil; t = t.Next() { + count++ + } + return count +} + +// MemberIDs returns a snapshot of the ThreadIDs (in PID namespace pidns) for +// all tasks in tg. +func (tg *ThreadGroup) MemberIDs(pidns *PIDNamespace) []ThreadID { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + + var tasks []ThreadID + for t := tg.tasks.Front(); t != nil; t = t.Next() { + if id, ok := pidns.tids[t]; ok { + tasks = append(tasks, id) + } + } + return tasks +} + +// ID returns tg's leader's thread ID in its own PID namespace. If tg's leader +// is dead, ID returns 0. +func (tg *ThreadGroup) ID() ThreadID { + tg.pidns.owner.mu.RLock() + defer tg.pidns.owner.mu.RUnlock() + return tg.pidns.tids[tg.leader] +} + +// A taskNode defines the relationship between a task and the rest of the +// system. The comments on threadGroupNode also apply to taskNode. +type taskNode struct { + // tg is the thread group that this task belongs to. The tg pointer is + // immutable. + tg *ThreadGroup `state:"wait"` + + // taskEntry links into tg.tasks. Note that this means that + // Task.Next/Prev/SetNext/SetPrev refer to sibling tasks in the same thread + // group. See threadGroupNode.tasks for synchronization info. + taskEntry + + // parent is the task's parent. parent may be nil. + // + // parent is protected by the TaskSet mutex. + parent *Task + + // children is this task's children. + // + // children is protected by the TaskSet mutex. + children map[*Task]struct{} + + // If childPIDNamespace is not nil, all new tasks created by this task will + // be members of childPIDNamespace rather than this one. (As a corollary, + // this task becomes unable to create sibling tasks in the same thread + // group.) + // + // childPIDNamespace is exclusive to the task goroutine. + childPIDNamespace *PIDNamespace +} + +// ThreadGroup returns the thread group containing t. +func (t *Task) ThreadGroup() *ThreadGroup { + return t.tg +} + +// PIDNamespace returns the PID namespace containing t. +func (t *Task) PIDNamespace() *PIDNamespace { + return t.tg.pidns +} + +// TaskSet returns the TaskSet containing t. +func (t *Task) TaskSet() *TaskSet { + return t.tg.pidns.owner +} + +// Timekeeper returns the system Timekeeper. +func (t *Task) Timekeeper() *Timekeeper { + return t.k.timekeeper +} + +// Parent returns t's parent. +func (t *Task) Parent() *Task { + return t.parent +} + +// ThreadID returns t's thread ID in its own PID namespace. If the task is +// dead, ThreadID returns 0. +func (t *Task) ThreadID() ThreadID { + return t.tg.pidns.IDOfTask(t) +} diff --git a/pkg/sentry/kernel/time/BUILD b/pkg/sentry/kernel/time/BUILD new file mode 100644 index 000000000..84f31b2dc --- /dev/null +++ b/pkg/sentry/kernel/time/BUILD @@ -0,0 +1,32 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_stateify( + name = "time_state", + srcs = [ + "time.go", + ], + out = "time_state.go", + package = "time", +) + +go_library( + name = "time", + srcs = [ + "context.go", + "time.go", + "time_state.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/abi/linux", + "//pkg/log", + "//pkg/sentry/context", + "//pkg/state", + "//pkg/syserror", + "//pkg/waiter", + ], +) diff --git a/pkg/sentry/kernel/time/context.go b/pkg/sentry/kernel/time/context.go new file mode 100644 index 000000000..ac4dc01d8 --- /dev/null +++ b/pkg/sentry/kernel/time/context.go @@ -0,0 +1,44 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package time + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/context" +) + +// contextID is the time package's type for context.Context.Value keys. +type contextID int + +const ( + // CtxRealtimeClock is a Context.Value key for the current real time. + CtxRealtimeClock contextID = iota +) + +// RealtimeClockFromContext returns the real time clock associated with context +// ctx. +func RealtimeClockFromContext(ctx context.Context) Clock { + if v := ctx.Value(CtxRealtimeClock); v != nil { + return v.(Clock) + } + return nil +} + +// NowFromContext returns the current real time associated with context ctx. +func NowFromContext(ctx context.Context) Time { + if clk := RealtimeClockFromContext(ctx); clk != nil { + return clk.Now() + } + panic("encountered context without RealtimeClock") +} diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go new file mode 100644 index 000000000..c223c2f19 --- /dev/null +++ b/pkg/sentry/kernel/time/time.go @@ -0,0 +1,649 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package time defines the Timer type, which provides a periodic timer that +// works by sampling a user-provided clock. +package time + +import ( + "fmt" + "math" + "sync" + "time" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// Events that may be generated by a Clock. +const ( + // ClockEventSet occurs when a Clock undergoes a discontinuous change. + ClockEventSet waiter.EventMask = 1 << iota + + // ClockEventRateIncrease occurs when the rate at which a Clock advances + // increases significantly, such that values returned by previous calls to + // Clock.WallTimeUntil may be too large. + ClockEventRateIncrease +) + +// Time represents an instant in time with nanosecond precision. +// +// Time may represent time with respect to any clock and may not have any +// meaning in the real world. +type Time struct { + ns int64 +} + +var ( + // MinTime is the zero time instant, the lowest possible time that can + // be represented by Time. + MinTime = Time{ns: math.MinInt64} + + // MaxTime is the highest possible time that can be represented by + // Time. + MaxTime = Time{ns: math.MaxInt64} + + // ZeroTime represents the zero time in an unspecified Clock's domain. + ZeroTime = Time{ns: 0} +) + +const ( + // MinDuration is the minimum duration representable by time.Duration. + MinDuration = time.Duration(math.MinInt64) + + // MaxDuration is the maximum duration representable by time.Duration. + MaxDuration = time.Duration(math.MaxInt64) +) + +// FromNanoseconds returns a Time representing the point ns nanoseconds after +// an unspecified Clock's zero time. +func FromNanoseconds(ns int64) Time { + return Time{ns} +} + +// FromSeconds returns a Time representing the point s seconds after an +// unspecified Clock's zero time. +func FromSeconds(s int64) Time { + if s > math.MaxInt64/time.Second.Nanoseconds() { + return MaxTime + } + return Time{s * 1e9} +} + +// FromUnix converts from Unix seconds and nanoseconds to Time, assuming a real +// time Unix clock domain. +func FromUnix(s int64, ns int64) Time { + if s > math.MaxInt64/time.Second.Nanoseconds() { + return MaxTime + } + t := s * 1e9 + if t > math.MaxInt64-ns { + return MaxTime + } + return Time{t + ns} +} + +// FromTimespec converts from Linux Timespec to Time. +func FromTimespec(ts linux.Timespec) Time { + return Time{ts.ToNsecCapped()} +} + +// FromTimeval converts a Linux Timeval to Time. +func FromTimeval(tv linux.Timeval) Time { + return Time{tv.ToNsecCapped()} +} + +// Nanoseconds returns nanoseconds elapsed since the zero time in t's Clock +// domain. If t represents walltime, this is nanoseconds since the Unix epoch. +func (t Time) Nanoseconds() int64 { + return t.ns +} + +// Seconds returns seconds elapsed since the zero time in t's Clock domain. If +// t represents walltime, this is seconds since Unix epoch. +func (t Time) Seconds() int64 { + return t.Nanoseconds() / time.Second.Nanoseconds() +} + +// Timespec converts Time to a Linux timespec. +func (t Time) Timespec() linux.Timespec { + return linux.NsecToTimespec(t.Nanoseconds()) +} + +// Unix returns the (seconds, nanoseconds) representation of t such that +// seconds*1e9 + nanoseconds = t. +func (t Time) Unix() (s int64, ns int64) { + s = t.ns / 1e9 + ns = t.ns % 1e9 + return +} + +// TimeT converts Time to a Linux time_t. +func (t Time) TimeT() linux.TimeT { + return linux.NsecToTimeT(t.Nanoseconds()) +} + +// Timeval converts Time to a Linux timeval. +func (t Time) Timeval() linux.Timeval { + return linux.NsecToTimeval(t.Nanoseconds()) +} + +// Add adds the duration of d to t. +func (t Time) Add(d time.Duration) Time { + if t.ns > 0 && d.Nanoseconds() > math.MaxInt64-int64(t.ns) { + return MaxTime + } + if t.ns < 0 && d.Nanoseconds() < math.MinInt64-int64(t.ns) { + return MinTime + } + return Time{int64(t.ns) + d.Nanoseconds()} +} + +// AddTime adds the duration of u to t. +func (t Time) AddTime(u Time) Time { + return t.Add(time.Duration(u.ns)) +} + +// Equal reports whether the two times represent the same instant in time. +func (t Time) Equal(u Time) bool { + return t.ns == u.ns +} + +// Before reports whether the instant t is before the instant u. +func (t Time) Before(u Time) bool { + return t.ns < u.ns +} + +// After reports whether the instant t is after the instant u. +func (t Time) After(u Time) bool { + return t.ns > u.ns +} + +// Sub returns the duration of t - u. +// +// N.B. This measure may not make sense for every Time returned by ktime.Clock. +// Callers who need wall time duration can use ktime.Clock.WallTimeUntil to +// estimate that wall time. +func (t Time) Sub(u Time) time.Duration { + dur := time.Duration(int64(t.ns)-int64(u.ns)) * time.Nanosecond + switch { + case u.Add(dur).Equal(t): + return dur + case t.Before(u): + return MinDuration + default: + return MaxDuration + } +} + +// IsMin returns whether t represents the lowest possible time instant. +func (t Time) IsMin() bool { + return t == MinTime +} + +// IsZero returns whether t represents the zero time instant in t's Clock domain. +func (t Time) IsZero() bool { + return t == ZeroTime +} + +// String returns the time represented in nanoseconds as a string. +func (t Time) String() string { + return fmt.Sprintf("%dns", t.Nanoseconds()) +} + +// A Clock is an abstract time source. +type Clock interface { + // Now returns the current time in nanoseconds according to the Clock. + Now() Time + + // WallTimeUntil returns the estimated wall time until Now will return a + // value greater than or equal to t, given that a recent call to Now + // returned now. If t has already passed, WallTimeUntil may return 0 or a + // negative value. + // + // WallTimeUntil must be abstract to support Clocks that do not represent + // wall time (e.g. thread group execution timers). Clocks that represent + // wall times may embed the WallRateClock type to obtain an appropriate + // trivial implementation of WallTimeUntil. + // + // WallTimeUntil is used to determine when associated Timers should next + // check for expirations. Returning too small a value may result in + // spurious Timer goroutine wakeups, while returning too large a value may + // result in late expirations. Implementations should usually err on the + // side of underestimating. + WallTimeUntil(t, now Time) time.Duration + + // Waitable methods may be used to subscribe to Clock events. Waiters will + // not be preserved by Save and must be re-established during restore. + // + // Since Clock events are transient, implementations of + // waiter.Waitable.Readiness should return 0. + waiter.Waitable +} + +// WallRateClock implements Clock.WallTimeUntil for Clocks that elapse at the +// same rate as wall time. +type WallRateClock struct{} + +// WallTimeUntil implements Clock.WallTimeUntil. +func (WallRateClock) WallTimeUntil(t, now Time) time.Duration { + return t.Sub(now) +} + +// NoClockEvents implements waiter.Waitable for Clocks that do not generate +// events. +type NoClockEvents struct{} + +// Readiness implements waiter.Waitable.Readiness. +func (NoClockEvents) Readiness(mask waiter.EventMask) waiter.EventMask { + return 0 +} + +// EventRegister implements waiter.Waitable.EventRegister. +func (NoClockEvents) EventRegister(e *waiter.Entry, mask waiter.EventMask) { +} + +// EventUnregister implements waiter.Waitable.EventUnregister. +func (NoClockEvents) EventUnregister(e *waiter.Entry) { +} + +// ClockEventsQueue implements waiter.Waitable by wrapping waiter.Queue and +// defining waiter.Waitable.Readiness as required by Clock. +type ClockEventsQueue struct { + waiter.Queue +} + +// Readiness implements waiter.Waitable.Readiness. +func (ClockEventsQueue) Readiness(mask waiter.EventMask) waiter.EventMask { + return 0 +} + +// A TimerListener receives expirations from a Timer. +type TimerListener interface { + // Notify is called when its associated Timer expires. exp is the number of + // expirations. + // + // Notify is called with the associated Timer's mutex locked, so Notify + // must not take any locks that precede Timer.mu in lock order. + // + // Preconditions: exp > 0. + Notify(exp uint64) + + // Destroy is called when the timer is destroyed. + Destroy() +} + +// Setting contains user-controlled mutable Timer properties. +type Setting struct { + // Enabled is true if the timer is running. + Enabled bool + + // Next is the time in nanoseconds of the next expiration. + Next Time + + // Period is the time in nanoseconds between expirations. If Period is + // zero, the timer will not automatically restart after expiring. + // + // Invariant: Period >= 0. + Period time.Duration +} + +// SettingFromSpec converts a (value, interval) pair to a Setting based on a +// reading from c. value is interpreted as a time relative to c.Now(). +func SettingFromSpec(value time.Duration, interval time.Duration, c Clock) (Setting, error) { + if value < 0 { + return Setting{}, syserror.EINVAL + } + if value == 0 { + return Setting{Period: interval}, nil + } + return Setting{ + Enabled: true, + Next: c.Now().Add(value), + Period: interval, + }, nil +} + +// SettingFromAbsSpec converts a (value, interval) pair to a Setting based on a +// reading from c. value is interpreted as an absolute time. +func SettingFromAbsSpec(value Time, interval time.Duration) (Setting, error) { + if value.Before(ZeroTime) { + return Setting{}, syserror.EINVAL + } + if value.IsZero() { + return Setting{Period: interval}, nil + } + return Setting{ + Enabled: true, + Next: value, + Period: interval, + }, nil +} + +// SpecFromSetting converts a timestamp and a Setting to a (relative value, +// interval) pair, as used by most Linux syscalls that return a struct +// itimerval or struct itimerspec. +func SpecFromSetting(now Time, s Setting) (value, period time.Duration) { + if !s.Enabled { + return 0, s.Period + } + return s.Next.Sub(now), s.Period +} + +// advancedTo returns an updated Setting and a number of expirations after +// the associated Clock indicates a time of now. +// +// Settings may be created by successive calls to advancedTo with decreasing +// values of now (i.e. time may appear to go backward). Supporting this is +// required to support non-monotonic clocks, as well as allowing +// Timer.clock.Now() to be called without holding Timer.mu. +func (s Setting) advancedTo(now Time) (Setting, uint64) { + if !s.Enabled { + return s, 0 + } + if s.Next.After(now) { + return s, 0 + } + if s.Period == 0 { + s.Enabled = false + return s, 1 + } + exp := 1 + uint64(now.Sub(s.Next).Nanoseconds())/uint64(s.Period) + s.Next = s.Next.Add(time.Duration(uint64(s.Period) * exp)) + return s, exp +} + +// Timer is an optionally-periodic timer driven by sampling a user-specified +// Clock. Timer's semantics support the requirements of Linux's interval timers +// (setitimer(2), timer_create(2), timerfd_create(2)). +// +// Timers should be created using NewTimer and must be cleaned up by calling +// Timer.Destroy when no longer used. +type Timer struct { + // clock is the time source. clock is immutable. + clock Clock + + // listener is notified of expirations. listener is immutable. + listener TimerListener + + // mu protects the following mutable fields. + mu sync.Mutex `state:"nosave"` + + // setting is the timer setting. setting is protected by mu. + setting Setting + + // paused is true if the Timer is paused. paused is protected by mu. + paused bool + + // kicker is used to wake the Timer goroutine. The kicker pointer is + // immutable, but its state is protected by mu. + kicker *time.Timer `state:"nosave"` + + // entry is registered with clock.EventRegister. entry is immutable. + // + // Per comment in Clock, entry must be re-registered after restore; per + // comment in Timer.Load, this is done in Timer.Resume. + entry waiter.Entry `state:"nosave"` + + // events is the channel that will be notified whenever entry receives an + // event. It is also closed by Timer.Destroy to instruct the Timer + // goroutine to exit. + events chan struct{} `state:"nosave"` +} + +// timerTickEvents are Clock events that require the Timer goroutine to Tick +// prematurely. +const timerTickEvents = ClockEventSet | ClockEventRateIncrease + +// NewTimer returns a new Timer that will obtain time from clock and send +// expirations to listener. The Timer is initially stopped and has no first +// expiration or period configured. +func NewTimer(clock Clock, listener TimerListener) *Timer { + t := &Timer{ + clock: clock, + listener: listener, + } + t.init() + return t +} + +// After waits for the duration to elapse according to clock and then sends a +// notification on the returned channel. The timer is started immediately and +// will fire exactly once. The second return value is the start time used with +// the duration. +// +// Callers must call Timer.Destroy. +func After(clock Clock, duration time.Duration) (*Timer, Time, <-chan struct{}) { + notifier, tchan := NewChannelNotifier() + t := NewTimer(clock, notifier) + now := clock.Now() + + t.Swap(Setting{ + Enabled: true, + Period: 0, + Next: now.Add(duration), + }) + return t, now, tchan +} + +// init initializes Timer state that is not preserved across save/restore. If +// init has already been called, calling it again is a no-op. +// +// Preconditions: t.mu must be locked, or the caller must have exclusive access +// to t. +func (t *Timer) init() { + if t.kicker != nil { + return + } + // If t.kicker is nil, the Timer goroutine can't be running, so we can't + // race with it. + t.kicker = time.NewTimer(0) + t.entry, t.events = waiter.NewChannelEntry(nil) + t.clock.EventRegister(&t.entry, timerTickEvents) + go t.runGoroutine() // S/R-SAFE: synchronized by t.mu +} + +// Destroy releases resources owned by the Timer. A Destroyed Timer must not be +// used again; in particular, a Destroyed Timer should not be Saved. +func (t *Timer) Destroy() { + // Stop the Timer, ensuring that the Timer goroutine will not call + // t.kicker.Reset, before calling t.kicker.Stop. + t.mu.Lock() + t.setting.Enabled = false + t.mu.Unlock() + t.kicker.Stop() + // Unregister t.entry, ensuring that the Clock will not send to t.events, + // before closing t.events to instruct the Timer goroutine to exit. + t.clock.EventUnregister(&t.entry) + close(t.events) + t.listener.Destroy() +} + +func (t *Timer) runGoroutine() { + for { + select { + case <-t.kicker.C: + case _, ok := <-t.events: + if !ok { + // Channel closed by Destroy. + return + } + } + t.Tick() + } +} + +// Tick requests that the Timer immediately check for expirations and +// re-evaluate when it should next check for expirations. +func (t *Timer) Tick() { + now := t.clock.Now() + t.mu.Lock() + defer t.mu.Unlock() + if t.paused { + return + } + s, exp := t.setting.advancedTo(now) + t.setting = s + if exp > 0 { + t.listener.Notify(exp) + } + t.resetKickerLocked(now) +} + +// Pause pauses the Timer, ensuring that it does not generate any further +// expirations until Resume is called. If the Timer is already paused, Pause +// has no effect. +func (t *Timer) Pause() { + t.mu.Lock() + defer t.mu.Unlock() + t.paused = true + // t.kicker may be nil if we were restored but never resumed. + if t.kicker != nil { + t.kicker.Stop() + } +} + +// Resume ends the effect of Pause. If the Timer is not paused, Resume has no +// effect. +func (t *Timer) Resume() { + t.mu.Lock() + defer t.mu.Unlock() + if !t.paused { + return + } + t.paused = false + + // Lazily initialize the Timer. We can't call Timer.init until Timer.Resume + // because save/restore will restore Timers before + // kernel.Timekeeper.SetClocks() has been called, so if t.clock is backed + // by a kernel.Timekeeper then the Timer goroutine will panic if it calls + // t.clock.Now(). + t.init() + + // Kick the Timer goroutine in case it was already initialized, but the + // Timer goroutine was sleeping. + t.kicker.Reset(0) +} + +// Get returns a snapshot of the Timer's current Setting and the time +// (according to the Timer's Clock) at which the snapshot was taken. +// +// Preconditions: The Timer must not be paused (since its Setting cannot +// be advanced to the current time while it is paused.) +func (t *Timer) Get() (Time, Setting) { + now := t.clock.Now() + t.mu.Lock() + defer t.mu.Unlock() + if t.paused { + panic(fmt.Sprintf("Timer.Get called on paused Timer %p", t)) + } + s, exp := t.setting.advancedTo(now) + t.setting = s + if exp > 0 { + t.listener.Notify(exp) + } + t.resetKickerLocked(now) + return now, s +} + +// Swap atomically changes the Timer's Setting and returns the Timer's previous +// Setting and the time (according to the Timer's Clock) at which the snapshot +// was taken. Setting s.Enabled to true starts the Timer, while setting +// s.Enabled to false stops it. +// +// Preconditions: The Timer must not be paused. +func (t *Timer) Swap(s Setting) (Time, Setting) { + return t.SwapAnd(s, nil) +} + +// SwapAnd atomically changes the Timer's Setting, calls f if it is not nil, +// and returns the Timer's previous Setting and the time (according to the +// Timer's Clock) at which the Setting was changed. Setting s.Enabled to true +// starts the timer, while setting s.Enabled to false stops it. +// +// Preconditions: The Timer must not be paused. f cannot call any Timer methods +// since it is called with the Timer mutex locked. +func (t *Timer) SwapAnd(s Setting, f func()) (Time, Setting) { + now := t.clock.Now() + t.mu.Lock() + defer t.mu.Unlock() + if t.paused { + panic(fmt.Sprintf("Timer.SwapAnd called on paused Timer %p", t)) + } + oldS, oldExp := t.setting.advancedTo(now) + if oldExp > 0 { + t.listener.Notify(oldExp) + } + if f != nil { + f() + } + newS, newExp := s.advancedTo(now) + t.setting = newS + if newExp > 0 { + t.listener.Notify(newExp) + } + t.resetKickerLocked(now) + return now, oldS +} + +// Preconditions: t.mu must be locked. +func (t *Timer) resetKickerLocked(now Time) { + if t.setting.Enabled { + // Clock.WallTimeUntil may return a negative value. This is fine; + // time.when treats negative Durations as 0. + t.kicker.Reset(t.clock.WallTimeUntil(t.setting.Next, now)) + } + // We don't call t.kicker.Stop if !t.setting.Enabled because in most cases + // resetKickerLocked will be called from the Timer goroutine itself, in + // which case t.kicker has already fired and t.kicker.Stop will be an + // expensive no-op (time.Timer.Stop => time.stopTimer => runtime.stopTimer + // => runtime.deltimer). +} + +// Clock returns the Clock used by t. +func (t *Timer) Clock() Clock { + return t.clock +} + +// ChannelNotifier is a TimerListener that sends a message on an empty struct +// channel. +// +// ChannelNotifier cannot be saved or loaded. +type ChannelNotifier struct { + // tchan must be a buffered channel. + tchan chan struct{} +} + +// NewChannelNotifier creates a new channel notifier. +// +// If the notifier is used with a timer, Timer.Destroy will close the channel +// returned here. +func NewChannelNotifier() (TimerListener, <-chan struct{}) { + tchan := make(chan struct{}, 1) + return &ChannelNotifier{tchan}, tchan +} + +// Notify implements ktime.TimerListener.Notify. +func (c *ChannelNotifier) Notify(uint64) { + select { + case c.tchan <- struct{}{}: + default: + } +} + +// Destroy implements ktime.TimerListener.Destroy and will close the channel. +func (c *ChannelNotifier) Destroy() { + close(c.tchan) +} diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go new file mode 100644 index 000000000..3f16c1676 --- /dev/null +++ b/pkg/sentry/kernel/timekeeper.go @@ -0,0 +1,270 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "sync" + "time" + + "gvisor.googlesource.com/gvisor/pkg/log" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" +) + +// Timekeeper manages all of the kernel clocks. +type Timekeeper struct { + // clocks are the clock sources. + // + // These are not saved directly, as the new machine's clock may behave + // differently. + // + // It is set only once, by SetClocks. + clocks sentrytime.Clocks `state:"nosave"` + + // bootTime is the realtime when the system "booted". i.e., when + // SetClocks was called in the initial (not restored) run. + bootTime ktime.Time + + // monotonicOffset is the offset to apply to the monotonic clock output + // from clocks. + // + // It is set only once, by SetClocks. + monotonicOffset int64 `state:"nosave"` + + // restored indicates that this Timekeeper was restored from a state + // file. + restored bool `state:"nosave"` + + // saveMonotonic is the (offset) value of the monotonic clock at the + // time of save. + // + // It is only valid if restored is true. + // + // It is only used in SetClocks after restore to compute the new + // monotonicOffset. + saveMonotonic int64 + + // saveRealtime is the value of the realtime clock at the time of save. + // + // It is only valid if restored is true. + // + // It is only used in SetClocks after restore to compute the new + // monotonicOffset. + saveRealtime int64 + + // params manages the parameter page. + params *VDSOParamPage + + // mu protects destruction with stop and wg. + mu sync.Mutex `state:"nosave"` + + // stop is used to tell the update goroutine to exit. + stop chan struct{} `state:"nosave"` + + // wg is used to indicate that the update goroutine has exited. + wg sync.WaitGroup `state:"nosave"` +} + +// NewTimekeeper returns a Timekeeper that is automatically kept up-to-date. +// NewTimekeeper does not take ownership of paramPage. +// +// SetClocks must be called on the returned Timekeeper before it is usable. +func NewTimekeeper(platform platform.Platform, paramPage platform.FileRange) (*Timekeeper, error) { + return &Timekeeper{ + params: NewVDSOParamPage(platform, paramPage), + }, nil +} + +// SetClocks the backing clock source. +// +// SetClocks must be called before the Timekeeper is used, and it may not be +// called more than once, as changing the clock source without extra correction +// could cause time discontinuities. +// +// It must also be called after Load. +func (t *Timekeeper) SetClocks(c sentrytime.Clocks) { + // Update the params, marking them "not ready", as we may need to + // restart calibration on this new machine. + if t.restored { + if err := t.params.Write(func() vdsoParams { + return vdsoParams{} + }); err != nil { + panic("unable to reset VDSO params: " + err.Error()) + } + } + + if t.clocks != nil { + panic("SetClocks called on previously-initialized Timekeeper") + } + + t.clocks = c + + // Compute the offset of the monotonic clock from the base Clocks. + // + // In a fresh (not restored) sentry, monotonic time starts at zero. + // + // In a restored sentry, monotonic time jumps forward by approximately + // the same amount as real time. There are no guarantees here, we are + // just making a best-effort attempt to to make it appear that the app + // was simply not scheduled for a long period, rather than that the + // real time clock was changed. + // + // If real time went backwards, it remains the same. + wantMonotonic := int64(0) + + nowMonotonic, err := t.clocks.GetTime(sentrytime.Monotonic) + if err != nil { + panic("Unable to get current monotonic time: " + err.Error()) + } + + nowRealtime, err := t.clocks.GetTime(sentrytime.Realtime) + if err != nil { + panic("Unable to get current realtime: " + err.Error()) + } + + if t.restored { + wantMonotonic = t.saveMonotonic + elapsed := nowRealtime - t.saveRealtime + if elapsed > 0 { + wantMonotonic += elapsed + } + } + + t.monotonicOffset = wantMonotonic - nowMonotonic + + if !t.restored { + // Hold on to the initial "boot" time. + t.bootTime = ktime.FromNanoseconds(nowRealtime) + } + + t.mu.Lock() + defer t.mu.Unlock() + t.startUpdater() +} + +// startUpdater starts an update goroutine that keeps the clocks updated. +// +// mu must be held. +func (t *Timekeeper) startUpdater() { + if t.stop != nil { + // Timekeeper already started + return + } + t.stop = make(chan struct{}) + + // Keep the clocks up to date. + // + // Note that the Go runtime uses host CLOCK_MONOTONIC to service the + // timer, so it may run at a *slightly* different rate from the + // application CLOCK_MONOTONIC. That is fine, as we only need to update + // at approximately this rate. + timer := time.NewTicker(sentrytime.ApproxUpdateInterval) + t.wg.Add(1) + go func() { // S/R-SAFE: stopped during save. + for { + // Start with an update immediately, so the clocks are + // ready ASAP. + + // Call Update within a Write block to prevent the VDSO + // from using the old params between Update and + // Write. + if err := t.params.Write(func() vdsoParams { + monotonicParams, monotonicOk, realtimeParams, realtimeOk := t.clocks.Update() + + var p vdsoParams + if monotonicOk { + p.monotonicReady = 1 + p.monotonicBaseCycles = int64(monotonicParams.BaseCycles) + p.monotonicBaseRef = int64(monotonicParams.BaseRef) + t.monotonicOffset + p.monotonicFrequency = monotonicParams.Frequency + } + if realtimeOk { + p.realtimeReady = 1 + p.realtimeBaseCycles = int64(realtimeParams.BaseCycles) + p.realtimeBaseRef = int64(realtimeParams.BaseRef) + p.realtimeFrequency = realtimeParams.Frequency + } + + log.Debugf("Updating VDSO parameters: %+v", p) + + return p + }); err != nil { + log.Warningf("Unable to update VDSO parameter page: %v", err) + } + + select { + case <-timer.C: + case <-t.stop: + t.wg.Done() + return + } + } + }() +} + +// stopUpdater stops the update goroutine, blocking until it exits. +// +// mu must be held. +func (t *Timekeeper) stopUpdater() { + if t.stop == nil { + // Updater not running. + return + } + + close(t.stop) + t.wg.Wait() + t.stop = nil +} + +// Destroy destroys the Timekeeper, freeing all associated resources. +func (t *Timekeeper) Destroy() { + t.mu.Lock() + defer t.mu.Unlock() + + t.stopUpdater() +} + +// PauseUpdates stops clock parameter updates. This should only be used when +// Tasks are not running and thus cannot access the clock. +func (t *Timekeeper) PauseUpdates() { + t.mu.Lock() + defer t.mu.Unlock() + t.stopUpdater() +} + +// ResumeUpdates restarts clock parameter updates stopped by PauseUpdates. +func (t *Timekeeper) ResumeUpdates() { + t.mu.Lock() + defer t.mu.Unlock() + t.startUpdater() +} + +// GetTime returns the current time in nanoseconds. +func (t *Timekeeper) GetTime(c sentrytime.ClockID) (int64, error) { + if t.clocks == nil { + panic("Timekeeper used before initialized with SetClocks") + } + now, err := t.clocks.GetTime(c) + if err == nil && c == sentrytime.Monotonic { + now += t.monotonicOffset + } + return now, err +} + +// BootTime returns the system boot real time. +func (t *Timekeeper) BootTime() ktime.Time { + return t.bootTime +} diff --git a/pkg/sentry/kernel/timekeeper_state.go b/pkg/sentry/kernel/timekeeper_state.go new file mode 100644 index 000000000..aee983ac7 --- /dev/null +++ b/pkg/sentry/kernel/timekeeper_state.go @@ -0,0 +1,41 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/time" +) + +// beforeSave is invoked by stateify. +func (t *Timekeeper) beforeSave() { + if t.stop != nil { + panic("pauseUpdates must be called before Save") + } + + // N.B. we want the *offset* monotonic time. + var err error + if t.saveMonotonic, err = t.GetTime(time.Monotonic); err != nil { + panic("unable to get current monotonic time: " + err.Error()) + } + + if t.saveRealtime, err = t.GetTime(time.Realtime); err != nil { + panic("unable to get current realtime: " + err.Error()) + } +} + +// afterLoad is invoked by stateify. +func (t *Timekeeper) afterLoad() { + t.restored = true +} diff --git a/pkg/sentry/kernel/timekeeper_test.go b/pkg/sentry/kernel/timekeeper_test.go new file mode 100644 index 000000000..08bacba4f --- /dev/null +++ b/pkg/sentry/kernel/timekeeper_test.go @@ -0,0 +1,156 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "testing" + + "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// mockClocks is a sentrytime.Clocks that simply returns the times in the +// struct. +type mockClocks struct { + monotonic int64 + realtime int64 +} + +// Update implements sentrytime.Clocks.Update. It does nothing. +func (*mockClocks) Update() (monotonicParams sentrytime.Parameters, monotonicOk bool, realtimeParam sentrytime.Parameters, realtimeOk bool) { + return +} + +// Update implements sentrytime.Clocks.GetTime. +func (c *mockClocks) GetTime(id sentrytime.ClockID) (int64, error) { + switch id { + case sentrytime.Monotonic: + return c.monotonic, nil + case sentrytime.Realtime: + return c.realtime, nil + default: + return 0, syserror.EINVAL + } +} + +// stateTestClocklessTimekeeper returns a test Timekeeper which has not had +// SetClocks called. +func stateTestClocklessTimekeeper(tb testing.TB) *Timekeeper { + ctx := contexttest.Context(tb) + p := platform.FromContext(ctx) + fr, err := p.Memory().Allocate(usermem.PageSize, usage.Anonymous) + if err != nil { + tb.Fatalf("failed to allocate memory: %v", err) + } + return &Timekeeper{ + params: NewVDSOParamPage(p, fr), + } +} + +func stateTestTimekeeper(tb testing.TB) *Timekeeper { + t := stateTestClocklessTimekeeper(tb) + t.SetClocks(sentrytime.NewCalibratedClocks()) + return t +} + +// TestTimekeeperMonotonicZero tests that monotonic time starts at zero. +func TestTimekeeperMonotonicZero(t *testing.T) { + c := &mockClocks{ + monotonic: 100000, + } + + tk := stateTestClocklessTimekeeper(t) + tk.SetClocks(c) + defer tk.Destroy() + + now, err := tk.GetTime(sentrytime.Monotonic) + if err != nil { + t.Errorf("GetTime err got %v want nil", err) + } + if now != 0 { + t.Errorf("GetTime got %d want 0", now) + } + + c.monotonic += 10 + + now, err = tk.GetTime(sentrytime.Monotonic) + if err != nil { + t.Errorf("GetTime err got %v want nil", err) + } + if now != 10 { + t.Errorf("GetTime got %d want 10", now) + } +} + +// TestTimekeeperMonotonicJumpForward tests that monotonic time jumps forward +// after restore. +func TestTimekeeperMonotonicForward(t *testing.T) { + c := &mockClocks{ + monotonic: 900000, + realtime: 600000, + } + + tk := stateTestClocklessTimekeeper(t) + tk.restored = true + tk.saveMonotonic = 100000 + tk.saveRealtime = 400000 + tk.SetClocks(c) + defer tk.Destroy() + + // The monotonic clock should jump ahead by 200000 to 300000. + // + // The new system monotonic time (900000) is irrelevant to what the app + // sees. + now, err := tk.GetTime(sentrytime.Monotonic) + if err != nil { + t.Errorf("GetTime err got %v want nil", err) + } + if now != 300000 { + t.Errorf("GetTime got %d want 300000", now) + } +} + +// TestTimekeeperMonotonicJumpBackwards tests that monotonic time does not jump +// backwards when realtime goes backwards. +func TestTimekeeperMonotonicJumpBackwards(t *testing.T) { + c := &mockClocks{ + monotonic: 900000, + realtime: 400000, + } + + tk := stateTestClocklessTimekeeper(t) + tk.restored = true + tk.saveMonotonic = 100000 + tk.saveRealtime = 600000 + tk.SetClocks(c) + defer tk.Destroy() + + // The monotonic clock should remain at 100000. + // + // The new system monotonic time (900000) is irrelevant to what the app + // sees and we don't want to jump the monotonic clock backwards like + // realtime did. + now, err := tk.GetTime(sentrytime.Monotonic) + if err != nil { + t.Errorf("GetTime err got %v want nil", err) + } + if now != 100000 { + t.Errorf("GetTime got %d want 100000", now) + } +} diff --git a/pkg/sentry/kernel/timer.go b/pkg/sentry/kernel/timer.go new file mode 100644 index 000000000..03a3310be --- /dev/null +++ b/pkg/sentry/kernel/timer.go @@ -0,0 +1,282 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + "time" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/limits" + sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" +) + +// timekeeperClock is a ktime.Clock that reads time from a +// kernel.Timekeeper-managed clock. +type timekeeperClock struct { + tk *Timekeeper + c sentrytime.ClockID + + // Implements ktime.Clock.WallTimeUntil. + ktime.WallRateClock `state:"nosave"` + + // Implements waiter.Waitable. (We have no ability to detect + // discontinuities from external changes to CLOCK_REALTIME). + ktime.NoClockEvents `state:"nosave"` +} + +// Now implements ktime.Clock.Now. +func (tc *timekeeperClock) Now() ktime.Time { + now, err := tc.tk.GetTime(tc.c) + if err != nil { + panic(fmt.Sprintf("timekeeperClock(ClockID=%v)).Now: %v", tc.c, err)) + } + return ktime.FromNanoseconds(now) +} + +// tgClock is a ktime.Clock that measures the time a thread group has spent +// executing. +type tgClock struct { + tg *ThreadGroup + + // If includeSys is true, the tgClock includes both time spent executing + // application code as well as time spent in the sentry. Otherwise, the + // tgClock includes only time spent executing application code. + includeSys bool + + // Implements waiter.Waitable. + ktime.ClockEventsQueue `state:"nosave"` +} + +// UserCPUClock returns a ktime.Clock that measures the time that a thread +// group has spent executing. +func (tg *ThreadGroup) UserCPUClock() ktime.Clock { + return tg.tm.virtClock +} + +// CPUClock returns a ktime.Clock that measures the time that a thread group +// has spent executing, including sentry time. +func (tg *ThreadGroup) CPUClock() ktime.Clock { + return tg.tm.profClock +} + +// Now implements ktime.Clock.Now. +func (tgc *tgClock) Now() ktime.Time { + stats := tgc.tg.CPUStats() + if tgc.includeSys { + return ktime.FromNanoseconds((stats.UserTime + stats.SysTime).Nanoseconds()) + } + return ktime.FromNanoseconds(stats.UserTime.Nanoseconds()) +} + +// WallTimeUntil implements ktime.Clock.WallTimeUntil. +func (tgc *tgClock) WallTimeUntil(t, now ktime.Time) time.Duration { + // The assumption here is that the time spent in this process (not matter + // virtual or prof) should not exceed wall time * active tasks, since + // Task.exitThreadGroup stops accounting as it transitions to + // TaskExitInitiated. + tgc.tg.pidns.owner.mu.RLock() + n := tgc.tg.activeTasks + tgc.tg.pidns.owner.mu.RUnlock() + if n == 0 { + if t.Before(now) { + return 0 + } + // The timer tick raced with thread group exit, after which no more + // tasks can enter the thread group. So tgc.Now() will never advance + // again. Return a large delay; the timer should be stopped long before + // it comes again anyway. + return time.Hour + } + // This is a lower bound on the amount of time that can elapse before an + // associated timer expires, so returning this value tends to result in a + // sequence of closely-spaced ticks just before timer expiry. To avoid + // this, round up to the nearest ClockTick; CPU usage measurements are + // limited to this resolution anyway. + remaining := time.Duration(int64(t.Sub(now))/int64(n)) * time.Nanosecond + return ((remaining + (linux.ClockTick - time.Nanosecond)) / linux.ClockTick) * linux.ClockTick +} + +// taskClock is a ktime.Clock that measures the time that a task has spent +// executing. +type taskClock struct { + t *Task + + // If includeSys is true, the taskClock includes both time spent executing + // application code as well as time spent in the sentry. Otherwise, the + // taskClock includes only time spent executing application code. + includeSys bool + + // Implements waiter.Waitable. TimeUntil wouldn't change its estimation + // based on either of the clock events, so there's no event to be + // notified for. + ktime.NoClockEvents `state:"nosave"` + + // Implements ktime.Clock.WallTimeUntil. + // + // As an upper bound, a task's clock cannot advance faster than CPU + // time. It would have to execute at a rate of more than 1 task-second + // per 1 CPU-second, which isn't possible. + ktime.WallRateClock `state:"nosave"` +} + +// UserCPUClock returns a clock measuring the CPU time the task has spent +// executing application code. +func (t *Task) UserCPUClock() ktime.Clock { + return &taskClock{t: t, includeSys: false} +} + +// CPUClock returns a clock measuring the CPU time the task has spent executing +// application and "kernel" code. +func (t *Task) CPUClock() ktime.Clock { + return &taskClock{t: t, includeSys: true} +} + +// Now implements ktime.Clock.Now. +func (tc *taskClock) Now() ktime.Time { + stats := tc.t.CPUStats() + if tc.includeSys { + return ktime.FromNanoseconds((stats.UserTime + stats.SysTime).Nanoseconds()) + } + return ktime.FromNanoseconds(stats.UserTime.Nanoseconds()) +} + +// signalNotifier is a ktime.Listener that sends signals to a ThreadGroup. +type signalNotifier struct { + tg *ThreadGroup + signal linux.Signal + realTimer bool + includeSys bool +} + +// Notify implements ktime.TimerListener.Notify. +func (s *signalNotifier) Notify(exp uint64) { + // Since all signals sent using a signalNotifier are standard (not + // real-time) signals, we can ignore the number of expirations and send + // only a single signal. + if s.realTimer { + // real timer signal sent to leader. See kernel/time/itimer.c:it_real_fn + s.tg.SendSignal(sigPriv(s.signal)) + } else { + s.tg.SendTimerSignal(sigPriv(s.signal), s.includeSys) + } +} + +// Destroy implements ktime.TimerListener.Destroy. +func (s *signalNotifier) Destroy() {} + +// TimerManager is a collection of supported process cpu timers. +type TimerManager struct { + // Clocks used to drive thread group execution time timers. + virtClock *tgClock + profClock *tgClock + + RealTimer *ktime.Timer + VirtualTimer *ktime.Timer + ProfTimer *ktime.Timer + SoftLimitTimer *ktime.Timer + HardLimitTimer *ktime.Timer +} + +// newTimerManager returns a new instance of TimerManager. +func newTimerManager(tg *ThreadGroup, monotonicClock ktime.Clock) TimerManager { + virtClock := &tgClock{tg: tg, includeSys: false} + profClock := &tgClock{tg: tg, includeSys: true} + tm := TimerManager{ + virtClock: virtClock, + profClock: profClock, + RealTimer: ktime.NewTimer(monotonicClock, &signalNotifier{ + tg: tg, + signal: linux.SIGALRM, + realTimer: true, + includeSys: false, + }), + VirtualTimer: ktime.NewTimer(virtClock, &signalNotifier{ + tg: tg, + signal: linux.SIGVTALRM, + realTimer: false, + includeSys: false, + }), + ProfTimer: ktime.NewTimer(profClock, &signalNotifier{ + tg: tg, + signal: linux.SIGPROF, + realTimer: false, + includeSys: true, + }), + SoftLimitTimer: ktime.NewTimer(profClock, &signalNotifier{ + tg: tg, + signal: linux.SIGXCPU, + realTimer: false, + includeSys: true, + }), + HardLimitTimer: ktime.NewTimer(profClock, &signalNotifier{ + tg: tg, + signal: linux.SIGKILL, + realTimer: false, + includeSys: true, + }), + } + tm.applyCPULimits(tg.Limits().Get(limits.CPU)) + return tm +} + +// Save saves this TimerManger. + +// destroy destroys all timers. +func (tm *TimerManager) destroy() { + tm.RealTimer.Destroy() + tm.VirtualTimer.Destroy() + tm.ProfTimer.Destroy() + tm.SoftLimitTimer.Destroy() + tm.HardLimitTimer.Destroy() +} + +func (tm *TimerManager) applyCPULimits(l limits.Limit) { + tm.SoftLimitTimer.Swap(ktime.Setting{ + Enabled: l.Cur != limits.Infinity, + Next: ktime.FromNanoseconds((time.Duration(l.Cur) * time.Second).Nanoseconds()), + Period: time.Second, + }) + tm.HardLimitTimer.Swap(ktime.Setting{ + Enabled: l.Max != limits.Infinity, + Next: ktime.FromNanoseconds((time.Duration(l.Max) * time.Second).Nanoseconds()), + }) +} + +// kick is called when the number of threads in the thread group associated +// with tm increases. +func (tm *TimerManager) kick() { + tm.virtClock.Notify(ktime.ClockEventRateIncrease) + tm.profClock.Notify(ktime.ClockEventRateIncrease) +} + +// pause is to pause the timers and stop timer signal delivery. +func (tm *TimerManager) pause() { + tm.RealTimer.Pause() + tm.VirtualTimer.Pause() + tm.ProfTimer.Pause() + tm.SoftLimitTimer.Pause() + tm.HardLimitTimer.Pause() +} + +// resume is to resume the timers and continue timer signal delivery. +func (tm *TimerManager) resume() { + tm.RealTimer.Resume() + tm.VirtualTimer.Resume() + tm.ProfTimer.Resume() + tm.SoftLimitTimer.Resume() + tm.HardLimitTimer.Resume() +} diff --git a/pkg/sentry/kernel/uts_namespace.go b/pkg/sentry/kernel/uts_namespace.go new file mode 100644 index 000000000..58e9b4d1b --- /dev/null +++ b/pkg/sentry/kernel/uts_namespace.go @@ -0,0 +1,100 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" +) + +// UTSNamespace represents a UTS namespace, a holder of two system identifiers: +// the hostname and domain name. +type UTSNamespace struct { + // mu protects all fields below. + mu sync.Mutex `state:"nosave"` + hostName string + domainName string + + // userns is the user namespace associated with the UTSNamespace. + // Privileged operations on this UTSNamespace must have appropriate + // capabilities in userns. + // + // userns is immutable. + userns *auth.UserNamespace +} + +// NewUTSNamespace creates a new UTS namespace. +func NewUTSNamespace(hostName, domainName string, userns *auth.UserNamespace) *UTSNamespace { + return &UTSNamespace{ + hostName: hostName, + domainName: domainName, + userns: userns, + } +} + +// UTSNamespace returns the task's UTS namespace. +func (t *Task) UTSNamespace() *UTSNamespace { + t.mu.Lock() + defer t.mu.Unlock() + return t.utsns +} + +// HostName returns the host name of this UTS namespace. +func (u *UTSNamespace) HostName() string { + u.mu.Lock() + defer u.mu.Unlock() + return u.hostName +} + +// SetHostName sets the host name of this UTS namespace. +func (u *UTSNamespace) SetHostName(host string) { + u.mu.Lock() + defer u.mu.Unlock() + u.hostName = host +} + +// DomainName returns the domain name of this UTS namespace. +func (u *UTSNamespace) DomainName() string { + u.mu.Lock() + defer u.mu.Unlock() + return u.domainName +} + +// SetDomainName sets the domain name of this UTS namespace. +func (u *UTSNamespace) SetDomainName(domain string) { + u.mu.Lock() + defer u.mu.Unlock() + u.domainName = domain +} + +// UserNamespace returns the user namespace associated with this UTS namespace. +func (u *UTSNamespace) UserNamespace() *auth.UserNamespace { + u.mu.Lock() + defer u.mu.Unlock() + return u.userns +} + +// Clone makes a copy of this UTS namespace, associating the given user +// namespace. +func (u *UTSNamespace) Clone(userns *auth.UserNamespace) *UTSNamespace { + u.mu.Lock() + defer u.mu.Unlock() + return &UTSNamespace{ + hostName: u.hostName, + domainName: u.domainName, + userns: userns, + } +} diff --git a/pkg/sentry/kernel/vdso.go b/pkg/sentry/kernel/vdso.go new file mode 100644 index 000000000..0bacbea49 --- /dev/null +++ b/pkg/sentry/kernel/vdso.go @@ -0,0 +1,145 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "fmt" + + "gvisor.googlesource.com/gvisor/pkg/binary" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" +) + +// vdsoParams are the parameters exposed to the VDSO. +// +// They are exposed to the VDSO via a parameter page managed by VDSOParamPage, +// which also includes a sequence counter. +type vdsoParams struct { + monotonicReady uint64 + monotonicBaseCycles int64 + monotonicBaseRef int64 + monotonicFrequency uint64 + + realtimeReady uint64 + realtimeBaseCycles int64 + realtimeBaseRef int64 + realtimeFrequency uint64 +} + +// VDSOParamPage manages a VDSO parameter page. +// +// Its memory layout looks like: +// +// type page struct { +// // seq is a sequence counter that protects the fields below. +// seq uint64 +// vdsoParams +// } +// +// Everything in the struct is 8 bytes for easy alignment. +// +// It must be kept in sync with params in vdso/vdso_time.cc. +type VDSOParamPage struct { + // The parameter page is fr, allocated from platform.Memory(). + platform platform.Platform + fr platform.FileRange + + // seq is the current sequence count written to the page. + // + // A write is in progress if bit 1 of the counter is set. + // + // Timekeeper's updater goroutine may call Write before equality is + // checked in state_test_util tests, causing this field to change across + // save / restore. + seq uint64 +} + +// NewVDSOParamPage returns a VDSOParamPage. +// +// Preconditions: +// +// * fr is a single page allocated from platform.Memory(). VDSOParamPage does +// not take ownership of fr; it must remain allocated for the lifetime of the +// VDSOParamPage. +// +// * VDSOParamPage must be the only writer to fr. +// +// * platform.Memory().MapInternal(fr) must return a single safemem.Block. +func NewVDSOParamPage(platform platform.Platform, fr platform.FileRange) *VDSOParamPage { + return &VDSOParamPage{platform: platform, fr: fr} +} + +// access returns a mapping of the param page. +func (v *VDSOParamPage) access() (safemem.Block, error) { + bs, err := v.platform.Memory().MapInternal(v.fr, usermem.ReadWrite) + if err != nil { + return safemem.Block{}, err + } + if bs.NumBlocks() != 1 { + panic(fmt.Sprintf("Multiple blocks (%d) in VDSO param BlockSeq", bs.NumBlocks())) + } + return bs.Head(), nil +} + +// incrementSeq increments the sequence counter in the param page. +func (v *VDSOParamPage) incrementSeq(paramPage safemem.Block) error { + next := v.seq + 1 + old, err := safemem.SwapUint64(paramPage, next) + if err != nil { + return err + } + + if old != v.seq { + return fmt.Errorf("unexpected VDSOParamPage seq value: got %d expected %d. Application may hang or get incorrect time from the VDSO.", old, v.seq) + } + + v.seq = next + return nil +} + +// Write updates the VDSO parameters. +// +// Write starts a write block, calls f to get the new parameters, writes +// out the new parameters, then ends the write block. +func (v *VDSOParamPage) Write(f func() vdsoParams) error { + paramPage, err := v.access() + if err != nil { + return err + } + + // Write begin. + next := v.seq + 1 + if next%2 != 1 { + panic("Out-of-order sequence count") + } + + err = v.incrementSeq(paramPage) + if err != nil { + return err + } + + // Get the new params. + p := f() + buf := binary.Marshal(nil, usermem.ByteOrder, p) + + // Skip the sequence counter. + if _, err := safemem.Copy(paramPage.DropFirst(8), safemem.BlockFromSafeSlice(buf)); err != nil { + panic(fmt.Sprintf("Unable to get set VDSO parameters: %v", err)) + } + + // Write end. + return v.incrementSeq(paramPage) +} diff --git a/pkg/sentry/kernel/version.go b/pkg/sentry/kernel/version.go new file mode 100644 index 000000000..a9e84673f --- /dev/null +++ b/pkg/sentry/kernel/version.go @@ -0,0 +1,33 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +// Version defines the application-visible system version. +type Version struct { + // Operating system name (e.g. "Linux"). + Sysname string + + // Operating system release (e.g. "3.11.10-amd64"). + Release string + + // Operating system version. On Linux this takes the shape + // "#VERSION CONFIG_FLAGS TIMESTAMP" + // where: + // - VERSION is a sequence counter incremented on every successful build + // - CONFIG_FLAGS is a space-separated list of major enabled kernel features + // (e.g. "SMP" and "PREEMPT") + // - TIMESTAMP is the build timestamp as returned by `date` + Version string +} -- cgit v1.2.3 From 3d3deef573a54e031cb98038b9f617f5fac31044 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Tue, 1 May 2018 22:11:07 -0700 Subject: Implement SO_TIMESTAMP PiperOrigin-RevId: 195047018 Change-Id: I6d99528a00a2125f414e1e51e067205289ec9d3d --- pkg/dhcp/client.go | 4 +- pkg/dhcp/dhcp_test.go | 2 +- pkg/dhcp/server.go | 2 +- pkg/sentry/fs/host/socket_test.go | 2 +- pkg/sentry/kernel/kernel.go | 9 +++ pkg/sentry/socket/BUILD | 1 + pkg/sentry/socket/control/control.go | 35 +++++++++++ pkg/sentry/socket/epsocket/epsocket.go | 69 +++++++++++++++------- pkg/sentry/socket/hostinet/socket.go | 10 ++-- pkg/sentry/socket/netlink/socket.go | 16 ++--- pkg/sentry/socket/rpcinet/socket.go | 20 +++---- pkg/sentry/socket/socket.go | 12 +++- pkg/sentry/socket/unix/unix.go | 14 ++--- pkg/sentry/strace/socket.go | 29 ++++++++- pkg/sentry/syscalls/linux/sys_socket.go | 21 ++++--- pkg/tcpip/adapters/gonet/gonet.go | 4 +- pkg/tcpip/adapters/gonet/gonet_test.go | 2 +- pkg/tcpip/network/arp/arp_test.go | 2 +- pkg/tcpip/network/ipv4/icmp_test.go | 2 +- pkg/tcpip/sample/tun_tcp_connect/main.go | 4 +- pkg/tcpip/sample/tun_tcp_echo/main.go | 4 +- pkg/tcpip/stack/stack.go | 17 +++++- pkg/tcpip/stack/stack_test.go | 22 +++---- pkg/tcpip/stack/transport_test.go | 16 ++--- pkg/tcpip/tcpip.go | 48 +++++++++++++-- pkg/tcpip/transport/tcp/endpoint.go | 22 +++---- pkg/tcpip/transport/tcp/tcp_test.go | 46 +++++++-------- pkg/tcpip/transport/tcp/tcp_timestamp_test.go | 4 +- pkg/tcpip/transport/tcp/testing/context/context.go | 2 +- pkg/tcpip/transport/udp/endpoint.go | 37 ++++++++++-- pkg/tcpip/transport/udp/udp_test.go | 10 ++-- runsc/boot/loader.go | 7 ++- 32 files changed, 345 insertions(+), 150 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/dhcp/client.go b/pkg/dhcp/client.go index 9a4fd7ae4..37deb69ff 100644 --- a/pkg/dhcp/client.go +++ b/pkg/dhcp/client.go @@ -162,7 +162,7 @@ func (c *Client) Request(ctx context.Context, requestedAddr tcpip.Address) error // DHCPOFFER for { var addr tcpip.FullAddress - v, err := epin.Read(&addr) + v, _, err := epin.Read(&addr) if err == tcpip.ErrWouldBlock { select { case <-ch: @@ -216,7 +216,7 @@ func (c *Client) Request(ctx context.Context, requestedAddr tcpip.Address) error // DHCPACK for { var addr tcpip.FullAddress - v, err := epin.Read(&addr) + v, _, err := epin.Read(&addr) if err == tcpip.ErrWouldBlock { select { case <-ch: diff --git a/pkg/dhcp/dhcp_test.go b/pkg/dhcp/dhcp_test.go index d56b93997..ed884fcb6 100644 --- a/pkg/dhcp/dhcp_test.go +++ b/pkg/dhcp/dhcp_test.go @@ -36,7 +36,7 @@ func TestDHCP(t *testing.T) { } }() - s := stack.New([]string{ipv4.ProtocolName}, []string{udp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName}, []string{udp.ProtocolName}) const nicid tcpip.NICID = 1 if err := s.CreateNIC(nicid, id); err != nil { diff --git a/pkg/dhcp/server.go b/pkg/dhcp/server.go index d132d90b4..8816203a8 100644 --- a/pkg/dhcp/server.go +++ b/pkg/dhcp/server.go @@ -104,7 +104,7 @@ func (s *Server) reader(ctx context.Context) { for { var addr tcpip.FullAddress - v, err := s.ep.Read(&addr) + v, _, err := s.ep.Read(&addr) if err == tcpip.ErrWouldBlock { select { case <-ch: diff --git a/pkg/sentry/fs/host/socket_test.go b/pkg/sentry/fs/host/socket_test.go index 80c46dcfa..9b73c5173 100644 --- a/pkg/sentry/fs/host/socket_test.go +++ b/pkg/sentry/fs/host/socket_test.go @@ -142,7 +142,7 @@ func TestSocketSendMsgLen0(t *testing.T) { defer sfile.DecRef() s := sfile.FileOperations.(socket.Socket) - n, terr := s.SendMsg(nil, usermem.BytesIOSequence(nil), []byte{}, 0, unix.ControlMessages{}) + n, terr := s.SendMsg(nil, usermem.BytesIOSequence(nil), []byte{}, 0, socket.ControlMessages{}) if n != 0 { t.Fatalf("socket sendmsg() failed: %v wrote: %d", terr, n) } diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 0932965e0..25c8dd885 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -887,6 +887,15 @@ func (k *Kernel) SetExitError(err error) { } } +// NowNanoseconds implements tcpip.Clock.NowNanoseconds. +func (k *Kernel) NowNanoseconds() int64 { + now, err := k.timekeeper.GetTime(sentrytime.Realtime) + if err != nil { + panic("Kernel.NowNanoseconds: " + err.Error()) + } + return now +} + // SupervisorContext returns a Context with maximum privileges in k. It should // only be used by goroutines outside the control of the emulated kernel // defined by e. diff --git a/pkg/sentry/socket/BUILD b/pkg/sentry/socket/BUILD index 87e32df37..5500a676e 100644 --- a/pkg/sentry/socket/BUILD +++ b/pkg/sentry/socket/BUILD @@ -32,6 +32,7 @@ go_library( "//pkg/sentry/usermem", "//pkg/state", "//pkg/syserr", + "//pkg/tcpip", "//pkg/tcpip/transport/unix", ], ) diff --git a/pkg/sentry/socket/control/control.go b/pkg/sentry/socket/control/control.go index cb34cbc85..17ecdd11c 100644 --- a/pkg/sentry/socket/control/control.go +++ b/pkg/sentry/socket/control/control.go @@ -208,6 +208,31 @@ func putCmsg(buf []byte, msgType uint32, align uint, data []int32) []byte { return alignSlice(buf, align) } +func putCmsgStruct(buf []byte, msgType uint32, align uint, data interface{}) []byte { + if cap(buf)-len(buf) < linux.SizeOfControlMessageHeader { + return buf + } + ob := buf + + buf = putUint64(buf, uint64(linux.SizeOfControlMessageHeader)) + buf = putUint32(buf, linux.SOL_SOCKET) + buf = putUint32(buf, msgType) + + hdrBuf := buf + + buf = binary.Marshal(buf, usermem.ByteOrder, data) + + // Check if we went over. + if cap(buf) != cap(ob) { + return hdrBuf + } + + // Fix up length. + putUint64(ob, uint64(len(buf)-len(ob))) + + return alignSlice(buf, align) +} + // Credentials implements SCMCredentials.Credentials. func (c *scmCredentials) Credentials(t *kernel.Task) (kernel.ThreadID, auth.UID, auth.GID) { // "When a process's user and group IDs are passed over a UNIX domain @@ -261,6 +286,16 @@ func alignSlice(buf []byte, align uint) []byte { return buf[:aligned] } +// PackTimestamp packs a SO_TIMESTAMP socket control message. +func PackTimestamp(t *kernel.Task, timestamp int64, buf []byte) []byte { + return putCmsgStruct( + buf, + linux.SO_TIMESTAMP, + t.Arch().Width(), + linux.NsecToTimeval(timestamp), + ) +} + // Parse parses a raw socket control message into portable objects. func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (unix.ControlMessages, error) { var ( diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index 3fc3ea58f..5701ecfac 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -109,6 +109,7 @@ type SocketOperations struct { // readMu protects access to readView, control, and sender. readMu sync.Mutex `state:"nosave"` readView buffer.View + readCM tcpip.ControlMessages sender tcpip.FullAddress } @@ -210,12 +211,13 @@ func (s *SocketOperations) fetchReadView() *syserr.Error { s.readView = nil s.sender = tcpip.FullAddress{} - v, err := s.Endpoint.Read(&s.sender) + v, cms, err := s.Endpoint.Read(&s.sender) if err != nil { return syserr.TranslateNetstackError(err) } s.readView = v + s.readCM = cms return nil } @@ -230,7 +232,7 @@ func (s *SocketOperations) Read(ctx context.Context, _ *fs.File, dst usermem.IOS if dst.NumBytes() == 0 { return 0, nil } - n, _, _, err := s.nonBlockingRead(ctx, dst, false, false, false) + n, _, _, _, err := s.nonBlockingRead(ctx, dst, false, false, false) if err == syserr.ErrWouldBlock { return int64(n), syserror.ErrWouldBlock } @@ -552,6 +554,18 @@ func GetSockOpt(t *kernel.Task, s socket.Socket, ep commonEndpoint, family int, } return linux.NsecToTimeval(s.RecvTimeout()), nil + + case linux.SO_TIMESTAMP: + if outLen < sizeOfInt32 { + return nil, syserr.ErrInvalidArgument + } + + var v tcpip.TimestampOption + if err := ep.GetSockOpt(&v); err != nil { + return nil, syserr.TranslateNetstackError(err) + } + + return int32(v), nil } case syscall.SOL_TCP: @@ -659,6 +673,14 @@ func SetSockOpt(t *kernel.Task, s socket.Socket, ep commonEndpoint, level int, n binary.Unmarshal(optVal[:linux.SizeOfTimeval], usermem.ByteOrder, &v) s.SetRecvTimeout(v.ToNsecCapped()) return nil + + case linux.SO_TIMESTAMP: + if len(optVal) < sizeOfInt32 { + return syserr.ErrInvalidArgument + } + + v := usermem.ByteOrder.Uint32(optVal) + return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.TimestampOption(v))) } case syscall.SOL_TCP: @@ -823,7 +845,9 @@ func (s *SocketOperations) coalescingRead(ctx context.Context, dst usermem.IOSeq } // nonBlockingRead issues a non-blocking read. -func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSequence, peek, trunc, senderRequested bool) (int, interface{}, uint32, *syserr.Error) { +// +// TODO: Support timestamps for stream sockets. +func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSequence, peek, trunc, senderRequested bool) (int, interface{}, uint32, socket.ControlMessages, *syserr.Error) { isPacket := s.isPacketBased() // Fast path for regular reads from stream (e.g., TCP) endpoints. Note @@ -839,14 +863,14 @@ func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSe s.readMu.Lock() n, err := s.coalescingRead(ctx, dst, trunc) s.readMu.Unlock() - return n, nil, 0, err + return n, nil, 0, socket.ControlMessages{}, err } s.readMu.Lock() defer s.readMu.Unlock() if err := s.fetchReadView(); err != nil { - return 0, nil, 0, err + return 0, nil, 0, socket.ControlMessages{}, err } if !isPacket && peek && trunc { @@ -854,14 +878,14 @@ func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSe // amount that could be read. var rql tcpip.ReceiveQueueSizeOption if err := s.Endpoint.GetSockOpt(&rql); err != nil { - return 0, nil, 0, syserr.TranslateNetstackError(err) + return 0, nil, 0, socket.ControlMessages{}, syserr.TranslateNetstackError(err) } available := len(s.readView) + int(rql) bufLen := int(dst.NumBytes()) if available < bufLen { - return available, nil, 0, nil + return available, nil, 0, socket.ControlMessages{}, nil } - return bufLen, nil, 0, nil + return bufLen, nil, 0, socket.ControlMessages{}, nil } n, err := dst.CopyOut(ctx, s.readView) @@ -874,17 +898,18 @@ func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSe if peek { if l := len(s.readView); trunc && l > n { // isPacket must be true. - return l, addr, addrLen, syserr.FromError(err) + return l, addr, addrLen, socket.ControlMessages{IP: s.readCM}, syserr.FromError(err) } if isPacket || err != nil { - return int(n), addr, addrLen, syserr.FromError(err) + return int(n), addr, addrLen, socket.ControlMessages{IP: s.readCM}, syserr.FromError(err) } // We need to peek beyond the first message. dst = dst.DropFirst(n) num, err := dst.CopyOutFrom(ctx, safemem.FromVecReaderFunc{func(dsts [][]byte) (int64, error) { - n, err := s.Endpoint.Peek(dsts) + n, _, err := s.Endpoint.Peek(dsts) + // TODO: Handle peek timestamp. if err != nil { return int64(n), syserr.TranslateNetstackError(err).ToError() } @@ -895,7 +920,7 @@ func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSe // We got some data, so no need to return an error. err = nil } - return int(n), nil, 0, syserr.FromError(err) + return int(n), nil, 0, socket.ControlMessages{IP: s.readCM}, syserr.FromError(err) } var msgLen int @@ -908,15 +933,15 @@ func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSe } if trunc { - return msgLen, addr, addrLen, syserr.FromError(err) + return msgLen, addr, addrLen, socket.ControlMessages{IP: s.readCM}, syserr.FromError(err) } - return int(n), addr, addrLen, syserr.FromError(err) + return int(n), addr, addrLen, socket.ControlMessages{IP: s.readCM}, syserr.FromError(err) } // RecvMsg implements the linux syscall recvmsg(2) for sockets backed by // tcpip.Endpoint. -func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (n int, senderAddr interface{}, senderAddrLen uint32, controlMessages unix.ControlMessages, err *syserr.Error) { +func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (n int, senderAddr interface{}, senderAddrLen uint32, controlMessages socket.ControlMessages, err *syserr.Error) { trunc := flags&linux.MSG_TRUNC != 0 peek := flags&linux.MSG_PEEK != 0 @@ -924,7 +949,7 @@ func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags // Stream sockets ignore the sender address. senderRequested = false } - n, senderAddr, senderAddrLen, err = s.nonBlockingRead(t, dst, peek, trunc, senderRequested) + n, senderAddr, senderAddrLen, controlMessages, err = s.nonBlockingRead(t, dst, peek, trunc, senderRequested) if err != syserr.ErrWouldBlock || flags&linux.MSG_DONTWAIT != 0 { return } @@ -936,25 +961,25 @@ func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags defer s.EventUnregister(&e) for { - n, senderAddr, senderAddrLen, err = s.nonBlockingRead(t, dst, peek, trunc, senderRequested) + n, senderAddr, senderAddrLen, controlMessages, err = s.nonBlockingRead(t, dst, peek, trunc, senderRequested) if err != syserr.ErrWouldBlock { return } if err := t.BlockWithDeadline(ch, haveDeadline, deadline); err != nil { if err == syserror.ETIMEDOUT { - return 0, nil, 0, unix.ControlMessages{}, syserr.ErrTryAgain + return 0, nil, 0, socket.ControlMessages{}, syserr.ErrTryAgain } - return 0, nil, 0, unix.ControlMessages{}, syserr.FromError(err) + return 0, nil, 0, socket.ControlMessages{}, syserr.FromError(err) } } } // SendMsg implements the linux syscall sendmsg(2) for sockets backed by // tcpip.Endpoint. -func (s *SocketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages unix.ControlMessages) (int, *syserr.Error) { - // Reject control messages. - if !controlMessages.Empty() { +func (s *SocketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages socket.ControlMessages) (int, *syserr.Error) { + // Reject Unix control messages. + if !controlMessages.Unix.Empty() { return 0, syserr.ErrInvalidArgument } diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index defa3db2c..02fad1c60 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -57,6 +57,8 @@ type socketOperations struct { queue waiter.Queue } +var _ = socket.Socket(&socketOperations{}) + func newSocketFile(ctx context.Context, fd int, nonblock bool) (*fs.File, *syserr.Error) { s := &socketOperations{fd: fd} if err := fdnotifier.AddFD(int32(fd), &s.queue); err != nil { @@ -339,14 +341,14 @@ func (s *socketOperations) SetSockOpt(t *kernel.Task, level int, name int, opt [ } // RecvMsg implements socket.Socket.RecvMsg. -func (s *socketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (int, interface{}, uint32, unix.ControlMessages, *syserr.Error) { +func (s *socketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (int, interface{}, uint32, socket.ControlMessages, *syserr.Error) { // Whitelist flags. // // FIXME: We can't support MSG_ERRQUEUE because it uses ancillary // messages that netstack/tcpip/transport/unix doesn't understand. Kill the // Socket interface's dependence on netstack. if flags&^(syscall.MSG_DONTWAIT|syscall.MSG_PEEK|syscall.MSG_TRUNC) != 0 { - return 0, nil, 0, unix.ControlMessages{}, syserr.ErrInvalidArgument + return 0, nil, 0, socket.ControlMessages{}, syserr.ErrInvalidArgument } var senderAddr []byte @@ -411,11 +413,11 @@ func (s *socketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags } } - return int(n), senderAddr, uint32(len(senderAddr)), unix.ControlMessages{}, syserr.FromError(err) + return int(n), senderAddr, uint32(len(senderAddr)), socket.ControlMessages{}, syserr.FromError(err) } // SendMsg implements socket.Socket.SendMsg. -func (s *socketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages unix.ControlMessages) (int, *syserr.Error) { +func (s *socketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages socket.ControlMessages) (int, *syserr.Error) { // Whitelist flags. if flags&^(syscall.MSG_DONTWAIT|syscall.MSG_EOR|syscall.MSG_FASTOPEN|syscall.MSG_MORE|syscall.MSG_NOSIGNAL) != 0 { return 0, syserr.ErrInvalidArgument diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index 2d0e59ceb..0b8f528d0 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -305,7 +305,7 @@ func (s *Socket) GetPeerName(t *kernel.Task) (interface{}, uint32, *syserr.Error } // RecvMsg implements socket.Socket.RecvMsg. -func (s *Socket) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (int, interface{}, uint32, unix.ControlMessages, *syserr.Error) { +func (s *Socket) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (int, interface{}, uint32, socket.ControlMessages, *syserr.Error) { from := linux.SockAddrNetlink{ Family: linux.AF_NETLINK, PortID: 0, @@ -323,7 +323,7 @@ func (s *Socket) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, have if trunc { n = int64(r.MsgSize) } - return int(n), from, fromLen, unix.ControlMessages{}, syserr.FromError(err) + return int(n), from, fromLen, socket.ControlMessages{}, syserr.FromError(err) } // We'll have to block. Register for notification and keep trying to @@ -337,14 +337,14 @@ func (s *Socket) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, have if trunc { n = int64(r.MsgSize) } - return int(n), from, fromLen, unix.ControlMessages{}, syserr.FromError(err) + return int(n), from, fromLen, socket.ControlMessages{}, syserr.FromError(err) } if err := t.BlockWithDeadline(ch, haveDeadline, deadline); err != nil { if err == syserror.ETIMEDOUT { - return 0, nil, 0, unix.ControlMessages{}, syserr.ErrTryAgain + return 0, nil, 0, socket.ControlMessages{}, syserr.ErrTryAgain } - return 0, nil, 0, unix.ControlMessages{}, syserr.FromError(err) + return 0, nil, 0, socket.ControlMessages{}, syserr.FromError(err) } } } @@ -459,7 +459,7 @@ func (s *Socket) processMessages(ctx context.Context, buf []byte) *syserr.Error } // sendMsg is the core of message send, used for SendMsg and Write. -func (s *Socket) sendMsg(ctx context.Context, src usermem.IOSequence, to []byte, flags int, controlMessages unix.ControlMessages) (int, *syserr.Error) { +func (s *Socket) sendMsg(ctx context.Context, src usermem.IOSequence, to []byte, flags int, controlMessages socket.ControlMessages) (int, *syserr.Error) { dstPort := int32(0) if len(to) != 0 { @@ -506,12 +506,12 @@ func (s *Socket) sendMsg(ctx context.Context, src usermem.IOSequence, to []byte, } // SendMsg implements socket.Socket.SendMsg. -func (s *Socket) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages unix.ControlMessages) (int, *syserr.Error) { +func (s *Socket) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages socket.ControlMessages) (int, *syserr.Error) { return s.sendMsg(t, src, to, flags, controlMessages) } // Write implements fs.FileOperations.Write. func (s *Socket) Write(ctx context.Context, _ *fs.File, src usermem.IOSequence, _ int64) (int64, error) { - n, err := s.sendMsg(ctx, src, nil, 0, unix.ControlMessages{}) + n, err := s.sendMsg(ctx, src, nil, 0, socket.ControlMessages{}) return int64(n), err.ToError() } diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index 574d99ba5..15047df01 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -402,7 +402,7 @@ func rpcRecvMsg(t *kernel.Task, req *pb.SyscallRequest_Recvmsg) (*pb.RecvmsgResp } // RecvMsg implements socket.Socket.RecvMsg. -func (s *socketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (int, interface{}, uint32, unix.ControlMessages, *syserr.Error) { +func (s *socketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (int, interface{}, uint32, socket.ControlMessages, *syserr.Error) { req := &pb.SyscallRequest_Recvmsg{&pb.RecvmsgRequest{ Fd: s.fd, Length: uint32(dst.NumBytes()), @@ -414,10 +414,10 @@ func (s *socketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags res, err := rpcRecvMsg(t, req) if err == nil { n, e := dst.CopyOut(t, res.Data) - return int(n), res.Address.GetAddress(), res.Address.GetLength(), unix.ControlMessages{}, syserr.FromError(e) + return int(n), res.Address.GetAddress(), res.Address.GetLength(), socket.ControlMessages{}, syserr.FromError(e) } if err != syserr.ErrWouldBlock || flags&linux.MSG_DONTWAIT != 0 { - return 0, nil, 0, unix.ControlMessages{}, err + return 0, nil, 0, socket.ControlMessages{}, err } // We'll have to block. Register for notifications and keep trying to @@ -430,17 +430,17 @@ func (s *socketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags res, err := rpcRecvMsg(t, req) if err == nil { n, e := dst.CopyOut(t, res.Data) - return int(n), res.Address.GetAddress(), res.Address.GetLength(), unix.ControlMessages{}, syserr.FromError(e) + return int(n), res.Address.GetAddress(), res.Address.GetLength(), socket.ControlMessages{}, syserr.FromError(e) } if err != syserr.ErrWouldBlock { - return 0, nil, 0, unix.ControlMessages{}, err + return 0, nil, 0, socket.ControlMessages{}, err } if err := t.BlockWithDeadline(ch, haveDeadline, deadline); err != nil { if err == syserror.ETIMEDOUT { - return 0, nil, 0, unix.ControlMessages{}, syserr.ErrTryAgain + return 0, nil, 0, socket.ControlMessages{}, syserr.ErrTryAgain } - return 0, nil, 0, unix.ControlMessages{}, syserr.FromError(err) + return 0, nil, 0, socket.ControlMessages{}, syserr.FromError(err) } } } @@ -459,14 +459,14 @@ func rpcSendMsg(t *kernel.Task, req *pb.SyscallRequest_Sendmsg) (uint32, *syserr } // SendMsg implements socket.Socket.SendMsg. -func (s *socketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages unix.ControlMessages) (int, *syserr.Error) { +func (s *socketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages socket.ControlMessages) (int, *syserr.Error) { // Whitelist flags. if flags&^(syscall.MSG_DONTWAIT|syscall.MSG_EOR|syscall.MSG_FASTOPEN|syscall.MSG_MORE|syscall.MSG_NOSIGNAL) != 0 { return 0, syserr.ErrInvalidArgument } - // Reject control messages. - if !controlMessages.Empty() { + // Reject Unix control messages. + if !controlMessages.Unix.Empty() { return 0, syserr.ErrInvalidArgument } diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index be3026bfa..bd4858a34 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -31,9 +31,17 @@ import ( ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" + "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) +// ControlMessages represents the union of unix control messages and tcpip +// control messages. +type ControlMessages struct { + Unix unix.ControlMessages + IP tcpip.ControlMessages +} + // Socket is the interface containing socket syscalls used by the syscall layer // to redirect them to the appropriate implementation. type Socket interface { @@ -78,11 +86,11 @@ type Socket interface { // // senderAddrLen is the address length to be returned to the application, // not necessarily the actual length of the address. - RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (n int, senderAddr interface{}, senderAddrLen uint32, controlMessages unix.ControlMessages, err *syserr.Error) + RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (n int, senderAddr interface{}, senderAddrLen uint32, controlMessages ControlMessages, err *syserr.Error) // SendMsg implements the sendmsg(2) linux syscall. SendMsg does not take // ownership of the ControlMessage on error. - SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages unix.ControlMessages) (n int, err *syserr.Error) + SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages ControlMessages) (n int, err *syserr.Error) // SetRecvTimeout sets the timeout (in ns) for recv operations. Zero means // no timeout. diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index a4b414851..f83156c8e 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -358,10 +358,10 @@ func (s *SocketOperations) Write(ctx context.Context, _ *fs.File, src usermem.IO // SendMsg implements the linux syscall sendmsg(2) for unix sockets backed by // a unix.Endpoint. -func (s *SocketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages unix.ControlMessages) (int, *syserr.Error) { +func (s *SocketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages socket.ControlMessages) (int, *syserr.Error) { w := EndpointWriter{ Endpoint: s.ep, - Control: controlMessages, + Control: controlMessages.Unix, To: nil, } if len(to) > 0 { @@ -452,7 +452,7 @@ func (s *SocketOperations) Read(ctx context.Context, _ *fs.File, dst usermem.IOS // RecvMsg implements the linux syscall recvmsg(2) for sockets backed by // a unix.Endpoint. -func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (n int, senderAddr interface{}, senderAddrLen uint32, controlMessages unix.ControlMessages, err *syserr.Error) { +func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (n int, senderAddr interface{}, senderAddrLen uint32, controlMessages socket.ControlMessages, err *syserr.Error) { trunc := flags&linux.MSG_TRUNC != 0 peek := flags&linux.MSG_PEEK != 0 @@ -490,7 +490,7 @@ func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags if trunc { n = int64(r.MsgSize) } - return int(n), from, fromLen, r.Control, syserr.FromError(err) + return int(n), from, fromLen, socket.ControlMessages{Unix: r.Control}, syserr.FromError(err) } // We'll have to block. Register for notification and keep trying to @@ -509,14 +509,14 @@ func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags if trunc { n = int64(r.MsgSize) } - return int(n), from, fromLen, r.Control, syserr.FromError(err) + return int(n), from, fromLen, socket.ControlMessages{Unix: r.Control}, syserr.FromError(err) } if err := t.BlockWithDeadline(ch, haveDeadline, deadline); err != nil { if err == syserror.ETIMEDOUT { - return 0, nil, 0, unix.ControlMessages{}, syserr.ErrTryAgain + return 0, nil, 0, socket.ControlMessages{}, syserr.ErrTryAgain } - return 0, nil, 0, unix.ControlMessages{}, syserr.FromError(err) + return 0, nil, 0, socket.ControlMessages{}, syserr.FromError(err) } } } diff --git a/pkg/sentry/strace/socket.go b/pkg/sentry/strace/socket.go index 48c072e96..1a2e8573e 100644 --- a/pkg/sentry/strace/socket.go +++ b/pkg/sentry/strace/socket.go @@ -440,6 +440,7 @@ var SocketProtocol = map[int32]abi.ValueSet{ var controlMessageType = map[int32]string{ linux.SCM_RIGHTS: "SCM_RIGHTS", linux.SCM_CREDENTIALS: "SCM_CREDENTIALS", + linux.SO_TIMESTAMP: "SO_TIMESTAMP", } func cmsghdr(t *kernel.Task, addr usermem.Addr, length uint64, maxBytes uint64) string { @@ -477,7 +478,7 @@ func cmsghdr(t *kernel.Task, addr usermem.Addr, length uint64, maxBytes uint64) typ = fmt.Sprint(h.Type) } - if h.Length > uint64(len(buf)-i) { + if h.Length > uint64(len(buf)-i+linux.SizeOfControlMessageHeader) { strs = append(strs, fmt.Sprintf( "{level=%s, type=%s, length=%d, content extends beyond buffer}", level, @@ -546,6 +547,32 @@ func cmsghdr(t *kernel.Task, addr usermem.Addr, length uint64, maxBytes uint64) i += control.AlignUp(length, width) + case linux.SO_TIMESTAMP: + if length < linux.SizeOfTimeval { + strs = append(strs, fmt.Sprintf( + "{level=%s, type=%s, length=%d, content too short}", + level, + typ, + h.Length, + )) + i += control.AlignUp(length, width) + break + } + + var tv linux.Timeval + binary.Unmarshal(buf[i:i+linux.SizeOfTimeval], usermem.ByteOrder, &tv) + + strs = append(strs, fmt.Sprintf( + "{level=%s, type=%s, length=%d, Sec: %d, Usec: %d}", + level, + typ, + h.Length, + tv.Sec, + tv.Usec, + )) + + i += control.AlignUp(length, width) + default: panic("unreachable") } diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go index 70c618398..6258a1539 100644 --- a/pkg/sentry/syscalls/linux/sys_socket.go +++ b/pkg/sentry/syscalls/linux/sys_socket.go @@ -731,10 +731,11 @@ func recvSingleMsg(t *kernel.Task, s socket.Socket, msgPtr usermem.Addr, flags i // Fast path when no control message nor name buffers are provided. if msg.ControlLen == 0 && msg.NameLen == 0 { - n, _, _, _, err := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, false, 0) + n, _, _, cms, err := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, false, 0) if err != nil { return 0, syserror.ConvertIntr(err.ToError(), kernel.ERESTARTSYS) } + cms.Unix.Release() return uintptr(n), nil } @@ -745,17 +746,21 @@ func recvSingleMsg(t *kernel.Task, s socket.Socket, msgPtr usermem.Addr, flags i if e != nil { return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) } - defer cms.Release() + defer cms.Unix.Release() controlData := make([]byte, 0, msg.ControlLen) if cr, ok := s.(unix.Credentialer); ok && cr.Passcred() { - creds, _ := cms.Credentials.(control.SCMCredentials) + creds, _ := cms.Unix.Credentials.(control.SCMCredentials) controlData = control.PackCredentials(t, creds, controlData) } - if cms.Rights != nil { - controlData = control.PackRights(t, cms.Rights.(control.SCMRights), flags&linux.MSG_CMSG_CLOEXEC != 0, controlData) + if cms.IP.HasTimestamp { + controlData = control.PackTimestamp(t, cms.IP.Timestamp, controlData) + } + + if cms.Unix.Rights != nil { + controlData = control.PackRights(t, cms.Unix.Rights.(control.SCMRights), flags&linux.MSG_CMSG_CLOEXEC != 0, controlData) } // Copy the address to the caller. @@ -823,7 +828,7 @@ func recvFrom(t *kernel.Task, fd kdefs.FD, bufPtr usermem.Addr, bufLen uint64, f } n, sender, senderLen, cm, e := s.RecvMsg(t, dst, int(flags), haveDeadline, deadline, nameLenPtr != 0, 0) - cm.Release() + cm.Unix.Release() if e != nil { return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) } @@ -997,7 +1002,7 @@ func sendSingleMsg(t *kernel.Task, s socket.Socket, file *fs.File, msgPtr userme } // Call the syscall implementation. - n, e := s.SendMsg(t, src, to, int(flags), controlMessages) + n, e := s.SendMsg(t, src, to, int(flags), socket.ControlMessages{Unix: controlMessages}) err = handleIOError(t, n != 0, e.ToError(), kernel.ERESTARTSYS, "sendmsg", file) if err != nil { controlMessages.Release() @@ -1048,7 +1053,7 @@ func sendTo(t *kernel.Task, fd kdefs.FD, bufPtr usermem.Addr, bufLen uint64, fla } // Call the syscall implementation. - n, e := s.SendMsg(t, src, to, int(flags), control.New(t, s, nil)) + n, e := s.SendMsg(t, src, to, int(flags), socket.ControlMessages{Unix: control.New(t, s, nil)}) return uintptr(n), handleIOError(t, n != 0, e.ToError(), kernel.ERESTARTSYS, "sendto", file) } diff --git a/pkg/tcpip/adapters/gonet/gonet.go b/pkg/tcpip/adapters/gonet/gonet.go index 96a2d670d..5aa6b1aa2 100644 --- a/pkg/tcpip/adapters/gonet/gonet.go +++ b/pkg/tcpip/adapters/gonet/gonet.go @@ -268,7 +268,7 @@ type opErrorer interface { // commonRead implements the common logic between net.Conn.Read and // net.PacketConn.ReadFrom. func commonRead(ep tcpip.Endpoint, wq *waiter.Queue, deadline <-chan struct{}, addr *tcpip.FullAddress, errorer opErrorer) ([]byte, error) { - read, err := ep.Read(addr) + read, _, err := ep.Read(addr) if err == tcpip.ErrWouldBlock { // Create wait queue entry that notifies a channel. @@ -276,7 +276,7 @@ func commonRead(ep tcpip.Endpoint, wq *waiter.Queue, deadline <-chan struct{}, a wq.EventRegister(&waitEntry, waiter.EventIn) defer wq.EventUnregister(&waitEntry) for { - read, err = ep.Read(addr) + read, _, err = ep.Read(addr) if err != tcpip.ErrWouldBlock { break } diff --git a/pkg/tcpip/adapters/gonet/gonet_test.go b/pkg/tcpip/adapters/gonet/gonet_test.go index 2f86469eb..e3d0c6c84 100644 --- a/pkg/tcpip/adapters/gonet/gonet_test.go +++ b/pkg/tcpip/adapters/gonet/gonet_test.go @@ -47,7 +47,7 @@ func TestTimeouts(t *testing.T) { func newLoopbackStack() (*stack.Stack, *tcpip.Error) { // Create the stack and add a NIC. - s := stack.New([]string{ipv4.ProtocolName, ipv6.ProtocolName}, []string{tcp.ProtocolName, udp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName, ipv6.ProtocolName}, []string{tcp.ProtocolName, udp.ProtocolName}) if err := s.CreateNIC(NICID, loopback.New()); err != nil { return nil, err diff --git a/pkg/tcpip/network/arp/arp_test.go b/pkg/tcpip/network/arp/arp_test.go index 91ffdce4b..47b10e64e 100644 --- a/pkg/tcpip/network/arp/arp_test.go +++ b/pkg/tcpip/network/arp/arp_test.go @@ -32,7 +32,7 @@ type testContext struct { } func newTestContext(t *testing.T) *testContext { - s := stack.New([]string{ipv4.ProtocolName, arp.ProtocolName}, []string{ipv4.PingProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName, arp.ProtocolName}, []string{ipv4.PingProtocolName}) const defaultMTU = 65536 id, linkEP := channel.New(256, defaultMTU, stackLinkAddr) diff --git a/pkg/tcpip/network/ipv4/icmp_test.go b/pkg/tcpip/network/ipv4/icmp_test.go index 378fba74b..c55aa1835 100644 --- a/pkg/tcpip/network/ipv4/icmp_test.go +++ b/pkg/tcpip/network/ipv4/icmp_test.go @@ -26,7 +26,7 @@ type testContext struct { } func newTestContext(t *testing.T) *testContext { - s := stack.New([]string{ipv4.ProtocolName}, []string{ipv4.PingProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName}, []string{ipv4.PingProtocolName}) const defaultMTU = 65536 id, linkEP := channel.New(256, defaultMTU, "") diff --git a/pkg/tcpip/sample/tun_tcp_connect/main.go b/pkg/tcpip/sample/tun_tcp_connect/main.go index 332929c85..ef5c7ec60 100644 --- a/pkg/tcpip/sample/tun_tcp_connect/main.go +++ b/pkg/tcpip/sample/tun_tcp_connect/main.go @@ -113,7 +113,7 @@ func main() { // Create the stack with ipv4 and tcp protocols, then add a tun-based // NIC and ipv4 address. - s := stack.New([]string{ipv4.ProtocolName}, []string{tcp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName}, []string{tcp.ProtocolName}) mtu, err := rawfile.GetMTU(tunName) if err != nil { @@ -183,7 +183,7 @@ func main() { // connection from its side. wq.EventRegister(&waitEntry, waiter.EventIn) for { - v, err := ep.Read(nil) + v, _, err := ep.Read(nil) if err != nil { if err == tcpip.ErrClosedForReceive { break diff --git a/pkg/tcpip/sample/tun_tcp_echo/main.go b/pkg/tcpip/sample/tun_tcp_echo/main.go index 10cd701af..8c166f643 100644 --- a/pkg/tcpip/sample/tun_tcp_echo/main.go +++ b/pkg/tcpip/sample/tun_tcp_echo/main.go @@ -42,7 +42,7 @@ func echo(wq *waiter.Queue, ep tcpip.Endpoint) { defer wq.EventUnregister(&waitEntry) for { - v, err := ep.Read(nil) + v, _, err := ep.Read(nil) if err != nil { if err == tcpip.ErrWouldBlock { <-notifyCh @@ -99,7 +99,7 @@ func main() { // Create the stack with ip and tcp protocols, then add a tun-based // NIC and address. - s := stack.New([]string{ipv4.ProtocolName, ipv6.ProtocolName, arp.ProtocolName}, []string{tcp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName, ipv6.ProtocolName, arp.ProtocolName}, []string{tcp.ProtocolName}) mtu, err := rawfile.GetMTU(tunName) if err != nil { diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index 558ecdb72..b480bf812 100644 --- a/pkg/tcpip/stack/stack.go +++ b/pkg/tcpip/stack/stack.go @@ -270,6 +270,9 @@ type Stack struct { // If not nil, then any new endpoints will have this probe function // invoked everytime they receive a TCP segment. tcpProbeFunc TCPProbeFunc + + // clock is used to generate user-visible times. + clock tcpip.Clock } // New allocates a new networking stack with only the requested networking and @@ -279,7 +282,7 @@ type Stack struct { // SetNetworkProtocolOption/SetTransportProtocolOption methods provided by the // stack. Please refer to individual protocol implementations as to what options // are supported. -func New(network []string, transport []string) *Stack { +func New(clock tcpip.Clock, network []string, transport []string) *Stack { s := &Stack{ transportProtocols: make(map[tcpip.TransportProtocolNumber]*transportProtocolState), networkProtocols: make(map[tcpip.NetworkProtocolNumber]NetworkProtocol), @@ -287,6 +290,7 @@ func New(network []string, transport []string) *Stack { nics: make(map[tcpip.NICID]*NIC), linkAddrCache: newLinkAddrCache(ageLimit, resolutionTimeout, resolutionAttempts), PortManager: ports.NewPortManager(), + clock: clock, } // Add specified network protocols. @@ -388,6 +392,11 @@ func (s *Stack) SetTransportProtocolHandler(p tcpip.TransportProtocolNumber, h f } } +// NowNanoseconds implements tcpip.Clock.NowNanoseconds. +func (s *Stack) NowNanoseconds() int64 { + return s.clock.NowNanoseconds() +} + // Stats returns a snapshot of the current stats. // // NOTE: The underlying stats are updated using atomic instructions as a result @@ -474,6 +483,12 @@ func (s *Stack) CreateDisabledNIC(id tcpip.NICID, linkEP tcpip.LinkEndpointID) * return s.createNIC(id, "", linkEP, false) } +// CreateDisabledNamedNIC is a combination of CreateNamedNIC and +// CreateDisabledNIC. +func (s *Stack) CreateDisabledNamedNIC(id tcpip.NICID, name string, linkEP tcpip.LinkEndpointID) *tcpip.Error { + return s.createNIC(id, name, linkEP, false) +} + // EnableNIC enables the given NIC so that the link-layer endpoint can start // delivering packets to it. func (s *Stack) EnableNIC(id tcpip.NICID) *tcpip.Error { diff --git a/pkg/tcpip/stack/stack_test.go b/pkg/tcpip/stack/stack_test.go index b416065d7..ea7dccdc2 100644 --- a/pkg/tcpip/stack/stack_test.go +++ b/pkg/tcpip/stack/stack_test.go @@ -176,7 +176,7 @@ func TestNetworkReceive(t *testing.T) { // Create a stack with the fake network protocol, one nic, and two // addresses attached to it: 1 & 2. id, linkEP := channel.New(10, defaultMTU, "") - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) if err := s.CreateNIC(1, id); err != nil { t.Fatalf("CreateNIC failed: %v", err) } @@ -270,7 +270,7 @@ func TestNetworkSend(t *testing.T) { // address: 1. The route table sends all packets through the only // existing nic. id, linkEP := channel.New(10, defaultMTU, "") - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) if err := s.CreateNIC(1, id); err != nil { t.Fatalf("NewNIC failed: %v", err) } @@ -292,7 +292,7 @@ func TestNetworkSendMultiRoute(t *testing.T) { // Create a stack with the fake network protocol, two nics, and two // addresses per nic, the first nic has odd address, the second one has // even addresses. - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) id1, linkEP1 := channel.New(10, defaultMTU, "") if err := s.CreateNIC(1, id1); err != nil { @@ -371,7 +371,7 @@ func TestRoutes(t *testing.T) { // Create a stack with the fake network protocol, two nics, and two // addresses per nic, the first nic has odd address, the second one has // even addresses. - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) id1, _ := channel.New(10, defaultMTU, "") if err := s.CreateNIC(1, id1); err != nil { @@ -435,7 +435,7 @@ func TestRoutes(t *testing.T) { } func TestAddressRemoval(t *testing.T) { - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) id, linkEP := channel.New(10, defaultMTU, "") if err := s.CreateNIC(1, id); err != nil { @@ -479,7 +479,7 @@ func TestAddressRemoval(t *testing.T) { } func TestDelayedRemovalDueToRoute(t *testing.T) { - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) id, linkEP := channel.New(10, defaultMTU, "") if err := s.CreateNIC(1, id); err != nil { @@ -547,7 +547,7 @@ func TestDelayedRemovalDueToRoute(t *testing.T) { } func TestPromiscuousMode(t *testing.T) { - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) id, linkEP := channel.New(10, defaultMTU, "") if err := s.CreateNIC(1, id); err != nil { @@ -607,7 +607,7 @@ func TestAddressSpoofing(t *testing.T) { srcAddr := tcpip.Address("\x01") dstAddr := tcpip.Address("\x02") - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) id, _ := channel.New(10, defaultMTU, "") if err := s.CreateNIC(1, id); err != nil { @@ -648,7 +648,7 @@ func TestAddressSpoofing(t *testing.T) { // Set the subnet, then check that packet is delivered. func TestSubnetAcceptsMatchingPacket(t *testing.T) { - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) id, linkEP := channel.New(10, defaultMTU, "") if err := s.CreateNIC(1, id); err != nil { @@ -682,7 +682,7 @@ func TestSubnetAcceptsMatchingPacket(t *testing.T) { // Set destination outside the subnet, then check it doesn't get delivered. func TestSubnetRejectsNonmatchingPacket(t *testing.T) { - s := stack.New([]string{"fakeNet"}, nil) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, nil) id, linkEP := channel.New(10, defaultMTU, "") if err := s.CreateNIC(1, id); err != nil { @@ -714,7 +714,7 @@ func TestSubnetRejectsNonmatchingPacket(t *testing.T) { } func TestNetworkOptions(t *testing.T) { - s := stack.New([]string{"fakeNet"}, []string{}) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, []string{}) // Try an unsupported network protocol. if err := s.SetNetworkProtocolOption(tcpip.NetworkProtocolNumber(99999), fakeNetGoodOption(false)); err != tcpip.ErrUnknownProtocol { diff --git a/pkg/tcpip/stack/transport_test.go b/pkg/tcpip/stack/transport_test.go index 7e072e96e..b870ab375 100644 --- a/pkg/tcpip/stack/transport_test.go +++ b/pkg/tcpip/stack/transport_test.go @@ -46,8 +46,8 @@ func (*fakeTransportEndpoint) Readiness(mask waiter.EventMask) waiter.EventMask return mask } -func (*fakeTransportEndpoint) Read(*tcpip.FullAddress) (buffer.View, *tcpip.Error) { - return buffer.View{}, nil +func (*fakeTransportEndpoint) Read(*tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) { + return buffer.View{}, tcpip.ControlMessages{}, nil } func (f *fakeTransportEndpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, *tcpip.Error) { @@ -67,8 +67,8 @@ func (f *fakeTransportEndpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) return uintptr(len(v)), nil } -func (f *fakeTransportEndpoint) Peek([][]byte) (uintptr, *tcpip.Error) { - return 0, nil +func (f *fakeTransportEndpoint) Peek([][]byte) (uintptr, tcpip.ControlMessages, *tcpip.Error) { + return 0, tcpip.ControlMessages{}, nil } // SetSockOpt sets a socket option. Currently not supported. @@ -210,7 +210,7 @@ func (f *fakeTransportProtocol) Option(option interface{}) *tcpip.Error { func TestTransportReceive(t *testing.T) { id, linkEP := channel.New(10, defaultMTU, "") - s := stack.New([]string{"fakeNet"}, []string{"fakeTrans"}) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, []string{"fakeTrans"}) if err := s.CreateNIC(1, id); err != nil { t.Fatalf("CreateNIC failed: %v", err) } @@ -270,7 +270,7 @@ func TestTransportReceive(t *testing.T) { func TestTransportControlReceive(t *testing.T) { id, linkEP := channel.New(10, defaultMTU, "") - s := stack.New([]string{"fakeNet"}, []string{"fakeTrans"}) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, []string{"fakeTrans"}) if err := s.CreateNIC(1, id); err != nil { t.Fatalf("CreateNIC failed: %v", err) } @@ -336,7 +336,7 @@ func TestTransportControlReceive(t *testing.T) { func TestTransportSend(t *testing.T) { id, _ := channel.New(10, defaultMTU, "") - s := stack.New([]string{"fakeNet"}, []string{"fakeTrans"}) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, []string{"fakeTrans"}) if err := s.CreateNIC(1, id); err != nil { t.Fatalf("CreateNIC failed: %v", err) } @@ -373,7 +373,7 @@ func TestTransportSend(t *testing.T) { } func TestTransportOptions(t *testing.T) { - s := stack.New([]string{"fakeNet"}, []string{"fakeTrans"}) + s := stack.New(&tcpip.StdClock{}, []string{"fakeNet"}, []string{"fakeTrans"}) // Try an unsupported transport protocol. if err := s.SetTransportProtocolOption(tcpip.TransportProtocolNumber(99999), fakeTransportGoodOption(false)); err != tcpip.ErrUnknownProtocol { diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index f3a94f353..f9df1d989 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -23,6 +23,7 @@ import ( "fmt" "strconv" "strings" + "time" "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" "gvisor.googlesource.com/gvisor/pkg/waiter" @@ -80,6 +81,24 @@ var ( errSubnetAddressMasked = errors.New("subnet address has bits set outside the mask") ) +// A Clock provides the current time. +// +// Times returned by a Clock should always be used for application-visible +// time, but never for netstack internal timekeeping. +type Clock interface { + // NowNanoseconds returns the current real time as a number of + // nanoseconds since some epoch. + NowNanoseconds() int64 +} + +// StdClock implements Clock with the time package. +type StdClock struct{} + +// NowNanoseconds implements Clock.NowNanoseconds. +func (*StdClock) NowNanoseconds() int64 { + return time.Now().UnixNano() +} + // Address is a byte slice cast as a string that represents the address of a // network node. Or, in the case of unix endpoints, it may represent a path. type Address string @@ -210,6 +229,16 @@ func (s SlicePayload) Size() int { return len(s) } +// A ControlMessages contains socket control messages for IP sockets. +type ControlMessages struct { + // HasTimestamp indicates whether Timestamp is valid/set. + HasTimestamp bool + + // Timestamp is the time (in ns) that the last packed used to create + // the read data was received. + Timestamp int64 +} + // Endpoint is the interface implemented by transport protocols (e.g., tcp, udp) // that exposes functionality like read, write, connect, etc. to users of the // networking stack. @@ -219,9 +248,13 @@ type Endpoint interface { Close() // Read reads data from the endpoint and optionally returns the sender. - // This method does not block if there is no data pending. - // It will also either return an error or data, never both. - Read(*FullAddress) (buffer.View, *Error) + // + // This method does not block if there is no data pending. It will also + // either return an error or data, never both. + // + // A timestamp (in ns) is optionally returned. A zero value indicates + // that no timestamp was available. + Read(*FullAddress) (buffer.View, ControlMessages, *Error) // Write writes data to the endpoint's peer. This method does not block if // the data cannot be written. @@ -238,7 +271,10 @@ type Endpoint interface { // Peek reads data without consuming it from the endpoint. // // This method does not block if there is no data pending. - Peek([][]byte) (uintptr, *Error) + // + // A timestamp (in ns) is optionally returned. A zero value indicates + // that no timestamp was available. + Peek([][]byte) (uintptr, ControlMessages, *Error) // Connect connects the endpoint to its peer. Specifying a NIC is // optional. @@ -347,6 +383,10 @@ type ReuseAddressOption int // Only supported on Unix sockets. type PasscredOption int +// TimestampOption is used by SetSockOpt/GetSockOpt to specify whether +// SO_TIMESTAMP socket control messages are enabled. +type TimestampOption int + // TCPInfoOption is used by GetSockOpt to expose TCP statistics. // // TODO: Add and populate stat fields. diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index 5d62589d8..d84171b0c 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -374,7 +374,7 @@ func (e *endpoint) cleanup() { } // Read reads data from the endpoint. -func (e *endpoint) Read(*tcpip.FullAddress) (buffer.View, *tcpip.Error) { +func (e *endpoint) Read(*tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) { e.mu.RLock() // The endpoint can be read if it's connected, or if it's already closed // but has some pending unread data. Also note that a RST being received @@ -383,9 +383,9 @@ func (e *endpoint) Read(*tcpip.FullAddress) (buffer.View, *tcpip.Error) { if s := e.state; s != stateConnected && s != stateClosed && e.rcvBufUsed == 0 { e.mu.RUnlock() if s == stateError { - return buffer.View{}, e.hardError + return buffer.View{}, tcpip.ControlMessages{}, e.hardError } - return buffer.View{}, tcpip.ErrInvalidEndpointState + return buffer.View{}, tcpip.ControlMessages{}, tcpip.ErrInvalidEndpointState } e.rcvListMu.Lock() @@ -394,7 +394,7 @@ func (e *endpoint) Read(*tcpip.FullAddress) (buffer.View, *tcpip.Error) { e.mu.RUnlock() - return v, err + return v, tcpip.ControlMessages{}, err } func (e *endpoint) readLocked() (buffer.View, *tcpip.Error) { @@ -498,7 +498,7 @@ func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, *tc // Peek reads data without consuming it from the endpoint. // // This method does not block if there is no data pending. -func (e *endpoint) Peek(vec [][]byte) (uintptr, *tcpip.Error) { +func (e *endpoint) Peek(vec [][]byte) (uintptr, tcpip.ControlMessages, *tcpip.Error) { e.mu.RLock() defer e.mu.RUnlock() @@ -506,9 +506,9 @@ func (e *endpoint) Peek(vec [][]byte) (uintptr, *tcpip.Error) { // but has some pending unread data. if s := e.state; s != stateConnected && s != stateClosed { if s == stateError { - return 0, e.hardError + return 0, tcpip.ControlMessages{}, e.hardError } - return 0, tcpip.ErrInvalidEndpointState + return 0, tcpip.ControlMessages{}, tcpip.ErrInvalidEndpointState } e.rcvListMu.Lock() @@ -516,9 +516,9 @@ func (e *endpoint) Peek(vec [][]byte) (uintptr, *tcpip.Error) { if e.rcvBufUsed == 0 { if e.rcvClosed || e.state != stateConnected { - return 0, tcpip.ErrClosedForReceive + return 0, tcpip.ControlMessages{}, tcpip.ErrClosedForReceive } - return 0, tcpip.ErrWouldBlock + return 0, tcpip.ControlMessages{}, tcpip.ErrWouldBlock } // Make a copy of vec so we can modify the slide headers. @@ -534,7 +534,7 @@ func (e *endpoint) Peek(vec [][]byte) (uintptr, *tcpip.Error) { for len(v) > 0 { if len(vec) == 0 { - return num, nil + return num, tcpip.ControlMessages{}, nil } if len(vec[0]) == 0 { vec = vec[1:] @@ -549,7 +549,7 @@ func (e *endpoint) Peek(vec [][]byte) (uintptr, *tcpip.Error) { } } - return num, nil + return num, tcpip.ControlMessages{}, nil } // zeroReceiveWindow checks if the receive window to be announced now would be diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go index 118d861ba..3c21a1ec3 100644 --- a/pkg/tcpip/transport/tcp/tcp_test.go +++ b/pkg/tcpip/transport/tcp/tcp_test.go @@ -147,7 +147,7 @@ func TestSimpleReceive(t *testing.T) { c.WQ.EventRegister(&we, waiter.EventIn) defer c.WQ.EventUnregister(&we) - if _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { + if _, _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { t.Fatalf("Unexpected error from Read: %v", err) } @@ -169,7 +169,7 @@ func TestSimpleReceive(t *testing.T) { } // Receive data. - v, err := c.EP.Read(nil) + v, _, err := c.EP.Read(nil) if err != nil { t.Fatalf("Unexpected error from Read: %v", err) } @@ -199,7 +199,7 @@ func TestOutOfOrderReceive(t *testing.T) { c.WQ.EventRegister(&we, waiter.EventIn) defer c.WQ.EventUnregister(&we) - if _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { + if _, _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { t.Fatalf("Unexpected error from Read: %v", err) } @@ -226,7 +226,7 @@ func TestOutOfOrderReceive(t *testing.T) { // Wait 200ms and check that no data has been received. time.Sleep(200 * time.Millisecond) - if _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { + if _, _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { t.Fatalf("Unexpected error from Read: %v", err) } @@ -243,7 +243,7 @@ func TestOutOfOrderReceive(t *testing.T) { // Receive data. read := make([]byte, 0, 6) for len(read) < len(data) { - v, err := c.EP.Read(nil) + v, _, err := c.EP.Read(nil) if err != nil { if err == tcpip.ErrWouldBlock { // Wait for receive to be notified. @@ -284,7 +284,7 @@ func TestOutOfOrderFlood(t *testing.T) { opt := tcpip.ReceiveBufferSizeOption(10) c.CreateConnected(789, 30000, &opt) - if _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { + if _, _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { t.Fatalf("Unexpected error from Read: %v", err) } @@ -361,7 +361,7 @@ func TestRstOnCloseWithUnreadData(t *testing.T) { c.WQ.EventRegister(&we, waiter.EventIn) defer c.WQ.EventUnregister(&we) - if _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { + if _, _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { t.Fatalf("Unexpected error from Read: %v", err) } @@ -414,7 +414,7 @@ func TestFullWindowReceive(t *testing.T) { c.WQ.EventRegister(&we, waiter.EventIn) defer c.WQ.EventUnregister(&we) - _, err := c.EP.Read(nil) + _, _, err := c.EP.Read(nil) if err != tcpip.ErrWouldBlock { t.Fatalf("Unexpected error from Read: %v", err) } @@ -449,7 +449,7 @@ func TestFullWindowReceive(t *testing.T) { ) // Receive data and check it. - v, err := c.EP.Read(nil) + v, _, err := c.EP.Read(nil) if err != nil { t.Fatalf("Unexpected error from Read: %v", err) } @@ -487,7 +487,7 @@ func TestNoWindowShrinking(t *testing.T) { c.WQ.EventRegister(&we, waiter.EventIn) defer c.WQ.EventUnregister(&we) - _, err := c.EP.Read(nil) + _, _, err := c.EP.Read(nil) if err != tcpip.ErrWouldBlock { t.Fatalf("Unexpected error from Read: %v", err) } @@ -551,7 +551,7 @@ func TestNoWindowShrinking(t *testing.T) { // Receive data and check it. read := make([]byte, 0, 10) for len(read) < len(data) { - v, err := c.EP.Read(nil) + v, _, err := c.EP.Read(nil) if err != nil { t.Fatalf("Unexpected error from Read: %v", err) } @@ -954,7 +954,7 @@ func TestZeroScaledWindowReceive(t *testing.T) { } // Read some data. An ack should be sent in response to that. - v, err := c.EP.Read(nil) + v, _, err := c.EP.Read(nil) if err != nil { t.Fatalf("Unexpected error from Read: %v", err) } @@ -1337,7 +1337,7 @@ func TestReceiveOnResetConnection(t *testing.T) { loop: for { - switch _, err := c.EP.Read(nil); err { + switch _, _, err := c.EP.Read(nil); err { case nil: t.Fatalf("Unexpected success.") case tcpip.ErrWouldBlock: @@ -2293,7 +2293,7 @@ func TestReadAfterClosedState(t *testing.T) { c.WQ.EventRegister(&we, waiter.EventIn) defer c.WQ.EventUnregister(&we) - if _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { + if _, _, err := c.EP.Read(nil); err != tcpip.ErrWouldBlock { t.Fatalf("Unexpected error from Read: %v", err) } @@ -2345,7 +2345,7 @@ func TestReadAfterClosedState(t *testing.T) { // Check that peek works. peekBuf := make([]byte, 10) - n, err := c.EP.Peek([][]byte{peekBuf}) + n, _, err := c.EP.Peek([][]byte{peekBuf}) if err != nil { t.Fatalf("Unexpected error from Peek: %v", err) } @@ -2356,7 +2356,7 @@ func TestReadAfterClosedState(t *testing.T) { } // Receive data. - v, err := c.EP.Read(nil) + v, _, err := c.EP.Read(nil) if err != nil { t.Fatalf("Unexpected error from Read: %v", err) } @@ -2367,11 +2367,11 @@ func TestReadAfterClosedState(t *testing.T) { // Now that we drained the queue, check that functions fail with the // right error code. - if _, err := c.EP.Read(nil); err != tcpip.ErrClosedForReceive { + if _, _, err := c.EP.Read(nil); err != tcpip.ErrClosedForReceive { t.Fatalf("Unexpected return from Read: got %v, want %v", err, tcpip.ErrClosedForReceive) } - if _, err := c.EP.Peek([][]byte{peekBuf}); err != tcpip.ErrClosedForReceive { + if _, _, err := c.EP.Peek([][]byte{peekBuf}); err != tcpip.ErrClosedForReceive { t.Fatalf("Unexpected return from Peek: got %v, want %v", err, tcpip.ErrClosedForReceive) } } @@ -2479,7 +2479,7 @@ func checkSendBufferSize(t *testing.T, ep tcpip.Endpoint, v int) { } func TestDefaultBufferSizes(t *testing.T) { - s := stack.New([]string{ipv4.ProtocolName}, []string{tcp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName}, []string{tcp.ProtocolName}) // Check the default values. ep, err := s.NewEndpoint(tcp.ProtocolNumber, ipv4.ProtocolNumber, &waiter.Queue{}) @@ -2525,7 +2525,7 @@ func TestDefaultBufferSizes(t *testing.T) { } func TestMinMaxBufferSizes(t *testing.T) { - s := stack.New([]string{ipv4.ProtocolName}, []string{tcp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName}, []string{tcp.ProtocolName}) // Check the default values. ep, err := s.NewEndpoint(tcp.ProtocolNumber, ipv4.ProtocolNumber, &waiter.Queue{}) @@ -2575,7 +2575,7 @@ func TestSelfConnect(t *testing.T) { // it checks that if an endpoint binds to say 127.0.0.1:1000 then // connects to 127.0.0.1:1000, then it will be connected to itself, and // is able to send and receive data through the same endpoint. - s := stack.New([]string{ipv4.ProtocolName}, []string{tcp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName}, []string{tcp.ProtocolName}) id := loopback.New() if testing.Verbose() { @@ -2637,13 +2637,13 @@ func TestSelfConnect(t *testing.T) { // Read back what was written. wq.EventUnregister(&waitEntry) wq.EventRegister(&waitEntry, waiter.EventIn) - rd, err := ep.Read(nil) + rd, _, err := ep.Read(nil) if err != nil { if err != tcpip.ErrWouldBlock { t.Fatalf("Read failed: %v", err) } <-notifyCh - rd, err = ep.Read(nil) + rd, _, err = ep.Read(nil) if err != nil { t.Fatalf("Read failed: %v", err) } diff --git a/pkg/tcpip/transport/tcp/tcp_timestamp_test.go b/pkg/tcpip/transport/tcp/tcp_timestamp_test.go index d12081bb7..335262e43 100644 --- a/pkg/tcpip/transport/tcp/tcp_timestamp_test.go +++ b/pkg/tcpip/transport/tcp/tcp_timestamp_test.go @@ -95,7 +95,7 @@ func TestTimeStampEnabledConnect(t *testing.T) { // There should be 5 views to read and each of them should // contain the same data. for i := 0; i < 5; i++ { - got, err := c.EP.Read(nil) + got, _, err := c.EP.Read(nil) if err != nil { t.Fatalf("Unexpected error from Read: %v", err) } @@ -296,7 +296,7 @@ func TestSegmentDropWhenTimestampMissing(t *testing.T) { } // Issue a read and we should data. - got, err := c.EP.Read(nil) + got, _, err := c.EP.Read(nil) if err != nil { t.Fatalf("Unexpected error from Read: %v", err) } diff --git a/pkg/tcpip/transport/tcp/testing/context/context.go b/pkg/tcpip/transport/tcp/testing/context/context.go index 6a402d150..eb928553f 100644 --- a/pkg/tcpip/transport/tcp/testing/context/context.go +++ b/pkg/tcpip/transport/tcp/testing/context/context.go @@ -129,7 +129,7 @@ type Context struct { // New allocates and initializes a test context containing a new // stack and a link-layer endpoint. func New(t *testing.T, mtu uint32) *Context { - s := stack.New([]string{ipv4.ProtocolName, ipv6.ProtocolName}, []string{tcp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName, ipv6.ProtocolName}, []string{tcp.ProtocolName}) // Allow minimum send/receive buffer sizes to be 1 during tests. if err := s.SetTransportProtocolOption(tcp.ProtocolNumber, tcp.SendBufferSizeOption{1, tcp.DefaultBufferSize, tcp.DefaultBufferSize * 10}); err != nil { diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 80fa88c4c..f86fc6d5a 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -19,6 +19,8 @@ type udpPacket struct { udpPacketEntry senderAddress tcpip.FullAddress data buffer.VectorisedView `state:".(buffer.VectorisedView)"` + timestamp int64 + hasTimestamp bool // views is used as buffer for data when its length is large // enough to store a VectorisedView. views [8]buffer.View `state:"nosave"` @@ -52,6 +54,7 @@ type endpoint struct { rcvBufSizeMax int `state:".(int)"` rcvBufSize int rcvClosed bool + rcvTimestamp bool // The following fields are protected by the mu mutex. mu sync.RWMutex `state:"nosave"` @@ -134,7 +137,7 @@ func (e *endpoint) Close() { // Read reads data from the endpoint. This method does not block if // there is no data pending. -func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, *tcpip.Error) { +func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, tcpip.ControlMessages, *tcpip.Error) { e.rcvMu.Lock() if e.rcvList.Empty() { @@ -143,12 +146,13 @@ func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, *tcpip.Error) { err = tcpip.ErrClosedForReceive } e.rcvMu.Unlock() - return buffer.View{}, err + return buffer.View{}, tcpip.ControlMessages{}, err } p := e.rcvList.Front() e.rcvList.Remove(p) e.rcvBufSize -= p.data.Size() + ts := e.rcvTimestamp e.rcvMu.Unlock() @@ -156,7 +160,12 @@ func (e *endpoint) Read(addr *tcpip.FullAddress) (buffer.View, *tcpip.Error) { *addr = p.senderAddress } - return p.data.ToView(), nil + if ts && !p.hasTimestamp { + // Linux uses the current time. + p.timestamp = e.stack.NowNanoseconds() + } + + return p.data.ToView(), tcpip.ControlMessages{HasTimestamp: ts, Timestamp: p.timestamp}, nil } // prepareForWrite prepares the endpoint for sending data. In particular, it @@ -299,8 +308,8 @@ func (e *endpoint) Write(p tcpip.Payload, opts tcpip.WriteOptions) (uintptr, *tc } // Peek only returns data from a single datagram, so do nothing here. -func (e *endpoint) Peek([][]byte) (uintptr, *tcpip.Error) { - return 0, nil +func (e *endpoint) Peek([][]byte) (uintptr, tcpip.ControlMessages, *tcpip.Error) { + return 0, tcpip.ControlMessages{}, nil } // SetSockOpt sets a socket option. Currently not supported. @@ -322,6 +331,11 @@ func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error { } e.v6only = v != 0 + + case tcpip.TimestampOption: + e.rcvMu.Lock() + e.rcvTimestamp = v != 0 + e.rcvMu.Unlock() } return nil } @@ -370,6 +384,14 @@ func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error { } e.rcvMu.Unlock() return nil + + case *tcpip.TimestampOption: + e.rcvMu.Lock() + *o = 0 + if e.rcvTimestamp { + *o = 1 + } + e.rcvMu.Unlock() } return tcpip.ErrUnknownProtocolOption @@ -733,6 +755,11 @@ func (e *endpoint) HandlePacket(r *stack.Route, id stack.TransportEndpointID, vv e.rcvList.PushBack(pkt) e.rcvBufSize += vv.Size() + if e.rcvTimestamp { + pkt.timestamp = e.stack.NowNanoseconds() + pkt.hasTimestamp = true + } + e.rcvMu.Unlock() // Notify any waiters that there's data to be read now. diff --git a/pkg/tcpip/transport/udp/udp_test.go b/pkg/tcpip/transport/udp/udp_test.go index 65c567952..1eb9ecb80 100644 --- a/pkg/tcpip/transport/udp/udp_test.go +++ b/pkg/tcpip/transport/udp/udp_test.go @@ -56,7 +56,7 @@ type headers struct { } func newDualTestContext(t *testing.T, mtu uint32) *testContext { - s := stack.New([]string{ipv4.ProtocolName, ipv6.ProtocolName}, []string{udp.ProtocolName}) + s := stack.New(&tcpip.StdClock{}, []string{ipv4.ProtocolName, ipv6.ProtocolName}, []string{udp.ProtocolName}) id, linkEP := channel.New(256, mtu, "") if testing.Verbose() { @@ -260,12 +260,12 @@ func testV4Read(c *testContext) { defer c.wq.EventUnregister(&we) var addr tcpip.FullAddress - v, err := c.ep.Read(&addr) + v, _, err := c.ep.Read(&addr) if err == tcpip.ErrWouldBlock { // Wait for data to become available. select { case <-ch: - v, err = c.ep.Read(&addr) + v, _, err = c.ep.Read(&addr) if err != nil { c.t.Fatalf("Read failed: %v", err) } @@ -355,12 +355,12 @@ func TestV6ReadOnV6(t *testing.T) { defer c.wq.EventUnregister(&we) var addr tcpip.FullAddress - v, err := c.ep.Read(&addr) + v, _, err := c.ep.Read(&addr) if err == tcpip.ErrWouldBlock { // Wait for data to become available. select { case <-ch: - v, err = c.ep.Read(&addr) + v, _, err = c.ep.Read(&addr) if err != nil { c.t.Fatalf("Read failed: %v", err) } diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index a470cb054..d63a9028e 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -37,6 +37,7 @@ import ( slinux "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/time" "gvisor.googlesource.com/gvisor/pkg/sentry/watchdog" + "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/tcpip/link/sniffer" "gvisor.googlesource.com/gvisor/pkg/tcpip/network/arp" "gvisor.googlesource.com/gvisor/pkg/tcpip/network/ipv4" @@ -177,7 +178,7 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console // this point. Netns is configured before Run() is called. Netstack is // configured using a control uRPC message. Host network is configured inside // Run(). - networkStack := newEmptyNetworkStack(conf) + networkStack := newEmptyNetworkStack(conf, k) // Initiate the Kernel object, which is required by the Context passed // to createVFS in order to mount (among other things) procfs. @@ -337,7 +338,7 @@ func (l *Loader) WaitExit() kernel.ExitStatus { return l.k.GlobalInit().ExitStatus() } -func newEmptyNetworkStack(conf *Config) inet.Stack { +func newEmptyNetworkStack(conf *Config, clock tcpip.Clock) inet.Stack { switch conf.Network { case NetworkHost: return hostinet.NewStack() @@ -346,7 +347,7 @@ func newEmptyNetworkStack(conf *Config) inet.Stack { // NetworkNone sets up loopback using netstack. netProtos := []string{ipv4.ProtocolName, ipv6.ProtocolName, arp.ProtocolName} protoNames := []string{tcp.ProtocolName, udp.ProtocolName} - return &epsocket.Stack{stack.New(netProtos, protoNames)} + return &epsocket.Stack{stack.New(clock, netProtos, protoNames)} default: panic(fmt.Sprintf("invalid network configuration: %v", conf.Network)) -- cgit v1.2.3 From 2264ce7f62994fa0476f8e1eb9fe497d60547bda Mon Sep 17 00:00:00 2001 From: Zhengyu He Date: Wed, 2 May 2018 03:42:54 -0700 Subject: Use png for the run states diagram PiperOrigin-RevId: 195071508 Change-Id: I63314bf7529560e4c779ef07cc9399ad8d53f0a2 --- pkg/sentry/kernel/README.md | 2 +- pkg/sentry/kernel/g3doc/run_states.png | Bin 0 -> 234152 bytes 2 files changed, 1 insertion(+), 1 deletion(-) create mode 100644 pkg/sentry/kernel/g3doc/run_states.png (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/README.md b/pkg/sentry/kernel/README.md index 3306780d6..88760a9bb 100644 --- a/pkg/sentry/kernel/README.md +++ b/pkg/sentry/kernel/README.md @@ -87,7 +87,7 @@ kept separate from the main "app" state to reduce the size of the latter. 4. `SyscallReinvoke`, which does not correspond to anything in Linux, and exists solely to serve the autosave feature. -![dot -Tsvg -Goverlap=false -orun_states.svg run_states.dot](g3doc/run_states.dot "Task control flow graph") +![dot -Tpng -Goverlap=false -orun_states.png run_states.dot](g3doc/run_states.png "Task control flow graph") States before which a stop may occur are represented as implementations of the `taskRunState` interface named `run(state)`, allowing them to be saved and diff --git a/pkg/sentry/kernel/g3doc/run_states.png b/pkg/sentry/kernel/g3doc/run_states.png new file mode 100644 index 000000000..b63b60f02 Binary files /dev/null and b/pkg/sentry/kernel/g3doc/run_states.png differ -- cgit v1.2.3 From 04b79137babed361fb227e3ad579adb2df4bb188 Mon Sep 17 00:00:00 2001 From: Cyrille Hemidy Date: Thu, 3 May 2018 14:05:25 -0700 Subject: Fix misspellings. PiperOrigin-RevId: 195307689 Change-Id: I499f19af49875a43214797d63376f20ae788d2f4 --- pkg/log/log.go | 2 +- pkg/sentry/fs/file.go | 2 +- pkg/sentry/fs/fsutil/file.go | 2 +- pkg/sentry/fs/ramfs/dir.go | 4 ++-- pkg/sentry/fs/tty/line_discipline.go | 4 ++-- pkg/sentry/kernel/semaphore/semaphore.go | 2 +- pkg/sentry/kernel/task_exit.go | 2 +- pkg/sentry/mm/vma.go | 2 +- pkg/sentry/socket/rpcinet/socket.go | 2 +- pkg/sentry/strace/syscalls.go | 2 +- pkg/tcpip/header/ipv6.go | 2 +- pkg/tcpip/stack/stack.go | 2 +- pkg/tcpip/transport/tcp/snd.go | 2 +- runsc/sandbox/sandbox.go | 4 ++-- 14 files changed, 17 insertions(+), 17 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/log/log.go b/pkg/log/log.go index 110e0e196..cdfc0601a 100644 --- a/pkg/log/log.go +++ b/pkg/log/log.go @@ -149,7 +149,7 @@ func (t TestEmitter) Emit(level Level, timestamp time.Time, format string, v ... // Logger is a high-level logging interface. It is in fact, not used within the // log package. Rather it is provided for others to provide contextual loggers // that may append some addition information to log statement. BasicLogger -// satifies this interface, and may be passed around as a Logger. +// satisfies this interface, and may be passed around as a Logger. type Logger interface { // Debugf logs a debug statement. Debugf(format string, v ...interface{}) diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index de2e80bf0..f2683bbd2 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -376,7 +376,7 @@ func (r *FileReader) Read(buf []byte) (int, error) { return int(n), err } -// ReadAt implementes io.Reader.ReadAt. +// ReadAt implements io.Reader.ReadAt. func (r *FileReader) ReadAt(buf []byte, offset int64) (int, error) { n, err := r.File.Preadv(r.Ctx, usermem.BytesIOSequence(buf), offset) return int(n), err diff --git a/pkg/sentry/fs/fsutil/file.go b/pkg/sentry/fs/fsutil/file.go index a7329f1c9..b17f11a5a 100644 --- a/pkg/sentry/fs/fsutil/file.go +++ b/pkg/sentry/fs/fsutil/file.go @@ -34,7 +34,7 @@ func (NoopRelease) Release() {} // SeekWithDirCursor is used to implement fs.FileOperations.Seek. If dirCursor // is not nil and the seek was on a directory, the cursor will be updated. // -// Currenly only seeking to 0 on a directory is supported. +// Currently only seeking to 0 on a directory is supported. // // FIXME: Lift directory seeking limitations. func SeekWithDirCursor(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64, dirCursor *string) (int64, error) { diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index bf4cd8dfd..19d5612ed 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -103,7 +103,7 @@ func (d *Dir) addChildLocked(name string, inode *fs.Inode) { } // Given we're now adding this inode to the directory we must also - // increase its link count. Similiarly we decremented it in removeChildLocked. + // increase its link count. Similarly we decremented it in removeChildLocked. inode.AddLink() } @@ -144,7 +144,7 @@ func (d *Dir) removeChildLocked(ctx context.Context, name string) (*fs.Inode, er inode.NotifyStatusChange(ctx) // Given we're now removing this inode to the directory we must also - // decrease its link count. Similiarly it is increased in addChildLocked. + // decrease its link count. Similarly it is increased in addChildLocked. inode.DropLink() return inode, nil diff --git a/pkg/sentry/fs/tty/line_discipline.go b/pkg/sentry/fs/tty/line_discipline.go index fde4e7941..a3aa95ece 100644 --- a/pkg/sentry/fs/tty/line_discipline.go +++ b/pkg/sentry/fs/tty/line_discipline.go @@ -244,8 +244,8 @@ func (l *lineDiscipline) queueWrite(ctx context.Context, src usermem.IOSequence, return int64(n), err } -// transformOutput does ouput processing for one end of the pty. See -// drivers/tty/n_tty.c:do_output_char for an analagous kernel function. +// transformOutput does output processing for one end of the pty. See +// drivers/tty/n_tty.c:do_output_char for an analogous kernel function. // // Precondition: l.termiosMu must be held. func (l *lineDiscipline) transformOutput(buf []byte) *bytes.Buffer { diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index 19ad5d537..fb8c2f98c 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -298,7 +298,7 @@ func (s *Set) GetVal(num int32, creds *auth.Credentials) (int16, error) { } // ExecuteOps attempts to execute a list of operations to the set. It only -// suceeds when all operations can be applied. No changes are made if it fails. +// succeeds when all operations can be applied. No changes are made if it fails. // // On failure, it may return an error (retries are hopeless) or it may return // a channel that can be waited on before attempting again. diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index 3d49ae350..d6604f37b 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -125,7 +125,7 @@ func (t *Task) killLocked() { Signo: int32(linux.SIGKILL), // Linux just sets SIGKILL in the pending signal bitmask without // enqueueing an actual siginfo, such that - // kernel/signal.c:collect_signal() initalizes si_code to SI_USER. + // kernel/signal.c:collect_signal() initializes si_code to SI_USER. Code: arch.SignalInfoUser, }) t.interrupt() diff --git a/pkg/sentry/mm/vma.go b/pkg/sentry/mm/vma.go index b6af48cb7..61aaa3195 100644 --- a/pkg/sentry/mm/vma.go +++ b/pkg/sentry/mm/vma.go @@ -243,7 +243,7 @@ func (mm *MemoryManager) getVMAsLocked(ctx context.Context, ar usermem.AddrRange } // getVecVMAsLocked ensures that vmas exist for all addresses in ars, and -// support access to type of (at, ignorePermissions). It retuns the subset of +// support access to type of (at, ignorePermissions). It returns the subset of // ars for which vmas exist. If this is not equal to ars, it returns a non-nil // error explaining why. // diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index 15047df01..2911d3fd6 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -530,7 +530,7 @@ func (p *socketProvider) Socket(t *kernel.Task, stypeflags unix.SockType, protoc // Only accept TCP and UDP. // // Try to restrict the flags we will accept to minimize backwards - // incompatability with netstack. + // incompatibility with netstack. stype := int(stypeflags) & linux.SOCK_TYPE_MASK switch stype { case syscall.SOCK_STREAM: diff --git a/pkg/sentry/strace/syscalls.go b/pkg/sentry/strace/syscalls.go index d0e661706..eccee733e 100644 --- a/pkg/sentry/strace/syscalls.go +++ b/pkg/sentry/strace/syscalls.go @@ -82,7 +82,7 @@ const ( // PipeFDs is an array of two FDs, formatted after syscall execution. PipeFDs - // Uname is a pointer to a struct uname, formatted after syscall exection. + // Uname is a pointer to a struct uname, formatted after syscall execution. Uname // Stat is a pointer to a struct stat, formatted after syscall execution. diff --git a/pkg/tcpip/header/ipv6.go b/pkg/tcpip/header/ipv6.go index d8dc138b3..da0210539 100644 --- a/pkg/tcpip/header/ipv6.go +++ b/pkg/tcpip/header/ipv6.go @@ -60,7 +60,7 @@ const ( // IPv6ProtocolNumber is IPv6's network protocol number. IPv6ProtocolNumber tcpip.NetworkProtocolNumber = 0x86dd - // IPv6Version is the version of the ipv6 procotol. + // IPv6Version is the version of the ipv6 protocol. IPv6Version = 6 // IPv6MinimumMTU is the minimum MTU required by IPv6, per RFC 2460, diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index b480bf812..f0fbd8aad 100644 --- a/pkg/tcpip/stack/stack.go +++ b/pkg/tcpip/stack/stack.go @@ -165,7 +165,7 @@ type TCPSenderState struct { // window size from a segment. SndWndScale uint8 - // MaxSentAck is the highest acknowledgemnt number sent till now. + // MaxSentAck is the highest acknowledgement number sent till now. MaxSentAck seqnum.Value // FastRecovery holds the fast recovery state for the endpoint. diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index ad94aecd8..6c363a929 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -152,7 +152,7 @@ func newSender(ep *endpoint, iss, irs seqnum.Value, sndWnd seqnum.Size, mss uint // updateMaxPayloadSize updates the maximum payload size based on the given // MTU. If this is in response to "packet too big" control packets (indicated -// by the count argument), it also reduces the number of oustanding packets and +// by the count argument), it also reduces the number of outstanding packets and // attempts to retransmit the first packet above the MTU size. func (s *sender) updateMaxPayloadSize(mtu, count int) { m := mtu - header.TCPMinimumSize diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 64810b4ea..954824ada 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -58,7 +58,7 @@ func validateID(id string) error { // // Within a root directory, we maintain subdirectories for each sandbox named // with the sandbox id. The sandbox metadata is is stored as json within the -// sandbox directoy in a file named "meta.json". This metadata format is +// sandbox directory in a file named "meta.json". This metadata format is // defined by us, and is not part of the OCI spec. // // Sandboxes must write this metadata file after any change to their internal @@ -199,7 +199,7 @@ func Load(rootDir, id string) (*Sandbox, error) { // If the status is "Running" or "Created", check that the process // still exists, and set it to Stopped if it does not. // - // This is inherintly racey. + // This is inherently racey. if s.Status == Running || s.Status == Created { // Send signal 0 to check if process exists. if err := s.Signal(0); err != nil { -- cgit v1.2.3 From f47174f06b9904b830268d46a7e817053b6235c8 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Fri, 4 May 2018 14:15:24 -0700 Subject: Run gofmt -s on everything PiperOrigin-RevId: 195469901 Change-Id: I66d5c7a334bbb8b47e40d266a2661291c2d91c7f --- pkg/sentry/control/proc_test.go | 8 ++++---- pkg/sentry/kernel/semaphore/semaphore_test.go | 8 ++++---- runsc/fsgofer/fsgofer.go | 4 ++-- runsc/fsgofer/fsgofer_test.go | 8 ++++---- runsc/sandbox/sandbox.go | 4 ++-- 5 files changed, 16 insertions(+), 16 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/control/proc_test.go b/pkg/sentry/control/proc_test.go index 18286496f..22c826236 100644 --- a/pkg/sentry/control/proc_test.go +++ b/pkg/sentry/control/proc_test.go @@ -38,7 +38,7 @@ func TestProcessListTable(t *testing.T) { }, { pl: []*Process{ - &Process{ + { UID: 0, PID: 0, PPID: 0, @@ -47,7 +47,7 @@ func TestProcessListTable(t *testing.T) { Time: "0", Cmd: "zero", }, - &Process{ + { UID: 1, PID: 1, PPID: 1, @@ -83,7 +83,7 @@ func TestProcessListJSON(t *testing.T) { }, { pl: []*Process{ - &Process{ + { UID: 0, PID: 0, PPID: 0, @@ -92,7 +92,7 @@ func TestProcessListJSON(t *testing.T) { Time: "0", Cmd: "zero", }, - &Process{ + { UID: 1, PID: 1, PPID: 1, diff --git a/pkg/sentry/kernel/semaphore/semaphore_test.go b/pkg/sentry/kernel/semaphore/semaphore_test.go index 0386586ab..1c6a2e1e9 100644 --- a/pkg/sentry/kernel/semaphore/semaphore_test.go +++ b/pkg/sentry/kernel/semaphore/semaphore_test.go @@ -57,7 +57,7 @@ func TestBasic(t *testing.T) { ctx := contexttest.Context(t) set := &Set{ID: 123, sems: make([]sem, 1)} ops := []linux.Sembuf{ - linux.Sembuf{SemOp: 1}, + {SemOp: 1}, } executeOps(ctx, t, set, ops, false) @@ -78,7 +78,7 @@ func TestWaitForZero(t *testing.T) { ctx := contexttest.Context(t) set := &Set{ID: 123, sems: make([]sem, 1)} ops := []linux.Sembuf{ - linux.Sembuf{SemOp: 0}, + {SemOp: 0}, } executeOps(ctx, t, set, ops, false) @@ -117,7 +117,7 @@ func TestNoWait(t *testing.T) { ctx := contexttest.Context(t) set := &Set{ID: 123, sems: make([]sem, 1)} ops := []linux.Sembuf{ - linux.Sembuf{SemOp: 1}, + {SemOp: 1}, } executeOps(ctx, t, set, ops, false) @@ -146,7 +146,7 @@ func TestUnregister(t *testing.T) { } ops := []linux.Sembuf{ - linux.Sembuf{SemOp: -1}, + {SemOp: -1}, } chs := make([]chan struct{}, 0, 5) for i := 0; i < 5; i++ { diff --git a/runsc/fsgofer/fsgofer.go b/runsc/fsgofer/fsgofer.go index 5ddc75a9d..be2ac5f3c 100644 --- a/runsc/fsgofer/fsgofer.go +++ b/runsc/fsgofer/fsgofer.go @@ -614,8 +614,8 @@ func (l *localFile) SetAttr(valid p9.SetAttrMask, attr p9.SetAttr) error { if valid.ATime || valid.MTime { utimes := [2]syscall.Timespec{ - syscall.Timespec{Sec: 0, Nsec: linux.UTIME_OMIT}, - syscall.Timespec{Sec: 0, Nsec: linux.UTIME_OMIT}, + {Sec: 0, Nsec: linux.UTIME_OMIT}, + {Sec: 0, Nsec: linux.UTIME_OMIT}, } if valid.ATime { if valid.ATimeNotSystemTime { diff --git a/runsc/fsgofer/fsgofer_test.go b/runsc/fsgofer/fsgofer_test.go index 7d834d596..58d04aefa 100644 --- a/runsc/fsgofer/fsgofer_test.go +++ b/runsc/fsgofer/fsgofer_test.go @@ -39,12 +39,12 @@ var ( allConfs []Config rwConfs = []Config{ - Config{ROMount: false, LazyOpenForWrite: false}, - Config{ROMount: false, LazyOpenForWrite: true}, + {ROMount: false, LazyOpenForWrite: false}, + {ROMount: false, LazyOpenForWrite: true}, } roConfs = []Config{ - Config{ROMount: true, LazyOpenForWrite: false}, - Config{ROMount: true, LazyOpenForWrite: true}, + {ROMount: true, LazyOpenForWrite: false}, + {ROMount: true, LazyOpenForWrite: true}, } ) diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 13bf5d800..0354a64b9 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -493,8 +493,8 @@ func (s *Sandbox) createSandboxProcess(conf *boot.Config, binPath string, common // namespace for these. log.Infof("Sandbox will be started in empty IPC and UTS namespaces") nss := []specs.LinuxNamespace{ - specs.LinuxNamespace{Type: specs.IPCNamespace}, - specs.LinuxNamespace{Type: specs.UTSNamespace}, + {Type: specs.IPCNamespace}, + {Type: specs.UTSNamespace}, } if conf.Platform == boot.PlatformPtrace { -- cgit v1.2.3 From 7c8c3705ea5d891a3d6126090b1f49d8bae44177 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Mon, 7 May 2018 16:37:08 -0700 Subject: Fix misspellings PiperOrigin-RevId: 195742598 Change-Id: Ibd4a8e4394e268c87700b6d1e50b4b37dfce5182 --- pkg/cpuid/cpuid.go | 2 +- pkg/sentry/fs/fdpipe/pipe_opener.go | 2 +- pkg/sentry/fs/file_overlay.go | 2 +- pkg/sentry/fs/inode.go | 2 +- pkg/sentry/fs/inotify.go | 2 +- pkg/sentry/kernel/auth/credentials.go | 2 +- pkg/sentry/kernel/pipe/pipe.go | 2 +- pkg/sentry/loader/loader.go | 2 +- pkg/sentry/platform/kvm/address_space.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables.go | 20 ++++++++++---------- .../platform/ring0/pagetables/pagetables_amd64.go | 2 +- pkg/sentry/usage/memory.go | 2 +- pkg/tcpip/header/ipv4.go | 2 +- runsc/sandbox/sandbox_test.go | 2 +- 14 files changed, 23 insertions(+), 23 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/cpuid/cpuid.go b/pkg/cpuid/cpuid.go index aa248dd98..b486ab037 100644 --- a/pkg/cpuid/cpuid.go +++ b/pkg/cpuid/cpuid.go @@ -720,7 +720,7 @@ func HostID(axArg, cxArg uint32) (ax, bx, cx, dx uint32) // HostFeatureSet uses cpuid to get host values and construct a feature set // that matches that of the host machine. Note that there are several places -// where there appear to be some unecessary assignments between register names +// where there appear to be some unnecessary assignments between register names // (ax, bx, cx, or dx) and featureBlockN variables. This is to explicitly show // where the different feature blocks come from, to make the code easier to // inspect and read. diff --git a/pkg/sentry/fs/fdpipe/pipe_opener.go b/pkg/sentry/fs/fdpipe/pipe_opener.go index a0d59575f..945cfaf08 100644 --- a/pkg/sentry/fs/fdpipe/pipe_opener.go +++ b/pkg/sentry/fs/fdpipe/pipe_opener.go @@ -94,7 +94,7 @@ func unwrapError(err error) error { // TryOpen uses a NonBlockingOpener to try to open a host pipe, respecting the fs.FileFlags. func (p *pipeOpenState) TryOpen(ctx context.Context, opener NonBlockingOpener, flags fs.FileFlags) (*pipeOperations, error) { switch { - // Reject invalid configurations so they don't accidently succeed below. + // Reject invalid configurations so they don't accidentally succeed below. case !flags.Read && !flags.Write: return nil, syscall.EINVAL diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index 0c6e622b9..c27c5946e 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -151,7 +151,7 @@ func (f *overlayFileOperations) Seek(ctx context.Context, file *File, whence See // If this was a seek on a directory, we must update the cursor. if seekDir && whence == SeekSet && offset == 0 { - // Currenly only seeking to 0 on a directory is supported. + // Currently only seeking to 0 on a directory is supported. // FIXME: Lift directory seeking limitations. f.dirCursor = "" } diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index b624f4182..6c8e6f188 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -26,7 +26,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) -// Inode is a file system object that can be simulatenously referenced by different +// Inode is a file system object that can be simultaneously referenced by different // components of the VFS (Dirent, fs.File, etc). type Inode struct { // AtomicRefCount is our reference count. diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 9f50cb800..a87be8590 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -316,7 +316,7 @@ func (i *Inotify) RmWatch(wd int32) error { // The watch is now isolated and we can safely drop the instance lock. We // need to do so because watch.destroy() acquires Watch.mu, which cannot be - // aquired with Inotify.mu held. + // acquired with Inotify.mu held. i.mu.Unlock() // Generate the event for the removal. diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go index b832b28fe..f6fb05285 100644 --- a/pkg/sentry/kernel/auth/credentials.go +++ b/pkg/sentry/kernel/auth/credentials.go @@ -126,7 +126,7 @@ func NewUserCredentials(kuid KUID, kgid KGID, extraKGIDs []KGID, capabilities *T creds.InheritableCaps = capabilities.InheritableCaps // // TODO: Support ambient capabilities. } else { - // If no capabilities are specified, grant the same capabilites + // If no capabilities are specified, grant the same capabilities // that NewRootCredentials does. creds.PermittedCaps = AllCapabilities creds.EffectiveCaps = AllCapabilities diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index 1656c6ff3..9a21df5b4 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -284,7 +284,7 @@ func (p *Pipe) rReadinessLocked() waiter.EventMask { ready |= waiter.EventIn } if !p.HasWriters() && p.hadWriter { - // POLLHUP must be supressed until the pipe has had at least one writer + // POLLHUP must be suppressed until the pipe has had at least one writer // at some point. Otherwise a reader thread may poll and immediately get // a POLLHUP before the writer ever opens the pipe, which the reader may // interpret as the writer opening then closing the pipe. diff --git a/pkg/sentry/loader/loader.go b/pkg/sentry/loader/loader.go index 94c281b72..4ed796493 100644 --- a/pkg/sentry/loader/loader.go +++ b/pkg/sentry/loader/loader.go @@ -118,7 +118,7 @@ func allocStack(ctx context.Context, m *mm.MemoryManager, a arch.Context) (*arch const ( // maxLoaderAttempts is the maximum number of attempts to try to load - // an interpreter scripts, to prevent loops. 6 (inital + 5 changes) is + // an interpreter scripts, to prevent loops. 6 (initial + 5 changes) is // what the Linux kernel allows (fs/exec.c:search_binary_handler). maxLoaderAttempts = 6 ) diff --git a/pkg/sentry/platform/kvm/address_space.go b/pkg/sentry/platform/kvm/address_space.go index 791f038b0..a4b9198cc 100644 --- a/pkg/sentry/platform/kvm/address_space.go +++ b/pkg/sentry/platform/kvm/address_space.go @@ -178,7 +178,7 @@ func (as *addressSpace) MapFile(addr usermem.Addr, fd int, fr platform.FileRange // we create distinct mappings for each address space. Unfortunately, // there's not a better way to manage this here. The file underlying // this fd can change at any time, so we can't actually index the file - // and share between address space. Oh well. It's all refering to the + // and share between address space. Oh well. It's all referring to the // same physical pages, hopefully we don't run out of address space. if fd != int(as.filemem.File().Fd()) { // N.B. precommit is ignored for host files. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables.go b/pkg/sentry/platform/ring0/pagetables/pagetables.go index 3cbf0bfa5..ee7f27601 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables.go @@ -44,18 +44,18 @@ type PageTables struct { // root is the pagetable root. root *Node - // translater is the translater passed at creation. - translater Translater + // translator is the translator passed at creation. + translator Translator // archPageTables includes architecture-specific features. archPageTables - // allNodes is a set of nodes indexed by translater address. + // allNodes is a set of nodes indexed by translator address. allNodes map[uintptr]*Node } -// Translater translates to guest physical addresses. -type Translater interface { +// Translator translates to guest physical addresses. +type Translator interface { // TranslateToPhysical translates the given pointer object into a // "physical" address. We do not require that it translates back, the // reverse mapping is maintained internally. @@ -63,9 +63,9 @@ type Translater interface { } // New returns new PageTables. -func New(t Translater, opts Opts) *PageTables { +func New(t Translator, opts Opts) *PageTables { p := &PageTables{ - translater: t, + translator: t, allNodes: make(map[uintptr]*Node), } p.root = p.allocNode() @@ -80,7 +80,7 @@ func New(t Translater, opts Opts) *PageTables { // managing multiple sets of pagetables. func (p *PageTables) New() *PageTables { np := &PageTables{ - translater: p.translater, + translator: p.translator, allNodes: make(map[uintptr]*Node), } np.root = np.allocNode() @@ -90,7 +90,7 @@ func (p *PageTables) New() *PageTables { // setPageTable sets the given index as a page table. func (p *PageTables) setPageTable(n *Node, index int, child *Node) { - phys := p.translater.TranslateToPhysical(child.PTEs()) + phys := p.translator.TranslateToPhysical(child.PTEs()) p.allNodes[phys] = child pte := &n.PTEs()[index] pte.setPageTable(phys) @@ -188,6 +188,6 @@ func (p *PageTables) Lookup(addr usermem.Addr) (physical uintptr, accessType use // allocNode allocates a new page. func (p *PageTables) allocNode() *Node { n := new(Node) - n.physical = p.translater.TranslateToPhysical(n.PTEs()) + n.physical = p.translator.TranslateToPhysical(n.PTEs()) return n } diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go index b89665c96..a2050b99c 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go @@ -301,7 +301,7 @@ func (p *PageTables) iterateRange(startAddr, endAddr uintptr, alloc bool, fn fun } // This level has 2-MB huge pages. If this - // region is contined in a single PMD entry? + // region is contained in a single PMD entry? // As above, we can skip allocating a new page. if start&(pmdSize-1) == 0 && end-start >= pmdSize { pmdEntry.SetSuper() diff --git a/pkg/sentry/usage/memory.go b/pkg/sentry/usage/memory.go index 5d1b3a595..4a1527b5f 100644 --- a/pkg/sentry/usage/memory.go +++ b/pkg/sentry/usage/memory.go @@ -95,7 +95,7 @@ type MemoryStats struct { // categories not backed by platform memory. For details about how this works, // see the memory accounting docs. // -// N.B. Please keep the struct in sync with the API. Noteably, changes to this +// N.B. Please keep the struct in sync with the API. Notably, changes to this // struct requires a version bump and addition of compatibility logic in the // control server. As a special-case, adding fields without re-ordering existing // ones do not require a version bump because the mapped page we use is diff --git a/pkg/tcpip/header/ipv4.go b/pkg/tcpip/header/ipv4.go index cb0d42093..6e2a3d6f4 100644 --- a/pkg/tcpip/header/ipv4.go +++ b/pkg/tcpip/header/ipv4.go @@ -81,7 +81,7 @@ const ( // IPv4ProtocolNumber is IPv4's network protocol number. IPv4ProtocolNumber tcpip.NetworkProtocolNumber = 0x0800 - // IPv4Version is the version of the ipv4 procotol. + // IPv4Version is the version of the ipv4 protocol. IPv4Version = 4 ) diff --git a/runsc/sandbox/sandbox_test.go b/runsc/sandbox/sandbox_test.go index 6e3125b7b..a46212173 100644 --- a/runsc/sandbox/sandbox_test.go +++ b/runsc/sandbox/sandbox_test.go @@ -131,7 +131,7 @@ func waitForProcessList(s *sandbox.Sandbox, expected []*control.Process) error { return fmt.Errorf("sandbox got process list: %s, want: %s", procListToString(got), procListToString(expected)) } -// TestLifecycle tests the basic Create/Start/Signal/Destory sandbox lifecycle. +// TestLifecycle tests the basic Create/Start/Signal/Destroy sandbox lifecycle. // It verifies after each step that the sandbox can be loaded from disk, and // has the correct status. func TestLifecycle(t *testing.T) { -- cgit v1.2.3 From d0d01a18963ed7cfc29e5b8334e30b1234b6048b Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Tue, 8 May 2018 09:38:55 -0700 Subject: Fix format string type in test PiperOrigin-RevId: 195831778 Change-Id: I413dc909cedc18fbf5320a4f75d876f1be133c6c --- pkg/sentry/kernel/fd_map_test.go | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/fd_map_test.go b/pkg/sentry/kernel/fd_map_test.go index e1ac900e8..95123aef3 100644 --- a/pkg/sentry/kernel/fd_map_test.go +++ b/pkg/sentry/kernel/fd_map_test.go @@ -119,16 +119,18 @@ func TestDescriptorFlags(t *testing.T) { limitSet := limits.NewLimitSet() limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}) - if err := f.NewFDAt(2, file, FDFlags{CloseOnExec: true}, limitSet); err != nil { + origFlags := FDFlags{CloseOnExec: true} + + if err := f.NewFDAt(2, file, origFlags, limitSet); err != nil { t.Fatalf("f.NewFDAt(2, r, FDFlags{}): got %v, wanted nil", err) } - newFile, flags := f.GetDescriptor(2) + newFile, newFlags := f.GetDescriptor(2) if newFile == nil { t.Fatalf("f.GetFile(2): got a %v, wanted nil", newFile) } - if !flags.CloseOnExec { - t.Fatalf("new File flags %d don't match original %d\n", flags, 0) + if newFlags != origFlags { + t.Fatalf("new File flags %+v don't match original %+v", newFlags, origFlags) } } -- cgit v1.2.3 From b4765f782d91443ab0415dc00e727d783632e2ad Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Tue, 8 May 2018 09:51:07 -0700 Subject: Fix warning: redundant if ...; err != nil check, just return error instead. This warning is produced by golint. PiperOrigin-RevId: 195833381 Change-Id: Idd6a7e57e3cfdf00819f2374b19fc113585dc1e1 --- pkg/bpf/decoder.go | 5 +---- pkg/compressio/compressio.go | 5 +---- pkg/p9/local_server/local_server.go | 6 +----- pkg/sentry/fs/fdpipe/pipe.go | 5 +---- pkg/sentry/fs/host/fs_test.go | 6 +----- pkg/sentry/fs/host/socket.go | 5 +---- pkg/sentry/kernel/task_signals.go | 10 ++-------- pkg/sentry/syscalls/linux/sys_file.go | 7 ++----- pkg/tcpip/stack/transport_demuxer.go | 6 +----- 9 files changed, 11 insertions(+), 44 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/bpf/decoder.go b/pkg/bpf/decoder.go index 6873ffa5c..ef41e9edc 100644 --- a/pkg/bpf/decoder.go +++ b/pkg/bpf/decoder.go @@ -161,10 +161,7 @@ func decodeAlu(inst linux.BPFInstruction, w *bytes.Buffer) error { default: return fmt.Errorf("invalid BPF ALU instruction: %v", inst) } - if err := decodeSource(inst, w); err != nil { - return err - } - return nil + return decodeSource(inst, w) } func decodeSource(inst linux.BPFInstruction, w *bytes.Buffer) error { diff --git a/pkg/compressio/compressio.go b/pkg/compressio/compressio.go index e0d36aee9..ef8cbd2a5 100644 --- a/pkg/compressio/compressio.go +++ b/pkg/compressio/compressio.go @@ -184,10 +184,7 @@ func handleResult(r result, callback func(*chunk) error) error { if r.err != nil { return r.err } - if err := callback(r.chunk); err != nil { - return err - } - return nil + return callback(r.chunk) } // schedule schedules the given buffers. diff --git a/pkg/p9/local_server/local_server.go b/pkg/p9/local_server/local_server.go index 5b1e97711..7a3e4cffe 100644 --- a/pkg/p9/local_server/local_server.go +++ b/pkg/p9/local_server/local_server.go @@ -246,11 +246,7 @@ func (l *local) Symlink(oldname string, newname string, _ p9.UID, _ p9.GID) (p9. // // Not properly implemented. func (l *local) Link(target p9.File, newname string) error { - if err := os.Link(target.(*local).path, path.Join(l.path, newname)); err != nil { - return err - } - - return nil + return os.Link(target.(*local).path, path.Join(l.path, newname)) } // Mknod implements p9.File.Mknod. diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index f7bbd4aff..7b318e35f 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -90,10 +90,7 @@ func (p *pipeOperations) init() error { if err := syscall.SetNonblock(p.file.FD(), true); err != nil { return err } - if err := fdnotifier.AddFD(int32(p.file.FD()), &p.Queue); err != nil { - return err - } - return nil + return fdnotifier.AddFD(int32(p.file.FD()), &p.Queue) } // EventRegister implements waiter.Waitable.EventRegister. diff --git a/pkg/sentry/fs/host/fs_test.go b/pkg/sentry/fs/host/fs_test.go index c000afc49..b08125ca8 100644 --- a/pkg/sentry/fs/host/fs_test.go +++ b/pkg/sentry/fs/host/fs_test.go @@ -141,11 +141,7 @@ func createTestDirs(ctx context.Context, t *testing.T, m *fs.MountNamespace) err return err } - if err := symlinks.CreateLink(ctx, r, "/symlinks", "recursive"); err != nil { - return err - } - - return nil + return symlinks.CreateLink(ctx, r, "/symlinks", "recursive") } // allPaths returns a slice of all paths of entries visible in the rootfs. diff --git a/pkg/sentry/fs/host/socket.go b/pkg/sentry/fs/host/socket.go index 8e36ed7ee..467633052 100644 --- a/pkg/sentry/fs/host/socket.go +++ b/pkg/sentry/fs/host/socket.go @@ -69,10 +69,7 @@ func (e *endpoint) init() error { } e.stype = unix.SockType(stype) - if err := fdnotifier.AddFD(int32(e.fd), &e.queue); err != nil { - return err - } - return nil + return fdnotifier.AddFD(int32(e.fd), &e.queue) } // newEndpoint creates a new host endpoint. diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 2340256b0..e4ef7fd67 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -127,10 +127,7 @@ func (t *Task) dequeueSignalLocked() *arch.SignalInfo { if info := t.pendingSignals.dequeue(t.tr.SignalMask); info != nil { return info } - if info := t.tg.pendingSignals.dequeue(t.tr.SignalMask); info != nil { - return info - } - return nil + return t.tg.pendingSignals.dequeue(t.tr.SignalMask) } // TakeSignal returns a pending signal not blocked by mask. Signal handlers are @@ -144,10 +141,7 @@ func (t *Task) TakeSignal(mask linux.SignalSet) *arch.SignalInfo { if info := t.pendingSignals.dequeue(mask); info != nil { return info } - if info := t.tg.pendingSignals.dequeue(mask); info != nil { - return info - } - return nil + return t.tg.pendingSignals.dequeue(mask) } // discardSpecificLocked removes all instances of the given signal from all diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index a2dbba7e0..5fbacc15e 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -460,14 +460,11 @@ func accessAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, resolve bool, m creds: creds, } - if err := d.Inode.CheckPermission(ctx, fs.PermMask{ + return d.Inode.CheckPermission(ctx, fs.PermMask{ Read: mode&rOK != 0, Write: mode&wOK != 0, Execute: mode&xOK != 0, - }); err != nil { - return err - } - return nil + }) }) } diff --git a/pkg/tcpip/stack/transport_demuxer.go b/pkg/tcpip/stack/transport_demuxer.go index 3c0d7aa31..7bb853622 100644 --- a/pkg/tcpip/stack/transport_demuxer.go +++ b/pkg/tcpip/stack/transport_demuxer.go @@ -158,9 +158,5 @@ func (d *transportDemuxer) findEndpointLocked(eps *transportEndpoints, vv *buffe // Try to find a match with only the local port. nid.LocalAddress = "" - if ep := eps.endpoints[nid]; ep != nil { - return ep - } - - return nil + return eps.endpoints[nid] } -- cgit v1.2.3 From 174161013de22be6a42b02ee06611a9de9e20b18 Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Tue, 8 May 2018 11:36:11 -0700 Subject: Capture restore file system corruption errors in exit error. PiperOrigin-RevId: 195850822 Change-Id: I4d7bdd8fe129c5ed461b73e1d7458be2cf5680c2 --- pkg/sentry/fs/fdpipe/pipe_state.go | 9 +++--- pkg/sentry/fs/fs.go | 63 ++++++++++++++++++++++++++++++++++++-- pkg/sentry/fs/gofer/file_state.go | 9 ++++-- pkg/sentry/fs/gofer/inode_state.go | 23 ++++++++------ pkg/sentry/fs/host/inode_state.go | 6 ++-- pkg/sentry/kernel/kernel.go | 4 ++- 6 files changed, 90 insertions(+), 24 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/fdpipe/pipe_state.go b/pkg/sentry/fs/fdpipe/pipe_state.go index 8996a2178..99c40d8ed 100644 --- a/pkg/sentry/fs/fdpipe/pipe_state.go +++ b/pkg/sentry/fs/fdpipe/pipe_state.go @@ -63,7 +63,7 @@ func (p *pipeOperations) loadFlags(flags fs.FileFlags) { // afterLoad is invoked by stateify. func (p *pipeOperations) afterLoad() { - load := func() { + load := func() error { if !p.flags.Read { readPipeOperationsLoading.Wait() } else { @@ -75,14 +75,15 @@ func (p *pipeOperations) afterLoad() { Write: p.flags.Write, }) if err != nil { - panic(fmt.Sprintf("unable to open pipe %v: %v", p, err)) + return fmt.Errorf("unable to open pipe %v: %v", p, err) } if err := p.init(); err != nil { - panic(fmt.Sprintf("unable to initialize pipe %v: %v", p, err)) + return fmt.Errorf("unable to initialize pipe %v: %v", p, err) } + return nil } // Do background opening of pipe ends. Note for write-only pipe ends we // have to do it asynchronously to avoid blocking the restore. - fs.Async(load) + fs.Async(fs.CatchError(load)) } diff --git a/pkg/sentry/fs/fs.go b/pkg/sentry/fs/fs.go index f54f767d3..6ec9ff446 100644 --- a/pkg/sentry/fs/fs.go +++ b/pkg/sentry/fs/fs.go @@ -55,11 +55,19 @@ package fs import ( "sync" + + "gvisor.googlesource.com/gvisor/pkg/log" ) -// work is a sync.WaitGroup that can be used to queue asynchronous operations -// via Do. Callers can use Barrier to ensure no operations are outstanding. -var work sync.WaitGroup +var ( + // work is a sync.WaitGroup that can be used to queue asynchronous + // operations via Do. Callers can use Barrier to ensure no operations + // are outstanding. + work sync.WaitGroup + + // asyncError is used to store up to one asynchronous execution error. + asyncError = make(chan error, 1) +) // AsyncBarrier waits for all outstanding asynchronous work to complete. func AsyncBarrier() { @@ -75,6 +83,43 @@ func Async(f func()) { }() } +// AsyncErrorBarrier waits for all outstanding asynchronous work to complete, or +// the first async error to arrive. Other unfinished async executions will +// continue in the background. Other past and future async errors are ignored. +func AsyncErrorBarrier() error { + wait := make(chan struct{}, 1) + go func() { // S/R-SAFE: Does not touch persistent state. + work.Wait() + wait <- struct{}{} + }() + select { + case <-wait: + select { + case err := <-asyncError: + return err + default: + return nil + } + case err := <-asyncError: + return err + } +} + +// CatchError tries to capture the potential async error returned by the +// function. At most one async error will be captured globally so excessive +// errors will be dropped. +func CatchError(f func() error) func() { + return func() { + if err := f(); err != nil { + select { + case asyncError <- err: + default: + log.Warningf("excessive async error dropped: %v", err) + } + } + } +} + // ErrSaveRejection indicates a failed save due to unsupported file system state // such as dangling open fd, etc. type ErrSaveRejection struct { @@ -86,3 +131,15 @@ type ErrSaveRejection struct { func (e ErrSaveRejection) Error() string { return "save rejected due to unsupported file system state: " + e.Err.Error() } + +// ErrCorruption indicates a failed restore due to external file system state in +// corruption. +type ErrCorruption struct { + // Err is the wrapped error. + Err error +} + +// Error returns a sensible description of the save rejection error. +func (e ErrCorruption) Error() string { + return "restore failed due to external file system state in corruption: " + e.Err.Error() +} diff --git a/pkg/sentry/fs/gofer/file_state.go b/pkg/sentry/fs/gofer/file_state.go index 1d63e33ec..715af8f16 100644 --- a/pkg/sentry/fs/gofer/file_state.go +++ b/pkg/sentry/fs/gofer/file_state.go @@ -15,13 +15,15 @@ package gofer import ( + "fmt" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" ) // afterLoad is invoked by stateify. func (f *fileOperations) afterLoad() { - load := func() { + load := func() error { f.inodeOperations.fileState.waitForLoad() // Manually load the open handles. @@ -29,9 +31,10 @@ func (f *fileOperations) afterLoad() { // TODO: Context is not plumbed to save/restore. f.handles, err = newHandles(context.Background(), f.inodeOperations.fileState.file, f.flags) if err != nil { - panic("failed to re-open handle: " + err.Error()) + return fmt.Errorf("failed to re-open handle: %v", err) } f.inodeOperations.fileState.setHandlesForCachedIO(f.flags, f.handles) + return nil } - fs.Async(load) + fs.Async(fs.CatchError(load)) } diff --git a/pkg/sentry/fs/gofer/inode_state.go b/pkg/sentry/fs/gofer/inode_state.go index 997a7d1c1..82d1dd4da 100644 --- a/pkg/sentry/fs/gofer/inode_state.go +++ b/pkg/sentry/fs/gofer/inode_state.go @@ -15,6 +15,7 @@ package gofer import ( + "errors" "fmt" "strings" @@ -83,7 +84,7 @@ func (i *inodeFileState) loadLoading(_ struct{}) { // afterLoad is invoked by stateify. func (i *inodeFileState) afterLoad() { - load := func() { + load := func() error { // See comment on i.loading(). defer i.loading.Unlock() @@ -92,14 +93,14 @@ func (i *inodeFileState) afterLoad() { if !ok { // This should be impossible, see assertion in // beforeSave. - panic(fmt.Sprintf("failed to find path for inode number %d. Device %s contains %s", i.sattr.InodeID, i.s.connID, fs.InodeMappings(i.s.inodeMappings))) + return fmt.Errorf("failed to find path for inode number %d. Device %s contains %s", i.sattr.InodeID, i.s.connID, fs.InodeMappings(i.s.inodeMappings)) } // TODO: Context is not plumbed to save/restore. ctx := &dummyClockContext{context.Background()} var err error _, i.file, err = i.s.attach.walk(ctx, strings.Split(name, "/")) if err != nil { - panic(fmt.Sprintf("failed to walk to %q: %v", name, err)) + return fmt.Errorf("failed to walk to %q: %v", name, err) } // Remap the saved inode number into the gofer device using the @@ -107,10 +108,10 @@ func (i *inodeFileState) afterLoad() { // environment. qid, mask, attrs, err := i.file.getAttr(ctx, p9.AttrMaskAll()) if err != nil { - panic(fmt.Sprintf("failed to get file attributes of %s: %v", name, err)) + return fmt.Errorf("failed to get file attributes of %s: %v", name, err) } if !mask.RDev { - panic(fmt.Sprintf("file %s lacks device", name)) + return fs.ErrCorruption{fmt.Errorf("file %s lacks device", name)} } i.key = device.MultiDeviceKey{ Device: attrs.RDev, @@ -118,24 +119,26 @@ func (i *inodeFileState) afterLoad() { Inode: qid.Path, } if !goferDevice.Load(i.key, i.sattr.InodeID) { - panic(fmt.Sprintf("gofer device %s -> %d conflict in gofer device mappings: %s", i.key, i.sattr.InodeID, goferDevice)) + return fs.ErrCorruption{fmt.Errorf("gofer device %s -> %d conflict in gofer device mappings: %s", i.key, i.sattr.InodeID, goferDevice)} } if i.sattr.Type == fs.RegularFile { env, ok := fs.CurrentRestoreEnvironment() if !ok { - panic("missing restore environment") + return errors.New("missing restore environment") } uattr := unstable(ctx, mask, attrs, i.s.mounter, i.s.client) if env.ValidateFileSize && uattr.Size != i.savedUAttr.Size { - panic(fmt.Errorf("file size has changed for %s: previously %d, now %d", i.s.inodeMappings[i.sattr.InodeID], i.savedUAttr.Size, uattr.Size)) + return fs.ErrCorruption{fmt.Errorf("file size has changed for %s: previously %d, now %d", i.s.inodeMappings[i.sattr.InodeID], i.savedUAttr.Size, uattr.Size)} } if env.ValidateFileTimestamp && uattr.ModificationTime != i.savedUAttr.ModificationTime { - panic(fmt.Errorf("file modification time has changed for %s: previously %v, now %v", i.s.inodeMappings[i.sattr.InodeID], i.savedUAttr.ModificationTime, uattr.ModificationTime)) + return fs.ErrCorruption{fmt.Errorf("file modification time has changed for %s: previously %v, now %v", i.s.inodeMappings[i.sattr.InodeID], i.savedUAttr.ModificationTime, uattr.ModificationTime)} } i.savedUAttr = nil } + + return nil } - fs.Async(load) + fs.Async(fs.CatchError(load)) } diff --git a/pkg/sentry/fs/host/inode_state.go b/pkg/sentry/fs/host/inode_state.go index 80066512a..135c75fd5 100644 --- a/pkg/sentry/fs/host/inode_state.go +++ b/pkg/sentry/fs/host/inode_state.go @@ -59,7 +59,7 @@ func (i *inodeFileState) afterLoad() { // saved filesystem are no longer unique on this filesystem. // Since this violates the contract that filesystems cannot // change across save and restore, error out. - panic(fmt.Sprintf("host %s conflict in host device mappings: %s", key, hostFileDevice)) + panic(fs.ErrCorruption{fmt.Errorf("host %s conflict in host device mappings: %s", key, hostFileDevice)}) } if !i.descriptor.donated && i.sattr.Type == fs.RegularFile { @@ -69,10 +69,10 @@ func (i *inodeFileState) afterLoad() { } uattr := unstableAttr(i.mops, &s) if env.ValidateFileSize && uattr.Size != i.savedUAttr.Size { - panic(fmt.Errorf("file size has changed for %s: previously %d, now %d", i.mops.inodeMappings[i.sattr.InodeID], i.savedUAttr.Size, uattr.Size)) + panic(fs.ErrCorruption{fmt.Errorf("file size has changed for %s: previously %d, now %d", i.mops.inodeMappings[i.sattr.InodeID], i.savedUAttr.Size, uattr.Size)}) } if env.ValidateFileTimestamp && uattr.ModificationTime != i.savedUAttr.ModificationTime { - panic(fmt.Errorf("file modification time has changed for %s: previously %v, now %v", i.mops.inodeMappings[i.sattr.InodeID], i.savedUAttr.ModificationTime, uattr.ModificationTime)) + panic(fs.ErrCorruption{fmt.Errorf("file modification time has changed for %s: previously %v, now %v", i.mops.inodeMappings[i.sattr.InodeID], i.savedUAttr.ModificationTime, uattr.ModificationTime)}) } i.savedUAttr = nil } diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 25c8dd885..536461bbd 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -418,7 +418,9 @@ func (k *Kernel) LoadFrom(r io.Reader, p platform.Platform, net inet.Stack) erro // Ensure that all pending asynchronous work is complete: // - namedpipe opening // - inode file opening - fs.AsyncBarrier() + if err := fs.AsyncErrorBarrier(); err != nil { + return err + } log.Infof("Overall load took [%s]", time.Since(loadStart)) -- cgit v1.2.3 From 8deabbaae1fc45b042d551891080deef866dc0f8 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Fri, 11 May 2018 12:23:25 -0700 Subject: Remove error return from AddressSpace.Release() PiperOrigin-RevId: 196291289 Change-Id: Ie3487be029850b0b410b82416750853a6c4a2b00 --- pkg/sentry/kernel/task_usermem.go | 4 +--- pkg/sentry/mm/address_space.go | 17 ++++++----------- pkg/sentry/platform/kvm/address_space.go | 3 +-- pkg/sentry/platform/platform.go | 2 +- pkg/sentry/platform/ptrace/subprocess.go | 3 +-- 5 files changed, 10 insertions(+), 19 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_usermem.go b/pkg/sentry/kernel/task_usermem.go index 7a62ab674..54964dd0d 100644 --- a/pkg/sentry/kernel/task_usermem.go +++ b/pkg/sentry/kernel/task_usermem.go @@ -39,9 +39,7 @@ func (t *Task) Activate() { // Deactivate relinquishes the task's active address space. func (t *Task) Deactivate() { if mm := t.MemoryManager(); mm != nil { - if err := mm.Deactivate(); err != nil { - panic("unable to deactivate mm: " + err.Error()) - } + mm.Deactivate() } } diff --git a/pkg/sentry/mm/address_space.go b/pkg/sentry/mm/address_space.go index 4dd67b1ea..27554f163 100644 --- a/pkg/sentry/mm/address_space.go +++ b/pkg/sentry/mm/address_space.go @@ -114,12 +114,12 @@ func (mm *MemoryManager) Activate() error { } } -// Deactivate releases a release to the MemoryManager. -func (mm *MemoryManager) Deactivate() error { +// Deactivate releases a reference to the MemoryManager. +func (mm *MemoryManager) Deactivate() { // Fast path: this is not the last goroutine to deactivate the // MemoryManager. if atomicbitops.DecUnlessOneInt32(&mm.active) { - return nil + return } mm.activeMu.Lock() @@ -128,26 +128,21 @@ func (mm *MemoryManager) Deactivate() error { // Still active? if atomic.AddInt32(&mm.active, -1) > 0 { mm.activeMu.Unlock() - return nil + return } // Can we hold on to the address space? if !mm.p.CooperativelySchedulesAddressSpace() { mm.activeMu.Unlock() - return nil + return } // Release the address space. - if err := mm.as.Release(); err != nil { - atomic.StoreInt32(&mm.active, 1) - mm.activeMu.Unlock() - return err - } + mm.as.Release() // Lost it. mm.as = nil mm.activeMu.Unlock() - return nil } // mapASLocked maps addresses in ar into mm.as. If precommit is true, mappings diff --git a/pkg/sentry/platform/kvm/address_space.go b/pkg/sentry/platform/kvm/address_space.go index a4b9198cc..173885867 100644 --- a/pkg/sentry/platform/kvm/address_space.go +++ b/pkg/sentry/platform/kvm/address_space.go @@ -200,8 +200,7 @@ func (as *addressSpace) Unmap(addr usermem.Addr, length uint64) { } // Release releases the page tables. -func (as *addressSpace) Release() error { +func (as *addressSpace) Release() { as.Unmap(0, ^uint64(0)) as.pageTables.Release() - return nil } diff --git a/pkg/sentry/platform/platform.go b/pkg/sentry/platform/platform.go index 6219dada7..1c385bc5a 100644 --- a/pkg/sentry/platform/platform.go +++ b/pkg/sentry/platform/platform.go @@ -205,7 +205,7 @@ type AddressSpace interface { // Release releases this address space. After releasing, a new AddressSpace // must be acquired via platform.NewAddressSpace(). - Release() error + Release() // AddressSpaceIO methods are supported iff the associated platform's // Platform.SupportsAddressSpaceIO() == true. AddressSpaces for which this diff --git a/pkg/sentry/platform/ptrace/subprocess.go b/pkg/sentry/platform/ptrace/subprocess.go index 0d6a38f15..035ebc332 100644 --- a/pkg/sentry/platform/ptrace/subprocess.go +++ b/pkg/sentry/platform/ptrace/subprocess.go @@ -242,14 +242,13 @@ func (s *subprocess) unmap() { // Therefore we simply unmap everything in the subprocess and return it to the // globalPool. This has the added benefit of reducing creation time for new // subprocesses. -func (s *subprocess) Release() error { +func (s *subprocess) Release() { go func() { // S/R-SAFE: Platform. s.unmap() globalPool.mu.Lock() globalPool.available = append(globalPool.available, s) globalPool.mu.Unlock() }() - return nil } // newThread creates a new traced thread. -- cgit v1.2.3 From 96c28a43682e8a665142da5b8b0734198fff3a00 Mon Sep 17 00:00:00 2001 From: Kevin Krakauer Date: Tue, 15 May 2018 14:55:29 -0700 Subject: sentry: Replaces saving of inet.Stack with retrieval via context. Previously, inet.Stack was referenced in 2 structs in sentry/socket that can be saved/restored. If an app is saved and restored on another machine, it may try to use the old stack, which will have been replaced by a new stack on the new machine. PiperOrigin-RevId: 196733985 Change-Id: I6a8cfe73b5d7a90749734677dada635ab3389cb9 --- pkg/sentry/fs/context.go | 2 +- pkg/sentry/inet/BUILD | 2 ++ pkg/sentry/inet/context.go | 35 +++++++++++++++++++++++++++++ pkg/sentry/kernel/task.go | 3 +++ pkg/sentry/socket/epsocket/epsocket.go | 25 ++++++++++++++------- pkg/sentry/socket/netlink/route/protocol.go | 21 +++++++---------- 6 files changed, 66 insertions(+), 22 deletions(-) create mode 100644 pkg/sentry/inet/context.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/context.go b/pkg/sentry/fs/context.go index b521bce75..da46ad77f 100644 --- a/pkg/sentry/fs/context.go +++ b/pkg/sentry/fs/context.go @@ -20,7 +20,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" ) -// contextID is the kernel package's type for context.Context.Value keys. +// contextID is the fs package's type for context.Context.Value keys. type contextID int const ( diff --git a/pkg/sentry/inet/BUILD b/pkg/sentry/inet/BUILD index 207cdb692..1150ced57 100644 --- a/pkg/sentry/inet/BUILD +++ b/pkg/sentry/inet/BUILD @@ -17,12 +17,14 @@ go_stateify( go_library( name = "inet", srcs = [ + "context.go", "inet.go", "inet_state.go", "test_stack.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/inet", deps = [ + "//pkg/sentry/context", "//pkg/state", ], ) diff --git a/pkg/sentry/inet/context.go b/pkg/sentry/inet/context.go new file mode 100644 index 000000000..370381f41 --- /dev/null +++ b/pkg/sentry/inet/context.go @@ -0,0 +1,35 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package inet + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/context" +) + +// contextID is the inet package's type for context.Context.Value keys. +type contextID int + +const ( + // CtxStack is a Context.Value key for a network stack. + CtxStack contextID = iota +) + +// StackFromContext returns the network stack associated with ctx. +func StackFromContext(ctx context.Context) Stack { + if v := ctx.Value(CtxStack); v != nil { + return v.(Stack) + } + return nil +} diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 3d2e035e9..490f795c2 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/bpf" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/inet" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" @@ -560,6 +561,8 @@ func (t *Task) Value(key interface{}) interface{} { return t.creds case fs.CtxRoot: return t.FSContext().RootDirectory() + case inet.CtxStack: + return t.NetworkContext() case ktime.CtxRealtimeClock: return t.k.RealtimeClock() case limits.CtxLimits: diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index a45dcd551..3e4887e16 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -33,6 +33,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/binary" + "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" @@ -102,7 +103,6 @@ type SocketOperations struct { *waiter.Queue family int - stack inet.Stack Endpoint tcpip.Endpoint skType unix.SockType @@ -119,7 +119,6 @@ func New(t *kernel.Task, family int, skType unix.SockType, queue *waiter.Queue, return fs.NewFile(t, dirent, fs.FileFlags{Read: true, Write: true}, &SocketOperations{ Queue: queue, family: family, - stack: t.NetworkContext(), Endpoint: endpoint, skType: skType, }) @@ -1042,7 +1041,12 @@ func (s *SocketOperations) interfaceIoctl(ctx context.Context, io usermem.IO, ar ) // Find the relevant device. - for index, iface = range s.stack.Interfaces() { + stack := inet.StackFromContext(ctx) + if stack == nil { + log.Warningf("Couldn't find a network stack.") + return syserr.ErrInvalidArgument + } + for index, iface = range stack.Interfaces() { if iface.Name == ifr.Name() { found = true break @@ -1074,7 +1078,7 @@ func (s *SocketOperations) interfaceIoctl(ctx context.Context, io usermem.IO, ar case syscall.SIOCGIFADDR: // Copy the IPv4 address out. - for _, addr := range s.stack.InterfaceAddrs()[index] { + for _, addr := range stack.InterfaceAddrs()[index] { // This ioctl is only compatible with AF_INET addresses. if addr.Family != linux.AF_INET { continue @@ -1109,7 +1113,7 @@ func (s *SocketOperations) interfaceIoctl(ctx context.Context, io usermem.IO, ar case syscall.SIOCGIFNETMASK: // Gets the network mask of a device. - for _, addr := range s.stack.InterfaceAddrs()[index] { + for _, addr := range stack.InterfaceAddrs()[index] { // This ioctl is only compatible with AF_INET addresses. if addr.Family != linux.AF_INET { continue @@ -1189,15 +1193,20 @@ func (s *SocketOperations) ifconfIoctl(ctx context.Context, io usermem.IO, ifc * // If Ptr is NULL, return the necessary buffer size via Len. // Otherwise, write up to Len bytes starting at Ptr containing ifreq // structs. + stack := inet.StackFromContext(ctx) + if stack == nil { + log.Warningf("Couldn't find a network stack.") + return syserr.ErrInvalidArgument.ToError() + } if ifc.Ptr == 0 { - ifc.Len = int32(len(s.stack.Interfaces())) * int32(linux.SizeOfIFReq) + ifc.Len = int32(len(stack.Interfaces())) * int32(linux.SizeOfIFReq) return nil } max := ifc.Len ifc.Len = 0 - for key, ifaceAddrs := range s.stack.InterfaceAddrs() { - iface := s.stack.Interfaces()[key] + for key, ifaceAddrs := range stack.InterfaceAddrs() { + iface := stack.Interfaces()[key] for _, ifaceAddr := range ifaceAddrs { // Don't write past the end of the buffer. if ifc.Len+int32(linux.SizeOfIFReq) > max { diff --git a/pkg/sentry/socket/netlink/route/protocol.go b/pkg/sentry/socket/netlink/route/protocol.go index d611519d4..e8030c518 100644 --- a/pkg/sentry/socket/netlink/route/protocol.go +++ b/pkg/sentry/socket/netlink/route/protocol.go @@ -43,20 +43,13 @@ func typeKind(typ uint16) commandKind { } // Protocol implements netlink.Protocol. -type Protocol struct { - // stack is the network stack that this provider describes. - // - // May be nil. - stack inet.Stack -} +type Protocol struct{} var _ netlink.Protocol = (*Protocol)(nil) // NewProtocol creates a NETLINK_ROUTE netlink.Protocol. func NewProtocol(t *kernel.Task) (netlink.Protocol, *syserr.Error) { - return &Protocol{ - stack: t.NetworkContext(), - }, nil + return &Protocol{}, nil } // Protocol implements netlink.Protocol.Protocol. @@ -83,12 +76,13 @@ func (p *Protocol) dumpLinks(ctx context.Context, hdr linux.NetlinkMessageHeader // We always send back an NLMSG_DONE. ms.Multi = true - if p.stack == nil { + stack := inet.StackFromContext(ctx) + if stack == nil { // No network devices. return nil } - for id, i := range p.stack.Interfaces() { + for id, i := range stack.Interfaces() { m := ms.AddMessage(linux.NetlinkMessageHeader{ Type: linux.RTM_NEWLINK, }) @@ -124,12 +118,13 @@ func (p *Protocol) dumpAddrs(ctx context.Context, hdr linux.NetlinkMessageHeader // We always send back an NLMSG_DONE. ms.Multi = true - if p.stack == nil { + stack := inet.StackFromContext(ctx) + if stack == nil { // No network devices. return nil } - for id, as := range p.stack.InterfaceAddrs() { + for id, as := range stack.InterfaceAddrs() { for _, a := range as { m := ms.AddMessage(linux.NetlinkMessageHeader{ Type: linux.RTM_NEWADDR, -- cgit v1.2.3 From 8878a66a565733493e702199b284cd7855f80bf0 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Thu, 17 May 2018 15:05:15 -0700 Subject: Implement sysv shm. PiperOrigin-RevId: 197058289 Change-Id: I3946c25028b7e032be4894d61acb48ac0c24d574 --- pkg/abi/linux/BUILD | 1 + pkg/abi/linux/shm.go | 75 +++++ pkg/refs/refcounter.go | 8 +- pkg/sentry/context/context.go | 20 ++ pkg/sentry/fs/dirent_refs_test.go | 62 ++-- pkg/sentry/kernel/BUILD | 1 + pkg/sentry/kernel/ipc_namespace.go | 15 +- pkg/sentry/kernel/shm/BUILD | 40 +++ pkg/sentry/kernel/shm/device.go | 20 ++ pkg/sentry/kernel/shm/shm.go | 630 +++++++++++++++++++++++++++++++++++ pkg/sentry/kernel/task.go | 3 + pkg/sentry/kernel/task_clone.go | 4 +- pkg/sentry/mm/BUILD | 2 + pkg/sentry/mm/shm.go | 66 ++++ pkg/sentry/syscalls/linux/BUILD | 2 + pkg/sentry/syscalls/linux/linux64.go | 8 +- pkg/sentry/syscalls/linux/sys_shm.go | 155 +++++++++ runsc/boot/loader.go | 2 +- 18 files changed, 1072 insertions(+), 42 deletions(-) create mode 100644 pkg/abi/linux/shm.go create mode 100644 pkg/sentry/kernel/shm/BUILD create mode 100644 pkg/sentry/kernel/shm/device.go create mode 100644 pkg/sentry/kernel/shm/shm.go create mode 100644 pkg/sentry/mm/shm.go create mode 100644 pkg/sentry/syscalls/linux/sys_shm.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index a428e61a3..693ce0fdd 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -51,6 +51,7 @@ go_library( "sched.go", "seccomp.go", "sem.go", + "shm.go", "signal.go", "socket.go", "time.go", diff --git a/pkg/abi/linux/shm.go b/pkg/abi/linux/shm.go new file mode 100644 index 000000000..9149ed094 --- /dev/null +++ b/pkg/abi/linux/shm.go @@ -0,0 +1,75 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package linux + +// shmat(2) flags. Source: include/uapi/linux/shm.h +const ( + SHM_RDONLY = 010000 // Read-only access. + SHM_RND = 020000 // Round attach address to SHMLBA boundary. + SHM_REMAP = 040000 // Take-over region on attach. + SHM_EXEC = 0100000 // Execution access. +) + +// IPCPerm.Mode upper byte flags. Source: include/linux/shm.h +const ( + SHM_DEST = 01000 // Segment will be destroyed on last detach. + SHM_LOCKED = 02000 // Segment will not be swapped. + SHM_HUGETLB = 04000 // Segment will use huge TLB pages. + SHM_NORESERVE = 010000 // Don't check for reservations. +) + +// Additional Linux-only flags for shmctl(2). Source: include/uapi/linux/shm.h +const ( + SHM_LOCK = 11 + SHM_UNLOCK = 12 + SHM_STAT = 13 + SHM_INFO = 14 +) + +// ShmidDS is equivalent to struct shmid64_ds. Source: +// include/uapi/asm-generic/shmbuf.h +type ShmidDS struct { + ShmPerm IPCPerm + ShmSegsz uint64 + ShmAtime TimeT + ShmDtime TimeT + ShmCtime TimeT + ShmCpid int32 + ShmLpid int32 + ShmNattach uint64 + + Unused4 uint64 + Unused5 uint64 +} + +// ShmParams is equivalent to struct shminfo. Source: include/uapi/linux/shm.h +type ShmParams struct { + ShmMax uint64 + ShmMin uint64 + ShmMni uint64 + ShmSeg uint64 + ShmAll uint64 +} + +// ShmInfo is equivalent to struct shm_info. Source: include/uapi/linux/shm.h +type ShmInfo struct { + UsedIDs int32 // Number of currently existing segments. + _ [4]byte + ShmTot uint64 // Total number of shared memory pages. + ShmRss uint64 // Number of resident shared memory pages. + ShmSwp uint64 // Number of swapped shared memory pages. + SwapAttempts uint64 // Unused since Linux 2.4. + SwapSuccesses uint64 // Unused since Linux 2.4. +} diff --git a/pkg/refs/refcounter.go b/pkg/refs/refcounter.go index 1036553c7..3162001e1 100644 --- a/pkg/refs/refcounter.go +++ b/pkg/refs/refcounter.go @@ -194,9 +194,11 @@ type AtomicRefCount struct { weakRefs ilist.List `state:"nosave"` } -// TestReadRefs returns the current reference count of r. Use only for tests. -func (r *AtomicRefCount) TestReadRefs() int64 { - return atomic.LoadInt64(&r.refCount) +// ReadRefs returns the current number of references. The returned count is +// inherently racy and is unsafe to use without external synchronization. +func (r *AtomicRefCount) ReadRefs() int64 { + // Account for the internal -1 offset on refcounts. + return atomic.LoadInt64(&r.refCount) + 1 } // IncRef increments this object's reference count. While the count is kept diff --git a/pkg/sentry/context/context.go b/pkg/sentry/context/context.go index e0dffafba..598c5b4ff 100644 --- a/pkg/sentry/context/context.go +++ b/pkg/sentry/context/context.go @@ -20,6 +20,26 @@ import ( "gvisor.googlesource.com/gvisor/pkg/log" ) +type contextID int + +// Globally accessible values from a context. These keys are defined in the +// context package to resolve dependency cycles by not requiring the caller to +// import packages usually required to get these information. +const ( + // CtxThreadGroupID is the current thread group ID when a context represents + // a task context. The value is represented as an int32. + CtxThreadGroupID contextID = iota +) + +// ThreadGroupIDFromContext returns the current thread group ID when ctx +// represents a task context. +func ThreadGroupIDFromContext(ctx Context) (tgid int32, ok bool) { + if tgid := ctx.Value(CtxThreadGroupID); tgid != nil { + return tgid.(int32), true + } + return 0, false +} + // A Context represents a thread of execution (hereafter "goroutine" to reflect // Go idiosyncrasy). It carries state associated with the goroutine across API // boundaries. diff --git a/pkg/sentry/fs/dirent_refs_test.go b/pkg/sentry/fs/dirent_refs_test.go index 8ce9ba02d..f9dcba316 100644 --- a/pkg/sentry/fs/dirent_refs_test.go +++ b/pkg/sentry/fs/dirent_refs_test.go @@ -33,8 +33,8 @@ func TestWalkPositive(t *testing.T) { ctx := contexttest.Context(t) root := NewDirent(newMockDirInode(ctx, nil), "root") - if got := root.TestReadRefs(); got != 0 { - t.Fatalf("root has a ref count of %d, want %d", got, 0) + if got := root.ReadRefs(); got != 1 { + t.Fatalf("root has a ref count of %d, want %d", got, 1) } name := "d" @@ -43,22 +43,22 @@ func TestWalkPositive(t *testing.T) { t.Fatalf("root.walk(root, %q) got %v, want nil", name, err) } - if got := root.TestReadRefs(); got != 1 { - t.Fatalf("root has a ref count of %d, want %d", got, 1) + if got := root.ReadRefs(); got != 2 { + t.Fatalf("root has a ref count of %d, want %d", got, 2) } - if got := d.TestReadRefs(); got != 0 { - t.Fatalf("child name = %q has a ref count of %d, want %d", d.name, got, 0) + if got := d.ReadRefs(); got != 1 { + t.Fatalf("child name = %q has a ref count of %d, want %d", d.name, got, 1) } d.DecRef() - if got := root.TestReadRefs(); got != 0 { - t.Fatalf("root has a ref count of %d, want %d", got, 0) + if got := root.ReadRefs(); got != 1 { + t.Fatalf("root has a ref count of %d, want %d", got, 1) } - if got := d.TestReadRefs(); got != -1 { - t.Fatalf("child name = %q has a ref count of %d, want %d", d.name, got, -1) + if got := d.ReadRefs(); got != 0 { + t.Fatalf("child name = %q has a ref count of %d, want %d", d.name, got, 0) } root.flush() @@ -76,8 +76,8 @@ func TestWalkNegative(t *testing.T) { root := NewDirent(NewEmptyDir(ctx, nil), "root") mn := root.Inode.InodeOperations.(*mockInodeOperationsLookupNegative) - if got := root.TestReadRefs(); got != 0 { - t.Fatalf("root has a ref count of %d, want %d", got, 0) + if got := root.ReadRefs(); got != 1 { + t.Fatalf("root has a ref count of %d, want %d", got, 1) } name := "d" @@ -88,7 +88,7 @@ func TestWalkNegative(t *testing.T) { } } - if got := root.TestReadRefs(); got != 0 { + if got := root.ReadRefs(); got != 1 { t.Fatalf("root has a ref count of %d, want %d", got, 1) } @@ -110,14 +110,14 @@ func TestWalkNegative(t *testing.T) { t.Fatalf("root found positive child at %q, want negative", name) } - if got := child.(*Dirent).TestReadRefs(); got != 1 { - t.Fatalf("child has a ref count of %d, want %d", got, 1) + if got := child.(*Dirent).ReadRefs(); got != 2 { + t.Fatalf("child has a ref count of %d, want %d", got, 2) } child.DecRef() - if got := child.(*Dirent).TestReadRefs(); got != 0 { - t.Fatalf("child has a ref count of %d, want %d", got, 0) + if got := child.(*Dirent).ReadRefs(); got != 1 { + t.Fatalf("child has a ref count of %d, want %d", got, 1) } if got := len(root.children); got != 1 { @@ -126,7 +126,7 @@ func TestWalkNegative(t *testing.T) { root.DecRef() - if got := root.TestReadRefs(); got != -1 { + if got := root.ReadRefs(); got != 0 { t.Fatalf("root has a ref count of %d, want %d", got, 0) } @@ -184,12 +184,12 @@ func TestHashNegativeToPositive(t *testing.T) { t.Fatalf("got negative Dirent, want positive") } - if got := d.TestReadRefs(); got != 0 { - t.Fatalf("child %q has a ref count of %d, want %d", name, got, 0) + if got := d.ReadRefs(); got != 1 { + t.Fatalf("child %q has a ref count of %d, want %d", name, got, 1) } - if got := root.TestReadRefs(); got != 1 { - t.Fatalf("root has a ref count of %d, want %d", got, 1) + if got := root.ReadRefs(); got != 2 { + t.Fatalf("root has a ref count of %d, want %d", got, 2) } if got := len(root.children); got != 1 { @@ -291,12 +291,12 @@ func TestCreateExtraRefs(t *testing.T) { { desc: "Create caching", root: NewDirent(NewEmptyDir(ctx, NewDirentCache(1)), "root"), - refs: 1, + refs: 2, }, { desc: "Create not caching", root: NewDirent(NewEmptyDir(ctx, nil), "root"), - refs: 0, + refs: 1, }, } { t.Run(test.desc, func(t *testing.T) { @@ -307,7 +307,7 @@ func TestCreateExtraRefs(t *testing.T) { } d := f.Dirent - if got := d.TestReadRefs(); got != test.refs { + if got := d.ReadRefs(); got != test.refs { t.Errorf("dirent has a ref count of %d, want %d", got, test.refs) } }) @@ -347,8 +347,8 @@ func TestRemoveExtraRefs(t *testing.T) { t.Fatalf("root.Remove(root, %q) failed: %v", name, err) } - if got := d.TestReadRefs(); got != 0 { - t.Fatalf("dirent has a ref count of %d, want %d", got, 0) + if got := d.ReadRefs(); got != 1 { + t.Fatalf("dirent has a ref count of %d, want %d", got, 1) } d.DecRef() @@ -406,11 +406,11 @@ func TestRenameExtraRefs(t *testing.T) { newParent.flush() // Expect to have only active references. - if got := renamed.TestReadRefs(); got != 0 { - t.Errorf("renamed has ref count %d, want only active references %d", got, 0) + if got := renamed.ReadRefs(); got != 1 { + t.Errorf("renamed has ref count %d, want only active references %d", got, 1) } - if got := replaced.TestReadRefs(); got != 0 { - t.Errorf("replaced has ref count %d, want only active references %d", got, 0) + if got := replaced.ReadRefs(); got != 1 { + t.Errorf("replaced has ref count %d, want only active references %d", got, 1) } }) } diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 62794cff5..377c94e4c 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -184,6 +184,7 @@ go_library( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/sched", "//pkg/sentry/kernel/semaphore", + "//pkg/sentry/kernel/shm", "//pkg/sentry/kernel/time", "//pkg/sentry/limits", "//pkg/sentry/loader", diff --git a/pkg/sentry/kernel/ipc_namespace.go b/pkg/sentry/kernel/ipc_namespace.go index 78737f58f..3049fead4 100644 --- a/pkg/sentry/kernel/ipc_namespace.go +++ b/pkg/sentry/kernel/ipc_namespace.go @@ -15,18 +15,26 @@ package kernel import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/semaphore" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/shm" ) // IPCNamespace represents an IPC namespace. type IPCNamespace struct { + // User namespace which owns this IPC namespace. Immutable. + userNS *auth.UserNamespace + semaphores *semaphore.Registry + shms *shm.Registry } // NewIPCNamespace creates a new IPC namespace. -func NewIPCNamespace() *IPCNamespace { +func NewIPCNamespace(userNS *auth.UserNamespace) *IPCNamespace { return &IPCNamespace{ + userNS: userNS, semaphores: semaphore.NewRegistry(), + shms: shm.NewRegistry(userNS), } } @@ -35,6 +43,11 @@ func (i *IPCNamespace) SemaphoreRegistry() *semaphore.Registry { return i.semaphores } +// ShmRegistry returns the shm segment registry for this namespace. +func (i *IPCNamespace) ShmRegistry() *shm.Registry { + return i.shms +} + // IPCNamespace returns the task's IPC namespace. func (t *Task) IPCNamespace() *IPCNamespace { t.mu.Lock() diff --git a/pkg/sentry/kernel/shm/BUILD b/pkg/sentry/kernel/shm/BUILD new file mode 100644 index 000000000..182cc1c76 --- /dev/null +++ b/pkg/sentry/kernel/shm/BUILD @@ -0,0 +1,40 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_stateify") + +go_stateify( + name = "shm_state", + srcs = [ + "shm.go", + ], + out = "shm_autogen_state.go", + package = "shm", +) + +go_library( + name = "shm", + srcs = [ + "device.go", + "shm.go", + "shm_autogen_state.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/shm", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/abi/linux", + "//pkg/log", + "//pkg/refs", + "//pkg/sentry/context", + "//pkg/sentry/device", + "//pkg/sentry/fs", + "//pkg/sentry/kernel/auth", + "//pkg/sentry/kernel/time", + "//pkg/sentry/memmap", + "//pkg/sentry/platform", + "//pkg/sentry/usage", + "//pkg/sentry/usermem", + "//pkg/state", + "//pkg/syserror", + ], +) diff --git a/pkg/sentry/kernel/shm/device.go b/pkg/sentry/kernel/shm/device.go new file mode 100644 index 000000000..b0dacdbe0 --- /dev/null +++ b/pkg/sentry/kernel/shm/device.go @@ -0,0 +1,20 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package shm + +import "gvisor.googlesource.com/gvisor/pkg/sentry/device" + +// shmDevice is the kernel shm device. +var shmDevice = device.NewAnonDevice() diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go new file mode 100644 index 000000000..7217e8103 --- /dev/null +++ b/pkg/sentry/kernel/shm/shm.go @@ -0,0 +1,630 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package shm implements sysv shared memory segments. +// +// Known missing features: +// +// - SHM_LOCK/SHM_UNLOCK are no-ops. The sentry currently doesn't implement +// memory locking in general. +// +// - SHM_HUGETLB and related flags for shmget(2) are ignored. There's no easy +// way to implement hugetlb support on a per-map basis, and it has no impact +// on correctness. +// +// - SHM_NORESERVE for shmget(2) is ignored, the sentry doesn't implement swap +// so it's meaningless to reserve space for swap. +// +// - No per-process segment size enforcement. This feature probably isn't used +// much anyways, since Linux sets the per-process limits to the system-wide +// limits by default. +// +// Lock ordering: mm.mappingMu -> shm registry lock -> shm lock +package shm + +import ( + "fmt" + "math" + "sync" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/refs" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// Various limits for shared memory segments. +const ( + // shmsTotalMaxPages is the system-wide limit on all shared memory segments, measured + // in number of pages. + shmsTotalMaxPages = math.MaxInt64 // SHMALL + + // shmMaxSize is the maximum size of a single segment, in bytes. + shmMaxSize = math.MaxInt64 // SHMMAX + + // shmMinSize is the minimum specifiable size of a segment, effectively + // yielding a size rounded up to the next page size. Measured in bytes. + shmMinSize = 1 // SHMMIN + + // shmsTotalMax is the maximum number of segments on the system. + shmsTotalMax = 4096 // SHMMNI +) + +// Registry tracks all shared memory segments in an IPC namespace. The registry +// provides the mechanisms for creating and finding segments, and reporting +// global shm parameters. +type Registry struct { + // userNS owns the IPC namespace this registry belong to. Immutable. + userNS *auth.UserNamespace + + mu sync.Mutex `state:"nosave"` + + // shms maps segment ids to segments. Protected by mu. + shms map[int32]*Shm + + // Sum of the sizes of all existing segments rounded up to page size, in + // units of page size. Protected by mu. + totalPages uint64 + + // lastIDUsed is protected by mu. + lastIDUsed int32 +} + +// NewRegistry creates a new shm registry. +func NewRegistry(userNS *auth.UserNamespace) *Registry { + return &Registry{ + userNS: userNS, + shms: make(map[int32]*Shm), + } +} + +// FindByID looks up a segment given an ID. +func (r *Registry) FindByID(id int32) *Shm { + r.mu.Lock() + defer r.mu.Unlock() + return r.shms[id] +} + +// Precondition: Caller must hold r.mu. +func (r *Registry) findByKey(key int32) *Shm { + for _, v := range r.shms { + if v.key == key { + return v + } + } + return nil +} + +// FindOrCreate looks up or creates a segment in the registry. It's functionally +// analogous to open(2). +func (r *Registry) FindOrCreate(ctx context.Context, pid, key int32, size uint64, mode linux.FileMode, private, create, exclusive bool) (*Shm, error) { + if create && (size < shmMinSize || size > shmMaxSize) { + // "A new segment was to be created and size is less than SHMMIN or + // greater than SHMMAX." - man shmget(2) + return nil, syserror.EINVAL + } + + r.mu.Lock() + defer r.mu.Unlock() + + if len(r.shms) >= shmsTotalMax { + // "All possible shared memory IDs have been taken (SHMMNI) ..." + // - man shmget(2) + return nil, syserror.ENOSPC + } + + if !private { + // Look up an existing segment. + if shm := r.findByKey(key); shm != nil { + shm.mu.Lock() + defer shm.mu.Unlock() + + // Check that caller can access the segment. + if !shm.checkPermissions(ctx, fs.PermsFromMode(mode)) { + // "The user does not have permission to access the shared + // memory segment, and does not have the CAP_IPC_OWNER + // capability in the user namespace that governs its IPC + // namespace." - man shmget(2) + return nil, syserror.EACCES + } + + if size > shm.size { + // "A segment for the given key exists, but size is greater than + // the size of that segment." - man shmget(2) + return nil, syserror.EINVAL + } + + if create && exclusive { + // "IPC_CREAT and IPC_EXCL were specified in shmflg, but a + // shared memory segment already exists for key." + // - man shmget(2) + return nil, syserror.EEXIST + } + + return shm, nil + } + + if !create { + // "No segment exists for the given key, and IPC_CREAT was not + // specified." - man shmget(2) + return nil, syserror.ENOENT + } + } + + var sizeAligned uint64 + if val, ok := usermem.Addr(size).RoundUp(); ok { + sizeAligned = uint64(val) + } else { + return nil, syserror.EINVAL + } + + if numPages := sizeAligned / usermem.PageSize; r.totalPages+numPages > shmsTotalMaxPages { + // "... allocating a segment of the requested size would cause the + // system to exceed the system-wide limit on shared memory (SHMALL)." + // - man shmget(2) + return nil, syserror.ENOSPC + } + + // Need to create a new segment. + creator := fs.FileOwnerFromContext(ctx) + perms := fs.FilePermsFromMode(mode) + return r.newShm(ctx, pid, key, creator, perms, size) +} + +// newShm creates a new segment in the registry. +func (r *Registry) newShm(ctx context.Context, pid, key int32, creator fs.FileOwner, perms fs.FilePermissions, size uint64) (*Shm, error) { + p := platform.FromContext(ctx) + if p == nil { + panic(fmt.Sprintf("context.Context %T lacks non-nil value for key %T", ctx, platform.CtxPlatform)) + } + + effectiveSize := uint64(usermem.Addr(size).MustRoundUp()) + fr, err := p.Memory().Allocate(effectiveSize, usage.Anonymous) + if err != nil { + return nil, err + } + + shm := &Shm{ + p: p, + registry: r, + creator: creator, + size: size, + effectiveSize: effectiveSize, + fr: fr, + key: key, + perms: perms, + owner: creator, + creatorPID: pid, + changeTime: ktime.NowFromContext(ctx), + } + + // Find the next available ID. + for id := r.lastIDUsed + 1; id != r.lastIDUsed; id++ { + // Handle wrap around. + if id < 0 { + id = 0 + continue + } + if r.shms[id] == nil { + r.lastIDUsed = id + r.shms[id] = shm + shm.ID = id + + r.totalPages += effectiveSize / usermem.PageSize + + return shm, nil + } + } + + log.Warningf("Shm ids exhuasted, they may be leaking") + return nil, syserror.ENOSPC +} + +// IPCInfo reports global parameters for sysv shared memory segments on this +// system. See shmctl(IPC_INFO). +func (r *Registry) IPCInfo() *linux.ShmParams { + return &linux.ShmParams{ + ShmMax: shmMaxSize, + ShmMin: shmMinSize, + ShmMni: shmsTotalMax, + ShmSeg: shmsTotalMax, // Linux also sets this to SHMMNI. + ShmAll: shmsTotalMaxPages, + } +} + +// ShmInfo reports linux-specific global parameters for sysv shared memory +// segments on this system. See shmctl(SHM_INFO). +func (r *Registry) ShmInfo() *linux.ShmInfo { + r.mu.Lock() + defer r.mu.Unlock() + + return &linux.ShmInfo{ + UsedIDs: int32(r.lastIDUsed), + ShmTot: r.totalPages, + ShmRss: r.totalPages, // We could probably get a better estimate from memory accounting. + ShmSwp: 0, // No reclaim at the moment. + } +} + +// remove unregisters a segment from this registry, preventing it from being +// discovered in the future. Caller is responsible for ensuring s is destroyed. +// +// Precondition: To preserve lock ordering, caller must not hold s.mu. +func (r *Registry) remove(s *Shm) { + r.mu.Lock() + defer r.mu.Unlock() + delete(r.shms, s.ID) + r.totalPages -= s.effectiveSize / usermem.PageSize +} + +// Shm represents a single shared memory segment. +// +// Shm segment are backed directly by an allocation from platform +// memory. Segments are always mapped as a whole, greatly simplifying how +// mappings are tracked. However note that mremap and munmap calls may cause the +// vma for a segment to become fragmented; which requires special care when +// unmapping a segment. See mm/shm.go. +// +// Segments persist until they are explicitly marked for destruction via +// shmctl(SHM_RMID). +// +// Shm implements memmap.Mappable and memmap.MappingIdentity. +type Shm struct { + // AtomicRefCount tracks the number of references to this segment from + // maps. A segment always holds a reference to itself, until it's marked for + // destruction. + refs.AtomicRefCount + + p platform.Platform + + // registry points to the shm registry containing this segment. Immutable. + registry *Registry + + // ID is the kernel identifier for this segment. Immutable. + ID int32 + + // creator is the user that created the segment. Immutable. + creator fs.FileOwner + + // size is the requested size of the segment at creation, in + // bytes. Immutable. + size uint64 + + // effectiveSize of the segment, rounding up to the next page + // boundary. Immutable. + // + // Invariant: effectiveSize must be a multiple of usermem.PageSize. + effectiveSize uint64 + + // fr is the offset into platform.Memory() that backs this contents of this + // segment. Immutable. + fr platform.FileRange + + // key is the public identifier for this segment. + key int32 + + // mu protects all fields below. + mu sync.Mutex `state:"nosave"` + + // perms is the access permissions for the segment. + perms fs.FilePermissions + + // owner of this segment. + owner fs.FileOwner + // attachTime is updated on every successful shmat. + attachTime ktime.Time + // detachTime is updated on every successful shmdt. + detachTime ktime.Time + // changeTime is updated on every successful changes to the segment via + // shmctl(IPC_SET). + changeTime ktime.Time + + // creatorPID is the PID of the process that created the segment. + creatorPID int32 + // lastAttachDetachPID is the pid of the process that issued the last shmat + // or shmdt syscall. + lastAttachDetachPID int32 + + // pendingDestruction indicates the segment was marked as destroyed through + // shmctl(IPC_RMID). When marked as destroyed, the segment will not be found + // in the registry and can no longer be attached. When the last user + // detaches from the segment, it is destroyed. Protected by mu. + pendingDestruction bool +} + +// MappedName implements memmap.MappingIdentity.MappedName. +func (s *Shm) MappedName(ctx context.Context) string { + return fmt.Sprintf("SYSV%08d", s.key) +} + +// DeviceID implements memmap.MappingIdentity.DeviceID. +func (s *Shm) DeviceID() uint64 { + return shmDevice.DeviceID() +} + +// InodeID implements memmap.MappingIdentity.InodeID. +func (s *Shm) InodeID() uint64 { + // "shmid gets reported as "inode#" in /proc/pid/maps. proc-ps tools use + // this. Changing this will break them." -- Linux, ipc/shm.c:newseg() + return uint64(s.ID) +} + +// DecRef overrides refs.RefCount.DecRef with a destructor. +func (s *Shm) DecRef() { + s.DecRefWithDestructor(s.destroy) +} + +// Msync implements memmap.MappingIdentity.Msync. Msync is a no-op for shm +// segments. +func (s *Shm) Msync(context.Context, memmap.MappableRange) error { + return nil +} + +// AddMapping implements memmap.Mappable.AddMapping. +func (s *Shm) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) error { + s.mu.Lock() + defer s.mu.Unlock() + s.attachTime = ktime.NowFromContext(ctx) + if pid, ok := context.ThreadGroupIDFromContext(ctx); ok { + s.lastAttachDetachPID = pid + } else { + // AddMapping is called during a syscall, so ctx should always be a task + // context. + log.Warningf("Adding mapping to shm %+v but couldn't get the current pid; not updating the last attach pid", s) + } + return nil +} + +// RemoveMapping implements memmap.Mappable.RemoveMapping. +func (s *Shm) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) { + s.mu.Lock() + defer s.mu.Unlock() + // TODO: RemoveMapping may be called during task exit, when ctx + // is context.Background. Gracefully handle missing clocks. Failing to + // update the detach time in these cases is ok, since no one can observe the + // omission. + if clock := ktime.RealtimeClockFromContext(ctx); clock != nil { + s.detachTime = clock.Now() + } + + // If called from a non-task context we also won't have a threadgroup + // id. Silently skip updating the lastAttachDetachPid in that case. + if pid, ok := context.ThreadGroupIDFromContext(ctx); ok { + s.lastAttachDetachPID = pid + } else { + log.Debugf("Couldn't obtain pid when removing mapping to shm %+v, not updating the last detach pid.", s) + } +} + +// CopyMapping implements memmap.Mappable.CopyMapping. +func (s *Shm) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error { + return nil +} + +// Translate implements memmap.Mappable.Translate. +func (s *Shm) Translate(ctx context.Context, required, optional memmap.MappableRange, at usermem.AccessType) ([]memmap.Translation, error) { + var err error + if required.End > s.fr.Length() { + err = &memmap.BusError{syserror.EFAULT} + } + if source := optional.Intersect(memmap.MappableRange{0, s.fr.Length()}); source.Length() != 0 { + return []memmap.Translation{ + { + Source: source, + File: s.p.Memory(), + Offset: s.fr.Start + source.Start, + }, + }, err + } + return nil, err +} + +// InvalidateUnsavable implements memmap.Mappable.InvalidateUnsavable. +func (s *Shm) InvalidateUnsavable(ctx context.Context) error { + return nil +} + +// AttachOpts describes various flags passed to shmat(2). +type AttachOpts struct { + Execute bool + Readonly bool + Remap bool +} + +// ConfigureAttach creates an mmap configuration for the segment with the +// requested attach options. +// +// ConfigureAttach returns with a ref on s on success. The caller should drop +// this once the map is installed. This reference prevents s from being +// destroyed before the returned configuration is used. +func (s *Shm) ConfigureAttach(ctx context.Context, addr usermem.Addr, opts AttachOpts) (memmap.MMapOpts, error) { + s.mu.Lock() + defer s.mu.Unlock() + if s.pendingDestruction && s.ReadRefs() == 0 { + return memmap.MMapOpts{}, syserror.EIDRM + } + + if !s.checkPermissions(ctx, fs.PermMask{ + Read: true, + Write: !opts.Readonly, + Execute: opts.Execute, + }) { + // "The calling process does not have the required permissions for the + // requested attach type, and does not have the CAP_IPC_OWNER capability + // in the user namespace that governs its IPC namespace." - man shmat(2) + return memmap.MMapOpts{}, syserror.EACCES + } + s.IncRef() + return memmap.MMapOpts{ + Length: s.size, + Offset: 0, + Addr: addr, + Fixed: opts.Remap, + Perms: usermem.AccessType{ + Read: true, + Write: !opts.Readonly, + Execute: opts.Execute, + }, + MaxPerms: usermem.AnyAccess, + Mappable: s, + MappingIdentity: s, + }, nil +} + +// EffectiveSize returns the size of the underlying shared memory segment. This +// may be larger than the requested size at creation, due to rounding to page +// boundaries. +func (s *Shm) EffectiveSize() uint64 { + return s.effectiveSize +} + +// IPCStat returns information about a shm. See shmctl(IPC_STAT). +func (s *Shm) IPCStat(ctx context.Context) (*linux.ShmidDS, error) { + s.mu.Lock() + defer s.mu.Unlock() + + // "The caller must have read permission on the shared memory segment." + // - man shmctl(2) + if !s.checkPermissions(ctx, fs.PermMask{Read: true}) { + // "IPC_STAT or SHM_STAT is requested and shm_perm.mode does not allow + // read access for shmid, and the calling process does not have the + // CAP_IPC_OWNER capability in the user namespace that governs its IPC + // namespace." - man shmctl(2) + return nil, syserror.EACCES + } + + var mode uint16 + if s.pendingDestruction { + mode |= linux.SHM_DEST + } + creds := auth.CredentialsFromContext(ctx) + + nattach := uint64(s.ReadRefs()) + // Don't report the self-reference we keep prior to being marked for + // destruction. However, also don't report a count of -1 for segments marked + // as destroyed, with no mappings. + if !s.pendingDestruction { + nattach-- + } + + ds := &linux.ShmidDS{ + ShmPerm: linux.IPCPerm{ + Key: uint32(s.key), + UID: uint32(creds.UserNamespace.MapFromKUID(s.owner.UID)), + GID: uint32(creds.UserNamespace.MapFromKGID(s.owner.GID)), + CUID: uint32(creds.UserNamespace.MapFromKUID(s.creator.UID)), + CGID: uint32(creds.UserNamespace.MapFromKGID(s.creator.GID)), + Mode: mode | uint16(s.perms.LinuxMode()), + Seq: 0, // IPC sequences not supported. + }, + ShmSegsz: s.size, + ShmAtime: s.attachTime.TimeT(), + ShmDtime: s.detachTime.TimeT(), + ShmCtime: s.changeTime.TimeT(), + ShmCpid: s.creatorPID, + ShmLpid: s.lastAttachDetachPID, + ShmNattach: nattach, + } + + return ds, nil +} + +// Set modifies attributes for a segment. See shmctl(IPC_SET). +func (s *Shm) Set(ctx context.Context, ds *linux.ShmidDS) error { + s.mu.Lock() + defer s.mu.Unlock() + + if !s.checkOwnership(ctx) { + return syserror.EPERM + } + + creds := auth.CredentialsFromContext(ctx) + uid := creds.UserNamespace.MapToKUID(auth.UID(ds.ShmPerm.UID)) + gid := creds.UserNamespace.MapToKGID(auth.GID(ds.ShmPerm.GID)) + if !uid.Ok() || !gid.Ok() { + return syserror.EINVAL + } + + // User may only modify the lower 9 bits of the mode. All the other bits are + // always 0 for the underlying inode. + mode := linux.FileMode(ds.ShmPerm.Mode & 0x1ff) + s.perms = fs.FilePermsFromMode(mode) + + s.owner.UID = uid + s.owner.GID = gid + + s.changeTime = ktime.NowFromContext(ctx) + return nil +} + +func (s *Shm) destroy() { + s.registry.remove(s) + s.p.Memory().DecRef(s.fr) +} + +// MarkDestroyed marks a shm for destruction. The shm is actually destroyed once +// it has no references. See shmctl(IPC_RMID). +func (s *Shm) MarkDestroyed() { + s.mu.Lock() + defer s.mu.Unlock() + // Prevent the segment from being found in the registry. + s.key = linux.IPC_PRIVATE + s.pendingDestruction = true + s.DecRef() +} + +// checkOwnership verifies whether a segment may be accessed by ctx as an +// owner. See ipc/util.c:ipcctl_pre_down_nolock() in Linux. +// +// Precondition: Caller must hold s.mu. +func (s *Shm) checkOwnership(ctx context.Context) bool { + creds := auth.CredentialsFromContext(ctx) + if s.owner.UID == creds.EffectiveKUID || s.creator.UID == creds.EffectiveKUID { + return true + } + + // Tasks with CAP_SYS_ADMIN may bypass ownership checks. Strangely, Linux + // doesn't use CAP_IPC_OWNER for this despite CAP_IPC_OWNER being documented + // for use to "override IPC ownership checks". + return creds.HasCapabilityIn(linux.CAP_SYS_ADMIN, s.registry.userNS) +} + +// checkPermissions verifies whether a segment is accessible by ctx for access +// described by req. See ipc/util.c:ipcperms() in Linux. +// +// Precondition: Caller must hold s.mu. +func (s *Shm) checkPermissions(ctx context.Context, req fs.PermMask) bool { + creds := auth.CredentialsFromContext(ctx) + + p := s.perms.Other + if s.owner.UID == creds.EffectiveKUID { + p = s.perms.User + } else if creds.InGroup(s.owner.GID) { + p = s.perms.Group + } + if p.SupersetOf(req) { + return true + } + + // Tasks with CAP_IPC_OWNER may bypass permission checks. + return creds.HasCapabilityIn(linux.CAP_IPC_OWNER, s.registry.userNS) +} diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 490f795c2..7763050a5 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -21,6 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/bpf" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/inet" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" @@ -559,6 +560,8 @@ func (t *Task) Value(key interface{}) interface{} { return t case auth.CtxCredentials: return t.creds + case context.CtxThreadGroupID: + return int32(t.ThreadGroup().ID()) case fs.CtxRoot: return t.FSContext().RootDirectory() case inet.CtxStack: diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 3a74abdfb..0c2427952 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -197,7 +197,7 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { if opts.NewIPCNamespace { // Note that "If CLONE_NEWIPC is set, then create the process in a new IPC // namespace" - ipcns = NewIPCNamespace() + ipcns = NewIPCNamespace(userns) } tc, err := t.tc.Fork(t, !opts.NewAddressSpace) @@ -449,7 +449,7 @@ func (t *Task) Unshare(opts *SharingOptions) error { } // Note that "If CLONE_NEWIPC is set, then create the process in a new IPC // namespace" - t.ipcns = NewIPCNamespace() + t.ipcns = NewIPCNamespace(t.creds.UserNamespace) } if opts.NewFiles { oldFDMap := t.tr.FDMap diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index 39bde2be3..258389bb2 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -107,6 +107,7 @@ go_library( "pma_set.go", "proc_pid_maps.go", "save_restore.go", + "shm.go", "special_mappable.go", "syscalls.go", "vma.go", @@ -123,6 +124,7 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/proc/seqfile", + "//pkg/sentry/kernel/shm", "//pkg/sentry/limits", "//pkg/sentry/memmap", "//pkg/sentry/platform", diff --git a/pkg/sentry/mm/shm.go b/pkg/sentry/mm/shm.go new file mode 100644 index 000000000..bab137a5a --- /dev/null +++ b/pkg/sentry/mm/shm.go @@ -0,0 +1,66 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package mm + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/shm" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// DetachShm unmaps a sysv shared memory segment. +func (mm *MemoryManager) DetachShm(ctx context.Context, addr usermem.Addr) error { + if addr != addr.RoundDown() { + // "... shmaddr is not aligned on a page boundary." - man shmdt(2) + return syserror.EINVAL + } + + var detached *shm.Shm + mm.mappingMu.Lock() + defer mm.mappingMu.Unlock() + + // Find and remove the first vma containing an address >= addr that maps a + // segment originally attached at addr. + vseg := mm.vmas.LowerBoundSegment(addr) + for vseg.Ok() { + vma := vseg.ValuePtr() + if shm, ok := vma.mappable.(*shm.Shm); ok && vseg.Start() >= addr && uint64(vseg.Start()-addr) == vma.off { + detached = shm + vseg = mm.unmapLocked(ctx, vseg.Range()).NextSegment() + break + } else { + vseg = vseg.NextSegment() + } + } + + if detached == nil { + // There is no shared memory segment attached at addr. + return syserror.EINVAL + } + + // Remove all vmas that could have been created by the same attach. + end := addr + usermem.Addr(detached.EffectiveSize()) + for vseg.Ok() && vseg.End() <= end { + vma := vseg.ValuePtr() + if vma.mappable == detached && uint64(vseg.Start()-addr) == vma.off { + vseg = mm.unmapLocked(ctx, vseg.Range()).NextSegment() + } else { + vseg = vseg.NextSegment() + } + } + + return nil +} diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index bc67ebf30..f9e0a4be3 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -44,6 +44,7 @@ go_library( "sys_rusage.go", "sys_sched.go", "sys_sem.go", + "sys_shm.go", "sys_signal.go", "sys_socket.go", "sys_stat.go", @@ -84,6 +85,7 @@ go_library( "//pkg/sentry/kernel/pipe", "//pkg/sentry/kernel/sched", "//pkg/sentry/kernel/semaphore", + "//pkg/sentry/kernel/shm", "//pkg/sentry/kernel/time", "//pkg/sentry/limits", "//pkg/sentry/memmap", diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index 44db2d582..237c61007 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -75,9 +75,9 @@ var AMD64 = &kernel.SyscallTable{ 26: Msync, 27: Mincore, 28: Madvise, - // 29: Shmget, TODO - // 30: Shmat, TODO - // 31: Shmctl, TODO + 29: Shmget, + 30: Shmat, + 31: Shmctl, 32: Dup, 33: Dup2, 34: Pause, @@ -113,7 +113,7 @@ var AMD64 = &kernel.SyscallTable{ 64: Semget, 65: Semop, 66: Semctl, - // 67: Shmdt, TODO + 67: Shmdt, // 68: Msgget, TODO // 69: Msgsnd, TODO // 70: Msgrcv, TODO diff --git a/pkg/sentry/syscalls/linux/sys_shm.go b/pkg/sentry/syscalls/linux/sys_shm.go new file mode 100644 index 000000000..48ff1d5f0 --- /dev/null +++ b/pkg/sentry/syscalls/linux/sys_shm.go @@ -0,0 +1,155 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package linux + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/shm" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// Shmget implements shmget(2). +func Shmget(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + key := args[0].Int() + size := uint64(args[1].SizeT()) + flag := args[2].Int() + + private := key == linux.IPC_PRIVATE + create := flag&linux.IPC_CREAT == linux.IPC_CREAT + exclusive := flag&linux.IPC_EXCL == linux.IPC_EXCL + mode := linux.FileMode(flag & 0777) + + pid := int32(t.ThreadGroup().ID()) + r := t.IPCNamespace().ShmRegistry() + segment, err := r.FindOrCreate(t, pid, key, size, mode, private, create, exclusive) + if err != nil { + return 0, nil, err + } + return uintptr(segment.ID), nil, nil +} + +// findSegment retrives a shm segment by the given id. +func findSegment(t *kernel.Task, id int32) (*shm.Shm, error) { + r := t.IPCNamespace().ShmRegistry() + segment := r.FindByID(id) + if segment == nil { + // No segment with provided id. + return nil, syserror.EINVAL + } + return segment, nil +} + +// Shmat implements shmat(2). +func Shmat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + id := args[0].Int() + addr := args[1].Pointer() + flag := args[2].Int() + + segment, err := findSegment(t, id) + if err != nil { + return 0, nil, syserror.EINVAL + } + + opts, err := segment.ConfigureAttach(t, addr, shm.AttachOpts{ + Execute: flag&linux.SHM_EXEC == linux.SHM_EXEC, + Readonly: flag&linux.SHM_RDONLY == linux.SHM_RDONLY, + Remap: flag&linux.SHM_REMAP == linux.SHM_REMAP, + }) + if err != nil { + return 0, nil, err + } + defer segment.DecRef() + addr, err = t.MemoryManager().MMap(t, opts) + return uintptr(addr), nil, err +} + +// Shmdt implements shmdt(2). +func Shmdt(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + addr := args[0].Pointer() + err := t.MemoryManager().DetachShm(t, addr) + return 0, nil, err +} + +// Shmctl implements shmctl(2). +func Shmctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + id := args[0].Int() + cmd := args[1].Int() + buf := args[2].Pointer() + + r := t.IPCNamespace().ShmRegistry() + + switch cmd { + case linux.SHM_STAT: + // Technically, we should be treating id as "an index into the kernel's + // internal array that maintains information about all shared memory + // segments on the system". Since we don't track segments in an array, + // we'll just pretend the shmid is the index and do the same thing as + // IPC_STAT. Linux also uses the index as the shmid. + fallthrough + case linux.IPC_STAT: + segment, err := findSegment(t, id) + if err != nil { + return 0, nil, syserror.EINVAL + } + + stat, err := segment.IPCStat(t) + if err == nil { + _, err = t.CopyOut(buf, stat) + } + return 0, nil, err + + case linux.IPC_INFO: + params := r.IPCInfo() + _, err := t.CopyOut(buf, params) + return 0, nil, err + + case linux.SHM_INFO: + info := r.ShmInfo() + _, err := t.CopyOut(buf, info) + return 0, nil, err + } + + // Remaining commands refer to a specific segment. + segment, err := findSegment(t, id) + if err != nil { + return 0, nil, syserror.EINVAL + } + + switch cmd { + case linux.IPC_SET: + var ds linux.ShmidDS + _, err = t.CopyIn(buf, &ds) + if err != nil { + return 0, nil, err + } + err = segment.Set(t, &ds) + return 0, nil, err + + case linux.IPC_RMID: + segment.MarkDestroyed() + return 0, nil, nil + + case linux.SHM_LOCK, linux.SHM_UNLOCK: + // We currently do not support memmory locking anywhere. + // mlock(2)/munlock(2) are currently stubbed out as no-ops so do the + // same here. + return 0, nil, nil + + default: + return 0, nil, syserror.EINVAL + } +} diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 0ff54d349..566f2eb46 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -146,7 +146,7 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console // not configurable from runtime spec. utsns := kernel.NewUTSNamespace(spec.Hostname, "", creds.UserNamespace) - ipcns := kernel.NewIPCNamespace() + ipcns := kernel.NewIPCNamespace(creds.UserNamespace) if err := enableStrace(conf); err != nil { return nil, fmt.Errorf("failed to enable strace: %v", err) -- cgit v1.2.3 From b904250b862c5c14da84e08b6a5400c7bf2458b0 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Thu, 17 May 2018 15:37:19 -0700 Subject: Fix capability check for sysv semaphores. Capabilities for sysv sem operations were being checked against the current task's user namespace. They should be checked against the user namespace owning the ipc namespace for the sems instead, per ipc/util.c:ipcperms(). PiperOrigin-RevId: 197063111 Change-Id: Iba29486b316f2e01ee331dda4e48a6ab7960d589 --- pkg/sentry/kernel/ipc_namespace.go | 2 +- pkg/sentry/kernel/semaphore/semaphore.go | 15 ++++++++++++--- pkg/sentry/kernel/semaphore/semaphore_test.go | 2 +- 3 files changed, 14 insertions(+), 5 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/ipc_namespace.go b/pkg/sentry/kernel/ipc_namespace.go index 3049fead4..a86bda77b 100644 --- a/pkg/sentry/kernel/ipc_namespace.go +++ b/pkg/sentry/kernel/ipc_namespace.go @@ -33,7 +33,7 @@ type IPCNamespace struct { func NewIPCNamespace(userNS *auth.UserNamespace) *IPCNamespace { return &IPCNamespace{ userNS: userNS, - semaphores: semaphore.NewRegistry(), + semaphores: semaphore.NewRegistry(userNS), shms: shm.NewRegistry(userNS), } } diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index fb8c2f98c..e9027dc14 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -43,6 +43,8 @@ const ( // Registry maintains a set of semaphores that can be found by key or ID. type Registry struct { + // userNS owning the ipc name this registry belongs to. Immutable. + userNS *auth.UserNamespace // mu protects all fields below. mu sync.Mutex `state:"nosave"` semaphores map[int32]*Set @@ -51,6 +53,9 @@ type Registry struct { // Set represents a set of semaphores that can be operated atomically. type Set struct { + // registry owning this sem set. Immutable. + registry *Registry + // Id is a handle that identifies the set. ID int32 @@ -90,8 +95,11 @@ type waiter struct { } // NewRegistry creates a new semaphore set registry. -func NewRegistry() *Registry { - return &Registry{semaphores: make(map[int32]*Set)} +func NewRegistry(userNS *auth.UserNamespace) *Registry { + return &Registry{ + userNS: userNS, + semaphores: make(map[int32]*Set), + } } // FindOrCreate searches for a semaphore set that matches 'key'. If not found, @@ -175,6 +183,7 @@ func (r *Registry) RemoveID(id int32, creds *auth.Credentials) error { func (r *Registry) newSet(ctx context.Context, key int32, owner, creator fs.FileOwner, perms fs.FilePermissions, nsems int32) (*Set, error) { set := &Set{ + registry: r, key: key, owner: owner, creator: owner, @@ -415,7 +424,7 @@ func (s *Set) checkCredentials(creds *auth.Credentials) bool { } func (s *Set) checkCapability(creds *auth.Credentials) bool { - return creds.HasCapability(linux.CAP_IPC_OWNER) && creds.UserNamespace.MapFromKUID(s.owner.UID).Ok() + return creds.HasCapabilityIn(linux.CAP_IPC_OWNER, s.registry.userNS) && creds.UserNamespace.MapFromKUID(s.owner.UID).Ok() } func (s *Set) checkPerms(creds *auth.Credentials, reqPerms fs.PermMask) bool { diff --git a/pkg/sentry/kernel/semaphore/semaphore_test.go b/pkg/sentry/kernel/semaphore/semaphore_test.go index 1c6a2e1e9..f9eb382e9 100644 --- a/pkg/sentry/kernel/semaphore/semaphore_test.go +++ b/pkg/sentry/kernel/semaphore/semaphore_test.go @@ -136,7 +136,7 @@ func TestNoWait(t *testing.T) { func TestUnregister(t *testing.T) { ctx := contexttest.Context(t) - r := NewRegistry() + r := NewRegistry(auth.NewRootUserNamespace()) set, err := r.FindOrCreate(ctx, 123, 2, linux.FileMode(0x600), true, true, true) if err != nil { t.Fatalf("FindOrCreate() failed, err: %v", err) -- cgit v1.2.3 From 6728f09910bd9f7633f277fafe6945cfaa2abf42 Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Fri, 8 Jun 2018 15:00:29 -0700 Subject: Fix sigaltstack semantics. Walking off the bottom of the sigaltstack, for example with recursive faults, results in forced signal delivery, not resetting the stack or pushing signal stack to whatever happens to lie below the signal stack. PiperOrigin-RevId: 199856085 Change-Id: I0004d2523f0df35d18714de2685b3eaa147837e0 --- pkg/sentry/arch/arch.go | 2 +- pkg/sentry/arch/signal_amd64.go | 16 ++++++--- pkg/sentry/arch/signal_stack.go | 11 ++++-- pkg/sentry/kernel/task_signals.go | 60 ++++++++++++++++++++++----------- pkg/sentry/syscalls/linux/sys_signal.go | 16 ++++----- 5 files changed, 70 insertions(+), 35 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go index 021789e4b..0189e958d 100644 --- a/pkg/sentry/arch/arch.go +++ b/pkg/sentry/arch/arch.go @@ -158,7 +158,7 @@ type Context interface { // rt is true if SignalRestore is being entered from rt_sigreturn and // false if SignalRestore is being entered from sigreturn. // SignalRestore returns the thread's new signal mask. - SignalRestore(st *Stack, rt bool) (linux.SignalSet, error) + SignalRestore(st *Stack, rt bool) (linux.SignalSet, SignalStack, error) // CPUIDEmulate emulates a CPUID instruction according to current register state. CPUIDEmulate(l log.Logger) diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index 4040b530f..c1d743f38 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -377,6 +377,14 @@ func (c *context64) SignalSetup(st *Stack, act *SignalAct, info *SignalInfo, alt sp = frameBottom + usermem.Addr(frameSize) st.Bottom = sp + // Prior to proceeding, figure out if the frame will exhaust the range + // for the signal stack. This is not allowed, and should immediately + // force signal delivery (reverting to the default handler). + if act.IsOnStack() && alt.IsEnabled() && !alt.Contains(frameBottom) { + return syscall.EFAULT + } + + // Adjust the code. info.FixSignalCodeForUser() // Set up the stack frame. @@ -422,15 +430,15 @@ func (c *context64) SignalSetup(st *Stack, act *SignalAct, info *SignalInfo, alt // SignalRestore implements Context.SignalRestore. (Compare to Linux's // arch/x86/kernel/signal.c:sys_rt_sigreturn().) -func (c *context64) SignalRestore(st *Stack, rt bool) (linux.SignalSet, error) { +func (c *context64) SignalRestore(st *Stack, rt bool) (linux.SignalSet, SignalStack, error) { // Copy out the stack frame. var uc UContext64 if _, err := st.Pop(&uc); err != nil { - return 0, err + return 0, SignalStack{}, err } var info SignalInfo if _, err := st.Pop(&info); err != nil { - return 0, err + return 0, SignalStack{}, err } // Restore registers. @@ -472,5 +480,5 @@ func (c *context64) SignalRestore(st *Stack, rt bool) (linux.SignalSet, error) { log.Infof("sigreturn unable to restore application fpstate") } - return uc.Sigset, nil + return uc.Sigset, uc.Stack, nil } diff --git a/pkg/sentry/arch/signal_stack.go b/pkg/sentry/arch/signal_stack.go index 7c6531d79..ba43dd1d4 100644 --- a/pkg/sentry/arch/signal_stack.go +++ b/pkg/sentry/arch/signal_stack.go @@ -39,12 +39,19 @@ func (s SignalStack) Top() usermem.Addr { return usermem.Addr(s.Addr + s.Size) } -// SetOnStack marks this signal stack as in use. (This is only called on copies -// sent to user applications, so there's no corresponding ClearOnStack.) +// SetOnStack marks this signal stack as in use. +// +// Note that there is no corresponding ClearOnStack, and that this should only +// be called on copies that are serialized to userspace. func (s *SignalStack) SetOnStack() { s.Flags |= SignalStackFlagOnStack } +// Contains checks if the stack pointer is within this stack. +func (s *SignalStack) Contains(sp usermem.Addr) bool { + return usermem.Addr(s.Addr) < sp && sp <= usermem.Addr(s.Addr+s.Size) +} + // NativeSignalStack is a type that is equivalent to stack_t in the guest // architecture. type NativeSignalStack interface { diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index e4ef7fd67..91f6c0874 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -212,7 +212,9 @@ func (t *Task) deliverSignal(info *arch.SignalInfo, act arch.SignalAct) taskRunS // Try to deliver the signal to the user-configured handler. t.Debugf("Signal %d: delivering to handler", info.Signo) if err := t.deliverSignalToHandler(info, act); err != nil { - t.Warningf("Failed to deliver signal %+v to user handler: %v", info, err) + // This is not a warning, it can occur during normal operation. + t.Debugf("Failed to deliver signal %+v to user handler: %v", info, err) + // Send a forced SIGSEGV. If the signal that couldn't be delivered // was a SIGSEGV, force the handler to SIG_DFL. t.forceSignal(linux.SIGSEGV, linux.Signal(info.Signo) == linux.SIGSEGV /* unconditional */) @@ -241,7 +243,7 @@ func (t *Task) deliverSignalToHandler(info *arch.SignalInfo, act arch.SignalAct) alt := t.signalStack if act.IsOnStack() && alt.IsEnabled() { alt.SetOnStack() - if !t.OnSignalStack(alt) { + if !alt.Contains(sp) { sp = usermem.Addr(alt.Top()) } } @@ -275,18 +277,20 @@ var ctrlResume = &SyscallControl{ignoreReturn: true} // rt is true). func (t *Task) SignalReturn(rt bool) (*SyscallControl, error) { st := t.Stack() - sigset, err := t.Arch().SignalRestore(st, rt) + sigset, alt, err := t.Arch().SignalRestore(st, rt) if err != nil { return nil, err } + // Attempt to record the given signal stack. Note that we silently + // ignore failures here, as does Linux. Only an EFAULT may be + // generated, but SignalRestore has already deserialized the entire + // frame successfully. + t.SetSignalStack(alt) + // Restore our signal mask. SIGKILL and SIGSTOP should not be blocked. t.SetSignalMask(sigset &^ UnblockableSignals) - // TODO: sys_rt_sigreturn also calls restore_altstack from - // uc.stack, allowing the signal handler to implicitly mutate the signal - // stack. - return ctrlResume, nil } @@ -624,23 +628,41 @@ func (t *Task) SetSavedSignalMask(mask linux.SignalSet) { // SignalStack returns the task-private signal stack. func (t *Task) SignalStack() arch.SignalStack { - return t.signalStack + alt := t.signalStack + if t.onSignalStack(alt) { + alt.Flags |= arch.SignalStackFlagOnStack + } + return alt } -// OnSignalStack returns true if, when the task resumes running, it will run on -// the task-private signal stack. -func (t *Task) OnSignalStack(s arch.SignalStack) bool { +// onSignalStack returns true if the task is executing on the given signal stack. +func (t *Task) onSignalStack(alt arch.SignalStack) bool { sp := usermem.Addr(t.Arch().Stack()) - return usermem.Addr(s.Addr) <= sp && sp < usermem.Addr(s.Addr+s.Size) + return alt.Contains(sp) } -// SetSignalStack sets the task-private signal stack and clears the -// SignalStackFlagDisable, since we have a signal stack. -func (t *Task) SetSignalStack(alt arch.SignalStack) error { - // Mask out irrelevant parts: only disable matters. - alt.Flags &= arch.SignalStackFlagDisable - t.signalStack = alt - return nil +// SetSignalStack sets the task-private signal stack. +// +// This value may not be changed if the task is currently executing on the +// signal stack, i.e. if t.onSignalStack returns true. In this case, this +// function will return false. Otherwise, true is returned. +func (t *Task) SetSignalStack(alt arch.SignalStack) bool { + // Check that we're not executing on the stack. + if t.onSignalStack(t.signalStack) { + return false + } + + if alt.Flags&arch.SignalStackFlagDisable != 0 { + // Don't record anything beyond the flags. + t.signalStack = arch.SignalStack{ + Flags: arch.SignalStackFlagDisable, + } + } else { + // Mask out irrelevant parts: only disable matters. + alt.Flags &= arch.SignalStackFlagDisable + t.signalStack = alt + } + return true } // SetSignalAct atomically sets the thread group's signal action for signal sig diff --git a/pkg/sentry/syscalls/linux/sys_signal.go b/pkg/sentry/syscalls/linux/sys_signal.go index 93b3f531a..66ecb1299 100644 --- a/pkg/sentry/syscalls/linux/sys_signal.go +++ b/pkg/sentry/syscalls/linux/sys_signal.go @@ -315,25 +315,23 @@ func Sigaltstack(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.S setaddr := args[0].Pointer() oldaddr := args[1].Pointer() + alt := t.SignalStack() if oldaddr != 0 { - alt := t.SignalStack() - if t.OnSignalStack(alt) { - alt.Flags |= arch.SignalStackFlagOnStack - } if err := t.CopyOutSignalStack(oldaddr, &alt); err != nil { return 0, nil, err } } if setaddr != 0 { - if t.OnSignalStack(t.SignalStack()) { - return 0, nil, syserror.EPERM - } alt, err := t.CopyInSignalStack(setaddr) if err != nil { return 0, nil, err } - if err := t.SetSignalStack(alt); err != nil { - return 0, nil, err + // The signal stack cannot be changed if the task is currently + // on the stack. This is enforced at the lowest level because + // these semantics apply to changing the signal stack via a + // ucontext during a signal handler. + if !t.SetSignalStack(alt) { + return 0, nil, syserror.EPERM } } -- cgit v1.2.3 From 657db692b2241d89a324acc246b3c5230d8bd6ac Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Thu, 14 Jun 2018 11:34:15 -0700 Subject: Ignore expiration count in kernelCPUClockListener.Notify. PiperOrigin-RevId: 200590832 Change-Id: I35b817ecccc9414a742dee4815dfc67d0c7d0496 --- pkg/sentry/kernel/kernel.go | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 536461bbd..a17148af1 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -960,7 +960,13 @@ type kernelCPUClockListener struct { // Notify implements ktime.TimerListener.Notify. func (l kernelCPUClockListener) Notify(exp uint64) { - atomic.AddUint64(&l.k.cpuClock, exp) + // Only increment cpuClock by 1 regardless of the number of expirations. + // This approximately compensates for cases where thread throttling or bad + // Go runtime scheduling prevents the cpuClockTicker goroutine, and + // presumably task goroutines as well, from executing for a long period of + // time. It's also necessary to prevent CPU clocks from seeing large + // discontinuous jumps. + atomic.AddUint64(&l.k.cpuClock, 1) } // Destroy implements ktime.TimerListener.Destroy. -- cgit v1.2.3 From fa6db05e0ce828f2500651ca1226babbbf5edc80 Mon Sep 17 00:00:00 2001 From: Brian Geffon Date: Fri, 15 Jun 2018 12:54:38 -0700 Subject: FIFOs should support O_TRUNC as a no-op. PiperOrigin-RevId: 200759323 Change-Id: I683b2edcc2188304c4ca563e46af457e23625905 --- pkg/sentry/kernel/pipe/node.go | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index 5b47427ef..e418cf174 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -162,6 +162,18 @@ func (i *inodeOperations) waitFor(wakeupChan *chan struct{}, sleeper amutex.Slee } } +// Truncate implements fs.InodeOperations.Truncate +// +// This method is required to override the default i.InodeOperations.Truncate +// which may return ErrInvalidOperation, this allows open related +// syscalls to set the O_TRUNC flag without returning an error by +// calling Truncate directly during openat. The ftruncate and truncate +// system calls will check that the file is an actual file and return +// EINVAL because it's a PIPE, making this behavior consistent with linux. +func (i *inodeOperations) Truncate(context.Context, *fs.Inode, int64) error { + return nil +} + // newHandleLocked signals a new pipe reader or writer depending on where // 'wakeupChan' points. This unblocks any corresponding reader or writer // waiting for the other end of the channel to be opened, see Fifo.waitFor. -- cgit v1.2.3 From aa14a2c1be7f705927e9558f0e46ceca159e23e6 Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Tue, 19 Jun 2018 16:07:08 -0700 Subject: sentry: futex S/R optimization. No need to save thousands of zerovalue buckets. PiperOrigin-RevId: 201258598 Change-Id: I5d3ea7b6a5345117ab4f610332d5288ca550be33 --- pkg/sentry/kernel/futex/futex.go | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index b3ba57a2c..15e3e5e2c 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -197,7 +197,7 @@ func bucketIndexForAddr(addr uintptr) uintptr { // Manager holds futex state for a single virtual address space. type Manager struct { - buckets [bucketCount]bucket + buckets [bucketCount]bucket `state:"zerovalue"` } // NewManager returns an initialized futex manager. -- cgit v1.2.3 From 4e9f0e91d724b547e1ecaeeb210017f4c0b3fd0d Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Wed, 20 Jun 2018 11:01:32 -0700 Subject: sentry: pending signals S/R optimization. Almost all of the hundreds of pending signal queues are empty upon save. PiperOrigin-RevId: 201380318 Change-Id: I40747072435299de681d646e0862efac0637e172 --- pkg/sentry/kernel/BUILD | 8 +++++-- pkg/sentry/kernel/pending_signals.go | 4 ++-- pkg/sentry/kernel/pending_signals_state.go | 37 ++++++++++++++++++++++++++++++ 3 files changed, 45 insertions(+), 4 deletions(-) create mode 100644 pkg/sentry/kernel/pending_signals_state.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 377c94e4c..b2a55ddff 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -13,7 +13,7 @@ go_stateify( "ipc_namespace.go", "kernel.go", "pending_signals.go", - "pending_signals_list.go", + "pending_signals_state.go", "process_group_list.go", "ptrace.go", "rseq.go", @@ -46,7 +46,10 @@ go_stateify( "version.go", ], out = "kernel_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs"], + imports = [ + "gvisor.googlesource.com/gvisor/pkg/sentry/arch", + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", + ], package = "kernel", ) @@ -117,6 +120,7 @@ go_library( "kernel_state.go", "pending_signals.go", "pending_signals_list.go", + "pending_signals_state.go", "process_group_list.go", "ptrace.go", "rseq.go", diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go index d8701f47a..5dc0f266c 100644 --- a/pkg/sentry/kernel/pending_signals.go +++ b/pkg/sentry/kernel/pending_signals.go @@ -44,11 +44,11 @@ type pendingSignals struct { // Note that signals is zero-indexed, but signal 1 is the first valid // signal, so signals[0] contains signals with signo 1 etc. This offset is // usually handled by using Signal.index(). - signals [linux.SignalMaximum]pendingSignalQueue + signals [linux.SignalMaximum]pendingSignalQueue `state:".([]*arch.SignalInfo)"` // Bit i of pendingSet is set iff there is at least one signal with signo // i+1 pending. - pendingSet linux.SignalSet + pendingSet linux.SignalSet `state:"manual"` } // pendingSignalQueue holds a pendingSignalList for a single signal number. diff --git a/pkg/sentry/kernel/pending_signals_state.go b/pkg/sentry/kernel/pending_signals_state.go new file mode 100644 index 000000000..af61f6e8e --- /dev/null +++ b/pkg/sentry/kernel/pending_signals_state.go @@ -0,0 +1,37 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" +) + +// saveSignals is invoked by stateify. +func (p *pendingSignals) saveSignals() []*arch.SignalInfo { + var pending []*arch.SignalInfo + for _, q := range p.signals { + for ps := q.pendingSignalList.Front(); ps != nil; ps = ps.Next() { + pending = append(pending, ps.SignalInfo) + } + } + return pending +} + +// loadSignals is invoked by stateify. +func (p *pendingSignals) loadSignals(pending []*arch.SignalInfo) { + for _, si := range pending { + p.enqueue(si) + } +} -- cgit v1.2.3 From d571a4359cebbcf8a9b201bb125f1cdc9fb126e4 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Thu, 21 Jun 2018 10:52:33 -0700 Subject: Implement ioctl(FIOASYNC) FIOASYNC and friends are used to send signals when a file is ready for IO. This may or may not be needed by Nginx. While Nginx does use it, it is unclear if the code that uses it has any effect. PiperOrigin-RevId: 201550828 Change-Id: I7ba05a7db4eb2dfffde11e9bd9a35b65b98d7f50 --- pkg/abi/linux/BUILD | 1 + pkg/abi/linux/fcntl.go | 29 +++++++ pkg/abi/linux/ioctl.go | 5 ++ pkg/abi/linux/signal.go | 34 ++++++++ pkg/sentry/fs/file.go | 62 +++++++++++++-- pkg/sentry/fs/file_state.go | 10 --- pkg/sentry/fs/flags.go | 7 ++ pkg/sentry/kernel/fasync/BUILD | 18 +++++ pkg/sentry/kernel/fasync/fasync.go | 145 ++++++++++++++++++++++++++++++++++ pkg/sentry/kernel/sessions.go | 5 ++ pkg/sentry/syscalls/linux/BUILD | 1 + pkg/sentry/syscalls/linux/flags.go | 5 ++ pkg/sentry/syscalls/linux/sys_file.go | 78 ++++++++++++++++-- 13 files changed, 376 insertions(+), 24 deletions(-) create mode 100644 pkg/abi/linux/fcntl.go create mode 100644 pkg/sentry/kernel/fasync/BUILD create mode 100644 pkg/sentry/kernel/fasync/fasync.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index 693ce0fdd..5d00b66cc 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -31,6 +31,7 @@ go_library( "elf.go", "errors.go", "exec.go", + "fcntl.go", "file.go", "fs.go", "futex.go", diff --git a/pkg/abi/linux/fcntl.go b/pkg/abi/linux/fcntl.go new file mode 100644 index 000000000..f5dbe5199 --- /dev/null +++ b/pkg/abi/linux/fcntl.go @@ -0,0 +1,29 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package linux + +// Comands from linux/fcntl.h. +const ( + F_DUPFD = 0 + F_DUPFD_CLOEXEC = 1030 + F_GETFD = 1 + F_GETFL = 3 + F_GETOWN = 9 + F_SETFD = 2 + F_SETFL = 4 + F_SETLK = 6 + F_SETLKW = 7 + F_SETOWN = 8 +) diff --git a/pkg/abi/linux/ioctl.go b/pkg/abi/linux/ioctl.go index 35cefbdfc..3ef046562 100644 --- a/pkg/abi/linux/ioctl.go +++ b/pkg/abi/linux/ioctl.go @@ -29,6 +29,11 @@ const ( TIOCSPTLCK = 0x40045431 FIONCLEX = 0x00005450 FIOCLEX = 0x00005451 + FIOASYNC = 0x00005452 + FIOSETOWN = 0x00008901 + SIOCSPGRP = 0x00008902 + FIOGETOWN = 0x00008903 + SIOCGPGRP = 0x00008904 ) // ioctl(2) requests provided by uapi/linux/android/binder.h diff --git a/pkg/abi/linux/signal.go b/pkg/abi/linux/signal.go index cd09008b5..fed2a159f 100644 --- a/pkg/abi/linux/signal.go +++ b/pkg/abi/linux/signal.go @@ -175,3 +175,37 @@ const ( SA_NOMASK = SA_NODEFER SA_ONESHOT = SA_RESTARTHAND ) + +// Signal info types. +const ( + SI_MASK = 0xffff0000 + SI_KILL = 0 << 16 + SI_TIMER = 1 << 16 + SI_POLL = 2 << 16 + SI_FAULT = 3 << 16 + SI_CHLD = 4 << 16 + SI_RT = 5 << 16 + SI_MESGQ = 6 << 16 + SI_SYS = 7 << 16 +) + +// SIGPOLL si_codes. +const ( + // POLL_IN indicates that data input available. + POLL_IN = SI_POLL | 1 + + // POLL_OUT indicates that output buffers available. + POLL_OUT = SI_POLL | 2 + + // POLL_MSG indicates that an input message available. + POLL_MSG = SI_POLL | 3 + + // POLL_ERR indicates that there was an i/o error. + POLL_ERR = SI_POLL | 4 + + // POLL_PRI indicates that a high priority input available. + POLL_PRI = SI_POLL | 5 + + // POLL_HUP indicates that a device disconnected. + POLL_HUP = SI_POLL | 6 +) diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index f2683bbd2..6d93ef760 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -16,6 +16,7 @@ package fs import ( "math" + "sync" "sync/atomic" "gvisor.googlesource.com/gvisor/pkg/amutex" @@ -72,9 +73,15 @@ type File struct { // other files via the Dirent cache. Dirent *Dirent + // flagsMu protects flags and async below. + flagsMu sync.Mutex `state:"nosave"` + // flags are the File's flags. Setting or getting flags is fully atomic // and is not protected by mu (below). - flags atomic.Value `state:".(FileFlags)"` + flags FileFlags + + // async handles O_ASYNC notifications. + async FileAsync // mu is dual-purpose: first, to make read(2) and write(2) thread-safe // in conformity with POSIX, and second, to cancel operations before they @@ -99,8 +106,8 @@ func NewFile(ctx context.Context, dirent *Dirent, flags FileFlags, fops FileOper UniqueID: uniqueid.GlobalFromContext(ctx), Dirent: dirent, FileOperations: fops, + flags: flags, } - f.flags.Store(flags) f.mu.Init() return f } @@ -117,22 +124,40 @@ func (f *File) DecRef() { // Release a reference on the Dirent. f.Dirent.DecRef() + + f.flagsMu.Lock() + if f.flags.Async && f.async != nil { + f.async.Unregister(f) + } + f.flagsMu.Unlock() }) } // Flags atomically loads the File's flags. func (f *File) Flags() FileFlags { - return f.flags.Load().(FileFlags) + f.flagsMu.Lock() + flags := f.flags + f.flagsMu.Unlock() + return flags } // SetFlags atomically changes the File's flags to the values contained // in newFlags. See SettableFileFlags for values that can be set. func (f *File) SetFlags(newFlags SettableFileFlags) { - flags := f.flags.Load().(FileFlags) - flags.Direct = newFlags.Direct - flags.NonBlocking = newFlags.NonBlocking - flags.Append = newFlags.Append - f.flags.Store(flags) + f.flagsMu.Lock() + f.flags.Direct = newFlags.Direct + f.flags.NonBlocking = newFlags.NonBlocking + f.flags.Append = newFlags.Append + if f.async != nil { + if newFlags.Async && !f.flags.Async { + f.async.Register(f) + } + if !newFlags.Async && f.flags.Async { + f.async.Unregister(f) + } + } + f.flags.Async = newFlags.Async + f.flagsMu.Unlock() } // Offset atomically loads the File's offset. @@ -361,6 +386,27 @@ func (f *File) Msync(ctx context.Context, mr memmap.MappableRange) error { return f.Fsync(ctx, int64(mr.Start), int64(mr.End-1), SyncData) } +// A FileAsync sends signals to its owner when w is ready for IO. +type FileAsync interface { + Register(w waiter.Waitable) + Unregister(w waiter.Waitable) +} + +// Async gets the stored FileAsync or creates a new one with the supplied +// function. If the supplied function is nil, no FileAsync is created and the +// current value is returned. +func (f *File) Async(newAsync func() FileAsync) FileAsync { + f.flagsMu.Lock() + defer f.flagsMu.Unlock() + if f.async == nil && newAsync != nil { + f.async = newAsync() + if f.flags.Async { + f.async.Register(f) + } + } + return f.async +} + // FileReader implements io.Reader and io.ReaderAt. type FileReader struct { // Ctx is the context for the file reader. diff --git a/pkg/sentry/fs/file_state.go b/pkg/sentry/fs/file_state.go index 341cbda0b..3384737ab 100644 --- a/pkg/sentry/fs/file_state.go +++ b/pkg/sentry/fs/file_state.go @@ -18,13 +18,3 @@ package fs func (f *File) afterLoad() { f.mu.Init() } - -// saveFlags is invoked by stateify. -func (f *File) saveFlags() FileFlags { - return f.flags.Load().(FileFlags) -} - -// loadFlags is invoked by stateify. -func (f *File) loadFlags(flags FileFlags) { - f.flags.Store(flags) -} diff --git a/pkg/sentry/fs/flags.go b/pkg/sentry/fs/flags.go index dfa6a3d62..7a8eefd02 100644 --- a/pkg/sentry/fs/flags.go +++ b/pkg/sentry/fs/flags.go @@ -42,6 +42,9 @@ type FileFlags struct { // Directory indicates that this file must be a directory. Directory bool + + // Async indicates that this file sends signals on IO events. + Async bool } // SettableFileFlags is a subset of FileFlags above that can be changed @@ -55,6 +58,9 @@ type SettableFileFlags struct { // Append indicates this file is append only. Append bool + + // Async indicates that this file sends signals on IO events. + Async bool } // Settable returns the subset of f that are settable. @@ -63,5 +69,6 @@ func (f FileFlags) Settable() SettableFileFlags { Direct: f.Direct, NonBlocking: f.NonBlocking, Append: f.Append, + Async: f.Async, } } diff --git a/pkg/sentry/kernel/fasync/BUILD b/pkg/sentry/kernel/fasync/BUILD new file mode 100644 index 000000000..8d06e1182 --- /dev/null +++ b/pkg/sentry/kernel/fasync/BUILD @@ -0,0 +1,18 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("@io_bazel_rules_go//go:def.bzl", "go_library") + +go_library( + name = "fasync", + srcs = ["fasync.go"], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/fasync", + visibility = ["//:sandbox"], + deps = [ + "//pkg/abi/linux", + "//pkg/sentry/arch", + "//pkg/sentry/fs", + "//pkg/sentry/kernel", + "//pkg/sentry/kernel/auth", + "//pkg/waiter", + ], +) diff --git a/pkg/sentry/kernel/fasync/fasync.go b/pkg/sentry/kernel/fasync/fasync.go new file mode 100644 index 000000000..028d6766f --- /dev/null +++ b/pkg/sentry/kernel/fasync/fasync.go @@ -0,0 +1,145 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package fasync provides FIOASYNC related functionality. +package fasync + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// New creates a new FileAsync. +func New() fs.FileAsync { + return &FileAsync{} +} + +// FileAsync sends signals when the registered file is ready for IO. +type FileAsync struct { + mu sync.Mutex + e waiter.Entry + requester auth.Credentials + + // Only one of the following is allowed to be non-nil. + recipientPG *kernel.ProcessGroup + recipientTG *kernel.ThreadGroup + recipientT *kernel.Task +} + +// Callback sends a signal. +func (a *FileAsync) Callback(e *waiter.Entry) { + a.mu.Lock() + if a.e.Callback == nil { + return + } + t := a.recipientT + tg := a.recipientTG + if a.recipientPG != nil { + tg = a.recipientPG.Originator() + } + if tg != nil { + t = tg.Leader() + } + c := t.Credentials() + // Logic from sigio_perm in fs/fcntl.c. + if a.requester.EffectiveKUID == 0 || + a.requester.EffectiveKUID == c.SavedKUID || + a.requester.EffectiveKUID == c.RealKUID || + a.requester.RealKUID == c.SavedKUID || + a.requester.RealKUID == c.RealKUID { + t.SendSignal(&arch.SignalInfo{ + Signo: int32(linux.SIGIO), + // SEND_SIG_PRIV + Code: arch.SignalInfoKernel, + }) + } + a.mu.Unlock() +} + +// Register sets the file which will be monitored for IO events. +// +// The file must not be currently registered. +func (a *FileAsync) Register(w waiter.Waitable) { + a.mu.Lock() + defer a.mu.Unlock() + + if a.e.Callback != nil { + panic("registering already registered file") + } + + a.e.Callback = a + w.EventRegister(&a.e, waiter.EventIn|waiter.EventOut|waiter.EventErr|waiter.EventHUp) +} + +// Unregister stops monitoring a file. +// +// The file must be currently registered. +func (a *FileAsync) Unregister(w waiter.Waitable) { + a.mu.Lock() + defer a.mu.Unlock() + + if a.e.Callback == nil { + panic("unregistering unregistered file") + } + + w.EventUnregister(&a.e) + a.e.Callback = nil +} + +// Owner returns who is currently getting signals. All return values will be +// nil if no one is set to receive signals. +func (a *FileAsync) Owner() (*kernel.Task, *kernel.ThreadGroup, *kernel.ProcessGroup) { + a.mu.Lock() + defer a.mu.Unlock() + return a.recipientT, a.recipientTG, a.recipientPG +} + +// SetOwnerTask sets the owner (who will receive signals) to a specified task. +// Only this owner will receive signals. +func (a *FileAsync) SetOwnerTask(requester *kernel.Task, recipient *kernel.Task) { + a.mu.Lock() + defer a.mu.Unlock() + a.requester = requester.Credentials() + a.recipientT = recipient + a.recipientTG = nil + a.recipientPG = nil +} + +// SetOwnerThreadGroup sets the owner (who will receive signals) to a specified +// thread group. Only this owner will receive signals. +func (a *FileAsync) SetOwnerThreadGroup(requester *kernel.Task, recipient *kernel.ThreadGroup) { + a.mu.Lock() + defer a.mu.Unlock() + a.requester = requester.Credentials() + a.recipientT = nil + a.recipientTG = recipient + a.recipientPG = nil +} + +// SetOwnerProcessGroup sets the owner (who will receive signals) to a +// specified process group. Only this owner will receive signals. +func (a *FileAsync) SetOwnerProcessGroup(requester *kernel.Task, recipient *kernel.ProcessGroup) { + a.mu.Lock() + defer a.mu.Unlock() + a.requester = requester.Credentials() + a.recipientT = nil + a.recipientTG = nil + a.recipientPG = recipient +} diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index 53d8fb844..fa4c7b8f6 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -110,6 +110,11 @@ type ProcessGroup struct { processGroupEntry } +// Originator retuns the originator of the process group. +func (pg *ProcessGroup) Originator() *ThreadGroup { + return pg.originator +} + // incRefWithParent grabs a reference. // // This function is called when this ProcessGroup is being associated with some diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index 7cfd37fb1..d3f3cc459 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -82,6 +82,7 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/epoll", "//pkg/sentry/kernel/eventfd", + "//pkg/sentry/kernel/fasync", "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/pipe", "//pkg/sentry/kernel/sched", diff --git a/pkg/sentry/syscalls/linux/flags.go b/pkg/sentry/syscalls/linux/flags.go index 82bfd7c2a..3d39a20f4 100644 --- a/pkg/sentry/syscalls/linux/flags.go +++ b/pkg/sentry/syscalls/linux/flags.go @@ -61,6 +61,9 @@ func flagsToLinux(flags fs.FileFlags) (mask uint) { if flags.Directory { mask |= syscall.O_DIRECTORY } + if flags.Async { + mask |= syscall.O_ASYNC + } switch { case flags.Read && flags.Write: mask |= syscall.O_RDWR @@ -82,6 +85,7 @@ func linuxToFlags(mask uint) (flags fs.FileFlags) { Write: (mask & syscall.O_ACCMODE) != syscall.O_RDONLY, Append: mask&syscall.O_APPEND != 0, Directory: mask&syscall.O_DIRECTORY != 0, + Async: mask&syscall.O_ASYNC != 0, } } @@ -91,5 +95,6 @@ func linuxToSettableFlags(mask uint) fs.SettableFileFlags { Direct: mask&syscall.O_DIRECT != 0, NonBlocking: mask&syscall.O_NONBLOCK != 0, Append: mask&syscall.O_APPEND != 0, + Async: mask&syscall.O_ASYNC != 0, } } diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index e2980842f..490649f87 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -25,6 +25,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/lock" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/fasync" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" @@ -528,6 +529,33 @@ func Ioctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall file.SetFlags(flags.Settable()) return 0, nil, nil + case linux.FIOASYNC: + var set int32 + if _, err := t.CopyIn(args[2].Pointer(), &set); err != nil { + return 0, nil, err + } + flags := file.Flags() + if set != 0 { + flags.Async = true + } else { + flags.Async = false + } + file.SetFlags(flags.Settable()) + return 0, nil, nil + + case linux.FIOSETOWN, linux.SIOCSPGRP: + var set int32 + if _, err := t.CopyIn(args[2].Pointer(), &set); err != nil { + return 0, nil, err + } + fSetOwn(t, file, set) + return 0, nil, nil + + case linux.FIOGETOWN, linux.SIOCGPGRP: + who := fGetOwn(t, file) + _, err := t.CopyOut(args[2].Pointer(), &who) + return 0, nil, err + default: ret, err := file.FileOperations.Ioctl(t, t.MemoryManager(), args) if err != nil { @@ -725,6 +753,39 @@ func Dup3(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallC return uintptr(newfd), nil, nil } +func fGetOwn(t *kernel.Task, file *fs.File) int32 { + ma := file.Async(nil) + if ma == nil { + return 0 + } + a := ma.(*fasync.FileAsync) + ot, otg, opg := a.Owner() + switch { + case ot != nil: + return int32(t.PIDNamespace().IDOfTask(ot)) + case otg != nil: + return int32(t.PIDNamespace().IDOfThreadGroup(otg)) + case opg != nil: + return int32(-t.PIDNamespace().IDOfProcessGroup(opg)) + default: + return 0 + } +} + +// fSetOwn sets the file's owner with the semantics of F_SETOWN in Linux. +// +// If who is positive, it represents a PID. If negative, it represents a PGID. +// If the PID or PGID is invalid, the owner is silently unset. +func fSetOwn(t *kernel.Task, file *fs.File, who int32) { + a := file.Async(fasync.New).(*fasync.FileAsync) + if who < 0 { + pg := t.PIDNamespace().ProcessGroupWithID(kernel.ProcessGroupID(-who)) + a.SetOwnerProcessGroup(t, pg) + } + tg := t.PIDNamespace().ThreadGroupWithID(kernel.ThreadID(who)) + a.SetOwnerThreadGroup(t, tg) +} + // Fcntl implements linux syscall fcntl(2). func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { fd := kdefs.FD(args[0].Int()) @@ -737,7 +798,7 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall defer file.DecRef() switch cmd { - case syscall.F_DUPFD, syscall.F_DUPFD_CLOEXEC: + case linux.F_DUPFD, linux.F_DUPFD_CLOEXEC: from := kdefs.FD(args[2].Int()) fdFlags := kernel.FDFlags{CloseOnExec: cmd == syscall.F_DUPFD_CLOEXEC} fd, err := t.FDMap().NewFDFrom(from, file, fdFlags, t.ThreadGroup().Limits()) @@ -745,19 +806,19 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall return 0, nil, err } return uintptr(fd), nil, nil - case syscall.F_GETFD: + case linux.F_GETFD: return uintptr(fdFlagsToLinux(flags)), nil, nil - case syscall.F_SETFD: + case linux.F_SETFD: flags := args[2].Uint() t.FDMap().SetFlags(fd, kernel.FDFlags{ CloseOnExec: flags&syscall.FD_CLOEXEC != 0, }) - case syscall.F_GETFL: + case linux.F_GETFL: return uintptr(flagsToLinux(file.Flags())), nil, nil - case syscall.F_SETFL: + case linux.F_SETFL: flags := uint(args[2].Uint()) file.SetFlags(linuxToSettableFlags(flags)) - case syscall.F_SETLK, syscall.F_SETLKW: + case linux.F_SETLK, linux.F_SETLKW: // In Linux the file system can choose to provide lock operations for an inode. // Normally pipe and socket types lack lock operations. We diverge and use a heavy // hammer by only allowing locks on files and directories. @@ -854,6 +915,11 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall default: return 0, nil, syserror.EINVAL } + case linux.F_GETOWN: + return uintptr(fGetOwn(t, file)), nil, nil + case linux.F_SETOWN: + fSetOwn(t, file, args[2].Int()) + return 0, nil, nil default: // Everything else is not yet supported. return 0, nil, syserror.EINVAL -- cgit v1.2.3 From 2dedbc7211fb6b7f8b86148e6627054e781eaa87 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Thu, 21 Jun 2018 14:53:05 -0700 Subject: Drop return from SendExternalSignal SendExternalSignal is no longer called before CreateProcess, so it can enforce this simplified precondition. StartForwarding, and after Kernel.Start. PiperOrigin-RevId: 201591170 Change-Id: Ib7022ef7895612d7d82a00942ab59fa433c4d6e9 --- pkg/sentry/kernel/kernel.go | 7 +++---- pkg/sentry/kernel/signal.go | 13 +++++-------- pkg/sentry/sighandling/sighandling.go | 14 +++++++++++--- 3 files changed, 19 insertions(+), 15 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index a17148af1..5662b8f08 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -760,12 +760,11 @@ func (k *Kernel) Unpause() { // // context is used only for debugging to describe how the signal was received. // -// Returns false if signal could not be sent because the Kernel is not fully -// initialized yet. -func (k *Kernel) SendExternalSignal(info *arch.SignalInfo, context string) bool { +// Preconditions: Kernel must have an init process. +func (k *Kernel) SendExternalSignal(info *arch.SignalInfo, context string) { k.extMu.Lock() defer k.extMu.Unlock() - return k.sendExternalSignal(info, context) + k.sendExternalSignal(info, context) } // FeatureSet returns the FeatureSet. diff --git a/pkg/sentry/kernel/signal.go b/pkg/sentry/kernel/signal.go index 8edd05cdf..e3a2a777a 100644 --- a/pkg/sentry/kernel/signal.go +++ b/pkg/sentry/kernel/signal.go @@ -15,6 +15,8 @@ package kernel import ( + "fmt" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" @@ -33,13 +35,11 @@ const SignalPanic = linux.SIGUSR2 // // context is used only for debugging to differentiate these cases. // -// Returns false if signal could not be sent because the Kernel is not fully -// initialized yet. -func (k *Kernel) sendExternalSignal(info *arch.SignalInfo, context string) bool { +// Preconditions: Kernel must have an init process. +func (k *Kernel) sendExternalSignal(info *arch.SignalInfo, context string) { switch linux.Signal(info.Signo) { case platform.SignalInterrupt: // Assume that a call to platform.Context.Interrupt() misfired. - return true case SignalPanic: // SignalPanic is also specially handled in sentry setup to ensure that @@ -50,13 +50,10 @@ func (k *Kernel) sendExternalSignal(info *arch.SignalInfo, context string) bool default: log.Infof("Received external signal %d in %s context", info.Signo, context) if k.globalInit == nil { - log.Warningf("Received external signal %d before init created", info.Signo) - return false + panic(fmt.Sprintf("Received external signal %d before init created", info.Signo)) } k.globalInit.SendSignal(info) } - - return true } // sigPriv returns a SignalInfo representing a signal sent by the sentry. (The diff --git a/pkg/sentry/sighandling/sighandling.go b/pkg/sentry/sighandling/sighandling.go index ef6f7f617..25295440c 100644 --- a/pkg/sentry/sighandling/sighandling.go +++ b/pkg/sentry/sighandling/sighandling.go @@ -16,6 +16,7 @@ package sighandling import ( + "fmt" "os" "os/signal" "reflect" @@ -65,7 +66,9 @@ func forwardSignals(k *kernel.Kernel, sigchans []chan os.Signal, start, stop cha // Otherwise, it was a signal on channel N. Index 0 represents the stop // channel, so index N represents the channel for signal N. - if !started || !k.SendExternalSignal(&arch.SignalInfo{Signo: int32(index)}, "sentry") { + signal := linux.Signal(index) + + if !started { // Kernel is not ready to receive signals. // // Kill ourselves if this signal would have killed the @@ -78,11 +81,16 @@ func forwardSignals(k *kernel.Kernel, sigchans []chan os.Signal, start, stop cha // TODO: Convert Go's runtime.raise from // tkill to tgkill so PrepareForwarding doesn't need to // be called until after filter installation. - switch linux.Signal(index) { + switch signal { case linux.SIGHUP, linux.SIGINT, linux.SIGTERM: - dieFromSignal(linux.Signal(index)) + dieFromSignal(signal) + panic(fmt.Sprintf("Failed to die from signal %d", signal)) + default: + continue } } + + k.SendExternalSignal(&arch.SignalInfo{Signo: int32(signal)}, "sentry") } // Close all individual channels. -- cgit v1.2.3 From 0e434b66a625b937d90e4ebe632de4546101be5a Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Thu, 21 Jun 2018 15:18:47 -0700 Subject: netstack: tcp socket connected state S/R support. PiperOrigin-RevId: 201596247 Change-Id: Id22f47b2cdcbe14aa0d930f7807ba75f91a56724 --- pkg/sentry/kernel/BUILD | 5 +- pkg/sentry/kernel/kernel.go | 6 + pkg/sentry/kernel/kernel_state.go | 31 +++++ pkg/tcpip/stack/stack_global_state.go | 2 +- pkg/tcpip/tcpip.go | 36 ++++++ pkg/tcpip/transport/tcp/BUILD | 7 ++ pkg/tcpip/transport/tcp/accept.go | 11 +- pkg/tcpip/transport/tcp/connect.go | 43 ++++++- pkg/tcpip/transport/tcp/endpoint.go | 72 +++++++++--- pkg/tcpip/transport/tcp/endpoint_state.go | 185 ++++++++++++++++++++---------- pkg/tcpip/transport/tcp/segment.go | 8 +- pkg/tcpip/transport/tcp/segment_queue.go | 4 +- pkg/tcpip/transport/tcp/segment_state.go | 41 +++++++ pkg/tcpip/transport/tcp/snd.go | 4 +- pkg/tcpip/transport/tcp/snd_state.go | 39 +++++++ 15 files changed, 403 insertions(+), 91 deletions(-) create mode 100644 pkg/sentry/kernel/kernel_state.go create mode 100644 pkg/tcpip/transport/tcp/segment_state.go create mode 100644 pkg/tcpip/transport/tcp/snd_state.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index b2a55ddff..07568b47c 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -12,6 +12,7 @@ go_stateify( "fs_context.go", "ipc_namespace.go", "kernel.go", + "kernel_state.go", "pending_signals.go", "pending_signals_state.go", "process_group_list.go", @@ -45,10 +46,11 @@ go_stateify( "vdso.go", "version.go", ], - out = "kernel_state.go", + out = "kernel_autogen_state.go", imports = [ "gvisor.googlesource.com/gvisor/pkg/sentry/arch", "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", + "gvisor.googlesource.com/gvisor/pkg/tcpip", ], package = "kernel", ) @@ -117,6 +119,7 @@ go_library( "fs_context.go", "ipc_namespace.go", "kernel.go", + "kernel_autogen_state.go", "kernel_state.go", "pending_signals.go", "pending_signals_list.go", diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 5662b8f08..64439cd9d 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -57,6 +57,7 @@ import ( sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/state" + "gvisor.googlesource.com/gvisor/pkg/tcpip" ) // Kernel represents an emulated Linux kernel. It must be initialized by calling @@ -158,6 +159,9 @@ type Kernel struct { // exitErr is the error causing the sandbox to exit, if any. It is // protected by extMu. exitErr error + + // danglingEndpoints is used to save / restore tcpip.DanglingEndpoints. + danglingEndpoints struct{} `state:".([]tcpip.Endpoint)"` } // InitKernelArgs holds arguments to Init. @@ -422,6 +426,8 @@ func (k *Kernel) LoadFrom(r io.Reader, p platform.Platform, net inet.Stack) erro return err } + tcpip.AsyncLoading.Wait() + log.Infof("Overall load took [%s]", time.Since(loadStart)) // Applications may size per-cpu structures based on k.applicationCores, so diff --git a/pkg/sentry/kernel/kernel_state.go b/pkg/sentry/kernel/kernel_state.go new file mode 100644 index 000000000..bb2d5102d --- /dev/null +++ b/pkg/sentry/kernel/kernel_state.go @@ -0,0 +1,31 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/tcpip" +) + +// saveDanglingEndpoints is invoked by stateify. +func (k *Kernel) saveDanglingEndpoints() []tcpip.Endpoint { + return tcpip.GetDanglingEndpoints() +} + +// loadDanglingEndpoints is invoked by stateify. +func (k *Kernel) loadDanglingEndpoints(es []tcpip.Endpoint) { + for _, e := range es { + tcpip.AddDanglingEndpoint(e) + } +} diff --git a/pkg/tcpip/stack/stack_global_state.go b/pkg/tcpip/stack/stack_global_state.go index 030ae98d1..260d7d05c 100644 --- a/pkg/tcpip/stack/stack_global_state.go +++ b/pkg/tcpip/stack/stack_global_state.go @@ -5,5 +5,5 @@ package stack // StackFromEnv is the global stack created in restore run. -// FIXME: remove this variable once tcpip S/R is fully supported. +// FIXME var StackFromEnv *Stack diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index cf25a086d..17fa0efb7 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -23,6 +23,7 @@ import ( "fmt" "strconv" "strings" + "sync" "time" "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" @@ -552,3 +553,38 @@ type ProtocolAddress struct { // Address is a network address. Address Address } + +// danglingEndpointsMu protects access to danglingEndpoints. +var danglingEndpointsMu sync.Mutex + +// danglingEndpoints tracks all dangling endpoints no longer owned by the app. +var danglingEndpoints = make(map[Endpoint]struct{}) + +// GetDanglingEndpoints returns all dangling endpoints. +func GetDanglingEndpoints() []Endpoint { + es := make([]Endpoint, 0, len(danglingEndpoints)) + danglingEndpointsMu.Lock() + for e, _ := range danglingEndpoints { + es = append(es, e) + } + danglingEndpointsMu.Unlock() + return es +} + +// AddDanglingEndpoint adds a dangling endpoint. +func AddDanglingEndpoint(e Endpoint) { + danglingEndpointsMu.Lock() + danglingEndpoints[e] = struct{}{} + danglingEndpointsMu.Unlock() +} + +// DeleteDanglingEndpoint removes a dangling endpoint. +func DeleteDanglingEndpoint(e Endpoint) { + danglingEndpointsMu.Lock() + delete(danglingEndpoints, e) + danglingEndpointsMu.Unlock() +} + +// AsyncLoading is the global barrier for asynchronous endpoint loading +// activities. +var AsyncLoading sync.WaitGroup diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index f38f58e87..d129aa285 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -10,11 +10,16 @@ go_stateify( "endpoint.go", "endpoint_state.go", "rcv.go", + "segment.go", "segment_heap.go", + "segment_queue.go", + "segment_state.go", "snd.go", + "snd_state.go", "tcp_segment_list.go", ], out = "tcp_state.go", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], package = "tcp", ) @@ -43,7 +48,9 @@ go_library( "segment.go", "segment_heap.go", "segment_queue.go", + "segment_state.go", "snd.go", + "snd_state.go", "tcp_segment_list.go", "tcp_state.go", "timer.go", diff --git a/pkg/tcpip/transport/tcp/accept.go b/pkg/tcpip/transport/tcp/accept.go index 85adeef0e..410dfdad4 100644 --- a/pkg/tcpip/transport/tcp/accept.go +++ b/pkg/tcpip/transport/tcp/accept.go @@ -68,7 +68,8 @@ func encodeMSS(mss uint16) uint32 { // to go above a threshold. var synRcvdCount struct { sync.Mutex - value uint64 + value uint64 + pending sync.WaitGroup } // listenContext is used by a listening endpoint to store state used while @@ -102,6 +103,7 @@ func incSynRcvdCount() bool { return false } + synRcvdCount.pending.Add(1) synRcvdCount.value++ return true @@ -115,6 +117,7 @@ func decSynRcvdCount() { defer synRcvdCount.Unlock() synRcvdCount.value-- + synRcvdCount.pending.Done() } // newListenContext creates a new listen context. @@ -292,7 +295,7 @@ func (e *endpoint) handleListenSegment(ctx *listenContext, s *segment) { opts := parseSynSegmentOptions(s) if incSynRcvdCount() { s.incRef() - go e.handleSynSegment(ctx, s, &opts) // S/R-FIXME + go e.handleSynSegment(ctx, s, &opts) // S/R-SAFE: synRcvdCount is the barrier. } else { cookie := ctx.createCookie(s.id, s.sequenceNumber, encodeMSS(opts.MSS)) // Send SYN with window scaling because we currently @@ -381,10 +384,12 @@ func (e *endpoint) protocolListenLoop(rcvWnd seqnum.Size) *tcpip.Error { return nil } if n¬ifyDrain != 0 { - for s := e.segmentQueue.dequeue(); s != nil; s = e.segmentQueue.dequeue() { + for !e.segmentQueue.empty() { + s := e.segmentQueue.dequeue() e.handleListenSegment(ctx, s) s.decRef() } + synRcvdCount.pending.Wait() close(e.drainDone) <-e.undrain } diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go index 9aaabe0b1..d9f87c793 100644 --- a/pkg/tcpip/transport/tcp/connect.go +++ b/pkg/tcpip/transport/tcp/connect.go @@ -443,7 +443,8 @@ func (h *handshake) execute() *tcpip.Error { return tcpip.ErrAborted } if n¬ifyDrain != 0 { - for s := h.ep.segmentQueue.dequeue(); s != nil; s = h.ep.segmentQueue.dequeue() { + for !h.ep.segmentQueue.empty() { + s := h.ep.segmentQueue.dequeue() err := h.handleSegment(s) s.decRef() if err != nil { @@ -813,15 +814,13 @@ func (e *endpoint) handleSegments() *tcpip.Error { // protocolMainLoop is the main loop of the TCP protocol. It runs in its own // goroutine and is responsible for sending segments and handling received // segments. -func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { +func (e *endpoint) protocolMainLoop(handshake bool) *tcpip.Error { var closeTimer *time.Timer var closeWaker sleep.Waker defer func() { // e.mu is expected to be hold upon entering this section. - e.completeWorkerLocked() - if e.snd != nil { e.snd.resendTimer.cleanup() } @@ -830,6 +829,8 @@ func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { closeTimer.Stop() } + e.completeWorkerLocked() + if e.drainDone != nil { close(e.drainDone) } @@ -840,7 +841,7 @@ func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { e.waiterQueue.Notify(waiter.EventHUp | waiter.EventErr | waiter.EventIn | waiter.EventOut) }() - if !passive { + if handshake { // This is an active connection, so we must initiate the 3-way // handshake, and then inform potential waiters about its // completion. @@ -945,6 +946,17 @@ func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { closeWaker.Assert() }) } + + if n¬ifyDrain != 0 { + for !e.segmentQueue.empty() { + if err := e.handleSegments(); err != nil { + return err + } + } + close(e.drainDone) + <-e.undrain + } + return nil }, }, @@ -956,6 +968,27 @@ func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { s.AddWaker(funcs[i].w, i) } + // The following assertions and notifications are needed for restored + // endpoints. Fresh newly created endpoints have empty states and should + // not invoke any. + e.segmentQueue.mu.Lock() + if !e.segmentQueue.list.Empty() { + e.newSegmentWaker.Assert() + } + e.segmentQueue.mu.Unlock() + + e.rcvListMu.Lock() + if !e.rcvList.Empty() { + e.waiterQueue.Notify(waiter.EventIn) + } + e.rcvListMu.Unlock() + + e.mu.RLock() + if e.workerCleanup { + e.notifyProtocolGoroutine(notifyClose) + } + e.mu.RUnlock() + // Main loop. Handle segments until both send and receive ends of the // connection have completed. for !e.rcv.closed || !e.snd.closed || e.snd.sndUna != e.snd.sndNxtList { diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index b21c2b4ab..706977618 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -69,7 +69,7 @@ type endpoint struct { // change throughout the lifetime of the endpoint. stack *stack.Stack `state:"manual"` netProto tcpip.NetworkProtocolNumber - waiterQueue *waiter.Queue + waiterQueue *waiter.Queue `state:"wait"` // lastError represents the last error that the endpoint reported; // access to it is protected by the following mutex. @@ -82,8 +82,8 @@ type endpoint struct { // // Once the peer has closed its send side, rcvClosed is set to true // to indicate to users that no more data is coming. - rcvListMu sync.Mutex `state:"nosave"` - rcvList segmentList + rcvListMu sync.Mutex `state:"nosave"` + rcvList segmentList `state:"wait"` rcvClosed bool rcvBufSize int rcvBufUsed int @@ -91,8 +91,8 @@ type endpoint struct { // The following fields are protected by the mutex. mu sync.RWMutex `state:"nosave"` id stack.TransportEndpointID - state endpointState - isPortReserved bool `state:"manual"` + state endpointState `state:".(endpointState)"` + isPortReserved bool `state:"manual"` isRegistered bool boundNICID tcpip.NICID `state:"manual"` route stack.Route `state:"manual"` @@ -118,7 +118,7 @@ type endpoint struct { // workerCleanup specifies if the worker goroutine must perform cleanup // before exitting. This can only be set to true when workerRunning is // also true, and they're both protected by the mutex. - workerCleanup bool `state:"zerovalue"` + workerCleanup bool // sendTSOk is used to indicate when the TS Option has been negotiated. // When sendTSOk is true every non-RST segment should carry a TS as per @@ -153,7 +153,7 @@ type endpoint struct { // segmentQueue is used to hand received segments to the protocol // goroutine. Segments are queued as long as the queue is not full, // and dropped when it is. - segmentQueue segmentQueue `state:"zerovalue"` + segmentQueue segmentQueue `state:"wait"` // The following fields are used to manage the send buffer. When // segments are ready to be sent, they are added to sndQueue and the @@ -166,7 +166,7 @@ type endpoint struct { sndBufUsed int sndClosed bool sndBufInQueue seqnum.Size - sndQueue segmentList + sndQueue segmentList `state:"wait"` sndWaker sleep.Waker `state:"manual"` sndCloseWaker sleep.Waker `state:"manual"` @@ -188,17 +188,21 @@ type endpoint struct { // notifyFlags is a bitmask of flags used to indicate to the protocol // goroutine what it was notified; this is only accessed atomically. - notifyFlags uint32 `state:"zerovalue"` + notifyFlags uint32 `state:"nosave"` // acceptedChan is used by a listening endpoint protocol goroutine to // send newly accepted connections to the endpoint so that they can be // read by Accept() calls. - acceptedChan chan *endpoint `state:".(endpointChan)"` + acceptedChan chan *endpoint `state:"manual"` + + // acceptedEndpoints is only used to save / restore the channel buffer. + // FIXME + acceptedEndpoints []*endpoint // The following are only used from the protocol goroutine, and // therefore don't need locks to protect them. - rcv *receiver - snd *sender + rcv *receiver `state:"wait"` + snd *sender `state:"wait"` // The goroutine drain completion notification channel. drainDone chan struct{} `state:"nosave"` @@ -211,6 +215,7 @@ type endpoint struct { probe stack.TCPProbeFunc `state:"nosave"` // The following are only used to assist the restore run to re-connect. + bindAddress tcpip.Address connectingAddress tcpip.Address } @@ -344,6 +349,7 @@ func (e *endpoint) Close() { // Either perform the local cleanup or kick the worker to make sure it // knows it needs to cleanup. + tcpip.AddDanglingEndpoint(e) if !e.workerRunning { e.cleanupLocked() } else { @@ -363,9 +369,12 @@ func (e *endpoint) cleanupLocked() { if e.acceptedChan != nil { close(e.acceptedChan) for n := range e.acceptedChan { + n.mu.Lock() n.resetConnectionLocked(tcpip.ErrConnectionAborted) + n.mu.Unlock() n.Close() } + e.acceptedChan = nil } e.workerCleanup = false @@ -374,6 +383,7 @@ func (e *endpoint) cleanupLocked() { } e.route.Release() + tcpip.DeleteDanglingEndpoint(e) } // Read reads data from the endpoint. @@ -786,6 +796,16 @@ func (e *endpoint) checkV4Mapped(addr *tcpip.FullAddress) (tcpip.NetworkProtocol // Connect connects the endpoint to its peer. func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { + return e.connect(addr, true, true) +} + +// connect connects the endpoint to its peer. In the normal non-S/R case, the +// new connection is expected to run the main goroutine and perform handshake. +// In restore of previously connected endpoints, both ends will be passively +// created (so no new handshaking is done); for stack-accepted connections not +// yet accepted by the app, they are restored without running the main goroutine +// here. +func (e *endpoint) connect(addr tcpip.FullAddress, handshake bool, run bool) *tcpip.Error { e.mu.Lock() defer e.mu.Unlock() @@ -897,9 +917,27 @@ func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { e.boundNICID = nicid e.effectiveNetProtos = netProtos e.connectingAddress = connectingAddr - e.workerRunning = true - go e.protocolMainLoop(false) // S/R-SAFE: will be drained before save. + // Connect in the restore phase does not perform handshake. Restore its + // connection setting here. + if !handshake { + e.segmentQueue.mu.Lock() + for _, l := range []segmentList{e.segmentQueue.list, e.sndQueue, e.snd.writeList} { + for s := l.Front(); s != nil; s = s.Next() { + s.id = e.id + s.route = r.Clone() + e.sndWaker.Assert() + } + } + e.segmentQueue.mu.Unlock() + e.snd.updateMaxPayloadSize(int(e.route.MTU()), 0) + e.state = stateConnected + } + + if run { + e.workerRunning = true + go e.protocolMainLoop(handshake) // S/R-SAFE: will be drained before save. + } return tcpip.ErrConnectStarted } @@ -971,6 +1009,9 @@ func (e *endpoint) Listen(backlog int) *tcpip.Error { if len(e.acceptedChan) > backlog { return tcpip.ErrInvalidEndpointState } + if cap(e.acceptedChan) == backlog { + return nil + } origChan := e.acceptedChan e.acceptedChan = make(chan *endpoint, backlog) close(origChan) @@ -1008,7 +1049,7 @@ func (e *endpoint) Listen(backlog int) *tcpip.Error { func (e *endpoint) startAcceptedLoop(waiterQueue *waiter.Queue) { e.waiterQueue = waiterQueue e.workerRunning = true - go e.protocolMainLoop(true) // S/R-FIXME + go e.protocolMainLoop(false) // S/R-SAFE: drained on save. } // Accept returns a new endpoint if a peer has established a connection @@ -1049,6 +1090,7 @@ func (e *endpoint) Bind(addr tcpip.FullAddress, commit func() *tcpip.Error) (ret return tcpip.ErrAlreadyBound } + e.bindAddress = addr.Addr netProto, err := e.checkV4Mapped(&addr) if err != nil { return err diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go index b1e249bff..38c97c796 100644 --- a/pkg/tcpip/transport/tcp/endpoint_state.go +++ b/pkg/tcpip/transport/tcp/endpoint_state.go @@ -9,6 +9,7 @@ import ( "sync" "gvisor.googlesource.com/gvisor/pkg/tcpip" + "gvisor.googlesource.com/gvisor/pkg/tcpip/header" "gvisor.googlesource.com/gvisor/pkg/tcpip/stack" ) @@ -22,7 +23,7 @@ func (e *endpoint) drainSegmentLocked() { e.undrain = make(chan struct{}) e.mu.Unlock() - e.notificationWaker.Assert() + e.notifyProtocolGoroutine(notifyDrain) <-e.drainDone e.mu.Lock() @@ -38,37 +39,98 @@ func (e *endpoint) beforeSave() { switch e.state { case stateInitial, stateBound: - case stateListen: - if !e.segmentQueue.empty() { - e.drainSegmentLocked() + case stateListen, stateConnecting, stateConnected: + if e.state == stateConnected && !e.workerRunning { + // The endpoint must be in acceptedChan. + break } - case stateConnecting: e.drainSegmentLocked() - if e.state != stateConnected { + if e.state != stateClosed && e.state != stateError { + if !e.workerRunning { + panic("endpoint has no worker running in listen, connecting, or connected state") + } break } fallthrough - case stateConnected: - // FIXME - panic(tcpip.ErrSaveRejection{fmt.Errorf("endpoint cannot be saved in connected state: local %v:%v, remote %v:%v", e.id.LocalAddress, e.id.LocalPort, e.id.RemoteAddress, e.id.RemotePort)}) case stateClosed, stateError: if e.workerRunning { - panic(fmt.Sprintf("endpoint still has worker running in closed or error state")) + panic("endpoint still has worker running in closed or error state") } default: panic(fmt.Sprintf("endpoint in unknown state %v", e.state)) } + + if e.waiterQueue != nil && !e.waiterQueue.IsEmpty() { + panic("endpoint still has waiters upon save") + } + + if !((e.state == stateBound || e.state == stateListen) == e.isPortReserved) { + panic("endpoint port must and must only be reserved in bound or listen state") + } + + if e.acceptedChan != nil { + close(e.acceptedChan) + e.acceptedEndpoints = make([]*endpoint, len(e.acceptedChan), cap(e.acceptedChan)) + i := 0 + for ep := range e.acceptedChan { + e.acceptedEndpoints[i] = ep + i++ + } + if i != len(e.acceptedEndpoints) { + panic("endpoint acceptedChan buffer got consumed by background context") + } + } +} + +// saveState is invoked by stateify. +func (e *endpoint) saveState() endpointState { + return e.state +} + +// Endpoint loading must be done in the following ordering by their state, to +// avoid dangling connecting w/o listening peer, and to avoid conflicts in port +// reservation. +var connectedLoading sync.WaitGroup +var listenLoading sync.WaitGroup +var connectingLoading sync.WaitGroup + +// Bound endpoint loading happens last. + +// loadState is invoked by stateify. +func (e *endpoint) loadState(state endpointState) { + // This is to ensure that the loading wait groups include all applicable + // endpoints before any asynchronous calls to the Wait() methods. + switch state { + case stateConnected: + connectedLoading.Add(1) + case stateListen: + listenLoading.Add(1) + case stateConnecting: + connectingLoading.Add(1) + } + e.state = state } // afterLoad is invoked by stateify. func (e *endpoint) afterLoad() { + // We load acceptedChan buffer indirectly here. Note that closed + // endpoints might not need to allocate the channel. + // FIXME + if cap(e.acceptedEndpoints) > 0 { + e.acceptedChan = make(chan *endpoint, cap(e.acceptedEndpoints)) + for _, ep := range e.acceptedEndpoints { + e.acceptedChan <- ep + } + e.acceptedEndpoints = nil + } + e.stack = stack.StackFromEnv e.segmentQueue.setLimit(2 * e.rcvBufSize) e.workMu.Init() state := e.state switch state { - case stateInitial, stateBound, stateListen, stateConnecting: + case stateInitial, stateBound, stateListen, stateConnecting, stateConnected: var ss SendBufferSizeOption if err := e.stack.TransportProtocolOption(ProtocolNumber, &ss); err == nil { if e.sndBufSize < ss.Min || e.sndBufSize > ss.Max { @@ -80,65 +142,72 @@ func (e *endpoint) afterLoad() { } } - switch state { - case stateBound, stateListen, stateConnecting: + bind := func() { e.state = stateInitial - if err := e.Bind(tcpip.FullAddress{Addr: e.id.LocalAddress, Port: e.id.LocalPort}, nil); err != nil { + if len(e.bindAddress) == 0 { + e.bindAddress = e.id.LocalAddress + } + if err := e.Bind(tcpip.FullAddress{Addr: e.bindAddress, Port: e.id.LocalPort}, nil); err != nil { panic("endpoint binding failed: " + err.String()) } } switch state { - case stateListen: - backlog := cap(e.acceptedChan) - e.acceptedChan = nil - if err := e.Listen(backlog); err != nil { - panic("endpoint listening failed: " + err.String()) + case stateConnected: + bind() + if len(e.connectingAddress) == 0 { + // This endpoint is accepted by netstack but not yet by + // the app. If the endpoint is IPv6 but the remote + // address is IPv4, we need to connect as IPv6 so that + // dual-stack mode can be properly activated. + if e.netProto == header.IPv6ProtocolNumber && len(e.id.RemoteAddress) != header.IPv6AddressSize { + e.connectingAddress = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xff\xff" + e.id.RemoteAddress + } else { + e.connectingAddress = e.id.RemoteAddress + } } - } - - switch state { - case stateConnecting: - if err := e.Connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}); err != tcpip.ErrConnectStarted { + if err := e.connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}, false, e.workerRunning); err != tcpip.ErrConnectStarted { panic("endpoint connecting failed: " + err.String()) } + connectedLoading.Done() + case stateListen: + tcpip.AsyncLoading.Add(1) + go func() { + connectedLoading.Wait() + bind() + backlog := cap(e.acceptedChan) + if err := e.Listen(backlog); err != nil { + panic("endpoint listening failed: " + err.String()) + } + listenLoading.Done() + tcpip.AsyncLoading.Done() + }() + case stateConnecting: + tcpip.AsyncLoading.Add(1) + go func() { + connectedLoading.Wait() + listenLoading.Wait() + bind() + if err := e.Connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}); err != tcpip.ErrConnectStarted { + panic("endpoint connecting failed: " + err.String()) + } + connectingLoading.Done() + tcpip.AsyncLoading.Done() + }() + case stateBound: + tcpip.AsyncLoading.Add(1) + go func() { + connectedLoading.Wait() + listenLoading.Wait() + connectingLoading.Wait() + bind() + tcpip.AsyncLoading.Done() + }() + case stateClosed, stateError: + tcpip.DeleteDanglingEndpoint(e) } } -// saveAcceptedChan is invoked by stateify. -func (e *endpoint) saveAcceptedChan() endpointChan { - if e.acceptedChan == nil { - return endpointChan{} - } - close(e.acceptedChan) - buffer := make([]*endpoint, 0, len(e.acceptedChan)) - for ep := range e.acceptedChan { - buffer = append(buffer, ep) - } - if len(buffer) != cap(buffer) { - panic("endpoint.acceptedChan buffer got consumed by background context") - } - c := cap(e.acceptedChan) - e.acceptedChan = nil - return endpointChan{buffer: buffer, cap: c} -} - -// loadAcceptedChan is invoked by stateify. -func (e *endpoint) loadAcceptedChan(c endpointChan) { - if c.cap == 0 { - return - } - e.acceptedChan = make(chan *endpoint, c.cap) - for _, ep := range c.buffer { - e.acceptedChan <- ep - } -} - -type endpointChan struct { - buffer []*endpoint - cap int -} - // saveLastError is invoked by stateify. func (e *endpoint) saveLastError() string { if e.lastError == nil { diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index 07e4bfd73..c5bff5f4f 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -29,9 +29,9 @@ const ( type segment struct { segmentEntry refCnt int32 - id stack.TransportEndpointID - route stack.Route `state:"manual"` - data buffer.VectorisedView + id stack.TransportEndpointID `state:"manual"` + route stack.Route `state:"manual"` + data buffer.VectorisedView `state:".(buffer.VectorisedView)"` // views is used as buffer for data when its length is large // enough to store a VectorisedView. views [8]buffer.View @@ -45,7 +45,7 @@ type segment struct { // parsedOptions stores the parsed values from the options in the segment. parsedOptions header.TCPOptions - options []byte + options []byte `state:".([]byte)"` } func newSegment(r *stack.Route, id stack.TransportEndpointID, vv *buffer.VectorisedView) *segment { diff --git a/pkg/tcpip/transport/tcp/segment_queue.go b/pkg/tcpip/transport/tcp/segment_queue.go index c4a7f7d5b..a5e7b2ebf 100644 --- a/pkg/tcpip/transport/tcp/segment_queue.go +++ b/pkg/tcpip/transport/tcp/segment_queue.go @@ -12,8 +12,8 @@ import ( // segmentQueue is a bounded, thread-safe queue of TCP segments. type segmentQueue struct { - mu sync.Mutex - list segmentList + mu sync.Mutex `state:"nosave"` + list segmentList `state:"wait"` limit int used int } diff --git a/pkg/tcpip/transport/tcp/segment_state.go b/pkg/tcpip/transport/tcp/segment_state.go new file mode 100644 index 000000000..e5243200b --- /dev/null +++ b/pkg/tcpip/transport/tcp/segment_state.go @@ -0,0 +1,41 @@ +// Copyright 2018 The Netstack Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package tcp + +import ( + "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" +) + +// saveData is invoked by stateify. +func (s *segment) saveData() buffer.VectorisedView { + // We cannot save s.data directly as s.data.views may alias to s.views, + // which is not allowed by state framework (in-struct pointer). + return s.data.Clone(nil) +} + +// loadData is invoked by stateify. +func (s *segment) loadData(data buffer.VectorisedView) { + // NOTE: We cannot do the s.data = data.Clone(s.views[:]) optimization + // here because data.views is not guaranteed to be loaded by now. Plus, + // data.views will be allocated anyway so there really is little point + // of utilizing s.views for data.views. + s.data = data +} + +// saveOptions is invoked by stateify. +func (s *segment) saveOptions() []byte { + // We cannot save s.options directly as it may point to s.data's trimmed + // tail, which is not allowed by state framework (in-struct pointer). + b := make([]byte, 0, cap(s.options)) + return append(b, s.options...) +} + +// loadOptions is invoked by stateify. +func (s *segment) loadOptions(options []byte) { + // NOTE: We cannot point s.options back into s.data's trimmed tail. But + // it is OK as they do not need to aliased. Plus, options is already + // allocated so there is no cost here. + s.options = options +} diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index 95bea4d88..a98aca293 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -28,7 +28,7 @@ type sender struct { ep *endpoint // lastSendTime is the timestamp when the last packet was sent. - lastSendTime time.Time + lastSendTime time.Time `state:".(unixTime)"` // dupAckCount is the number of duplicated acks received. It is used for // fast retransmit. @@ -71,7 +71,7 @@ type sender struct { rttMeasureSeqNum seqnum.Value // rttMeasureTime is the time when the rttMeasureSeqNum was sent. - rttMeasureTime time.Time + rttMeasureTime time.Time `state:".(unixTime)"` closed bool writeNext *segment diff --git a/pkg/tcpip/transport/tcp/snd_state.go b/pkg/tcpip/transport/tcp/snd_state.go new file mode 100644 index 000000000..d68773a7c --- /dev/null +++ b/pkg/tcpip/transport/tcp/snd_state.go @@ -0,0 +1,39 @@ +// Copyright 2018 The Netstack Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package tcp + +import ( + "time" +) + +type unixTime struct { + second int64 + nano int64 +} + +// saveLastSendTime is invoked by stateify. +func (s *sender) saveLastSendTime() unixTime { + return unixTime{s.lastSendTime.Unix(), s.lastSendTime.UnixNano()} +} + +// loadLastSendTime is invoked by stateify. +func (s *sender) loadLastSendTime(unix unixTime) { + s.lastSendTime = time.Unix(unix.second, unix.nano) +} + +// saveRttMeasureTime is invoked by stateify. +func (s *sender) saveRttMeasureTime() unixTime { + return unixTime{s.rttMeasureTime.Unix(), s.rttMeasureTime.UnixNano()} +} + +// loadRttMeasureTime is invoked by stateify. +func (s *sender) loadRttMeasureTime(unix unixTime) { + s.rttMeasureTime = time.Unix(unix.second, unix.nano) +} + +// afterLoad is invoked by stateify. +func (s *sender) afterLoad() { + s.resendTimer.init(&s.resendWaker) +} -- cgit v1.2.3 From 478f0ac0038afda267814fa154bcd32feb07c3b3 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Mon, 25 Jun 2018 15:22:04 -0700 Subject: Don't read FSContext.root without holding FSContext.mu IsChrooted still has the opportunity to race with another thread entering the FSContext into a chroot, but that is unchanged (and fine, AFAIK). PiperOrigin-RevId: 202029117 Change-Id: I38bce763b3a7715fa6ae98aa200a19d51a0235f1 --- pkg/sentry/kernel/task_resources.go | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_resources.go b/pkg/sentry/kernel/task_resources.go index e529f0c2d..0389989e8 100644 --- a/pkg/sentry/kernel/task_resources.go +++ b/pkg/sentry/kernel/task_resources.go @@ -122,5 +122,7 @@ func (t *Task) AbstractSockets() *AbstractSocketNamespace { func (t *Task) IsChrooted() bool { realRoot := t.k.mounts.Root() defer realRoot.DecRef() - return t.tr.FSContext.root != realRoot + root := t.tr.FSContext.RootDirectory() + defer root.DecRef() + return root != realRoot } -- cgit v1.2.3 From 4ac79312b093f2831079d0d71846747a4996d9ad Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Mon, 25 Jun 2018 16:45:31 -0700 Subject: Don't read cwd or root without holding mu PiperOrigin-RevId: 202043090 Change-Id: I3c47fb3413ca8615d50d8a0503d72fcce9b09421 --- pkg/sentry/kernel/fs_context.go | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/fs_context.go b/pkg/sentry/kernel/fs_context.go index 9aa6fa951..d1ca9c09d 100644 --- a/pkg/sentry/kernel/fs_context.go +++ b/pkg/sentry/kernel/fs_context.go @@ -114,11 +114,14 @@ func (f *FSContext) SetWorkingDirectory(d *fs.Dirent) { if d == nil { panic("FSContext.SetWorkingDirectory called with nil dirent") } + + f.mu.Lock() + defer f.mu.Unlock() + if f.cwd == nil { panic(fmt.Sprintf("FSContext.SetWorkingDirectory(%v)) called after destroy", d)) } - f.mu.Lock() - defer f.mu.Unlock() + old := f.cwd f.cwd = d d.IncRef() @@ -144,11 +147,14 @@ func (f *FSContext) SetRootDirectory(d *fs.Dirent) { if d == nil { panic("FSContext.SetRootDirectory called with nil dirent") } + + f.mu.Lock() + defer f.mu.Unlock() + if f.root == nil { panic(fmt.Sprintf("FSContext.SetRootDirectory(%v)) called after destroy", d)) } - f.mu.Lock() - defer f.mu.Unlock() + old := f.root f.root = d d.IncRef() -- cgit v1.2.3 From 51c1e510ab79607d80d6b81c2ae8ab308c323a58 Mon Sep 17 00:00:00 2001 From: Brian Geffon Date: Tue, 26 Jun 2018 10:32:22 -0700 Subject: Automated rollback of changelist 201596247 PiperOrigin-RevId: 202151720 Change-Id: I0491172c436bbb32b977f557953ba0bc41cfe299 --- pkg/sentry/kernel/BUILD | 5 +- pkg/sentry/kernel/kernel.go | 6 - pkg/sentry/kernel/kernel_state.go | 31 ----- pkg/tcpip/stack/stack_global_state.go | 2 +- pkg/tcpip/tcpip.go | 36 ------ pkg/tcpip/transport/tcp/BUILD | 7 -- pkg/tcpip/transport/tcp/accept.go | 11 +- pkg/tcpip/transport/tcp/connect.go | 43 +------ pkg/tcpip/transport/tcp/endpoint.go | 72 +++--------- pkg/tcpip/transport/tcp/endpoint_state.go | 185 ++++++++++-------------------- pkg/tcpip/transport/tcp/segment.go | 8 +- pkg/tcpip/transport/tcp/segment_queue.go | 4 +- pkg/tcpip/transport/tcp/segment_state.go | 41 ------- pkg/tcpip/transport/tcp/snd.go | 4 +- pkg/tcpip/transport/tcp/snd_state.go | 39 ------- 15 files changed, 91 insertions(+), 403 deletions(-) delete mode 100644 pkg/sentry/kernel/kernel_state.go delete mode 100644 pkg/tcpip/transport/tcp/segment_state.go delete mode 100644 pkg/tcpip/transport/tcp/snd_state.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 07568b47c..b2a55ddff 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -12,7 +12,6 @@ go_stateify( "fs_context.go", "ipc_namespace.go", "kernel.go", - "kernel_state.go", "pending_signals.go", "pending_signals_state.go", "process_group_list.go", @@ -46,11 +45,10 @@ go_stateify( "vdso.go", "version.go", ], - out = "kernel_autogen_state.go", + out = "kernel_state.go", imports = [ "gvisor.googlesource.com/gvisor/pkg/sentry/arch", "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", - "gvisor.googlesource.com/gvisor/pkg/tcpip", ], package = "kernel", ) @@ -119,7 +117,6 @@ go_library( "fs_context.go", "ipc_namespace.go", "kernel.go", - "kernel_autogen_state.go", "kernel_state.go", "pending_signals.go", "pending_signals_list.go", diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 64439cd9d..5662b8f08 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -57,7 +57,6 @@ import ( sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/state" - "gvisor.googlesource.com/gvisor/pkg/tcpip" ) // Kernel represents an emulated Linux kernel. It must be initialized by calling @@ -159,9 +158,6 @@ type Kernel struct { // exitErr is the error causing the sandbox to exit, if any. It is // protected by extMu. exitErr error - - // danglingEndpoints is used to save / restore tcpip.DanglingEndpoints. - danglingEndpoints struct{} `state:".([]tcpip.Endpoint)"` } // InitKernelArgs holds arguments to Init. @@ -426,8 +422,6 @@ func (k *Kernel) LoadFrom(r io.Reader, p platform.Platform, net inet.Stack) erro return err } - tcpip.AsyncLoading.Wait() - log.Infof("Overall load took [%s]", time.Since(loadStart)) // Applications may size per-cpu structures based on k.applicationCores, so diff --git a/pkg/sentry/kernel/kernel_state.go b/pkg/sentry/kernel/kernel_state.go deleted file mode 100644 index bb2d5102d..000000000 --- a/pkg/sentry/kernel/kernel_state.go +++ /dev/null @@ -1,31 +0,0 @@ -// Copyright 2018 Google Inc. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package kernel - -import ( - "gvisor.googlesource.com/gvisor/pkg/tcpip" -) - -// saveDanglingEndpoints is invoked by stateify. -func (k *Kernel) saveDanglingEndpoints() []tcpip.Endpoint { - return tcpip.GetDanglingEndpoints() -} - -// loadDanglingEndpoints is invoked by stateify. -func (k *Kernel) loadDanglingEndpoints(es []tcpip.Endpoint) { - for _, e := range es { - tcpip.AddDanglingEndpoint(e) - } -} diff --git a/pkg/tcpip/stack/stack_global_state.go b/pkg/tcpip/stack/stack_global_state.go index 260d7d05c..030ae98d1 100644 --- a/pkg/tcpip/stack/stack_global_state.go +++ b/pkg/tcpip/stack/stack_global_state.go @@ -5,5 +5,5 @@ package stack // StackFromEnv is the global stack created in restore run. -// FIXME +// FIXME: remove this variable once tcpip S/R is fully supported. var StackFromEnv *Stack diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index 17fa0efb7..cf25a086d 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -23,7 +23,6 @@ import ( "fmt" "strconv" "strings" - "sync" "time" "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" @@ -553,38 +552,3 @@ type ProtocolAddress struct { // Address is a network address. Address Address } - -// danglingEndpointsMu protects access to danglingEndpoints. -var danglingEndpointsMu sync.Mutex - -// danglingEndpoints tracks all dangling endpoints no longer owned by the app. -var danglingEndpoints = make(map[Endpoint]struct{}) - -// GetDanglingEndpoints returns all dangling endpoints. -func GetDanglingEndpoints() []Endpoint { - es := make([]Endpoint, 0, len(danglingEndpoints)) - danglingEndpointsMu.Lock() - for e, _ := range danglingEndpoints { - es = append(es, e) - } - danglingEndpointsMu.Unlock() - return es -} - -// AddDanglingEndpoint adds a dangling endpoint. -func AddDanglingEndpoint(e Endpoint) { - danglingEndpointsMu.Lock() - danglingEndpoints[e] = struct{}{} - danglingEndpointsMu.Unlock() -} - -// DeleteDanglingEndpoint removes a dangling endpoint. -func DeleteDanglingEndpoint(e Endpoint) { - danglingEndpointsMu.Lock() - delete(danglingEndpoints, e) - danglingEndpointsMu.Unlock() -} - -// AsyncLoading is the global barrier for asynchronous endpoint loading -// activities. -var AsyncLoading sync.WaitGroup diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index d129aa285..f38f58e87 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -10,16 +10,11 @@ go_stateify( "endpoint.go", "endpoint_state.go", "rcv.go", - "segment.go", "segment_heap.go", - "segment_queue.go", - "segment_state.go", "snd.go", - "snd_state.go", "tcp_segment_list.go", ], out = "tcp_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], package = "tcp", ) @@ -48,9 +43,7 @@ go_library( "segment.go", "segment_heap.go", "segment_queue.go", - "segment_state.go", "snd.go", - "snd_state.go", "tcp_segment_list.go", "tcp_state.go", "timer.go", diff --git a/pkg/tcpip/transport/tcp/accept.go b/pkg/tcpip/transport/tcp/accept.go index 410dfdad4..85adeef0e 100644 --- a/pkg/tcpip/transport/tcp/accept.go +++ b/pkg/tcpip/transport/tcp/accept.go @@ -68,8 +68,7 @@ func encodeMSS(mss uint16) uint32 { // to go above a threshold. var synRcvdCount struct { sync.Mutex - value uint64 - pending sync.WaitGroup + value uint64 } // listenContext is used by a listening endpoint to store state used while @@ -103,7 +102,6 @@ func incSynRcvdCount() bool { return false } - synRcvdCount.pending.Add(1) synRcvdCount.value++ return true @@ -117,7 +115,6 @@ func decSynRcvdCount() { defer synRcvdCount.Unlock() synRcvdCount.value-- - synRcvdCount.pending.Done() } // newListenContext creates a new listen context. @@ -295,7 +292,7 @@ func (e *endpoint) handleListenSegment(ctx *listenContext, s *segment) { opts := parseSynSegmentOptions(s) if incSynRcvdCount() { s.incRef() - go e.handleSynSegment(ctx, s, &opts) // S/R-SAFE: synRcvdCount is the barrier. + go e.handleSynSegment(ctx, s, &opts) // S/R-FIXME } else { cookie := ctx.createCookie(s.id, s.sequenceNumber, encodeMSS(opts.MSS)) // Send SYN with window scaling because we currently @@ -384,12 +381,10 @@ func (e *endpoint) protocolListenLoop(rcvWnd seqnum.Size) *tcpip.Error { return nil } if n¬ifyDrain != 0 { - for !e.segmentQueue.empty() { - s := e.segmentQueue.dequeue() + for s := e.segmentQueue.dequeue(); s != nil; s = e.segmentQueue.dequeue() { e.handleListenSegment(ctx, s) s.decRef() } - synRcvdCount.pending.Wait() close(e.drainDone) <-e.undrain } diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go index d9f87c793..9aaabe0b1 100644 --- a/pkg/tcpip/transport/tcp/connect.go +++ b/pkg/tcpip/transport/tcp/connect.go @@ -443,8 +443,7 @@ func (h *handshake) execute() *tcpip.Error { return tcpip.ErrAborted } if n¬ifyDrain != 0 { - for !h.ep.segmentQueue.empty() { - s := h.ep.segmentQueue.dequeue() + for s := h.ep.segmentQueue.dequeue(); s != nil; s = h.ep.segmentQueue.dequeue() { err := h.handleSegment(s) s.decRef() if err != nil { @@ -814,13 +813,15 @@ func (e *endpoint) handleSegments() *tcpip.Error { // protocolMainLoop is the main loop of the TCP protocol. It runs in its own // goroutine and is responsible for sending segments and handling received // segments. -func (e *endpoint) protocolMainLoop(handshake bool) *tcpip.Error { +func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { var closeTimer *time.Timer var closeWaker sleep.Waker defer func() { // e.mu is expected to be hold upon entering this section. + e.completeWorkerLocked() + if e.snd != nil { e.snd.resendTimer.cleanup() } @@ -829,8 +830,6 @@ func (e *endpoint) protocolMainLoop(handshake bool) *tcpip.Error { closeTimer.Stop() } - e.completeWorkerLocked() - if e.drainDone != nil { close(e.drainDone) } @@ -841,7 +840,7 @@ func (e *endpoint) protocolMainLoop(handshake bool) *tcpip.Error { e.waiterQueue.Notify(waiter.EventHUp | waiter.EventErr | waiter.EventIn | waiter.EventOut) }() - if handshake { + if !passive { // This is an active connection, so we must initiate the 3-way // handshake, and then inform potential waiters about its // completion. @@ -946,17 +945,6 @@ func (e *endpoint) protocolMainLoop(handshake bool) *tcpip.Error { closeWaker.Assert() }) } - - if n¬ifyDrain != 0 { - for !e.segmentQueue.empty() { - if err := e.handleSegments(); err != nil { - return err - } - } - close(e.drainDone) - <-e.undrain - } - return nil }, }, @@ -968,27 +956,6 @@ func (e *endpoint) protocolMainLoop(handshake bool) *tcpip.Error { s.AddWaker(funcs[i].w, i) } - // The following assertions and notifications are needed for restored - // endpoints. Fresh newly created endpoints have empty states and should - // not invoke any. - e.segmentQueue.mu.Lock() - if !e.segmentQueue.list.Empty() { - e.newSegmentWaker.Assert() - } - e.segmentQueue.mu.Unlock() - - e.rcvListMu.Lock() - if !e.rcvList.Empty() { - e.waiterQueue.Notify(waiter.EventIn) - } - e.rcvListMu.Unlock() - - e.mu.RLock() - if e.workerCleanup { - e.notifyProtocolGoroutine(notifyClose) - } - e.mu.RUnlock() - // Main loop. Handle segments until both send and receive ends of the // connection have completed. for !e.rcv.closed || !e.snd.closed || e.snd.sndUna != e.snd.sndNxtList { diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index 706977618..b21c2b4ab 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -69,7 +69,7 @@ type endpoint struct { // change throughout the lifetime of the endpoint. stack *stack.Stack `state:"manual"` netProto tcpip.NetworkProtocolNumber - waiterQueue *waiter.Queue `state:"wait"` + waiterQueue *waiter.Queue // lastError represents the last error that the endpoint reported; // access to it is protected by the following mutex. @@ -82,8 +82,8 @@ type endpoint struct { // // Once the peer has closed its send side, rcvClosed is set to true // to indicate to users that no more data is coming. - rcvListMu sync.Mutex `state:"nosave"` - rcvList segmentList `state:"wait"` + rcvListMu sync.Mutex `state:"nosave"` + rcvList segmentList rcvClosed bool rcvBufSize int rcvBufUsed int @@ -91,8 +91,8 @@ type endpoint struct { // The following fields are protected by the mutex. mu sync.RWMutex `state:"nosave"` id stack.TransportEndpointID - state endpointState `state:".(endpointState)"` - isPortReserved bool `state:"manual"` + state endpointState + isPortReserved bool `state:"manual"` isRegistered bool boundNICID tcpip.NICID `state:"manual"` route stack.Route `state:"manual"` @@ -118,7 +118,7 @@ type endpoint struct { // workerCleanup specifies if the worker goroutine must perform cleanup // before exitting. This can only be set to true when workerRunning is // also true, and they're both protected by the mutex. - workerCleanup bool + workerCleanup bool `state:"zerovalue"` // sendTSOk is used to indicate when the TS Option has been negotiated. // When sendTSOk is true every non-RST segment should carry a TS as per @@ -153,7 +153,7 @@ type endpoint struct { // segmentQueue is used to hand received segments to the protocol // goroutine. Segments are queued as long as the queue is not full, // and dropped when it is. - segmentQueue segmentQueue `state:"wait"` + segmentQueue segmentQueue `state:"zerovalue"` // The following fields are used to manage the send buffer. When // segments are ready to be sent, they are added to sndQueue and the @@ -166,7 +166,7 @@ type endpoint struct { sndBufUsed int sndClosed bool sndBufInQueue seqnum.Size - sndQueue segmentList `state:"wait"` + sndQueue segmentList sndWaker sleep.Waker `state:"manual"` sndCloseWaker sleep.Waker `state:"manual"` @@ -188,21 +188,17 @@ type endpoint struct { // notifyFlags is a bitmask of flags used to indicate to the protocol // goroutine what it was notified; this is only accessed atomically. - notifyFlags uint32 `state:"nosave"` + notifyFlags uint32 `state:"zerovalue"` // acceptedChan is used by a listening endpoint protocol goroutine to // send newly accepted connections to the endpoint so that they can be // read by Accept() calls. - acceptedChan chan *endpoint `state:"manual"` - - // acceptedEndpoints is only used to save / restore the channel buffer. - // FIXME - acceptedEndpoints []*endpoint + acceptedChan chan *endpoint `state:".(endpointChan)"` // The following are only used from the protocol goroutine, and // therefore don't need locks to protect them. - rcv *receiver `state:"wait"` - snd *sender `state:"wait"` + rcv *receiver + snd *sender // The goroutine drain completion notification channel. drainDone chan struct{} `state:"nosave"` @@ -215,7 +211,6 @@ type endpoint struct { probe stack.TCPProbeFunc `state:"nosave"` // The following are only used to assist the restore run to re-connect. - bindAddress tcpip.Address connectingAddress tcpip.Address } @@ -349,7 +344,6 @@ func (e *endpoint) Close() { // Either perform the local cleanup or kick the worker to make sure it // knows it needs to cleanup. - tcpip.AddDanglingEndpoint(e) if !e.workerRunning { e.cleanupLocked() } else { @@ -369,12 +363,9 @@ func (e *endpoint) cleanupLocked() { if e.acceptedChan != nil { close(e.acceptedChan) for n := range e.acceptedChan { - n.mu.Lock() n.resetConnectionLocked(tcpip.ErrConnectionAborted) - n.mu.Unlock() n.Close() } - e.acceptedChan = nil } e.workerCleanup = false @@ -383,7 +374,6 @@ func (e *endpoint) cleanupLocked() { } e.route.Release() - tcpip.DeleteDanglingEndpoint(e) } // Read reads data from the endpoint. @@ -796,16 +786,6 @@ func (e *endpoint) checkV4Mapped(addr *tcpip.FullAddress) (tcpip.NetworkProtocol // Connect connects the endpoint to its peer. func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { - return e.connect(addr, true, true) -} - -// connect connects the endpoint to its peer. In the normal non-S/R case, the -// new connection is expected to run the main goroutine and perform handshake. -// In restore of previously connected endpoints, both ends will be passively -// created (so no new handshaking is done); for stack-accepted connections not -// yet accepted by the app, they are restored without running the main goroutine -// here. -func (e *endpoint) connect(addr tcpip.FullAddress, handshake bool, run bool) *tcpip.Error { e.mu.Lock() defer e.mu.Unlock() @@ -917,27 +897,9 @@ func (e *endpoint) connect(addr tcpip.FullAddress, handshake bool, run bool) *tc e.boundNICID = nicid e.effectiveNetProtos = netProtos e.connectingAddress = connectingAddr + e.workerRunning = true - // Connect in the restore phase does not perform handshake. Restore its - // connection setting here. - if !handshake { - e.segmentQueue.mu.Lock() - for _, l := range []segmentList{e.segmentQueue.list, e.sndQueue, e.snd.writeList} { - for s := l.Front(); s != nil; s = s.Next() { - s.id = e.id - s.route = r.Clone() - e.sndWaker.Assert() - } - } - e.segmentQueue.mu.Unlock() - e.snd.updateMaxPayloadSize(int(e.route.MTU()), 0) - e.state = stateConnected - } - - if run { - e.workerRunning = true - go e.protocolMainLoop(handshake) // S/R-SAFE: will be drained before save. - } + go e.protocolMainLoop(false) // S/R-SAFE: will be drained before save. return tcpip.ErrConnectStarted } @@ -1009,9 +971,6 @@ func (e *endpoint) Listen(backlog int) *tcpip.Error { if len(e.acceptedChan) > backlog { return tcpip.ErrInvalidEndpointState } - if cap(e.acceptedChan) == backlog { - return nil - } origChan := e.acceptedChan e.acceptedChan = make(chan *endpoint, backlog) close(origChan) @@ -1049,7 +1008,7 @@ func (e *endpoint) Listen(backlog int) *tcpip.Error { func (e *endpoint) startAcceptedLoop(waiterQueue *waiter.Queue) { e.waiterQueue = waiterQueue e.workerRunning = true - go e.protocolMainLoop(false) // S/R-SAFE: drained on save. + go e.protocolMainLoop(true) // S/R-FIXME } // Accept returns a new endpoint if a peer has established a connection @@ -1090,7 +1049,6 @@ func (e *endpoint) Bind(addr tcpip.FullAddress, commit func() *tcpip.Error) (ret return tcpip.ErrAlreadyBound } - e.bindAddress = addr.Addr netProto, err := e.checkV4Mapped(&addr) if err != nil { return err diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go index 38c97c796..b1e249bff 100644 --- a/pkg/tcpip/transport/tcp/endpoint_state.go +++ b/pkg/tcpip/transport/tcp/endpoint_state.go @@ -9,7 +9,6 @@ import ( "sync" "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/header" "gvisor.googlesource.com/gvisor/pkg/tcpip/stack" ) @@ -23,7 +22,7 @@ func (e *endpoint) drainSegmentLocked() { e.undrain = make(chan struct{}) e.mu.Unlock() - e.notifyProtocolGoroutine(notifyDrain) + e.notificationWaker.Assert() <-e.drainDone e.mu.Lock() @@ -39,98 +38,37 @@ func (e *endpoint) beforeSave() { switch e.state { case stateInitial, stateBound: - case stateListen, stateConnecting, stateConnected: - if e.state == stateConnected && !e.workerRunning { - // The endpoint must be in acceptedChan. - break + case stateListen: + if !e.segmentQueue.empty() { + e.drainSegmentLocked() } + case stateConnecting: e.drainSegmentLocked() - if e.state != stateClosed && e.state != stateError { - if !e.workerRunning { - panic("endpoint has no worker running in listen, connecting, or connected state") - } + if e.state != stateConnected { break } fallthrough + case stateConnected: + // FIXME + panic(tcpip.ErrSaveRejection{fmt.Errorf("endpoint cannot be saved in connected state: local %v:%v, remote %v:%v", e.id.LocalAddress, e.id.LocalPort, e.id.RemoteAddress, e.id.RemotePort)}) case stateClosed, stateError: if e.workerRunning { - panic("endpoint still has worker running in closed or error state") + panic(fmt.Sprintf("endpoint still has worker running in closed or error state")) } default: panic(fmt.Sprintf("endpoint in unknown state %v", e.state)) } - - if e.waiterQueue != nil && !e.waiterQueue.IsEmpty() { - panic("endpoint still has waiters upon save") - } - - if !((e.state == stateBound || e.state == stateListen) == e.isPortReserved) { - panic("endpoint port must and must only be reserved in bound or listen state") - } - - if e.acceptedChan != nil { - close(e.acceptedChan) - e.acceptedEndpoints = make([]*endpoint, len(e.acceptedChan), cap(e.acceptedChan)) - i := 0 - for ep := range e.acceptedChan { - e.acceptedEndpoints[i] = ep - i++ - } - if i != len(e.acceptedEndpoints) { - panic("endpoint acceptedChan buffer got consumed by background context") - } - } -} - -// saveState is invoked by stateify. -func (e *endpoint) saveState() endpointState { - return e.state -} - -// Endpoint loading must be done in the following ordering by their state, to -// avoid dangling connecting w/o listening peer, and to avoid conflicts in port -// reservation. -var connectedLoading sync.WaitGroup -var listenLoading sync.WaitGroup -var connectingLoading sync.WaitGroup - -// Bound endpoint loading happens last. - -// loadState is invoked by stateify. -func (e *endpoint) loadState(state endpointState) { - // This is to ensure that the loading wait groups include all applicable - // endpoints before any asynchronous calls to the Wait() methods. - switch state { - case stateConnected: - connectedLoading.Add(1) - case stateListen: - listenLoading.Add(1) - case stateConnecting: - connectingLoading.Add(1) - } - e.state = state } // afterLoad is invoked by stateify. func (e *endpoint) afterLoad() { - // We load acceptedChan buffer indirectly here. Note that closed - // endpoints might not need to allocate the channel. - // FIXME - if cap(e.acceptedEndpoints) > 0 { - e.acceptedChan = make(chan *endpoint, cap(e.acceptedEndpoints)) - for _, ep := range e.acceptedEndpoints { - e.acceptedChan <- ep - } - e.acceptedEndpoints = nil - } - e.stack = stack.StackFromEnv e.segmentQueue.setLimit(2 * e.rcvBufSize) e.workMu.Init() state := e.state switch state { - case stateInitial, stateBound, stateListen, stateConnecting, stateConnected: + case stateInitial, stateBound, stateListen, stateConnecting: var ss SendBufferSizeOption if err := e.stack.TransportProtocolOption(ProtocolNumber, &ss); err == nil { if e.sndBufSize < ss.Min || e.sndBufSize > ss.Max { @@ -142,72 +80,65 @@ func (e *endpoint) afterLoad() { } } - bind := func() { + switch state { + case stateBound, stateListen, stateConnecting: e.state = stateInitial - if len(e.bindAddress) == 0 { - e.bindAddress = e.id.LocalAddress - } - if err := e.Bind(tcpip.FullAddress{Addr: e.bindAddress, Port: e.id.LocalPort}, nil); err != nil { + if err := e.Bind(tcpip.FullAddress{Addr: e.id.LocalAddress, Port: e.id.LocalPort}, nil); err != nil { panic("endpoint binding failed: " + err.String()) } } switch state { - case stateConnected: - bind() - if len(e.connectingAddress) == 0 { - // This endpoint is accepted by netstack but not yet by - // the app. If the endpoint is IPv6 but the remote - // address is IPv4, we need to connect as IPv6 so that - // dual-stack mode can be properly activated. - if e.netProto == header.IPv6ProtocolNumber && len(e.id.RemoteAddress) != header.IPv6AddressSize { - e.connectingAddress = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xff\xff" + e.id.RemoteAddress - } else { - e.connectingAddress = e.id.RemoteAddress - } + case stateListen: + backlog := cap(e.acceptedChan) + e.acceptedChan = nil + if err := e.Listen(backlog); err != nil { + panic("endpoint listening failed: " + err.String()) } - if err := e.connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}, false, e.workerRunning); err != tcpip.ErrConnectStarted { + } + + switch state { + case stateConnecting: + if err := e.Connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}); err != tcpip.ErrConnectStarted { panic("endpoint connecting failed: " + err.String()) } - connectedLoading.Done() - case stateListen: - tcpip.AsyncLoading.Add(1) - go func() { - connectedLoading.Wait() - bind() - backlog := cap(e.acceptedChan) - if err := e.Listen(backlog); err != nil { - panic("endpoint listening failed: " + err.String()) - } - listenLoading.Done() - tcpip.AsyncLoading.Done() - }() - case stateConnecting: - tcpip.AsyncLoading.Add(1) - go func() { - connectedLoading.Wait() - listenLoading.Wait() - bind() - if err := e.Connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}); err != tcpip.ErrConnectStarted { - panic("endpoint connecting failed: " + err.String()) - } - connectingLoading.Done() - tcpip.AsyncLoading.Done() - }() - case stateBound: - tcpip.AsyncLoading.Add(1) - go func() { - connectedLoading.Wait() - listenLoading.Wait() - connectingLoading.Wait() - bind() - tcpip.AsyncLoading.Done() - }() - case stateClosed, stateError: - tcpip.DeleteDanglingEndpoint(e) } } +// saveAcceptedChan is invoked by stateify. +func (e *endpoint) saveAcceptedChan() endpointChan { + if e.acceptedChan == nil { + return endpointChan{} + } + close(e.acceptedChan) + buffer := make([]*endpoint, 0, len(e.acceptedChan)) + for ep := range e.acceptedChan { + buffer = append(buffer, ep) + } + if len(buffer) != cap(buffer) { + panic("endpoint.acceptedChan buffer got consumed by background context") + } + c := cap(e.acceptedChan) + e.acceptedChan = nil + return endpointChan{buffer: buffer, cap: c} +} + +// loadAcceptedChan is invoked by stateify. +func (e *endpoint) loadAcceptedChan(c endpointChan) { + if c.cap == 0 { + return + } + e.acceptedChan = make(chan *endpoint, c.cap) + for _, ep := range c.buffer { + e.acceptedChan <- ep + } +} + +type endpointChan struct { + buffer []*endpoint + cap int +} + // saveLastError is invoked by stateify. func (e *endpoint) saveLastError() string { if e.lastError == nil { diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index c5bff5f4f..07e4bfd73 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -29,9 +29,9 @@ const ( type segment struct { segmentEntry refCnt int32 - id stack.TransportEndpointID `state:"manual"` - route stack.Route `state:"manual"` - data buffer.VectorisedView `state:".(buffer.VectorisedView)"` + id stack.TransportEndpointID + route stack.Route `state:"manual"` + data buffer.VectorisedView // views is used as buffer for data when its length is large // enough to store a VectorisedView. views [8]buffer.View @@ -45,7 +45,7 @@ type segment struct { // parsedOptions stores the parsed values from the options in the segment. parsedOptions header.TCPOptions - options []byte `state:".([]byte)"` + options []byte } func newSegment(r *stack.Route, id stack.TransportEndpointID, vv *buffer.VectorisedView) *segment { diff --git a/pkg/tcpip/transport/tcp/segment_queue.go b/pkg/tcpip/transport/tcp/segment_queue.go index a5e7b2ebf..c4a7f7d5b 100644 --- a/pkg/tcpip/transport/tcp/segment_queue.go +++ b/pkg/tcpip/transport/tcp/segment_queue.go @@ -12,8 +12,8 @@ import ( // segmentQueue is a bounded, thread-safe queue of TCP segments. type segmentQueue struct { - mu sync.Mutex `state:"nosave"` - list segmentList `state:"wait"` + mu sync.Mutex + list segmentList limit int used int } diff --git a/pkg/tcpip/transport/tcp/segment_state.go b/pkg/tcpip/transport/tcp/segment_state.go deleted file mode 100644 index e5243200b..000000000 --- a/pkg/tcpip/transport/tcp/segment_state.go +++ /dev/null @@ -1,41 +0,0 @@ -// Copyright 2018 The Netstack Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -package tcp - -import ( - "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" -) - -// saveData is invoked by stateify. -func (s *segment) saveData() buffer.VectorisedView { - // We cannot save s.data directly as s.data.views may alias to s.views, - // which is not allowed by state framework (in-struct pointer). - return s.data.Clone(nil) -} - -// loadData is invoked by stateify. -func (s *segment) loadData(data buffer.VectorisedView) { - // NOTE: We cannot do the s.data = data.Clone(s.views[:]) optimization - // here because data.views is not guaranteed to be loaded by now. Plus, - // data.views will be allocated anyway so there really is little point - // of utilizing s.views for data.views. - s.data = data -} - -// saveOptions is invoked by stateify. -func (s *segment) saveOptions() []byte { - // We cannot save s.options directly as it may point to s.data's trimmed - // tail, which is not allowed by state framework (in-struct pointer). - b := make([]byte, 0, cap(s.options)) - return append(b, s.options...) -} - -// loadOptions is invoked by stateify. -func (s *segment) loadOptions(options []byte) { - // NOTE: We cannot point s.options back into s.data's trimmed tail. But - // it is OK as they do not need to aliased. Plus, options is already - // allocated so there is no cost here. - s.options = options -} diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index a98aca293..95bea4d88 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -28,7 +28,7 @@ type sender struct { ep *endpoint // lastSendTime is the timestamp when the last packet was sent. - lastSendTime time.Time `state:".(unixTime)"` + lastSendTime time.Time // dupAckCount is the number of duplicated acks received. It is used for // fast retransmit. @@ -71,7 +71,7 @@ type sender struct { rttMeasureSeqNum seqnum.Value // rttMeasureTime is the time when the rttMeasureSeqNum was sent. - rttMeasureTime time.Time `state:".(unixTime)"` + rttMeasureTime time.Time closed bool writeNext *segment diff --git a/pkg/tcpip/transport/tcp/snd_state.go b/pkg/tcpip/transport/tcp/snd_state.go deleted file mode 100644 index d68773a7c..000000000 --- a/pkg/tcpip/transport/tcp/snd_state.go +++ /dev/null @@ -1,39 +0,0 @@ -// Copyright 2018 The Netstack Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -package tcp - -import ( - "time" -) - -type unixTime struct { - second int64 - nano int64 -} - -// saveLastSendTime is invoked by stateify. -func (s *sender) saveLastSendTime() unixTime { - return unixTime{s.lastSendTime.Unix(), s.lastSendTime.UnixNano()} -} - -// loadLastSendTime is invoked by stateify. -func (s *sender) loadLastSendTime(unix unixTime) { - s.lastSendTime = time.Unix(unix.second, unix.nano) -} - -// saveRttMeasureTime is invoked by stateify. -func (s *sender) saveRttMeasureTime() unixTime { - return unixTime{s.rttMeasureTime.Unix(), s.rttMeasureTime.UnixNano()} -} - -// loadRttMeasureTime is invoked by stateify. -func (s *sender) loadRttMeasureTime(unix unixTime) { - s.rttMeasureTime = time.Unix(unix.second, unix.nano) -} - -// afterLoad is invoked by stateify. -func (s *sender) afterLoad() { - s.resendTimer.init(&s.resendWaker) -} -- cgit v1.2.3 From 99afc982f1f0e40059e1446ea6f3cb725b1fbde7 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Wed, 27 Jun 2018 14:30:45 -0700 Subject: Call mm.CheckIORange() when copying in IOVecs. CheckIORange is analagous to Linux's access_ok() method, which is checked when copying in IOVecs in both lib/iov_iter.c:import_single_range() and lib/iov_iter.c:import_iovec() => fs/read_write.c:rw_copy_check_uvector(). gVisor copies in IOVecs via Task.SingleIOSequence() and Task.CopyInIovecs(). We were checking the address range bounds, but not whether the address is valid. To conform with linux, we should also check that the address is valid. For usual preadv/pwritev syscalls, the effect of this change is not noticeable, since we find out that the address is invalid before the syscall completes. For vectorized async-IO operations, however, this change is necessary because Linux returns EFAULT when the operation is submitted, but before it executes. Thus, we must validate the iovecs when copying them in. PiperOrigin-RevId: 202370092 Change-Id: I8759a63ccf7e6b90d90d30f78ab8935a0fcf4936 --- pkg/sentry/kernel/task_usermem.go | 17 +++++++++++------ pkg/sentry/mm/io.go | 18 +++++++++--------- 2 files changed, 20 insertions(+), 15 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_usermem.go b/pkg/sentry/kernel/task_usermem.go index 54964dd0d..2b4954869 100644 --- a/pkg/sentry/kernel/task_usermem.go +++ b/pkg/sentry/kernel/task_usermem.go @@ -184,6 +184,9 @@ func (t *Task) CopyOutIovecs(addr usermem.Addr, src usermem.AddrRangeSeq) error // - If the length of any AddrRange would cause its end to overflow, // CopyInIovecs returns EFAULT. // +// - If any AddrRange would include addresses outside the application address +// range, CopyInIovecs returns EFAULT. +// // - The combined length of all AddrRanges is limited to _MAX_RW_COUNT. If the // combined length of all AddrRanges would otherwise exceed this amount, ranges // beyond _MAX_RW_COUNT are silently truncated. @@ -218,7 +221,7 @@ func (t *Task) CopyInIovecs(addr usermem.Addr, numIovecs int) (usermem.AddrRange if length > math.MaxInt64 { return usermem.AddrRangeSeq{}, syserror.EINVAL } - ar, ok := base.ToRange(length) + ar, ok := t.MemoryManager().CheckIORange(base, int64(length)) if !ok { return usermem.AddrRangeSeq{}, syserror.EFAULT } @@ -251,18 +254,20 @@ func (t *Task) CopyInIovecs(addr usermem.Addr, numIovecs int) (usermem.AddrRange } // SingleIOSequence returns a usermem.IOSequence representing [addr, -// addr+length) in t's address space. If length exceeds _MAX_RW_COUNT, it is -// silently truncated. +// addr+length) in t's address space. If this contains addresses outside the +// application address range, it returns EFAULT. If length exceeds +// _MAX_RW_COUNT, the range is silently truncated. // // SingleIOSequence is analogous to Linux's // lib/iov_iter.c:import_single_range(). (Note that the non-vectorized read and -// write syscalls in Linux do not use import_single_range(), but are still -// truncated to _MAX_RW_COUNT by fs/read_write.c:rw_verify_area().) +// write syscalls in Linux do not use import_single_range(). However they check +// access_ok() in fs/read_write.c:vfs_read/vfs_write, and overflowing address +// ranges are truncated to _MAX_RW_COUNT by fs/read_write.c:rw_verify_area().) func (t *Task) SingleIOSequence(addr usermem.Addr, length int, opts usermem.IOOpts) (usermem.IOSequence, error) { if length > _MAX_RW_COUNT { length = _MAX_RW_COUNT } - ar, ok := addr.ToRange(uint64(length)) + ar, ok := t.MemoryManager().CheckIORange(addr, int64(length)) if !ok { return usermem.IOSequence{}, syserror.EFAULT } diff --git a/pkg/sentry/mm/io.go b/pkg/sentry/mm/io.go index cac81a59d..6741db594 100644 --- a/pkg/sentry/mm/io.go +++ b/pkg/sentry/mm/io.go @@ -60,11 +60,11 @@ const ( rwMapMinBytes = 512 ) -// checkIORange is similar to usermem.Addr.ToRange, but applies bounds checks +// CheckIORange is similar to usermem.Addr.ToRange, but applies bounds checks // consistent with Linux's arch/x86/include/asm/uaccess.h:access_ok(). // // Preconditions: length >= 0. -func (mm *MemoryManager) checkIORange(addr usermem.Addr, length int64) (usermem.AddrRange, bool) { +func (mm *MemoryManager) CheckIORange(addr usermem.Addr, length int64) (usermem.AddrRange, bool) { // Note that access_ok() constrains end even if length == 0. ar, ok := addr.ToRange(uint64(length)) return ar, (ok && ar.End <= mm.layout.MaxAddr) @@ -75,7 +75,7 @@ func (mm *MemoryManager) checkIORange(addr usermem.Addr, length int64) (usermem. func (mm *MemoryManager) checkIOVec(ars usermem.AddrRangeSeq) bool { for !ars.IsEmpty() { ar := ars.Head() - if _, ok := mm.checkIORange(ar.Start, int64(ar.Length())); !ok { + if _, ok := mm.CheckIORange(ar.Start, int64(ar.Length())); !ok { return false } ars = ars.Tail() @@ -101,7 +101,7 @@ func translateIOError(ctx context.Context, err error) error { // CopyOut implements usermem.IO.CopyOut. func (mm *MemoryManager) CopyOut(ctx context.Context, addr usermem.Addr, src []byte, opts usermem.IOOpts) (int, error) { - ar, ok := mm.checkIORange(addr, int64(len(src))) + ar, ok := mm.CheckIORange(addr, int64(len(src))) if !ok { return 0, syserror.EFAULT } @@ -144,7 +144,7 @@ func (mm *MemoryManager) asCopyOut(ctx context.Context, addr usermem.Addr, src [ // CopyIn implements usermem.IO.CopyIn. func (mm *MemoryManager) CopyIn(ctx context.Context, addr usermem.Addr, dst []byte, opts usermem.IOOpts) (int, error) { - ar, ok := mm.checkIORange(addr, int64(len(dst))) + ar, ok := mm.CheckIORange(addr, int64(len(dst))) if !ok { return 0, syserror.EFAULT } @@ -187,7 +187,7 @@ func (mm *MemoryManager) asCopyIn(ctx context.Context, addr usermem.Addr, dst [] // ZeroOut implements usermem.IO.ZeroOut. func (mm *MemoryManager) ZeroOut(ctx context.Context, addr usermem.Addr, toZero int64, opts usermem.IOOpts) (int64, error) { - ar, ok := mm.checkIORange(addr, toZero) + ar, ok := mm.CheckIORange(addr, toZero) if !ok { return 0, syserror.EFAULT } @@ -311,7 +311,7 @@ func (mm *MemoryManager) CopyInTo(ctx context.Context, ars usermem.AddrRangeSeq, // SwapUint32 implements usermem.IO.SwapUint32. func (mm *MemoryManager) SwapUint32(ctx context.Context, addr usermem.Addr, new uint32, opts usermem.IOOpts) (uint32, error) { - ar, ok := mm.checkIORange(addr, 4) + ar, ok := mm.CheckIORange(addr, 4) if !ok { return 0, syserror.EFAULT } @@ -353,7 +353,7 @@ func (mm *MemoryManager) SwapUint32(ctx context.Context, addr usermem.Addr, new // CompareAndSwapUint32 implements usermem.IO.CompareAndSwapUint32. func (mm *MemoryManager) CompareAndSwapUint32(ctx context.Context, addr usermem.Addr, old, new uint32, opts usermem.IOOpts) (uint32, error) { - ar, ok := mm.checkIORange(addr, 4) + ar, ok := mm.CheckIORange(addr, 4) if !ok { return 0, syserror.EFAULT } @@ -399,7 +399,7 @@ func (mm *MemoryManager) CompareAndSwapUint32(ctx context.Context, addr usermem. // Preconditions: mm.as != nil. ioar.Length() != 0. ioar.Contains(addr). func (mm *MemoryManager) handleASIOFault(ctx context.Context, addr usermem.Addr, ioar usermem.AddrRange, at usermem.AccessType) error { // Try to map all remaining pages in the I/O operation. This RoundUp can't - // overflow because otherwise it would have been caught by checkIORange. + // overflow because otherwise it would have been caught by CheckIORange. end, _ := ioar.End.RoundUp() ar := usermem.AddrRange{addr.RoundDown(), end} -- cgit v1.2.3 From 6b6852bceb12900f27a541682ddfe47893911c6e Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Wed, 27 Jun 2018 14:40:37 -0700 Subject: Fix semaphore data races PiperOrigin-RevId: 202371908 Change-Id: I72603b1d321878cae6404987c49e64732b676331 --- pkg/sentry/kernel/semaphore/semaphore.go | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index e9027dc14..a1ee83ce5 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -118,6 +118,9 @@ func (r *Registry) FindOrCreate(ctx context.Context, key, nsems int32, mode linu if !private { // Look up an existing semaphore. if set := r.findByKey(key); set != nil { + set.mu.Lock() + defer set.mu.Unlock() + // Check that caller can access semaphore set. creds := auth.CredentialsFromContext(ctx) if !set.checkPerms(creds, fs.PermsFromMode(mode)) { @@ -170,6 +173,9 @@ func (r *Registry) RemoveID(id int32, creds *auth.Credentials) error { return syserror.EINVAL } + set.mu.Lock() + defer set.mu.Unlock() + // "The effective user ID of the calling process must match the creator or // owner of the semaphore set, or the caller must be privileged." if !set.checkCredentials(creds) && !set.checkCapability(creds) { @@ -444,11 +450,9 @@ func (s *Set) checkPerms(creds *auth.Credentials, reqPerms fs.PermMask) bool { return s.checkCapability(creds) } +// destroy destroys the set. Caller must hold 's.mu'. func (s *Set) destroy() { - s.mu.Lock() - defer s.mu.Unlock() - - // Notify all waiters. Tney will fail on the next attempt to execute + // Notify all waiters. They will fail on the next attempt to execute // operations and return error. s.dead = true for _, s := range s.sems { -- cgit v1.2.3 From f93bd2cbe66817c300114630bb52702466e52129 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Thu, 28 Jun 2018 16:10:17 -0700 Subject: Hold t.mu while calling t.FSContext(). PiperOrigin-RevId: 202562686 Change-Id: I0f5be7cc9098e86fa31d016251c127cb91084b05 --- pkg/sentry/fs/proc/mounts.go | 15 ++++++++++++++- pkg/sentry/kernel/fs_context.go | 17 ++++++++++++----- pkg/sentry/kernel/task_resources.go | 4 +++- 3 files changed, 29 insertions(+), 7 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/proc/mounts.go b/pkg/sentry/fs/proc/mounts.go index b9988625e..108432f4e 100644 --- a/pkg/sentry/fs/proc/mounts.go +++ b/pkg/sentry/fs/proc/mounts.go @@ -28,9 +28,22 @@ import ( // forEachMountSource runs f for the process root mount and each mount that is a // descendant of the root. func forEachMountSource(t *kernel.Task, fn func(string, *fs.MountSource)) { + var fsctx *kernel.FSContext + t.WithMuLocked(func(t *kernel.Task) { + fsctx = t.FSContext() + }) + if fsctx == nil { + // The task has been destroyed. Nothing to show here. + return + } + // All mount points must be relative to the rootDir, and mounts outside // will be excluded. - rootDir := t.FSContext().RootDirectory() + rootDir := fsctx.RootDirectory() + if rootDir == nil { + // The task has been destroyed. Nothing to show here. + return + } defer rootDir.DecRef() if rootDir.Inode == nil { diff --git a/pkg/sentry/kernel/fs_context.go b/pkg/sentry/kernel/fs_context.go index d1ca9c09d..dbc097696 100644 --- a/pkg/sentry/kernel/fs_context.go +++ b/pkg/sentry/kernel/fs_context.go @@ -66,6 +66,11 @@ func newFSContext(root, cwd *fs.Dirent, umask uint) *FSContext { // (that return nil). This is because valid references may still be held via // proc files or other mechanisms. func (f *FSContext) destroy() { + // Hold f.mu so that we don't race with RootDirectory() and + // WorkingDirectory(). + f.mu.Lock() + defer f.mu.Unlock() + f.root.DecRef() f.root = nil @@ -94,9 +99,9 @@ func (f *FSContext) Fork() *FSContext { } // WorkingDirectory returns the current working directory. -// You should call DecRef on the returned Dirent when finished. // -// This will return nil if called after destroy(). +// This will return nil if called after destroy(), otherwise it will return a +// Dirent with a reference taken. func (f *FSContext) WorkingDirectory() *fs.Dirent { f.mu.Lock() defer f.mu.Unlock() @@ -129,13 +134,15 @@ func (f *FSContext) SetWorkingDirectory(d *fs.Dirent) { } // RootDirectory returns the current filesystem root. -// You should call DecRef on the returned Dirent when finished. // -// This will return nil if called after destroy(). +// This will return nil if called after destroy(), otherwise it will return a +// Dirent with a reference taken. func (f *FSContext) RootDirectory() *fs.Dirent { f.mu.Lock() defer f.mu.Unlock() - f.root.IncRef() + if f.root != nil { + f.root.IncRef() + } return f.root } diff --git a/pkg/sentry/kernel/task_resources.go b/pkg/sentry/kernel/task_resources.go index 0389989e8..4ca25664a 100644 --- a/pkg/sentry/kernel/task_resources.go +++ b/pkg/sentry/kernel/task_resources.go @@ -123,6 +123,8 @@ func (t *Task) IsChrooted() bool { realRoot := t.k.mounts.Root() defer realRoot.DecRef() root := t.tr.FSContext.RootDirectory() - defer root.DecRef() + if root != nil { + defer root.DecRef() + } return root != realRoot } -- cgit v1.2.3 From 41aeb680b1882c9416e25e100b5ff5eebead36de Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Mon, 9 Jul 2018 16:15:14 -0700 Subject: Inherit parent in clone(CLONE_THREAD) under TaskSet.mu. PiperOrigin-RevId: 203849534 Change-Id: I4d81513bfd32e0b7fc40c8a4c194eba7abc35a83 --- pkg/sentry/kernel/task_clone.go | 8 +++++--- pkg/sentry/kernel/task_start.go | 7 +++++++ 2 files changed, 12 insertions(+), 3 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 0c2427952..a61283267 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -220,18 +220,15 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { pidns = pidns.NewChild(userns) } tg := t.tg - parent := t.parent if opts.NewThreadGroup { sh := t.tg.signalHandlers if opts.NewSignalHandlers { sh = sh.Fork() } tg = NewThreadGroup(pidns, sh, opts.TerminationSignal, tg.limits.GetCopy(), t.k.monotonicClock) - parent = t } cfg := &TaskConfig{ Kernel: t.k, - Parent: parent, ThreadGroup: tg, TaskContext: tc, TaskResources: t.tr.Fork(!opts.NewFiles, !opts.NewFSContext), @@ -242,6 +239,11 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { UTSNamespace: utsns, IPCNamespace: ipcns, } + if opts.NewThreadGroup { + cfg.Parent = t + } else { + cfg.InheritParent = t + } if opts.NewNetworkNamespace { cfg.NetworkNamespaced = true } diff --git a/pkg/sentry/kernel/task_start.go b/pkg/sentry/kernel/task_start.go index 801cb3395..c97dee8fc 100644 --- a/pkg/sentry/kernel/task_start.go +++ b/pkg/sentry/kernel/task_start.go @@ -31,6 +31,10 @@ type TaskConfig struct { // Parent is the new task's parent. Parent may be nil. Parent *Task + // If InheritParent is not nil, use InheritParent's parent as the new + // task's parent. + InheritParent *Task + // ThreadGroup is the ThreadGroup the new task belongs to. *ThreadGroup @@ -133,6 +137,9 @@ func (ts *TaskSet) newTask(cfg *TaskConfig) (*Task, error) { // IDs). t.updateLogPrefixLocked() + if cfg.InheritParent != nil { + t.parent = cfg.InheritParent.parent + } if t.parent != nil { t.parent.children[t] = struct{}{} } -- cgit v1.2.3 From b1683df90bf81974e9e309ed66edaff30537c1be Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Tue, 10 Jul 2018 09:22:37 -0700 Subject: netstack: tcp socket connected state S/R support. PiperOrigin-RevId: 203958972 Change-Id: Ia6fe16547539296d48e2c6731edacdd96bd6e93c --- pkg/sentry/kernel/BUILD | 5 +- pkg/sentry/kernel/kernel.go | 6 + pkg/sentry/kernel/kernel_state.go | 31 +++++ pkg/tcpip/stack/stack_global_state.go | 2 +- pkg/tcpip/tcpip.go | 36 ++++++ pkg/tcpip/transport/tcp/BUILD | 7 ++ pkg/tcpip/transport/tcp/accept.go | 11 +- pkg/tcpip/transport/tcp/connect.go | 43 ++++++- pkg/tcpip/transport/tcp/endpoint.go | 72 ++++++++--- pkg/tcpip/transport/tcp/endpoint_state.go | 193 +++++++++++++++++++++--------- pkg/tcpip/transport/tcp/segment.go | 8 +- pkg/tcpip/transport/tcp/segment_queue.go | 4 +- pkg/tcpip/transport/tcp/segment_state.go | 51 ++++++++ pkg/tcpip/transport/tcp/snd.go | 4 +- pkg/tcpip/transport/tcp/snd_state.go | 49 ++++++++ 15 files changed, 430 insertions(+), 92 deletions(-) create mode 100644 pkg/sentry/kernel/kernel_state.go create mode 100644 pkg/tcpip/transport/tcp/segment_state.go create mode 100644 pkg/tcpip/transport/tcp/snd_state.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index b2a55ddff..07568b47c 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -12,6 +12,7 @@ go_stateify( "fs_context.go", "ipc_namespace.go", "kernel.go", + "kernel_state.go", "pending_signals.go", "pending_signals_state.go", "process_group_list.go", @@ -45,10 +46,11 @@ go_stateify( "vdso.go", "version.go", ], - out = "kernel_state.go", + out = "kernel_autogen_state.go", imports = [ "gvisor.googlesource.com/gvisor/pkg/sentry/arch", "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", + "gvisor.googlesource.com/gvisor/pkg/tcpip", ], package = "kernel", ) @@ -117,6 +119,7 @@ go_library( "fs_context.go", "ipc_namespace.go", "kernel.go", + "kernel_autogen_state.go", "kernel_state.go", "pending_signals.go", "pending_signals_list.go", diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 5662b8f08..64439cd9d 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -57,6 +57,7 @@ import ( sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/state" + "gvisor.googlesource.com/gvisor/pkg/tcpip" ) // Kernel represents an emulated Linux kernel. It must be initialized by calling @@ -158,6 +159,9 @@ type Kernel struct { // exitErr is the error causing the sandbox to exit, if any. It is // protected by extMu. exitErr error + + // danglingEndpoints is used to save / restore tcpip.DanglingEndpoints. + danglingEndpoints struct{} `state:".([]tcpip.Endpoint)"` } // InitKernelArgs holds arguments to Init. @@ -422,6 +426,8 @@ func (k *Kernel) LoadFrom(r io.Reader, p platform.Platform, net inet.Stack) erro return err } + tcpip.AsyncLoading.Wait() + log.Infof("Overall load took [%s]", time.Since(loadStart)) // Applications may size per-cpu structures based on k.applicationCores, so diff --git a/pkg/sentry/kernel/kernel_state.go b/pkg/sentry/kernel/kernel_state.go new file mode 100644 index 000000000..bb2d5102d --- /dev/null +++ b/pkg/sentry/kernel/kernel_state.go @@ -0,0 +1,31 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/tcpip" +) + +// saveDanglingEndpoints is invoked by stateify. +func (k *Kernel) saveDanglingEndpoints() []tcpip.Endpoint { + return tcpip.GetDanglingEndpoints() +} + +// loadDanglingEndpoints is invoked by stateify. +func (k *Kernel) loadDanglingEndpoints(es []tcpip.Endpoint) { + for _, e := range es { + tcpip.AddDanglingEndpoint(e) + } +} diff --git a/pkg/tcpip/stack/stack_global_state.go b/pkg/tcpip/stack/stack_global_state.go index 6d261ce96..b6c095efb 100644 --- a/pkg/tcpip/stack/stack_global_state.go +++ b/pkg/tcpip/stack/stack_global_state.go @@ -15,5 +15,5 @@ package stack // StackFromEnv is the global stack created in restore run. -// FIXME: remove this variable once tcpip S/R is fully supported. +// FIXME var StackFromEnv *Stack diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index 4107c0f78..eb1e4645d 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -33,6 +33,7 @@ import ( "fmt" "strconv" "strings" + "sync" "time" "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" @@ -562,3 +563,38 @@ type ProtocolAddress struct { // Address is a network address. Address Address } + +// danglingEndpointsMu protects access to danglingEndpoints. +var danglingEndpointsMu sync.Mutex + +// danglingEndpoints tracks all dangling endpoints no longer owned by the app. +var danglingEndpoints = make(map[Endpoint]struct{}) + +// GetDanglingEndpoints returns all dangling endpoints. +func GetDanglingEndpoints() []Endpoint { + es := make([]Endpoint, 0, len(danglingEndpoints)) + danglingEndpointsMu.Lock() + for e, _ := range danglingEndpoints { + es = append(es, e) + } + danglingEndpointsMu.Unlock() + return es +} + +// AddDanglingEndpoint adds a dangling endpoint. +func AddDanglingEndpoint(e Endpoint) { + danglingEndpointsMu.Lock() + danglingEndpoints[e] = struct{}{} + danglingEndpointsMu.Unlock() +} + +// DeleteDanglingEndpoint removes a dangling endpoint. +func DeleteDanglingEndpoint(e Endpoint) { + danglingEndpointsMu.Lock() + delete(danglingEndpoints, e) + danglingEndpointsMu.Unlock() +} + +// AsyncLoading is the global barrier for asynchronous endpoint loading +// activities. +var AsyncLoading sync.WaitGroup diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index 6cb0ebab2..6a2f42a12 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -10,11 +10,16 @@ go_stateify( "endpoint.go", "endpoint_state.go", "rcv.go", + "segment.go", "segment_heap.go", + "segment_queue.go", + "segment_state.go", "snd.go", + "snd_state.go", "tcp_segment_list.go", ], out = "tcp_state.go", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], package = "tcp", ) @@ -43,7 +48,9 @@ go_library( "segment.go", "segment_heap.go", "segment_queue.go", + "segment_state.go", "snd.go", + "snd_state.go", "tcp_segment_list.go", "tcp_state.go", "timer.go", diff --git a/pkg/tcpip/transport/tcp/accept.go b/pkg/tcpip/transport/tcp/accept.go index ae4359ff4..d6d2b4555 100644 --- a/pkg/tcpip/transport/tcp/accept.go +++ b/pkg/tcpip/transport/tcp/accept.go @@ -78,7 +78,8 @@ func encodeMSS(mss uint16) uint32 { // to go above a threshold. var synRcvdCount struct { sync.Mutex - value uint64 + value uint64 + pending sync.WaitGroup } // listenContext is used by a listening endpoint to store state used while @@ -112,6 +113,7 @@ func incSynRcvdCount() bool { return false } + synRcvdCount.pending.Add(1) synRcvdCount.value++ return true @@ -125,6 +127,7 @@ func decSynRcvdCount() { defer synRcvdCount.Unlock() synRcvdCount.value-- + synRcvdCount.pending.Done() } // newListenContext creates a new listen context. @@ -302,7 +305,7 @@ func (e *endpoint) handleListenSegment(ctx *listenContext, s *segment) { opts := parseSynSegmentOptions(s) if incSynRcvdCount() { s.incRef() - go e.handleSynSegment(ctx, s, &opts) // S/R-FIXME + go e.handleSynSegment(ctx, s, &opts) // S/R-SAFE: synRcvdCount is the barrier. } else { cookie := ctx.createCookie(s.id, s.sequenceNumber, encodeMSS(opts.MSS)) // Send SYN with window scaling because we currently @@ -391,10 +394,12 @@ func (e *endpoint) protocolListenLoop(rcvWnd seqnum.Size) *tcpip.Error { return nil } if n¬ifyDrain != 0 { - for s := e.segmentQueue.dequeue(); s != nil; s = e.segmentQueue.dequeue() { + for !e.segmentQueue.empty() { + s := e.segmentQueue.dequeue() e.handleListenSegment(ctx, s) s.decRef() } + synRcvdCount.pending.Wait() close(e.drainDone) <-e.undrain } diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go index afdea2b53..33bf4fc0b 100644 --- a/pkg/tcpip/transport/tcp/connect.go +++ b/pkg/tcpip/transport/tcp/connect.go @@ -453,7 +453,8 @@ func (h *handshake) execute() *tcpip.Error { return tcpip.ErrAborted } if n¬ifyDrain != 0 { - for s := h.ep.segmentQueue.dequeue(); s != nil; s = h.ep.segmentQueue.dequeue() { + for !h.ep.segmentQueue.empty() { + s := h.ep.segmentQueue.dequeue() err := h.handleSegment(s) s.decRef() if err != nil { @@ -823,15 +824,13 @@ func (e *endpoint) handleSegments() *tcpip.Error { // protocolMainLoop is the main loop of the TCP protocol. It runs in its own // goroutine and is responsible for sending segments and handling received // segments. -func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { +func (e *endpoint) protocolMainLoop(handshake bool) *tcpip.Error { var closeTimer *time.Timer var closeWaker sleep.Waker defer func() { // e.mu is expected to be hold upon entering this section. - e.completeWorkerLocked() - if e.snd != nil { e.snd.resendTimer.cleanup() } @@ -840,6 +839,8 @@ func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { closeTimer.Stop() } + e.completeWorkerLocked() + if e.drainDone != nil { close(e.drainDone) } @@ -850,7 +851,7 @@ func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { e.waiterQueue.Notify(waiter.EventHUp | waiter.EventErr | waiter.EventIn | waiter.EventOut) }() - if !passive { + if handshake { // This is an active connection, so we must initiate the 3-way // handshake, and then inform potential waiters about its // completion. @@ -960,6 +961,17 @@ func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { closeWaker.Assert() }) } + + if n¬ifyDrain != 0 { + for !e.segmentQueue.empty() { + if err := e.handleSegments(); err != nil { + return err + } + } + close(e.drainDone) + <-e.undrain + } + return nil }, }, @@ -971,6 +983,27 @@ func (e *endpoint) protocolMainLoop(passive bool) *tcpip.Error { s.AddWaker(funcs[i].w, i) } + // The following assertions and notifications are needed for restored + // endpoints. Fresh newly created endpoints have empty states and should + // not invoke any. + e.segmentQueue.mu.Lock() + if !e.segmentQueue.list.Empty() { + e.newSegmentWaker.Assert() + } + e.segmentQueue.mu.Unlock() + + e.rcvListMu.Lock() + if !e.rcvList.Empty() { + e.waiterQueue.Notify(waiter.EventIn) + } + e.rcvListMu.Unlock() + + e.mu.RLock() + if e.workerCleanup { + e.notifyProtocolGoroutine(notifyClose) + } + e.mu.RUnlock() + // Main loop. Handle segments until both send and receive ends of the // connection have completed. for !e.rcv.closed || !e.snd.closed || e.snd.sndUna != e.snd.sndNxtList { diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index cb105b863..8b9a81f6a 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -80,7 +80,7 @@ type endpoint struct { // change throughout the lifetime of the endpoint. stack *stack.Stack `state:"manual"` netProto tcpip.NetworkProtocolNumber - waiterQueue *waiter.Queue + waiterQueue *waiter.Queue `state:"wait"` // lastError represents the last error that the endpoint reported; // access to it is protected by the following mutex. @@ -95,8 +95,8 @@ type endpoint struct { // to indicate to users that no more data is coming. // // rcvListMu can be taken after the endpoint mu below. - rcvListMu sync.Mutex `state:"nosave"` - rcvList segmentList + rcvListMu sync.Mutex `state:"nosave"` + rcvList segmentList `state:"wait"` rcvClosed bool rcvBufSize int rcvBufUsed int @@ -104,8 +104,8 @@ type endpoint struct { // The following fields are protected by the mutex. mu sync.RWMutex `state:"nosave"` id stack.TransportEndpointID - state endpointState - isPortReserved bool `state:"manual"` + state endpointState `state:".(endpointState)"` + isPortReserved bool `state:"manual"` isRegistered bool boundNICID tcpip.NICID `state:"manual"` route stack.Route `state:"manual"` @@ -131,7 +131,7 @@ type endpoint struct { // workerCleanup specifies if the worker goroutine must perform cleanup // before exitting. This can only be set to true when workerRunning is // also true, and they're both protected by the mutex. - workerCleanup bool `state:"zerovalue"` + workerCleanup bool // sendTSOk is used to indicate when the TS Option has been negotiated. // When sendTSOk is true every non-RST segment should carry a TS as per @@ -166,7 +166,7 @@ type endpoint struct { // segmentQueue is used to hand received segments to the protocol // goroutine. Segments are queued as long as the queue is not full, // and dropped when it is. - segmentQueue segmentQueue `state:"zerovalue"` + segmentQueue segmentQueue `state:"wait"` // The following fields are used to manage the send buffer. When // segments are ready to be sent, they are added to sndQueue and the @@ -179,7 +179,7 @@ type endpoint struct { sndBufUsed int sndClosed bool sndBufInQueue seqnum.Size - sndQueue segmentList + sndQueue segmentList `state:"wait"` sndWaker sleep.Waker `state:"manual"` sndCloseWaker sleep.Waker `state:"manual"` @@ -201,17 +201,21 @@ type endpoint struct { // notifyFlags is a bitmask of flags used to indicate to the protocol // goroutine what it was notified; this is only accessed atomically. - notifyFlags uint32 `state:"zerovalue"` + notifyFlags uint32 `state:"nosave"` // acceptedChan is used by a listening endpoint protocol goroutine to // send newly accepted connections to the endpoint so that they can be // read by Accept() calls. - acceptedChan chan *endpoint `state:".(endpointChan)"` + acceptedChan chan *endpoint `state:"manual"` + + // acceptedEndpoints is only used to save / restore the channel buffer. + // FIXME + acceptedEndpoints []*endpoint // The following are only used from the protocol goroutine, and // therefore don't need locks to protect them. - rcv *receiver - snd *sender + rcv *receiver `state:"wait"` + snd *sender `state:"wait"` // The goroutine drain completion notification channel. drainDone chan struct{} `state:"nosave"` @@ -224,6 +228,7 @@ type endpoint struct { probe stack.TCPProbeFunc `state:"nosave"` // The following are only used to assist the restore run to re-connect. + bindAddress tcpip.Address connectingAddress tcpip.Address } @@ -357,6 +362,7 @@ func (e *endpoint) Close() { // Either perform the local cleanup or kick the worker to make sure it // knows it needs to cleanup. + tcpip.AddDanglingEndpoint(e) if !e.workerRunning { e.cleanupLocked() } else { @@ -376,9 +382,12 @@ func (e *endpoint) cleanupLocked() { if e.acceptedChan != nil { close(e.acceptedChan) for n := range e.acceptedChan { + n.mu.Lock() n.resetConnectionLocked(tcpip.ErrConnectionAborted) + n.mu.Unlock() n.Close() } + e.acceptedChan = nil } e.workerCleanup = false @@ -387,6 +396,7 @@ func (e *endpoint) cleanupLocked() { } e.route.Release() + tcpip.DeleteDanglingEndpoint(e) } // Read reads data from the endpoint. @@ -801,6 +811,16 @@ func (e *endpoint) checkV4Mapped(addr *tcpip.FullAddress) (tcpip.NetworkProtocol // Connect connects the endpoint to its peer. func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { + return e.connect(addr, true, true) +} + +// connect connects the endpoint to its peer. In the normal non-S/R case, the +// new connection is expected to run the main goroutine and perform handshake. +// In restore of previously connected endpoints, both ends will be passively +// created (so no new handshaking is done); for stack-accepted connections not +// yet accepted by the app, they are restored without running the main goroutine +// here. +func (e *endpoint) connect(addr tcpip.FullAddress, handshake bool, run bool) *tcpip.Error { e.mu.Lock() defer e.mu.Unlock() @@ -912,9 +932,27 @@ func (e *endpoint) Connect(addr tcpip.FullAddress) *tcpip.Error { e.boundNICID = nicid e.effectiveNetProtos = netProtos e.connectingAddress = connectingAddr - e.workerRunning = true - go e.protocolMainLoop(false) // S/R-SAFE: will be drained before save. + // Connect in the restore phase does not perform handshake. Restore its + // connection setting here. + if !handshake { + e.segmentQueue.mu.Lock() + for _, l := range []segmentList{e.segmentQueue.list, e.sndQueue, e.snd.writeList} { + for s := l.Front(); s != nil; s = s.Next() { + s.id = e.id + s.route = r.Clone() + e.sndWaker.Assert() + } + } + e.segmentQueue.mu.Unlock() + e.snd.updateMaxPayloadSize(int(e.route.MTU()), 0) + e.state = stateConnected + } + + if run { + e.workerRunning = true + go e.protocolMainLoop(handshake) // S/R-SAFE: will be drained before save. + } return tcpip.ErrConnectStarted } @@ -999,6 +1037,9 @@ func (e *endpoint) Listen(backlog int) *tcpip.Error { if len(e.acceptedChan) > backlog { return tcpip.ErrInvalidEndpointState } + if cap(e.acceptedChan) == backlog { + return nil + } origChan := e.acceptedChan e.acceptedChan = make(chan *endpoint, backlog) close(origChan) @@ -1036,7 +1077,7 @@ func (e *endpoint) Listen(backlog int) *tcpip.Error { func (e *endpoint) startAcceptedLoop(waiterQueue *waiter.Queue) { e.waiterQueue = waiterQueue e.workerRunning = true - go e.protocolMainLoop(true) // S/R-FIXME + go e.protocolMainLoop(false) // S/R-SAFE: drained on save. } // Accept returns a new endpoint if a peer has established a connection @@ -1077,6 +1118,7 @@ func (e *endpoint) Bind(addr tcpip.FullAddress, commit func() *tcpip.Error) (ret return tcpip.ErrAlreadyBound } + e.bindAddress = addr.Addr netProto, err := e.checkV4Mapped(&addr) if err != nil { return err diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go index aa4ccea75..43765d425 100644 --- a/pkg/tcpip/transport/tcp/endpoint_state.go +++ b/pkg/tcpip/transport/tcp/endpoint_state.go @@ -17,8 +17,10 @@ package tcp import ( "fmt" "sync" + "time" "gvisor.googlesource.com/gvisor/pkg/tcpip" + "gvisor.googlesource.com/gvisor/pkg/tcpip/header" "gvisor.googlesource.com/gvisor/pkg/tcpip/stack" ) @@ -32,7 +34,7 @@ func (e *endpoint) drainSegmentLocked() { e.undrain = make(chan struct{}) e.mu.Unlock() - e.notificationWaker.Assert() + e.notifyProtocolGoroutine(notifyDrain) <-e.drainDone e.mu.Lock() @@ -48,37 +50,103 @@ func (e *endpoint) beforeSave() { switch e.state { case stateInitial, stateBound: - case stateListen: - if !e.segmentQueue.empty() { - e.drainSegmentLocked() + case stateListen, stateConnecting, stateConnected: + if e.state == stateConnected && !e.workerRunning { + // The endpoint must be in acceptedChan. + break } - case stateConnecting: e.drainSegmentLocked() - if e.state != stateConnected { + if e.state != stateClosed && e.state != stateError { + if !e.workerRunning { + panic("endpoint has no worker running in listen, connecting, or connected state") + } break } fallthrough - case stateConnected: - // FIXME - panic(tcpip.ErrSaveRejection{fmt.Errorf("endpoint cannot be saved in connected state: local %v:%v, remote %v:%v", e.id.LocalAddress, e.id.LocalPort, e.id.RemoteAddress, e.id.RemotePort)}) - case stateClosed, stateError: + case stateError, stateClosed: + for e.state == stateError && e.workerRunning { + e.mu.Unlock() + time.Sleep(100 * time.Millisecond) + e.mu.Lock() + } if e.workerRunning { - panic(fmt.Sprintf("endpoint still has worker running in closed or error state")) + panic("endpoint still has worker running in closed or error state") } default: panic(fmt.Sprintf("endpoint in unknown state %v", e.state)) } + + if e.waiterQueue != nil && !e.waiterQueue.IsEmpty() { + panic("endpoint still has waiters upon save") + } + + if !((e.state == stateBound || e.state == stateListen) == e.isPortReserved) { + panic("endpoint port must and must only be reserved in bound or listen state") + } + + if e.acceptedChan != nil { + close(e.acceptedChan) + e.acceptedEndpoints = make([]*endpoint, len(e.acceptedChan), cap(e.acceptedChan)) + i := 0 + for ep := range e.acceptedChan { + e.acceptedEndpoints[i] = ep + i++ + } + if i != len(e.acceptedEndpoints) { + panic("endpoint acceptedChan buffer got consumed by background context") + } + } +} + +// saveState is invoked by stateify. +func (e *endpoint) saveState() endpointState { + return e.state +} + +// Endpoint loading must be done in the following ordering by their state, to +// avoid dangling connecting w/o listening peer, and to avoid conflicts in port +// reservation. +var connectedLoading sync.WaitGroup +var listenLoading sync.WaitGroup +var connectingLoading sync.WaitGroup + +// Bound endpoint loading happens last. + +// loadState is invoked by stateify. +func (e *endpoint) loadState(state endpointState) { + // This is to ensure that the loading wait groups include all applicable + // endpoints before any asynchronous calls to the Wait() methods. + switch state { + case stateConnected: + connectedLoading.Add(1) + case stateListen: + listenLoading.Add(1) + case stateConnecting: + connectingLoading.Add(1) + } + e.state = state } // afterLoad is invoked by stateify. func (e *endpoint) afterLoad() { + // We load acceptedChan buffer indirectly here. Note that closed + // endpoints might not need to allocate the channel. + // FIXME + if cap(e.acceptedEndpoints) > 0 { + e.acceptedChan = make(chan *endpoint, cap(e.acceptedEndpoints)) + for _, ep := range e.acceptedEndpoints { + e.acceptedChan <- ep + } + e.acceptedEndpoints = nil + } + e.stack = stack.StackFromEnv e.segmentQueue.setLimit(2 * e.rcvBufSize) e.workMu.Init() state := e.state switch state { - case stateInitial, stateBound, stateListen, stateConnecting: + case stateInitial, stateBound, stateListen, stateConnecting, stateConnected: var ss SendBufferSizeOption if err := e.stack.TransportProtocolOption(ProtocolNumber, &ss); err == nil { if e.sndBufSize < ss.Min || e.sndBufSize > ss.Max { @@ -90,65 +158,72 @@ func (e *endpoint) afterLoad() { } } - switch state { - case stateBound, stateListen, stateConnecting: + bind := func() { e.state = stateInitial - if err := e.Bind(tcpip.FullAddress{Addr: e.id.LocalAddress, Port: e.id.LocalPort}, nil); err != nil { + if len(e.bindAddress) == 0 { + e.bindAddress = e.id.LocalAddress + } + if err := e.Bind(tcpip.FullAddress{Addr: e.bindAddress, Port: e.id.LocalPort}, nil); err != nil { panic("endpoint binding failed: " + err.String()) } } switch state { - case stateListen: - backlog := cap(e.acceptedChan) - e.acceptedChan = nil - if err := e.Listen(backlog); err != nil { - panic("endpoint listening failed: " + err.String()) + case stateConnected: + bind() + if len(e.connectingAddress) == 0 { + // This endpoint is accepted by netstack but not yet by + // the app. If the endpoint is IPv6 but the remote + // address is IPv4, we need to connect as IPv6 so that + // dual-stack mode can be properly activated. + if e.netProto == header.IPv6ProtocolNumber && len(e.id.RemoteAddress) != header.IPv6AddressSize { + e.connectingAddress = "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xff\xff" + e.id.RemoteAddress + } else { + e.connectingAddress = e.id.RemoteAddress + } } - } - - switch state { - case stateConnecting: - if err := e.Connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}); err != tcpip.ErrConnectStarted { + if err := e.connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}, false, e.workerRunning); err != tcpip.ErrConnectStarted { panic("endpoint connecting failed: " + err.String()) } + connectedLoading.Done() + case stateListen: + tcpip.AsyncLoading.Add(1) + go func() { + connectedLoading.Wait() + bind() + backlog := cap(e.acceptedChan) + if err := e.Listen(backlog); err != nil { + panic("endpoint listening failed: " + err.String()) + } + listenLoading.Done() + tcpip.AsyncLoading.Done() + }() + case stateConnecting: + tcpip.AsyncLoading.Add(1) + go func() { + connectedLoading.Wait() + listenLoading.Wait() + bind() + if err := e.Connect(tcpip.FullAddress{NIC: e.boundNICID, Addr: e.connectingAddress, Port: e.id.RemotePort}); err != tcpip.ErrConnectStarted { + panic("endpoint connecting failed: " + err.String()) + } + connectingLoading.Done() + tcpip.AsyncLoading.Done() + }() + case stateBound: + tcpip.AsyncLoading.Add(1) + go func() { + connectedLoading.Wait() + listenLoading.Wait() + connectingLoading.Wait() + bind() + tcpip.AsyncLoading.Done() + }() + case stateClosed, stateError: + tcpip.DeleteDanglingEndpoint(e) } } -// saveAcceptedChan is invoked by stateify. -func (e *endpoint) saveAcceptedChan() endpointChan { - if e.acceptedChan == nil { - return endpointChan{} - } - close(e.acceptedChan) - buffer := make([]*endpoint, 0, len(e.acceptedChan)) - for ep := range e.acceptedChan { - buffer = append(buffer, ep) - } - if len(buffer) != cap(buffer) { - panic("endpoint.acceptedChan buffer got consumed by background context") - } - c := cap(e.acceptedChan) - e.acceptedChan = nil - return endpointChan{buffer: buffer, cap: c} -} - -// loadAcceptedChan is invoked by stateify. -func (e *endpoint) loadAcceptedChan(c endpointChan) { - if c.cap == 0 { - return - } - e.acceptedChan = make(chan *endpoint, c.cap) - for _, ep := range c.buffer { - e.acceptedChan <- ep - } -} - -type endpointChan struct { - buffer []*endpoint - cap int -} - // saveLastError is invoked by stateify. func (e *endpoint) saveLastError() string { if e.lastError == nil { diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index a90f6661d..40928ba2c 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -39,9 +39,9 @@ const ( type segment struct { segmentEntry refCnt int32 - id stack.TransportEndpointID - route stack.Route `state:"manual"` - data buffer.VectorisedView + id stack.TransportEndpointID `state:"manual"` + route stack.Route `state:"manual"` + data buffer.VectorisedView `state:".(buffer.VectorisedView)"` // views is used as buffer for data when its length is large // enough to store a VectorisedView. views [8]buffer.View @@ -55,7 +55,7 @@ type segment struct { // parsedOptions stores the parsed values from the options in the segment. parsedOptions header.TCPOptions - options []byte + options []byte `state:".([]byte)"` } func newSegment(r *stack.Route, id stack.TransportEndpointID, vv *buffer.VectorisedView) *segment { diff --git a/pkg/tcpip/transport/tcp/segment_queue.go b/pkg/tcpip/transport/tcp/segment_queue.go index 83f554ebd..2ddcf5f10 100644 --- a/pkg/tcpip/transport/tcp/segment_queue.go +++ b/pkg/tcpip/transport/tcp/segment_queue.go @@ -22,8 +22,8 @@ import ( // segmentQueue is a bounded, thread-safe queue of TCP segments. type segmentQueue struct { - mu sync.Mutex - list segmentList + mu sync.Mutex `state:"nosave"` + list segmentList `state:"wait"` limit int used int } diff --git a/pkg/tcpip/transport/tcp/segment_state.go b/pkg/tcpip/transport/tcp/segment_state.go new file mode 100644 index 000000000..22f0bbf18 --- /dev/null +++ b/pkg/tcpip/transport/tcp/segment_state.go @@ -0,0 +1,51 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package tcp + +import ( + "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" +) + +// saveData is invoked by stateify. +func (s *segment) saveData() buffer.VectorisedView { + // We cannot save s.data directly as s.data.views may alias to s.views, + // which is not allowed by state framework (in-struct pointer). + return s.data.Clone(nil) +} + +// loadData is invoked by stateify. +func (s *segment) loadData(data buffer.VectorisedView) { + // NOTE: We cannot do the s.data = data.Clone(s.views[:]) optimization + // here because data.views is not guaranteed to be loaded by now. Plus, + // data.views will be allocated anyway so there really is little point + // of utilizing s.views for data.views. + s.data = data +} + +// saveOptions is invoked by stateify. +func (s *segment) saveOptions() []byte { + // We cannot save s.options directly as it may point to s.data's trimmed + // tail, which is not allowed by state framework (in-struct pointer). + b := make([]byte, 0, cap(s.options)) + return append(b, s.options...) +} + +// loadOptions is invoked by stateify. +func (s *segment) loadOptions(options []byte) { + // NOTE: We cannot point s.options back into s.data's trimmed tail. But + // it is OK as they do not need to aliased. Plus, options is already + // allocated so there is no cost here. + s.options = options +} diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index a9892eb64..7dfbf6384 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -38,7 +38,7 @@ type sender struct { ep *endpoint // lastSendTime is the timestamp when the last packet was sent. - lastSendTime time.Time + lastSendTime time.Time `state:".(unixTime)"` // dupAckCount is the number of duplicated acks received. It is used for // fast retransmit. @@ -81,7 +81,7 @@ type sender struct { rttMeasureSeqNum seqnum.Value // rttMeasureTime is the time when the rttMeasureSeqNum was sent. - rttMeasureTime time.Time + rttMeasureTime time.Time `state:".(unixTime)"` closed bool writeNext *segment diff --git a/pkg/tcpip/transport/tcp/snd_state.go b/pkg/tcpip/transport/tcp/snd_state.go new file mode 100644 index 000000000..33c8867f4 --- /dev/null +++ b/pkg/tcpip/transport/tcp/snd_state.go @@ -0,0 +1,49 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package tcp + +import ( + "time" +) + +type unixTime struct { + second int64 + nano int64 +} + +// saveLastSendTime is invoked by stateify. +func (s *sender) saveLastSendTime() unixTime { + return unixTime{s.lastSendTime.Unix(), s.lastSendTime.UnixNano()} +} + +// loadLastSendTime is invoked by stateify. +func (s *sender) loadLastSendTime(unix unixTime) { + s.lastSendTime = time.Unix(unix.second, unix.nano) +} + +// saveRttMeasureTime is invoked by stateify. +func (s *sender) saveRttMeasureTime() unixTime { + return unixTime{s.rttMeasureTime.Unix(), s.rttMeasureTime.UnixNano()} +} + +// loadRttMeasureTime is invoked by stateify. +func (s *sender) loadRttMeasureTime(unix unixTime) { + s.rttMeasureTime = time.Unix(unix.second, unix.nano) +} + +// afterLoad is invoked by stateify. +func (s *sender) afterLoad() { + s.resendTimer.init(&s.resendWaker) +} -- cgit v1.2.3 From b9c469f37282129031a6036cfe43028faaeb1a96 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Wed, 11 Jul 2018 14:23:17 -0700 Subject: Move ptrace constants to abi/linux. PiperOrigin-RevId: 204188763 Change-Id: I5596ab7abb3ec9e210a7f57b3fc420e836fa43f3 --- pkg/abi/linux/BUILD | 1 + pkg/abi/linux/ptrace.go | 89 ++++++++++++++++++++++++++++++++++++ pkg/sentry/kernel/ptrace.go | 108 ++++++++++++++++++++------------------------ pkg/sentry/strace/ptrace.go | 80 ++++++++++++++++---------------- 4 files changed, 179 insertions(+), 99 deletions(-) create mode 100644 pkg/abi/linux/ptrace.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index 5d00b66cc..e164945cf 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -48,6 +48,7 @@ go_library( "netlink_route.go", "poll.go", "prctl.go", + "ptrace.go", "rusage.go", "sched.go", "seccomp.go", diff --git a/pkg/abi/linux/ptrace.go b/pkg/abi/linux/ptrace.go new file mode 100644 index 000000000..ba48d4d6d --- /dev/null +++ b/pkg/abi/linux/ptrace.go @@ -0,0 +1,89 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package linux + +// ptrace commands from include/uapi/linux/ptrace.h. +const ( + PTRACE_TRACEME = 0 + PTRACE_PEEKTEXT = 1 + PTRACE_PEEKDATA = 2 + PTRACE_PEEKUSR = 3 + PTRACE_POKETEXT = 4 + PTRACE_POKEDATA = 5 + PTRACE_POKEUSR = 6 + PTRACE_CONT = 7 + PTRACE_KILL = 8 + PTRACE_SINGLESTEP = 9 + PTRACE_ATTACH = 16 + PTRACE_DETACH = 17 + PTRACE_SYSCALL = 24 + PTRACE_SETOPTIONS = 0x4200 + PTRACE_GETEVENTMSG = 0x4201 + PTRACE_GETSIGINFO = 0x4202 + PTRACE_SETSIGINFO = 0x4203 + PTRACE_GETREGSET = 0x4204 + PTRACE_SETREGSET = 0x4205 + PTRACE_SEIZE = 0x4206 + PTRACE_INTERRUPT = 0x4207 + PTRACE_LISTEN = 0x4208 + PTRACE_PEEKSIGINFO = 0x4209 + PTRACE_GETSIGMASK = 0x420a + PTRACE_SETSIGMASK = 0x420b + PTRACE_SECCOMP_GET_FILTER = 0x420c + PTRACE_SECCOMP_GET_METADATA = 0x420d +) + +// ptrace commands from arch/x86/include/uapi/asm/ptrace-abi.h. +const ( + PTRACE_GETREGS = 12 + PTRACE_SETREGS = 13 + PTRACE_GETFPREGS = 14 + PTRACE_SETFPREGS = 15 + PTRACE_GETFPXREGS = 18 + PTRACE_SETFPXREGS = 19 + PTRACE_OLDSETOPTIONS = 21 + PTRACE_GET_THREAD_AREA = 25 + PTRACE_SET_THREAD_AREA = 26 + PTRACE_ARCH_PRCTL = 30 + PTRACE_SYSEMU = 31 + PTRACE_SYSEMU_SINGLESTEP = 32 + PTRACE_SINGLEBLOCK = 33 +) + +// ptrace event codes from include/uapi/linux/ptrace.h. +const ( + PTRACE_EVENT_FORK = 1 + PTRACE_EVENT_VFORK = 2 + PTRACE_EVENT_CLONE = 3 + PTRACE_EVENT_EXEC = 4 + PTRACE_EVENT_VFORK_DONE = 5 + PTRACE_EVENT_EXIT = 6 + PTRACE_EVENT_SECCOMP = 7 + PTRACE_EVENT_STOP = 128 +) + +// PTRACE_SETOPTIONS options from include/uapi/linux/ptrace.h. +const ( + PTRACE_O_TRACESYSGOOD = 1 + PTRACE_O_TRACEFORK = 1 << PTRACE_EVENT_FORK + PTRACE_O_TRACEVFORK = 1 << PTRACE_EVENT_VFORK + PTRACE_O_TRACECLONE = 1 << PTRACE_EVENT_CLONE + PTRACE_O_TRACEEXEC = 1 << PTRACE_EVENT_EXEC + PTRACE_O_TRACEVFORKDONE = 1 << PTRACE_EVENT_VFORK_DONE + PTRACE_O_TRACEEXIT = 1 << PTRACE_EVENT_EXIT + PTRACE_O_TRACESECCOMP = 1 << PTRACE_EVENT_SECCOMP + PTRACE_O_EXITKILL = 1 << 20 + PTRACE_O_SUSPEND_SECCOMP = 1 << 21 +) diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index 20b1c4cd4..f1c2c4bf0 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -16,7 +16,6 @@ package kernel import ( "fmt" - "syscall" "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" @@ -24,19 +23,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/syserror" ) -// ptrace constants from Linux's include/uapi/linux/ptrace.h. -const ( - _PTRACE_EVENT_SECCOMP = 7 - PTRACE_SEIZE = 0x4206 - PTRACE_INTERRUPT = 0x4207 - PTRACE_LISTEN = 0x4208 - PTRACE_PEEKSIGINFO = 0x4209 - PTRACE_GETSIGMASK = 0x420a - PTRACE_SETSIGMASK = 0x420b - _PTRACE_O_EXITKILL = 1 << 20 - _PTRACE_O_TRACESECCOMP = 1 << _PTRACE_EVENT_SECCOMP -) - // ptraceOptions are the subset of options controlling a task's ptrace behavior // that are set by ptrace(PTRACE_SETOPTIONS). type ptraceOptions struct { @@ -505,7 +491,7 @@ func (t *Task) ptraceSeccomp(data uint16) bool { return false } t.Debugf("Entering PTRACE_EVENT_SECCOMP stop") - t.ptraceEventLocked(_PTRACE_EVENT_SECCOMP, uint64(data)) + t.ptraceEventLocked(linux.PTRACE_EVENT_SECCOMP, uint64(data)) return true } @@ -587,19 +573,19 @@ func (t *Task) ptraceClone(kind ptraceCloneKind, child *Task, opts *CloneOptions case ptraceCloneKindClone: if t.ptraceOpts.TraceClone { t.Debugf("Entering PTRACE_EVENT_CLONE stop") - t.ptraceEventLocked(syscall.PTRACE_EVENT_CLONE, uint64(t.tg.pidns.tids[child])) + t.ptraceEventLocked(linux.PTRACE_EVENT_CLONE, uint64(t.tg.pidns.tids[child])) event = true } case ptraceCloneKindFork: if t.ptraceOpts.TraceFork { t.Debugf("Entering PTRACE_EVENT_FORK stop") - t.ptraceEventLocked(syscall.PTRACE_EVENT_FORK, uint64(t.tg.pidns.tids[child])) + t.ptraceEventLocked(linux.PTRACE_EVENT_FORK, uint64(t.tg.pidns.tids[child])) event = true } case ptraceCloneKindVfork: if t.ptraceOpts.TraceVfork { t.Debugf("Entering PTRACE_EVENT_VFORK stop") - t.ptraceEventLocked(syscall.PTRACE_EVENT_VFORK, uint64(t.tg.pidns.tids[child])) + t.ptraceEventLocked(linux.PTRACE_EVENT_VFORK, uint64(t.tg.pidns.tids[child])) event = true } default: @@ -657,7 +643,7 @@ func (t *Task) ptraceVforkDone(child ThreadID) bool { return false } t.Debugf("Entering PTRACE_EVENT_VFORK_DONE stop") - t.ptraceEventLocked(syscall.PTRACE_EVENT_VFORK_DONE, uint64(child)) + t.ptraceEventLocked(linux.PTRACE_EVENT_VFORK_DONE, uint64(child)) return true } @@ -680,7 +666,7 @@ func (t *Task) ptraceExec(oldTID ThreadID) { } if t.ptraceOpts.TraceExec { t.Debugf("Entering PTRACE_EVENT_EXEC stop") - t.ptraceEventLocked(syscall.PTRACE_EVENT_EXEC, uint64(oldTID)) + t.ptraceEventLocked(linux.PTRACE_EVENT_EXEC, uint64(oldTID)) return } // "If the PTRACE_O_TRACEEXEC option is not in effect for the execing @@ -714,7 +700,7 @@ func (t *Task) ptraceExit() { status := t.exitStatus.Status() t.tg.signalHandlers.mu.Unlock() t.Debugf("Entering PTRACE_EVENT_EXIT stop") - t.ptraceEventLocked(syscall.PTRACE_EVENT_EXIT, uint64(status)) + t.ptraceEventLocked(linux.PTRACE_EVENT_EXIT, uint64(status)) } // Preconditions: The TaskSet mutex must be locked. @@ -762,7 +748,7 @@ func (t *Task) ptraceKill(target *Task) error { // Ptrace implements the ptrace system call. func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { // PTRACE_TRACEME ignores all other arguments. - if req == syscall.PTRACE_TRACEME { + if req == linux.PTRACE_TRACEME { return t.ptraceTraceme() } // All other ptrace requests operate on a current or future tracee @@ -774,12 +760,12 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { // PTRACE_ATTACH (and PTRACE_SEIZE, which is unimplemented) do not require // that target is not already a tracee. - if req == syscall.PTRACE_ATTACH { + if req == linux.PTRACE_ATTACH { return t.ptraceAttach(target) } // PTRACE_KILL (and PTRACE_INTERRUPT, which is unimplemented) require that // the target is a tracee, but does not require that it is ptrace-stopped. - if req == syscall.PTRACE_KILL { + if req == linux.PTRACE_KILL { return t.ptraceKill(target) } // All other ptrace requests require that the target is a ptrace-stopped @@ -812,37 +798,37 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { // Resuming commands end the ptrace stop, but only if successful. switch req { - case syscall.PTRACE_DETACH: + case linux.PTRACE_DETACH: if err := t.ptraceDetach(target, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil - case syscall.PTRACE_CONT: + case linux.PTRACE_CONT: if err := target.ptraceUnstop(ptraceSyscallNone, false, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil - case syscall.PTRACE_SYSCALL: + case linux.PTRACE_SYSCALL: if err := target.ptraceUnstop(ptraceSyscallIntercept, false, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil - case syscall.PTRACE_SINGLESTEP: + case linux.PTRACE_SINGLESTEP: if err := target.ptraceUnstop(ptraceSyscallNone, true, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil - case syscall.PTRACE_SYSEMU: + case linux.PTRACE_SYSEMU: if err := target.ptraceUnstop(ptraceSyscallEmu, false, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil - case syscall.PTRACE_SYSEMU_SINGLESTEP: + case linux.PTRACE_SYSEMU_SINGLESTEP: if err := target.ptraceUnstop(ptraceSyscallEmu, true, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err @@ -853,7 +839,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { defer target.ptraceUnfreeze() switch req { - case syscall.PTRACE_PEEKTEXT, syscall.PTRACE_PEEKDATA: + case linux.PTRACE_PEEKTEXT, linux.PTRACE_PEEKDATA: // "At the system call level, the PTRACE_PEEKTEXT, PTRACE_PEEKDATA, and // PTRACE_PEEKUSER requests have a different API: they store the result // at the address specified by the data parameter, and the return value @@ -867,13 +853,13 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { _, err := t.CopyOut(data, word) return err - case syscall.PTRACE_POKETEXT, syscall.PTRACE_POKEDATA: + case linux.PTRACE_POKETEXT, linux.PTRACE_POKEDATA: _, err := usermem.CopyObjectOut(t, target.MemoryManager(), addr, t.Arch().Native(uintptr(data)), usermem.IOOpts{ IgnorePermissions: true, }) return err - case syscall.PTRACE_PEEKUSR: // aka PTRACE_PEEKUSER + case linux.PTRACE_PEEKUSR: // aka PTRACE_PEEKUSER n, err := target.Arch().PtracePeekUser(uintptr(addr)) if err != nil { return err @@ -881,10 +867,10 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { _, err = t.CopyOut(data, n) return err - case syscall.PTRACE_POKEUSR: // aka PTRACE_POKEUSER + case linux.PTRACE_POKEUSR: // aka PTRACE_POKEUSER return target.Arch().PtracePokeUser(uintptr(addr), uintptr(data)) - case syscall.PTRACE_GETREGS: + case linux.PTRACE_GETREGS: // "Copy the tracee's general-purpose ... registers ... to the address // data in the tracer. ... (addr is ignored.) Note that SPARC systems // have the meaning of data and addr reversed ..." @@ -898,7 +884,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { }) return err - case syscall.PTRACE_GETFPREGS: + case linux.PTRACE_GETFPREGS: _, err := target.Arch().PtraceGetFPRegs(&usermem.IOReadWriter{ Ctx: t, IO: t.MemoryManager(), @@ -909,7 +895,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { }) return err - case syscall.PTRACE_GETREGSET: + case linux.PTRACE_GETREGSET: // "Read the tracee's registers. addr specifies, in an // architecture-dependent way, the type of registers to be read. ... // data points to a struct iovec, which describes the destination @@ -934,7 +920,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { ar.End -= usermem.Addr(n) return t.CopyOutIovecs(data, usermem.AddrRangeSeqOf(ar)) - case syscall.PTRACE_SETREGS: + case linux.PTRACE_SETREGS: _, err := target.Arch().PtraceSetRegs(&usermem.IOReadWriter{ Ctx: t, IO: t.MemoryManager(), @@ -945,7 +931,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { }) return err - case syscall.PTRACE_SETFPREGS: + case linux.PTRACE_SETFPREGS: _, err := target.Arch().PtraceSetFPRegs(&usermem.IOReadWriter{ Ctx: t, IO: t.MemoryManager(), @@ -956,7 +942,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { }) return err - case syscall.PTRACE_SETREGSET: + case linux.PTRACE_SETREGSET: ars, err := t.CopyInIovecs(data, 1) if err != nil { return err @@ -976,7 +962,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { ar.End -= usermem.Addr(n) return t.CopyOutIovecs(data, usermem.AddrRangeSeqOf(ar)) - case syscall.PTRACE_GETSIGINFO: + case linux.PTRACE_GETSIGINFO: t.tg.pidns.owner.mu.RLock() defer t.tg.pidns.owner.mu.RUnlock() if target.ptraceSiginfo == nil { @@ -985,7 +971,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { _, err := t.CopyOut(data, target.ptraceSiginfo) return err - case syscall.PTRACE_SETSIGINFO: + case linux.PTRACE_SETSIGINFO: var info arch.SignalInfo if _, err := t.CopyIn(data, &info); err != nil { return err @@ -998,7 +984,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { target.ptraceSiginfo = &info return nil - case PTRACE_GETSIGMASK: + case linux.PTRACE_GETSIGMASK: if addr != linux.SignalSetSize { return syserror.EINVAL } @@ -1007,7 +993,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { _, err := t.CopyOut(data, target.tr.SignalMask) return err - case PTRACE_SETSIGMASK: + case linux.PTRACE_SETSIGMASK: if addr != linux.SignalSetSize { return syserror.EINVAL } @@ -1019,29 +1005,35 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { target.SetSignalMask(mask &^ UnblockableSignals) return nil - case syscall.PTRACE_SETOPTIONS: + case linux.PTRACE_SETOPTIONS: t.tg.pidns.owner.mu.Lock() defer t.tg.pidns.owner.mu.Unlock() - validOpts := uintptr(_PTRACE_O_EXITKILL | syscall.PTRACE_O_TRACESYSGOOD | syscall.PTRACE_O_TRACECLONE | - syscall.PTRACE_O_TRACEEXEC | syscall.PTRACE_O_TRACEEXIT | syscall.PTRACE_O_TRACEFORK | - _PTRACE_O_TRACESECCOMP | syscall.PTRACE_O_TRACEVFORK | syscall.PTRACE_O_TRACEVFORKDONE) + validOpts := uintptr(linux.PTRACE_O_EXITKILL | + linux.PTRACE_O_TRACESYSGOOD | + linux.PTRACE_O_TRACECLONE | + linux.PTRACE_O_TRACEEXEC | + linux.PTRACE_O_TRACEEXIT | + linux.PTRACE_O_TRACEFORK | + linux.PTRACE_O_TRACESECCOMP | + linux.PTRACE_O_TRACEVFORK | + linux.PTRACE_O_TRACEVFORKDONE) if uintptr(data)&^validOpts != 0 { return syserror.EINVAL } target.ptraceOpts = ptraceOptions{ - ExitKill: data&_PTRACE_O_EXITKILL != 0, - SysGood: data&syscall.PTRACE_O_TRACESYSGOOD != 0, - TraceClone: data&syscall.PTRACE_O_TRACECLONE != 0, - TraceExec: data&syscall.PTRACE_O_TRACEEXEC != 0, - TraceExit: data&syscall.PTRACE_O_TRACEEXIT != 0, - TraceFork: data&syscall.PTRACE_O_TRACEFORK != 0, - TraceSeccomp: data&_PTRACE_O_TRACESECCOMP != 0, - TraceVfork: data&syscall.PTRACE_O_TRACEVFORK != 0, - TraceVforkDone: data&syscall.PTRACE_O_TRACEVFORKDONE != 0, + ExitKill: data&linux.PTRACE_O_EXITKILL != 0, + SysGood: data&linux.PTRACE_O_TRACESYSGOOD != 0, + TraceClone: data&linux.PTRACE_O_TRACECLONE != 0, + TraceExec: data&linux.PTRACE_O_TRACEEXEC != 0, + TraceExit: data&linux.PTRACE_O_TRACEEXIT != 0, + TraceFork: data&linux.PTRACE_O_TRACEFORK != 0, + TraceSeccomp: data&linux.PTRACE_O_TRACESECCOMP != 0, + TraceVfork: data&linux.PTRACE_O_TRACEVFORK != 0, + TraceVforkDone: data&linux.PTRACE_O_TRACEVFORKDONE != 0, } return nil - case syscall.PTRACE_GETEVENTMSG: + case linux.PTRACE_GETEVENTMSG: t.tg.pidns.owner.mu.RLock() defer t.tg.pidns.owner.mu.RUnlock() _, err := t.CopyOut(usermem.Addr(data), target.ptraceEventMsg) diff --git a/pkg/sentry/strace/ptrace.go b/pkg/sentry/strace/ptrace.go index a0dabb27a..fcdb7e9f4 100644 --- a/pkg/sentry/strace/ptrace.go +++ b/pkg/sentry/strace/ptrace.go @@ -15,164 +15,162 @@ package strace import ( - "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" ) // PtraceRequestSet are the possible ptrace(2) requests. var PtraceRequestSet = abi.ValueSet{ { - Value: syscall.PTRACE_TRACEME, + Value: linux.PTRACE_TRACEME, Name: "PTRACE_TRACEME", }, { - Value: syscall.PTRACE_PEEKTEXT, + Value: linux.PTRACE_PEEKTEXT, Name: "PTRACE_PEEKTEXT", }, { - Value: syscall.PTRACE_PEEKDATA, + Value: linux.PTRACE_PEEKDATA, Name: "PTRACE_PEEKDATA", }, { - Value: syscall.PTRACE_PEEKUSR, + Value: linux.PTRACE_PEEKUSR, Name: "PTRACE_PEEKUSR", }, { - Value: syscall.PTRACE_POKETEXT, + Value: linux.PTRACE_POKETEXT, Name: "PTRACE_POKETEXT", }, { - Value: syscall.PTRACE_POKEDATA, + Value: linux.PTRACE_POKEDATA, Name: "PTRACE_POKEDATA", }, { - Value: syscall.PTRACE_POKEUSR, + Value: linux.PTRACE_POKEUSR, Name: "PTRACE_POKEUSR", }, { - Value: syscall.PTRACE_CONT, + Value: linux.PTRACE_CONT, Name: "PTRACE_CONT", }, { - Value: syscall.PTRACE_KILL, + Value: linux.PTRACE_KILL, Name: "PTRACE_KILL", }, { - Value: syscall.PTRACE_SINGLESTEP, + Value: linux.PTRACE_SINGLESTEP, Name: "PTRACE_SINGLESTEP", }, { - Value: syscall.PTRACE_ATTACH, + Value: linux.PTRACE_ATTACH, Name: "PTRACE_ATTACH", }, { - Value: syscall.PTRACE_DETACH, + Value: linux.PTRACE_DETACH, Name: "PTRACE_DETACH", }, { - Value: syscall.PTRACE_SYSCALL, + Value: linux.PTRACE_SYSCALL, Name: "PTRACE_SYSCALL", }, { - Value: syscall.PTRACE_SETOPTIONS, + Value: linux.PTRACE_SETOPTIONS, Name: "PTRACE_SETOPTIONS", }, { - Value: syscall.PTRACE_GETEVENTMSG, + Value: linux.PTRACE_GETEVENTMSG, Name: "PTRACE_GETEVENTMSG", }, { - Value: syscall.PTRACE_GETSIGINFO, + Value: linux.PTRACE_GETSIGINFO, Name: "PTRACE_GETSIGINFO", }, { - Value: syscall.PTRACE_SETSIGINFO, + Value: linux.PTRACE_SETSIGINFO, Name: "PTRACE_SETSIGINFO", }, { - Value: syscall.PTRACE_GETREGSET, + Value: linux.PTRACE_GETREGSET, Name: "PTRACE_GETREGSET", }, { - Value: syscall.PTRACE_SETREGSET, + Value: linux.PTRACE_SETREGSET, Name: "PTRACE_SETREGSET", }, { - Value: kernel.PTRACE_SEIZE, + Value: linux.PTRACE_SEIZE, Name: "PTRACE_SEIZE", }, { - Value: kernel.PTRACE_INTERRUPT, + Value: linux.PTRACE_INTERRUPT, Name: "PTRACE_INTERRUPT", }, { - Value: kernel.PTRACE_LISTEN, + Value: linux.PTRACE_LISTEN, Name: "PTRACE_LISTEN", }, { - Value: kernel.PTRACE_PEEKSIGINFO, + Value: linux.PTRACE_PEEKSIGINFO, Name: "PTRACE_PEEKSIGINFO", }, { - Value: kernel.PTRACE_GETSIGMASK, + Value: linux.PTRACE_GETSIGMASK, Name: "PTRACE_GETSIGMASK", }, { - Value: kernel.PTRACE_SETSIGMASK, + Value: linux.PTRACE_SETSIGMASK, Name: "PTRACE_SETSIGMASK", }, { - Value: syscall.PTRACE_GETREGS, + Value: linux.PTRACE_GETREGS, Name: "PTRACE_GETREGS", }, { - Value: syscall.PTRACE_SETREGS, + Value: linux.PTRACE_SETREGS, Name: "PTRACE_SETREGS", }, { - Value: syscall.PTRACE_GETFPREGS, + Value: linux.PTRACE_GETFPREGS, Name: "PTRACE_GETFPREGS", }, { - Value: syscall.PTRACE_SETFPREGS, + Value: linux.PTRACE_SETFPREGS, Name: "PTRACE_SETFPREGS", }, { - Value: syscall.PTRACE_GETFPXREGS, + Value: linux.PTRACE_GETFPXREGS, Name: "PTRACE_GETFPXREGS", }, { - Value: syscall.PTRACE_SETFPXREGS, + Value: linux.PTRACE_SETFPXREGS, Name: "PTRACE_SETFPXREGS", }, { - Value: syscall.PTRACE_OLDSETOPTIONS, + Value: linux.PTRACE_OLDSETOPTIONS, Name: "PTRACE_OLDSETOPTIONS", }, { - Value: syscall.PTRACE_GET_THREAD_AREA, + Value: linux.PTRACE_GET_THREAD_AREA, Name: "PTRACE_GET_THREAD_AREA", }, { - Value: syscall.PTRACE_SET_THREAD_AREA, + Value: linux.PTRACE_SET_THREAD_AREA, Name: "PTRACE_SET_THREAD_AREA", }, { - Value: syscall.PTRACE_ARCH_PRCTL, + Value: linux.PTRACE_ARCH_PRCTL, Name: "PTRACE_ARCH_PRCTL", }, { - Value: syscall.PTRACE_SYSEMU, + Value: linux.PTRACE_SYSEMU, Name: "PTRACE_SYSEMU", }, { - Value: syscall.PTRACE_SYSEMU_SINGLESTEP, + Value: linux.PTRACE_SYSEMU_SINGLESTEP, Name: "PTRACE_SYSEMU_SINGLESTEP", }, { - Value: syscall.PTRACE_SINGLEBLOCK, + Value: linux.PTRACE_SINGLEBLOCK, Name: "PTRACE_SINGLEBLOCK", }, } -- cgit v1.2.3 From 41e0b977e5ffc667750c0f706bb70173c5de2161 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Thu, 12 Jul 2018 10:36:16 -0700 Subject: Format documentation PiperOrigin-RevId: 204323728 Change-Id: I1ff9aa062ffa12583b2e38ec94c87db7a3711971 --- CONTRIBUTING.md | 74 +++++++++++--------- README.md | 119 ++++++++++++++++---------------- pkg/sentry/fs/README.md | 8 +-- pkg/sentry/fs/proc/README.md | 5 +- pkg/sentry/kernel/README.md | 80 ++++++++++----------- pkg/sentry/mm/README.md | 161 ++++++++++++++++++++++--------------------- pkg/sentry/usermem/README.md | 42 +++++------ 7 files changed, 252 insertions(+), 237 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index fa607113c..7ad19fb02 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -7,15 +7,16 @@ Before we can use your code, you must sign the online. The CLA is necessary mainly because you own the copyright to your changes, even after your contribution becomes part of our codebase, so we need your permission to use and distribute your code. We also need to be sure of -various other things—for instance that you'll tell us if you know that your -code infringes on other people's patents. You don't have to sign the CLA until -after you've submitted your code for review and a member has approved it, but -you must do it before we can put your code into our codebase. Before you start -working on a larger contribution, you should get in touch with us first through -the issue tracker with your idea so that we can help out and possibly guide you. +various other things—for instance that you'll tell us if you know that your code +infringes on other people's patents. You don't have to sign the CLA until after +you've submitted your code for review and a member has approved it, but you must +do it before we can put your code into our codebase. Before you start working on +a larger contribution, you should get in touch with us first through the issue +tracker with your idea so that we can help out and possibly guide you. Coordinating up front makes it much easier to avoid frustration later on. ### Coding Guidelines + All code should conform to the [Go style guidelines][gostyle]. As a secure runtime, we need to maintain the safety of all of code included in @@ -25,34 +26,41 @@ Definitions for the rules below: `core`: - * `//pkg/sentry/...` - * Transitive dependencies in `//pkg/...` +* `//pkg/sentry/...` +* Transitive dependencies in `//pkg/...` `runsc`: - * `//runsc/...` +* `//runsc/...` Rules: - * No cgo in `core` or `runsc`. The final binary must be a statically-linked +* No cgo in `core` or `runsc`. The final binary must be a statically-linked pure Go binary. - * Any files importing "unsafe" must have a name ending in `_unsafe.go`. +* Any files importing "unsafe" must have a name ending in `_unsafe.go`. + +* `core` may only depend on the following packages: + + * Itself. + * Go standard library. + * Except (transitively) package "net" (this will result in a non-cgo + binary). Use `//pkg/unet` instead. + * `@org_golang_x_sys//unix:go_default_library` (Go import + `golang.org/x/sys/unix`). + * Generated Go protobuf packages. + * `@com_github_golang_protobuf//proto:go_default_library` (Go import + `github.com/golang/protobuf/proto`). + * `@com_github_golang_protobuf//ptypes:go_default_library` (Go import + `github.com/golang/protobuf/ptypes`). - * `core` may only depend on the following packages: - * Itself. - * Go standard library. - * Except (transitively) package "net" (this will result in a non-cgo - binary). Use `//pkg/unet` instead. - * `@org_golang_x_sys//unix:go_default_library` (Go import `golang.org/x/sys/unix`). - * Generated Go protobuf packages. - * `@com_github_golang_protobuf//proto:go_default_library` (Go import `github.com/golang/protobuf/proto`). - * `@com_github_golang_protobuf//ptypes:go_default_library` (Go import `github.com/golang/protobuf/ptypes`). +* `runsc` may only depend on the following packages: - * `runsc` may only depend on the following packages: - * All packages allowed for `core`. - * `@com_github_google_subcommands//:go_default_library` (Go import `github.com/google/subcommands`). - * `@com_github_opencontainers_runtime_spec//specs_go:go_default_library` (Go import `github.com/opencontainers/runtime-spec/specs_go`). + * All packages allowed for `core`. + * `@com_github_google_subcommands//:go_default_library` (Go import + `github.com/google/subcommands`). + * `@com_github_opencontainers_runtime_spec//specs_go:go_default_library` + (Go import `github.com/opencontainers/runtime-spec/specs_go`). ### Code reviews @@ -66,8 +74,8 @@ To submit a patch, first clone the canonical repository. git clone https://gvisor.googlesource.com/gvisor ``` -From within the cloned directory, install the commit hooks (optional, but if -you don't you will need to generate Change-Ids manually in your commits). +From within the cloned directory, install the commit hooks (optional, but if you +don't you will need to generate Change-Ids manually in your commits). ``` curl -Lo `git rev-parse --git-dir`/hooks/commit-msg https://gerrit-review.googlesource.com/tools/hooks/commit-msg @@ -79,8 +87,8 @@ changes, remember to organize commits logically. Changes are not reviewed per branch (as with a pull request), they are reviewed per commit. Before posting a new patch, you will need to generate an appropriate -authentication cookie. Visit the [repository][repo] and click the -"Generate Password" link at the top of the page for instructions. +authentication cookie. Visit the [repository][repo] and click the "Generate +Password" link at the top of the page for instructions. To post a patch for review, push to a special "for" reference. @@ -90,17 +98,17 @@ git push origin HEAD:refs/for/master A change link will be generated for the commit, and a team member will review your change request, provide feedback (and submit when appropriate). To address -feedback, you may be required to amend your commit and repush (don't change -the Commit-Id in the commit message). This will generate a new version of -the change. +feedback, you may be required to amend your commit and repush (don't change the +Commit-Id in the commit message). This will generate a new version of the +change. When approved, the change will be submitted by a team member and automatically merged into the repository. ### The small print -Contributions made by corporations are covered by a different agreement than -the one above, the +Contributions made by corporations are covered by a different agreement than the +one above, the [Software Grant and Corporate Contributor License Agreement][gccla]. [gcla]: https://cla.developers.google.com/about/google-individual diff --git a/README.md b/README.md index ed989b0ed..624b8c062 100644 --- a/README.md +++ b/README.md @@ -1,10 +1,11 @@ # gVisor gVisor is a user-space kernel, written in Go, that implements a substantial -portion of the Linux system surface. It includes an [Open Container Initiative -(OCI)][oci] runtime called `runsc` that provides an isolation boundary between -the application and the host kernel. The `runsc` runtime integrates with Docker -and Kubernetes, making it simple to run sandboxed containers. +portion of the Linux system surface. It includes an +[Open Container Initiative (OCI)][oci] runtime called `runsc` that provides an +isolation boundary between the application and the host kernel. The `runsc` +runtime integrates with Docker and Kubernetes, making it simple to run sandboxed +containers. gVisor takes a distinct approach to container sandboxing and makes a different set of technical trade-offs compared to existing sandbox technologies, thus @@ -51,11 +52,11 @@ require a larger resource footprint and slower start-up times. [AppArmor][apparmor], allows the specification of a fine-grained security policy for an application or container. These schemes typically rely on hooks implemented inside the host kernel to enforce the rules. If the surface can be -made small enough (i.e. a sufficiently complete policy defined), then this is -an excellent way to sandbox applications and maintain native performance. -However, in practice it can be extremely difficult (if not impossible) to -reliably define a policy for arbitrary, previously unknown applications, -making this approach challenging to apply universally. +made small enough (i.e. a sufficiently complete policy defined), then this is an +excellent way to sandbox applications and maintain native performance. However, +in practice it can be extremely difficult (if not impossible) to reliably define +a policy for arbitrary, previously unknown applications, making this approach +challenging to apply universally. ![Rule-based execution](g3doc/Rule-Based-Execution.png "Rule-based execution") @@ -109,9 +110,9 @@ application to directly control the system calls it makes. In order to provide defense-in-depth and limit the host system surface, the gVisor container runtime is normally split into two separate processes. First, the *Sentry* process includes the kernel and is responsible for executing user -code and handling system calls. Second, file system operations that extend beyond -the sandbox (not internal proc or tmp files, pipes, etc.) are sent to a proxy, -called a *Gofer*, via a 9P connection. +code and handling system calls. Second, file system operations that extend +beyond the sandbox (not internal proc or tmp files, pipes, etc.) are sent to a +proxy, called a *Gofer*, via a 9P connection. ![Sentry](g3doc/Sentry-Gofer.png "Sentry and Gofer") @@ -138,17 +139,17 @@ isolation (see below). The Sentry requires a *platform* to implement basic context switching and memory mapping functionality. Today, gVisor supports two platforms: -* The **Ptrace** platform uses SYSEMU functionality to execute user code without - executing host system calls. This platform can run anywhere that `ptrace` - works (even VMs without nested virtualization). +* The **Ptrace** platform uses SYSEMU functionality to execute user code + without executing host system calls. This platform can run anywhere that + `ptrace` works (even VMs without nested virtualization). -* The **KVM** platform (experimental) allows the Sentry to act as both guest OS - and VMM, switching back and forth between the two worlds seamlessly. The KVM - platform can run on bare-metal or on a VM with nested virtualization enabled. - While there is no virtualized hardware layer -- the sandbox retains a process - model -- gVisor leverages virtualization extensions available on modern - processors in order to improve isolation and performance of address space - switches. +* The **KVM** platform (experimental) allows the Sentry to act as both guest + OS and VMM, switching back and forth between the two worlds seamlessly. The + KVM platform can run on bare-metal or on a VM with nested virtualization + enabled. While there is no virtualized hardware layer -- the sandbox retains + a process model -- gVisor leverages virtualization extensions available on + modern processors in order to improve isolation and performance of address + space switches. ### Performance @@ -172,8 +173,8 @@ binaries). The easiest way to get `runsc` is from the [latest nightly build][runsc-nightly]. After you download the binary, check it -against the SHA512 [checksum file][runsc-nightly-sha]. Older builds can be -found here: +against the SHA512 [checksum file][runsc-nightly-sha]. Older builds can be found +here: `https://storage.googleapis.com/gvisor/releases/nightly/${yyyy-mm-dd}/runsc` and `https://storage.googleapis.com/gvisor/releases/nightly/${yyyy-mm-dd}/runsc.sha512` @@ -193,8 +194,8 @@ sudo mv runsc /usr/local/bin Next, configure Docker to use `runsc` by adding a runtime entry to your Docker configuration (`/etc/docker/daemon.json`). You may have to create this file if -it does not exist. Also, some Docker versions also require you to [specify the -`storage-driver` field][docker-storage-driver]. +it does not exist. Also, some Docker versions also require you to +[specify the `storage-driver` field][docker-storage-driver]. In the end, the file should look something like: @@ -208,7 +209,8 @@ In the end, the file should look something like: } ``` -You must restart the Docker daemon after making changes to this file, typically this is done via: +You must restart the Docker daemon after making changes to this file, typically +this is done via: ``` sudo systemctl restart docker @@ -229,8 +231,8 @@ docker run --runtime=runsc -it ubuntu /bin/bash ### Kubernetes Support (Experimental) gVisor can run sandboxed containers in a Kubernetes cluster with cri-o, although -this is not recommended for production environments yet. Follow [these -instructions][cri-o-k8s] to run [cri-o][cri-o] on a node in a Kubernetes +this is not recommended for production environments yet. Follow +[these instructions][cri-o-k8s] to run [cri-o][cri-o] on a node in a Kubernetes cluster. Build `runsc` and put it on the node, and set it as the `runtime_untrusted_workload` in `/etc/crio/crio.conf`. @@ -251,11 +253,11 @@ gVisor currently requires x86\_64 Linux to build. Make sure the following dependencies are installed: -* [git][git] -* [Bazel][bazel] -* [Python][python] -* [Docker version 17.09.0 or greater][docker] -* Gold linker (e.g. `binutils-gold` package on Ubuntu) +* [git][git] +* [Bazel][bazel] +* [Python][python] +* [Docker version 17.09.0 or greater][docker] +* Gold linker (e.g. `binutils-gold` package on Ubuntu) #### Getting the source @@ -275,7 +277,6 @@ bazel build runsc sudo cp ./bazel-bin/runsc/linux_amd64_pure_stripped/runsc /usr/local/bin ``` - ### Testing The gVisor test suite can be run with Bazel: @@ -366,33 +367,33 @@ Then restart the Docker daemon. gVisor implements a large portion of the Linux surface and while we strive to make it broadly compatible, there are (and always will be) unimplemented features and bugs. The only real way to know if it will work is to try. If you -find a container that doesn’t work and there is no known issue, please [file a -bug][bug] indicating the full command you used to run the image. Providing the -debug logs is also helpful. +find a container that doesn’t work and there is no known issue, please +[file a bug][bug] indicating the full command you used to run the image. +Providing the debug logs is also helpful. ### What works? The following applications/images have been tested: -* elasticsearch -* golang -* httpd -* java8 -* jenkins -* mariadb -* memcached -* mongo -* mysql -* nginx -* node -* php -* postgres -* prometheus -* python -* redis -* registry -* tomcat -* wordpress +* elasticsearch +* golang +* httpd +* java8 +* jenkins +* mariadb +* memcached +* mongo +* mysql +* nginx +* node +* php +* postgres +* prometheus +* python +* redis +* registry +* tomcat +* wordpress ### My container runs fine with *runc* but fails with *runsc*. @@ -416,8 +417,8 @@ This bug is tracked in [bug #4](https://github.com/google/gvisor/issues/4). ## Technical details -We plan to release a full paper with technical details and will include it -here when available. +We plan to release a full paper with technical details and will include it here +when available. ## Community diff --git a/pkg/sentry/fs/README.md b/pkg/sentry/fs/README.md index 898271ee8..76638cdae 100644 --- a/pkg/sentry/fs/README.md +++ b/pkg/sentry/fs/README.md @@ -149,10 +149,10 @@ An `fs.File` references the following filesystem objects: fs.File -> fs.Dirent -> fs.Inode -> fs.MountedFilesystem ``` -The `fs.Inode` is restored using its `fs.MountedFilesystem`. The [Mount -points](#mount-points) section above describes how this happens in detail. The -`fs.Dirent` restores its pointer to an `fs.Inode`, pointers to parent and -children `fs.Dirents`, and the basename of the file. +The `fs.Inode` is restored using its `fs.MountedFilesystem`. The +[Mount points](#mount-points) section above describes how this happens in +detail. The `fs.Dirent` restores its pointer to an `fs.Inode`, pointers to +parent and children `fs.Dirents`, and the basename of the file. Otherwise an `fs.File` restores flags, an offset, and a unique identifier (only used internally). diff --git a/pkg/sentry/fs/proc/README.md b/pkg/sentry/fs/proc/README.md index 6ad7297d2..cec842403 100644 --- a/pkg/sentry/fs/proc/README.md +++ b/pkg/sentry/fs/proc/README.md @@ -6,6 +6,7 @@ procfs generally. inconsistency, please file a bug. [TOC] + ## Kernel data The following files are implemented: @@ -91,6 +92,7 @@ Num currently running processes | Always zero Total num processes | Always zero TODO: Populate the columns with accurate statistics. + ### meminfo ```bash @@ -122,7 +124,7 @@ Shmem: 0 kB Notable divergences: Field name | Notes -:---------------- | :-------------------------------------------------------- +:---------------- | :----------------------------------------------------- Buffers | Always zero, no block devices SwapCache | Always zero, no swap Inactive(anon) | Always zero, see SwapCache @@ -182,6 +184,7 @@ softirq 0 0 0 0 0 0 0 0 0 0 0 ``` All fields except for `btime` are always zero. + TODO: Populate with accurate fields. ### sys diff --git a/pkg/sentry/kernel/README.md b/pkg/sentry/kernel/README.md index 88760a9bb..427311be8 100644 --- a/pkg/sentry/kernel/README.md +++ b/pkg/sentry/kernel/README.md @@ -1,12 +1,12 @@ This package contains: -- A (partial) emulation of the "core Linux kernel", which governs task - execution and scheduling, system call dispatch, and signal handling. See - below for details. +- A (partial) emulation of the "core Linux kernel", which governs task + execution and scheduling, system call dispatch, and signal handling. See + below for details. -- The top-level interface for the sentry's Linux kernel emulation in general, - used by the `main` function of all versions of the sentry. This interface - revolves around the `Env` type (defined in `kernel.go`). +- The top-level interface for the sentry's Linux kernel emulation in general, + used by the `main` function of all versions of the sentry. This interface + revolves around the `Env` type (defined in `kernel.go`). # Background @@ -20,15 +20,15 @@ sentry's notion of a task unless otherwise specified.) At a high level, Linux application threads can be thought of as repeating a "run loop": -- Some amount of application code is executed in userspace. +- Some amount of application code is executed in userspace. -- A trap (explicit syscall invocation, hardware interrupt or exception, etc.) - causes control flow to switch to the kernel. +- A trap (explicit syscall invocation, hardware interrupt or exception, etc.) + causes control flow to switch to the kernel. -- Some amount of kernel code is executed in kernelspace, e.g. to handle the - cause of the trap. +- Some amount of kernel code is executed in kernelspace, e.g. to handle the + cause of the trap. -- The kernel "returns from the trap" into application code. +- The kernel "returns from the trap" into application code. Analogously, each task in the sentry is associated with a *task goroutine* that executes that task's run loop (`Task.run` in `task_run.go`). However, the @@ -38,24 +38,25 @@ state to, and resuming execution from, checkpoints. While in kernelspace, a Linux thread can be descheduled (cease execution) in a variety of ways: -- It can yield or be preempted, becoming temporarily descheduled but still - runnable. At present, the sentry delegates scheduling of runnable threads to - the Go runtime. +- It can yield or be preempted, becoming temporarily descheduled but still + runnable. At present, the sentry delegates scheduling of runnable threads to + the Go runtime. -- It can exit, becoming permanently descheduled. The sentry's equivalent is - returning from `Task.run`, terminating the task goroutine. +- It can exit, becoming permanently descheduled. The sentry's equivalent is + returning from `Task.run`, terminating the task goroutine. -- It can enter interruptible sleep, a state in which it can be woken by a - caller-defined wakeup or the receipt of a signal. In the sentry, interruptible - sleep (which is ambiguously referred to as *blocking*) is implemented by - making all events that can end blocking (including signal notifications) - communicated via Go channels and using `select` to multiplex wakeup sources; - see `task_block.go`. +- It can enter interruptible sleep, a state in which it can be woken by a + caller-defined wakeup or the receipt of a signal. In the sentry, + interruptible sleep (which is ambiguously referred to as *blocking*) is + implemented by making all events that can end blocking (including signal + notifications) communicated via Go channels and using `select` to multiplex + wakeup sources; see `task_block.go`. -- It can enter uninterruptible sleep, a state in which it can only be woken by a - caller-defined wakeup. Killable sleep is a closely related variant in which - the task can also be woken by SIGKILL. (These definitions also include Linux's - "group-stopped" (`TASK_STOPPED`) and "ptrace-stopped" (`TASK_TRACED`) states.) +- It can enter uninterruptible sleep, a state in which it can only be woken by + a caller-defined wakeup. Killable sleep is a closely related variant in + which the task can also be woken by SIGKILL. (These definitions also include + Linux's "group-stopped" (`TASK_STOPPED`) and "ptrace-stopped" + (`TASK_TRACED`) states.) To maximize compatibility with Linux, sentry checkpointing appears as a spurious signal-delivery interrupt on all tasks; interrupted system calls return `EINTR` @@ -71,21 +72,22 @@ through sleeping operations. We break the task's control flow graph into *states*, delimited by: -1. Points where uninterruptible and killable sleeps may occur. For example, -there exists a state boundary between signal dequeueing and signal delivery -because there may be an intervening ptrace signal-delivery-stop. +1. Points where uninterruptible and killable sleeps may occur. For example, + there exists a state boundary between signal dequeueing and signal delivery + because there may be an intervening ptrace signal-delivery-stop. -2. Points where sleep-induced branches may "rejoin" normal execution. For -example, the syscall exit state exists because it can be reached immediately -following a synchronous syscall, or after a task that is sleeping in `execve()` -or `vfork()` resumes execution. +2. Points where sleep-induced branches may "rejoin" normal execution. For + example, the syscall exit state exists because it can be reached immediately + following a synchronous syscall, or after a task that is sleeping in + `execve()` or `vfork()` resumes execution. -3. Points containing large branches. This is strictly for organizational -purposes. For example, the state that processes interrupt-signaled conditions is -kept separate from the main "app" state to reduce the size of the latter. +3. Points containing large branches. This is strictly for organizational + purposes. For example, the state that processes interrupt-signaled + conditions is kept separate from the main "app" state to reduce the size of + the latter. -4. `SyscallReinvoke`, which does not correspond to anything in Linux, and exists -solely to serve the autosave feature. +4. `SyscallReinvoke`, which does not correspond to anything in Linux, and + exists solely to serve the autosave feature. ![dot -Tpng -Goverlap=false -orun_states.png run_states.dot](g3doc/run_states.png "Task control flow graph") diff --git a/pkg/sentry/mm/README.md b/pkg/sentry/mm/README.md index 067733475..e485a5ca5 100644 --- a/pkg/sentry/mm/README.md +++ b/pkg/sentry/mm/README.md @@ -38,50 +38,50 @@ forces the kernel to create such a mapping to service the read. For a file, doing so consists of several logical phases: -1. The kernel allocates physical memory to store the contents of the required - part of the file, and copies file contents to the allocated memory. Supposing - that the kernel chooses the physical memory at physical address (PA) - 0x2fb000, the resulting state of the system is: +1. The kernel allocates physical memory to store the contents of the required + part of the file, and copies file contents to the allocated memory. + Supposing that the kernel chooses the physical memory at physical address + (PA) 0x2fb000, the resulting state of the system is: VMA: VA:0x400000 -> /tmp/foo:0x0 Filemap: /tmp/foo:0x0 -> PA:0x2fb000 - (In Linux the state of the mapping from file offset to physical memory is - stored in `struct address_space`, but to avoid confusion with other notions - of address space we will refer to this system as filemap, named after Linux - kernel source file `mm/filemap.c`.) + (In Linux the state of the mapping from file offset to physical memory is + stored in `struct address_space`, but to avoid confusion with other notions + of address space we will refer to this system as filemap, named after Linux + kernel source file `mm/filemap.c`.) -2. The kernel stores the effective mapping from virtual to physical address in a - *page table entry* (PTE) in the application's *page tables*, which are used - by the CPU's virtual memory hardware to perform address translation. The - resulting state of the system is: +2. The kernel stores the effective mapping from virtual to physical address in + a *page table entry* (PTE) in the application's *page tables*, which are + used by the CPU's virtual memory hardware to perform address translation. + The resulting state of the system is: VMA: VA:0x400000 -> /tmp/foo:0x0 Filemap: /tmp/foo:0x0 -> PA:0x2fb000 PTE: VA:0x400000 -----------------> PA:0x2fb000 - The PTE is required for the application to actually use the contents of the - mapped file as virtual memory. However, the PTE is derived from the VMA and - filemap state, both of which are independently mutable, such that mutations - to either will affect the PTE. For example: - - - The application may remove the VMA using the `munmap` system call. This - breaks the mapping from VA:0x400000 to /tmp/foo:0x0, and consequently the - mapping from VA:0x400000 to PA:0x2fb000. However, it does not necessarily - break the mapping from /tmp/foo:0x0 to PA:0x2fb000, so a future mapping of - the same file offset may reuse this physical memory. - - - The application may invalidate the file's contents by passing a length of 0 - to the `ftruncate` system call. This breaks the mapping from /tmp/foo:0x0 - to PA:0x2fb000, and consequently the mapping from VA:0x400000 to - PA:0x2fb000. However, it does not break the mapping from VA:0x400000 to - /tmp/foo:0x0, so future changes to the file's contents may again be made - visible at VA:0x400000 after another page fault results in the allocation - of a new physical address. - - Note that, in order to correctly break the mapping from VA:0x400000 to - PA:0x2fb000 in the latter case, filemap must also store a *reverse mapping* - from /tmp/foo:0x0 to VA:0x400000 so that it can locate and remove the PTE. + The PTE is required for the application to actually use the contents of the + mapped file as virtual memory. However, the PTE is derived from the VMA and + filemap state, both of which are independently mutable, such that mutations + to either will affect the PTE. For example: + + - The application may remove the VMA using the `munmap` system call. This + breaks the mapping from VA:0x400000 to /tmp/foo:0x0, and consequently + the mapping from VA:0x400000 to PA:0x2fb000. However, it does not + necessarily break the mapping from /tmp/foo:0x0 to PA:0x2fb000, so a + future mapping of the same file offset may reuse this physical memory. + + - The application may invalidate the file's contents by passing a length + of 0 to the `ftruncate` system call. This breaks the mapping from + /tmp/foo:0x0 to PA:0x2fb000, and consequently the mapping from + VA:0x400000 to PA:0x2fb000. However, it does not break the mapping from + VA:0x400000 to /tmp/foo:0x0, so future changes to the file's contents + may again be made visible at VA:0x400000 after another page fault + results in the allocation of a new physical address. + + Note that, in order to correctly break the mapping from VA:0x400000 to + PA:0x2fb000 in the latter case, filemap must also store a *reverse mapping* + from /tmp/foo:0x0 to VA:0x400000 so that it can locate and remove the PTE. [^mmap-anon]: Memory mappings to non-files are discussed in later sections. @@ -146,30 +146,30 @@ When the application first incurs a page fault on this address, the host kernel delivers information about the page fault to the sentry in a platform-dependent manner, and the sentry handles the fault: -1. The sentry allocates memory to store the contents of the required part of the - file, and copies file contents to the allocated memory. However, since the - sentry is implemented atop a host kernel, it does not configure mappings to - physical memory directly. Instead, mappable "memory" in the sentry is - represented by a host file descriptor and offset, since (as noted in - "Background") this is the memory mapping primitive provided by the host - kernel. In general, memory is allocated from a temporary host file using the - `filemem` package. Supposing that the sentry allocates offset 0x3000 from - host file "memory-file", the resulting state is: +1. The sentry allocates memory to store the contents of the required part of + the file, and copies file contents to the allocated memory. However, since + the sentry is implemented atop a host kernel, it does not configure mappings + to physical memory directly. Instead, mappable "memory" in the sentry is + represented by a host file descriptor and offset, since (as noted in + "Background") this is the memory mapping primitive provided by the host + kernel. In general, memory is allocated from a temporary host file using the + `filemem` package. Supposing that the sentry allocates offset 0x3000 from + host file "memory-file", the resulting state is: Sentry VMA: VA:0x400000 -> /tmp/foo:0x0 Sentry filemap: /tmp/foo:0x0 -> host:memory-file:0x3000 -2. The sentry stores the effective mapping from virtual address to host file in - a host VMA by invoking the `mmap` system call: +2. The sentry stores the effective mapping from virtual address to host file in + a host VMA by invoking the `mmap` system call: Sentry VMA: VA:0x400000 -> /tmp/foo:0x0 Sentry filemap: /tmp/foo:0x0 -> host:memory-file:0x3000 Host VMA: VA:0x400000 -----------------> host:memory-file:0x3000 -3. The sentry returns control to the application, which immediately incurs the - page fault again.[^mmap-populate] However, since a host VMA now exists for - the faulting virtual address, the host kernel now handles the page fault as - described in "Background": +3. The sentry returns control to the application, which immediately incurs the + page fault again.[^mmap-populate] However, since a host VMA now exists for + the faulting virtual address, the host kernel now handles the page fault as + described in "Background": Sentry VMA: VA:0x400000 -> /tmp/foo:0x0 Sentry filemap: /tmp/foo:0x0 -> host:memory-file:0x3000 @@ -183,12 +183,12 @@ independently mutable, and the desired state of host VMAs is derived from that state. [^mmap-populate]: The sentry could force the host kernel to establish PTEs when - it creates the host VMA by passing the `MAP_POPULATE` flag to - the `mmap` system call, but usually does not. This is because, - to reduce the number of page faults that require handling by - the sentry and (correspondingly) the number of host `mmap` - system calls, the sentry usually creates host VMAs that are - much larger than the single faulting page. + it creates the host VMA by passing the `MAP_POPULATE` flag to + the `mmap` system call, but usually does not. This is because, + to reduce the number of page faults that require handling by + the sentry and (correspondingly) the number of host `mmap` + system calls, the sentry usually creates host VMAs that are + much larger than the single faulting page. ## Private Mappings @@ -233,45 +233,46 @@ there is no shared zero page. In Linux: -- A virtual address space is represented by `struct mm_struct`. +- A virtual address space is represented by `struct mm_struct`. -- VMAs are represented by `struct vm_area_struct`, stored in `struct - mm_struct::mmap`. +- VMAs are represented by `struct vm_area_struct`, stored in `struct + mm_struct::mmap`. -- Mappings from file offsets to physical memory are stored in `struct - address_space`. +- Mappings from file offsets to physical memory are stored in `struct + address_space`. -- Reverse mappings from file offsets to virtual mappings are stored in `struct - address_space::i_mmap`. +- Reverse mappings from file offsets to virtual mappings are stored in `struct + address_space::i_mmap`. -- Physical memory pages are represented by a pointer to `struct page` or an - index called a *page frame number* (PFN), represented by `pfn_t`. +- Physical memory pages are represented by a pointer to `struct page` or an + index called a *page frame number* (PFN), represented by `pfn_t`. -- PTEs are represented by architecture-dependent type `pte_t`, stored in a table - hierarchy rooted at `struct mm_struct::pgd`. +- PTEs are represented by architecture-dependent type `pte_t`, stored in a + table hierarchy rooted at `struct mm_struct::pgd`. In the sentry: -- A virtual address space is represented by type [`mm.MemoryManager`][mm]. +- A virtual address space is represented by type [`mm.MemoryManager`][mm]. -- Sentry VMAs are represented by type [`mm.vma`][mm], stored in - `mm.MemoryManager.vmas`. +- Sentry VMAs are represented by type [`mm.vma`][mm], stored in + `mm.MemoryManager.vmas`. -- Mappings from sentry file offsets to host file offsets are abstracted through - interface method [`memmap.Mappable.Translate`][memmap]. +- Mappings from sentry file offsets to host file offsets are abstracted + through interface method [`memmap.Mappable.Translate`][memmap]. -- Reverse mappings from sentry file offsets to virtual mappings are abstracted - through interface methods [`memmap.Mappable.AddMapping` and - `memmap.Mappable.RemoveMapping`][memmap]. +- Reverse mappings from sentry file offsets to virtual mappings are abstracted + through interface methods + [`memmap.Mappable.AddMapping` and `memmap.Mappable.RemoveMapping`][memmap]. -- Host files that may be mapped into host VMAs are represented by type - [`platform.File`][platform]. +- Host files that may be mapped into host VMAs are represented by type + [`platform.File`][platform]. -- Host VMAs are represented in the sentry by type [`mm.pma`][mm] ("platform - mapping area"), stored in `mm.MemoryManager.pmas`. +- Host VMAs are represented in the sentry by type [`mm.pma`][mm] ("platform + mapping area"), stored in `mm.MemoryManager.pmas`. -- Creation and destruction of host VMAs is abstracted through interface methods - [`platform.AddressSpace.MapFile` and `platform.AddressSpace.Unmap`][platform]. +- Creation and destruction of host VMAs is abstracted through interface + methods + [`platform.AddressSpace.MapFile` and `platform.AddressSpace.Unmap`][platform]. [filemem]: https://gvisor.googlesource.com/gvisor/+/master/pkg/sentry/platform/filemem/filemem.go [memmap]: https://gvisor.googlesource.com/gvisor/+/master/pkg/sentry/memmap/memmap.go diff --git a/pkg/sentry/usermem/README.md b/pkg/sentry/usermem/README.md index 2ebd3bcc1..f6d2137eb 100644 --- a/pkg/sentry/usermem/README.md +++ b/pkg/sentry/usermem/README.md @@ -2,30 +2,30 @@ This package defines primitives for sentry access to application memory. Major types: -- The `IO` interface represents a virtual address space and provides I/O methods - on that address space. `IO` is the lowest-level primitive. The primary - implementation of the `IO` interface is `mm.MemoryManager`. +- The `IO` interface represents a virtual address space and provides I/O + methods on that address space. `IO` is the lowest-level primitive. The + primary implementation of the `IO` interface is `mm.MemoryManager`. -- `IOSequence` represents a collection of individually-contiguous address ranges - in a `IO` that is operated on sequentially, analogous to Linux's `struct - iov_iter`. +- `IOSequence` represents a collection of individually-contiguous address + ranges in a `IO` that is operated on sequentially, analogous to Linux's + `struct iov_iter`. Major usage patterns: -- Access to a task's virtual memory, subject to the application's memory - protections and while running on that task's goroutine, from a context that is - at or above the level of the `kernel` package (e.g. most syscall - implementations in `syscalls/linux`); use the `kernel.Task.Copy*` wrappers - defined in `kernel/task_usermem.go`. +- Access to a task's virtual memory, subject to the application's memory + protections and while running on that task's goroutine, from a context that + is at or above the level of the `kernel` package (e.g. most syscall + implementations in `syscalls/linux`); use the `kernel.Task.Copy*` wrappers + defined in `kernel/task_usermem.go`. -- Access to a task's virtual memory, from a context that is at or above the - level of the `kernel` package, but where any of the above constraints does not - hold (e.g. `PTRACE_POKEDATA`, which ignores application memory protections); - obtain the task's `mm.MemoryManager` by calling `kernel.Task.MemoryManager`, - and call its `IO` methods directly. +- Access to a task's virtual memory, from a context that is at or above the + level of the `kernel` package, but where any of the above constraints does + not hold (e.g. `PTRACE_POKEDATA`, which ignores application memory + protections); obtain the task's `mm.MemoryManager` by calling + `kernel.Task.MemoryManager`, and call its `IO` methods directly. -- Access to a task's virtual memory, from a context that is below the level of - the `kernel` package (e.g. filesystem I/O); clients must pass I/O arguments - from higher layers, usually in the form of an `IOSequence`. The - `kernel.Task.SingleIOSequence` and `kernel.Task.IovecsIOSequence` functions in - `kernel/task_usermem.go` are convenience functions for doing so. +- Access to a task's virtual memory, from a context that is below the level of + the `kernel` package (e.g. filesystem I/O); clients must pass I/O arguments + from higher layers, usually in the form of an `IOSequence`. The + `kernel.Task.SingleIOSequence` and `kernel.Task.IovecsIOSequence` functions + in `kernel/task_usermem.go` are convenience functions for doing so. -- cgit v1.2.3 From 1cd46c8dd1a92dd0ad3eeb60a763278f2e98d0b4 Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Thu, 12 Jul 2018 15:07:59 -0700 Subject: sentry: wait for restore clock instead of panicing in Timekeeper. PiperOrigin-RevId: 204372296 Change-Id: If1ed9843b93039806e0c65521f30177dc8036979 --- pkg/sentry/kernel/timekeeper.go | 25 ++++++++++++++++--------- pkg/sentry/kernel/timekeeper_state.go | 2 +- pkg/sentry/kernel/timekeeper_test.go | 4 ++-- 3 files changed, 19 insertions(+), 12 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go index 3f16c1676..4de8ac13b 100644 --- a/pkg/sentry/kernel/timekeeper.go +++ b/pkg/sentry/kernel/timekeeper.go @@ -44,14 +44,14 @@ type Timekeeper struct { // It is set only once, by SetClocks. monotonicOffset int64 `state:"nosave"` - // restored indicates that this Timekeeper was restored from a state - // file. - restored bool `state:"nosave"` + // restored, if non-nil, indicates that this Timekeeper was restored + // from a state file. The clocks are not set until restored is closed. + restored chan struct{} `state:"nosave"` // saveMonotonic is the (offset) value of the monotonic clock at the // time of save. // - // It is only valid if restored is true. + // It is only valid if restored is non-nil. // // It is only used in SetClocks after restore to compute the new // monotonicOffset. @@ -59,7 +59,7 @@ type Timekeeper struct { // saveRealtime is the value of the realtime clock at the time of save. // - // It is only valid if restored is true. + // It is only valid if restored is non-nil. // // It is only used in SetClocks after restore to compute the new // monotonicOffset. @@ -98,7 +98,7 @@ func NewTimekeeper(platform platform.Platform, paramPage platform.FileRange) (*T func (t *Timekeeper) SetClocks(c sentrytime.Clocks) { // Update the params, marking them "not ready", as we may need to // restart calibration on this new machine. - if t.restored { + if t.restored != nil { if err := t.params.Write(func() vdsoParams { return vdsoParams{} }); err != nil { @@ -135,7 +135,7 @@ func (t *Timekeeper) SetClocks(c sentrytime.Clocks) { panic("Unable to get current realtime: " + err.Error()) } - if t.restored { + if t.restored != nil { wantMonotonic = t.saveMonotonic elapsed := nowRealtime - t.saveRealtime if elapsed > 0 { @@ -145,7 +145,7 @@ func (t *Timekeeper) SetClocks(c sentrytime.Clocks) { t.monotonicOffset = wantMonotonic - nowMonotonic - if !t.restored { + if t.restored == nil { // Hold on to the initial "boot" time. t.bootTime = ktime.FromNanoseconds(nowRealtime) } @@ -153,6 +153,10 @@ func (t *Timekeeper) SetClocks(c sentrytime.Clocks) { t.mu.Lock() defer t.mu.Unlock() t.startUpdater() + + if t.restored != nil { + close(t.restored) + } } // startUpdater starts an update goroutine that keeps the clocks updated. @@ -255,7 +259,10 @@ func (t *Timekeeper) ResumeUpdates() { // GetTime returns the current time in nanoseconds. func (t *Timekeeper) GetTime(c sentrytime.ClockID) (int64, error) { if t.clocks == nil { - panic("Timekeeper used before initialized with SetClocks") + if t.restored == nil { + panic("Timekeeper used before initialized with SetClocks") + } + <-t.restored } now, err := t.clocks.GetTime(c) if err == nil && c == sentrytime.Monotonic { diff --git a/pkg/sentry/kernel/timekeeper_state.go b/pkg/sentry/kernel/timekeeper_state.go index aee983ac7..2e7fed4d8 100644 --- a/pkg/sentry/kernel/timekeeper_state.go +++ b/pkg/sentry/kernel/timekeeper_state.go @@ -37,5 +37,5 @@ func (t *Timekeeper) beforeSave() { // afterLoad is invoked by stateify. func (t *Timekeeper) afterLoad() { - t.restored = true + t.restored = make(chan struct{}) } diff --git a/pkg/sentry/kernel/timekeeper_test.go b/pkg/sentry/kernel/timekeeper_test.go index 08bacba4f..34a5cec27 100644 --- a/pkg/sentry/kernel/timekeeper_test.go +++ b/pkg/sentry/kernel/timekeeper_test.go @@ -107,7 +107,7 @@ func TestTimekeeperMonotonicForward(t *testing.T) { } tk := stateTestClocklessTimekeeper(t) - tk.restored = true + tk.restored = make(chan struct{}) tk.saveMonotonic = 100000 tk.saveRealtime = 400000 tk.SetClocks(c) @@ -135,7 +135,7 @@ func TestTimekeeperMonotonicJumpBackwards(t *testing.T) { } tk := stateTestClocklessTimekeeper(t) - tk.restored = true + tk.restored = make(chan struct{}) tk.saveMonotonic = 100000 tk.saveRealtime = 600000 tk.SetClocks(c) -- cgit v1.2.3 From 8f21c0bb2807888d812318def43c2405c9b13f5a Mon Sep 17 00:00:00 2001 From: Neel Natu Date: Mon, 16 Jul 2018 12:19:02 -0700 Subject: Add EventOperations.HostFD() This method allows an eventfd inside the Sentry to be registered with with the host kernel. Update comment about memory mapping host fds via CachingInodeOperations. PiperOrigin-RevId: 204784859 Change-Id: I55823321e2d84c17ae0f7efaabc6b55b852ae257 --- pkg/abi/linux/BUILD | 1 + pkg/abi/linux/eventfd.go | 22 +++++++ pkg/sentry/fs/fsutil/inode_cached.go | 3 +- pkg/sentry/kernel/eventfd/BUILD | 2 + pkg/sentry/kernel/eventfd/eventfd.go | 119 +++++++++++++++++++++++++++++++++-- 5 files changed, 139 insertions(+), 8 deletions(-) create mode 100644 pkg/abi/linux/eventfd.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index e164945cf..ae7e4378c 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -30,6 +30,7 @@ go_library( "dev.go", "elf.go", "errors.go", + "eventfd.go", "exec.go", "fcntl.go", "file.go", diff --git a/pkg/abi/linux/eventfd.go b/pkg/abi/linux/eventfd.go new file mode 100644 index 000000000..bc0fb44d2 --- /dev/null +++ b/pkg/abi/linux/eventfd.go @@ -0,0 +1,22 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package linux + +// Constants for eventfd2(2). +const ( + EFD_SEMAPHORE = 0x1 + EFD_CLOEXEC = O_CLOEXEC + EFD_NONBLOCK = O_NONBLOCK +) diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index 7c0f96ac2..cba642a8f 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -44,8 +44,7 @@ import ( // // CachingInodeOperations implements Mappable for the CachedFileObject: // -// - If CachedFileObject.FD returns a value >= 0 and the current platform shares -// a host fd table with the sentry, then the value of CachedFileObject.FD +// - If CachedFileObject.FD returns a value >= 0 then the file descriptor // will be memory mapped on the host. // // - Otherwise, the contents of CachedFileObject are buffered into memory diff --git a/pkg/sentry/kernel/eventfd/BUILD b/pkg/sentry/kernel/eventfd/BUILD index 2d5a3c693..561ced852 100644 --- a/pkg/sentry/kernel/eventfd/BUILD +++ b/pkg/sentry/kernel/eventfd/BUILD @@ -21,6 +21,7 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/eventfd", visibility = ["//pkg/sentry:internal"], deps = [ + "//pkg/abi/linux", "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/fs", @@ -30,6 +31,7 @@ go_library( "//pkg/state", "//pkg/syserror", "//pkg/waiter", + "//pkg/waiter/fdnotifier", ], ) diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index c9333719e..bd50bd9fe 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -21,6 +21,7 @@ import ( "sync" "syscall" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/anon" @@ -28,10 +29,12 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/waiter" + "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) // EventOperations represents an event with the semantics of Linux's file-based event -// notification (eventfd). +// notification (eventfd). Eventfds are usually internal to the Sentry but in certain +// situations they may be converted into a host-backed eventfd. type EventOperations struct { fsutil.NoopRelease `state:"nosave"` fsutil.PipeSeek `state:"nosave"` @@ -46,13 +49,16 @@ type EventOperations struct { // Queue is used to notify interested parties when the event object // becomes readable or writable. - waiter.Queue `state:"nosave"` + wq waiter.Queue `state:"nosave"` // val is the current value of the event counter. val uint64 // semMode specifies whether the event is in "semaphore" mode. semMode bool + + // hostfd indicates whether this eventfd is passed through to the host. + hostfd int } // New creates a new event object with the supplied initial value and mode. @@ -62,9 +68,48 @@ func New(ctx context.Context, initVal uint64, semMode bool) *fs.File { return fs.NewFile(ctx, dirent, fs.FileFlags{Read: true, Write: true}, &EventOperations{ val: initVal, semMode: semMode, + hostfd: -1, }) } +// HostFD returns the host eventfd associated with this event. +func (e *EventOperations) HostFD() (int, error) { + e.mu.Lock() + defer e.mu.Unlock() + if e.hostfd >= 0 { + return e.hostfd, nil + } + + flags := linux.EFD_NONBLOCK + if e.semMode { + flags |= linux.EFD_SEMAPHORE + } + + fd, _, err := syscall.Syscall(syscall.SYS_EVENTFD2, uintptr(e.val), uintptr(flags), 0) + if err != 0 { + return -1, err + } + + if err := fdnotifier.AddFD(int32(fd), &e.wq); err != nil { + syscall.Close(int(fd)) + return -1, err + } + + e.hostfd = int(fd) + return e.hostfd, nil +} + +// Release implements fs.FileOperations.Release. +func (e *EventOperations) Release() { + e.mu.Lock() + defer e.mu.Unlock() + if e.hostfd >= 0 { + fdnotifier.RemoveFD(int32(e.hostfd)) + syscall.Close(e.hostfd) + e.hostfd = -1 + } +} + // Read implements fs.FileOperations.Read. func (e *EventOperations) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, _ int64) (int64, error) { if dst.NumBytes() < 8 { @@ -87,9 +132,29 @@ func (e *EventOperations) Write(ctx context.Context, _ *fs.File, src usermem.IOS return 8, nil } +// Must be called with e.mu locked. +func (e *EventOperations) hostRead(ctx context.Context, dst usermem.IOSequence) error { + var buf [8]byte + + if _, err := syscall.Read(e.hostfd, buf[:]); err != nil { + if err == syscall.EWOULDBLOCK { + return syserror.ErrWouldBlock + } + return err + } + + _, err := dst.CopyOut(ctx, buf[:]) + return err +} + func (e *EventOperations) read(ctx context.Context, dst usermem.IOSequence) error { e.mu.Lock() + if e.hostfd >= 0 { + defer e.mu.Unlock() + return e.hostRead(ctx, dst) + } + // We can't complete the read if the value is currently zero. if e.val == 0 { e.mu.Unlock() @@ -112,7 +177,7 @@ func (e *EventOperations) read(ctx context.Context, dst usermem.IOSequence) erro // Notify writers. We do this even if we were already writable because // it is possible that a writer is waiting to write the maximum value // to the event. - e.Notify(waiter.EventOut) + e.wq.Notify(waiter.EventOut) var buf [8]byte usermem.ByteOrder.PutUint64(buf[:], val) @@ -120,6 +185,17 @@ func (e *EventOperations) read(ctx context.Context, dst usermem.IOSequence) erro return err } +// Must be called with e.mu locked. +func (e *EventOperations) hostWrite(val uint64) error { + var buf [8]byte + usermem.ByteOrder.PutUint64(buf[:], val) + _, err := syscall.Write(e.hostfd, buf[:]) + if err == syscall.EWOULDBLOCK { + return syserror.ErrWouldBlock + } + return err +} + func (e *EventOperations) write(ctx context.Context, src usermem.IOSequence) error { var buf [8]byte if _, err := src.CopyIn(ctx, buf[:]); err != nil { @@ -138,6 +214,11 @@ func (e *EventOperations) Signal(val uint64) error { e.mu.Lock() + if e.hostfd >= 0 { + defer e.mu.Unlock() + return e.hostWrite(val) + } + // We only allow writes that won't cause the value to go over the max // uint64 minus 1. if val > math.MaxUint64-1-e.val { @@ -149,16 +230,20 @@ func (e *EventOperations) Signal(val uint64) error { e.mu.Unlock() // Always trigger a notification. - e.Notify(waiter.EventIn) + e.wq.Notify(waiter.EventIn) return nil } // Readiness returns the ready events for the event fd. func (e *EventOperations) Readiness(mask waiter.EventMask) waiter.EventMask { - ready := waiter.EventMask(0) - e.mu.Lock() + if e.hostfd >= 0 { + defer e.mu.Unlock() + return fdnotifier.NonBlockingPoll(int32(e.hostfd), mask) + } + + ready := waiter.EventMask(0) if e.val > 0 { ready |= waiter.EventIn } @@ -170,3 +255,25 @@ func (e *EventOperations) Readiness(mask waiter.EventMask) waiter.EventMask { return mask & ready } + +// EventRegister implements waiter.Waitable.EventRegister. +func (e *EventOperations) EventRegister(entry *waiter.Entry, mask waiter.EventMask) { + e.wq.EventRegister(entry, mask) + + e.mu.Lock() + defer e.mu.Unlock() + if e.hostfd >= 0 { + fdnotifier.UpdateFD(int32(e.hostfd)) + } +} + +// EventUnregister implements waiter.Waitable.EventUnregister. +func (e *EventOperations) EventUnregister(entry *waiter.Entry) { + e.wq.EventUnregister(entry) + + e.mu.Lock() + defer e.mu.Unlock() + if e.hostfd >= 0 { + fdnotifier.UpdateFD(int32(e.hostfd)) + } +} -- cgit v1.2.3 From 29e00c943a61dfcfd4ac8d3f6f526eab641c44a6 Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Mon, 16 Jul 2018 22:02:03 -0700 Subject: Add CPUID faulting for ptrace and KVM. PiperOrigin-RevId: 204858314 Change-Id: I8252bf8de3232a7a27af51076139b585e73276d4 --- pkg/abi/linux/prctl.go | 9 +++--- pkg/sentry/kernel/task_run.go | 41 ++++++++++++-------------- pkg/sentry/platform/kvm/machine.go | 22 +++++++------- pkg/sentry/platform/kvm/machine_amd64.go | 22 ++++++++++++++ pkg/sentry/platform/platform.go | 7 +++++ pkg/sentry/platform/ptrace/ptrace.go | 15 +++++++--- pkg/sentry/platform/ptrace/subprocess_linux.go | 5 ++++ pkg/sentry/platform/ring0/kernel_amd64.go | 22 +++++++++++++- pkg/sentry/platform/ring0/x86.go | 14 ++++++--- 9 files changed, 111 insertions(+), 46 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/prctl.go b/pkg/abi/linux/prctl.go index 6c93601de..074ec03f0 100644 --- a/pkg/abi/linux/prctl.go +++ b/pkg/abi/linux/prctl.go @@ -65,8 +65,9 @@ const ( // From // Flags are used in syscall arch_prctl(2). const ( - ARCH_SET_GS = 0x1001 - ARCH_SET_FS = 0x1002 - ARCH_GET_FS = 0x1003 - ARCH_GET_GS = 0x1004 + ARCH_SET_GS = 0x1001 + ARCH_SET_FS = 0x1002 + ARCH_GET_FS = 0x1003 + ARCH_GET_GS = 0x1004 + ARCH_SET_CPUID = 0x1012 ) diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 94ce5582b..a03fa6ac0 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -221,6 +221,24 @@ func (*runApp) execute(t *Task) taskRunState { // loop to figure out why. return (*runApp)(nil) + case platform.ErrContextSignalCPUID: + // Is this a CPUID instruction? + expected := arch.CPUIDInstruction[:] + found := make([]byte, len(expected)) + _, err := t.CopyIn(usermem.Addr(t.Arch().IP()), &found) + if err == nil && bytes.Equal(expected, found) { + // Skip the cpuid instruction. + t.Arch().CPUIDEmulate(t) + t.Arch().SetIP(t.Arch().IP() + uintptr(len(expected))) + + // Resume execution. + return (*runApp)(nil) + } + + // The instruction at the given RIP was not a CPUID, and we + // fallthrough to the default signal deliver behavior below. + fallthrough + case platform.ErrContextSignal: // Looks like a signal has been delivered to us. If it's a synchronous // signal (SEGV, SIGBUS, etc.), it should be sent to the application @@ -266,28 +284,7 @@ func (*runApp) execute(t *Task) taskRunState { } switch sig { - case linux.SIGILL: - // N.B. The debug stuff here is arguably - // expensive. Don't fret. This gets called - // about 5 times for a typical application, if - // that. - t.Debugf("SIGILL @ %x", t.Arch().IP()) - - // Is this a CPUID instruction? - expected := arch.CPUIDInstruction[:] - found := make([]byte, len(expected)) - _, err := t.CopyIn(usermem.Addr(t.Arch().IP()), &found) - if err == nil && bytes.Equal(expected, found) { - // Skip the cpuid instruction. - t.Arch().CPUIDEmulate(t) - t.Arch().SetIP(t.Arch().IP() + uintptr(len(expected))) - break - } - - // Treat it like any other synchronous signal. - fallthrough - - case linux.SIGSEGV, linux.SIGBUS, linux.SIGFPE, linux.SIGTRAP: + case linux.SIGILL, linux.SIGSEGV, linux.SIGBUS, linux.SIGFPE, linux.SIGTRAP: // Synchronous signal. Send it to ourselves. Assume the signal is // legitimate and force it (work around the signal being ignored or // blocked) like Linux does. Conveniently, this is even the correct diff --git a/pkg/sentry/platform/kvm/machine.go b/pkg/sentry/platform/kvm/machine.go index abdc51431..68e099d1b 100644 --- a/pkg/sentry/platform/kvm/machine.go +++ b/pkg/sentry/platform/kvm/machine.go @@ -141,11 +141,6 @@ func (m *machine) newVCPU() *vCPU { panic(fmt.Sprintf("error setting signal mask: %v", err)) } - // Initialize architecture state. - if err := c.initArchState(); err != nil { - panic(fmt.Sprintf("error initialization vCPU state: %v", err)) - } - // Map the run data. runData, err := mapRunData(int(fd)) if err != nil { @@ -153,6 +148,11 @@ func (m *machine) newVCPU() *vCPU { } c.runData = runData + // Initialize architecture state. + if err := c.initArchState(); err != nil { + panic(fmt.Sprintf("error initialization vCPU state: %v", err)) + } + return c // Done. } @@ -168,12 +168,6 @@ func newMachine(vm int) (*machine, error) { PageTables: pagetables.New(newAllocator()), }) - // Initialize architecture state. - if err := m.initArchState(); err != nil { - m.Destroy() - return nil, err - } - // Apply the physical mappings. Note that these mappings may point to // guest physical addresses that are not actually available. These // physical pages are mapped on demand, see kernel_unsafe.go. @@ -221,6 +215,12 @@ func newMachine(vm int) (*machine, error) { } }) + // Initialize architecture state. + if err := m.initArchState(); err != nil { + m.Destroy() + return nil, err + } + // Ensure the machine is cleaned up properly. runtime.SetFinalizer(m, (*machine).Destroy) return m, nil diff --git a/pkg/sentry/platform/kvm/machine_amd64.go b/pkg/sentry/platform/kvm/machine_amd64.go index 9af4f3f3d..bcd29a947 100644 --- a/pkg/sentry/platform/kvm/machine_amd64.go +++ b/pkg/sentry/platform/kvm/machine_amd64.go @@ -19,6 +19,7 @@ package kvm import ( "fmt" "reflect" + "runtime/debug" "syscall" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" @@ -39,6 +40,21 @@ func (m *machine) initArchState() error { uintptr(reservedMemory-(3*usermem.PageSize))); errno != 0 { return errno } + + // Enable CPUID faulting, if possible. Note that this also serves as a + // basic platform sanity tests, since we will enter guest mode for the + // first time here. The recovery is necessary, since if we fail to read + // the platform info register, we will retry to host mode and + // ultimately need to handle a segmentation fault. + old := debug.SetPanicOnFault(true) + defer func() { + recover() + debug.SetPanicOnFault(old) + }() + m.retryInGuest(func() { + ring0.SetCPUIDFaulting(true) + }) + return nil } @@ -238,6 +254,12 @@ func (c *vCPU) SwitchToUser(switchOpts ring0.SwitchOpts) (*arch.SignalInfo, user Code: arch.SignalInfoKernel, } info.SetAddr(switchOpts.Registers.Rip) // Include address. + if vector == ring0.GeneralProtectionFault { + // When CPUID faulting is enabled, we will generate a #GP(0) when + // userspace executes a CPUID instruction. This is handled above, + // because we need to be able to map and read user memory. + return info, usermem.AccessType{}, platform.ErrContextSignalCPUID + } return info, usermem.AccessType{}, platform.ErrContextSignal case ring0.InvalidOpcode: diff --git a/pkg/sentry/platform/platform.go b/pkg/sentry/platform/platform.go index f2fe163e8..6eb2acbd7 100644 --- a/pkg/sentry/platform/platform.go +++ b/pkg/sentry/platform/platform.go @@ -154,6 +154,13 @@ var ( // Context was interrupted by a signal. ErrContextSignal = fmt.Errorf("interrupted by signal") + // ErrContextSignalCPUID is equivalent to ErrContextSignal, except that + // a check should be done for execution of the CPUID instruction. If + // the current instruction pointer is a CPUID instruction, then this + // should be emulated appropriately. If not, then the given signal + // should be handled per above. + ErrContextSignalCPUID = fmt.Errorf("interrupted by signal, possible CPUID") + // ErrContextInterrupt is returned by Context.Switch() to indicate that the // Context was interrupted by a call to Context.Interrupt(). ErrContextInterrupt = fmt.Errorf("interrupted by platform.Context.Interrupt()") diff --git a/pkg/sentry/platform/ptrace/ptrace.go b/pkg/sentry/platform/ptrace/ptrace.go index 05f8b1d05..a44f549a2 100644 --- a/pkg/sentry/platform/ptrace/ptrace.go +++ b/pkg/sentry/platform/ptrace/ptrace.go @@ -101,9 +101,11 @@ func (c *context) Switch(as platform.AddressSpace, ac arch.Context, cpu int32) ( s := as.(*subprocess) isSyscall := s.switchToApp(c, ac) - var faultSP *subprocess - var faultAddr usermem.Addr - var faultIP usermem.Addr + var ( + faultSP *subprocess + faultAddr usermem.Addr + faultIP usermem.Addr + ) if !isSyscall && linux.Signal(c.signalInfo.Signo) == linux.SIGSEGV { faultSP = s faultAddr = usermem.Addr(c.signalInfo.Addr()) @@ -161,7 +163,12 @@ func (c *context) Switch(as platform.AddressSpace, ac arch.Context, cpu int32) ( lastFaultIP == faultIP { at.Write = true } - return &c.signalInfo, at, platform.ErrContextSignal + + // Unfortunately, we have to unilaterally return ErrContextSignalCPUID + // here, in case this fault was generated by a CPUID exception. There + // is no way to distinguish between CPUID-generated faults and regular + // page faults. + return &c.signalInfo, at, platform.ErrContextSignalCPUID } // Interrupt interrupts the running guest application associated with this context. diff --git a/pkg/sentry/platform/ptrace/subprocess_linux.go b/pkg/sentry/platform/ptrace/subprocess_linux.go index b3f2ebb20..b212bbdfe 100644 --- a/pkg/sentry/platform/ptrace/subprocess_linux.go +++ b/pkg/sentry/platform/ptrace/subprocess_linux.go @@ -20,6 +20,7 @@ import ( "fmt" "syscall" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/platform/procid" ) @@ -85,6 +86,10 @@ func createStub() (*thread, error) { syscall.RawSyscall(syscall.SYS_EXIT, uintptr(errno), 0, 0) } + // Enable cpuid-faulting; this may fail on older kernels or hardware, + // so we just disregard the result. Host CPUID will be enabled. + syscall.RawSyscall(syscall.SYS_ARCH_PRCTL, linux.ARCH_SET_CPUID, 0, 0) + // Call the stub; should not return. stubCall(stubStart, ppid) panic("unreachable") diff --git a/pkg/sentry/platform/ring0/kernel_amd64.go b/pkg/sentry/platform/ring0/kernel_amd64.go index 117e86104..0d2b0f7dc 100644 --- a/pkg/sentry/platform/ring0/kernel_amd64.go +++ b/pkg/sentry/platform/ring0/kernel_amd64.go @@ -163,7 +163,6 @@ func IsCanonical(addr uint64) bool { // the case for amd64, but may not be the case for other architectures. // // Precondition: the Rip, Rsp, Fs and Gs registers must be canonical. - // //go:nosplit func (c *CPU) SwitchToUser(switchOpts SwitchOpts) (vector Vector) { @@ -237,6 +236,27 @@ func start(c *CPU) { wrmsr(_MSR_CSTAR, kernelFunc(sysenter)) } +// SetCPUIDFaulting sets CPUID faulting per the boolean value. +// +// True is returned if faulting could be set. +// +//go:nosplit +func SetCPUIDFaulting(on bool) bool { + // Per the SDM (Vol 3, Table 2-43), PLATFORM_INFO bit 31 denotes support + // for CPUID faulting, and we enable and disable via the MISC_FEATURES MSR. + if rdmsr(_MSR_PLATFORM_INFO)&_PLATFORM_INFO_CPUID_FAULT != 0 { + features := rdmsr(_MSR_MISC_FEATURES) + if on { + features |= _MISC_FEATURE_CPUID_TRAP + } else { + features &^= _MISC_FEATURE_CPUID_TRAP + } + wrmsr(_MSR_MISC_FEATURES, features) + return true // Setting successful. + } + return false +} + // ReadCR2 reads the current CR2 value. // //go:nosplit diff --git a/pkg/sentry/platform/ring0/x86.go b/pkg/sentry/platform/ring0/x86.go index 3d437a77c..f489fcecb 100644 --- a/pkg/sentry/platform/ring0/x86.go +++ b/pkg/sentry/platform/ring0/x86.go @@ -50,10 +50,16 @@ const ( _EFER_LMA = 0x400 _EFER_NX = 0x800 - _MSR_STAR = 0xc0000081 - _MSR_LSTAR = 0xc0000082 - _MSR_CSTAR = 0xc0000083 - _MSR_SYSCALL_MASK = 0xc0000084 + _MSR_STAR = 0xc0000081 + _MSR_LSTAR = 0xc0000082 + _MSR_CSTAR = 0xc0000083 + _MSR_SYSCALL_MASK = 0xc0000084 + _MSR_PLATFORM_INFO = 0xce + _MSR_MISC_FEATURES = 0x140 + + _PLATFORM_INFO_CPUID_FAULT = 1 << 31 + + _MISC_FEATURE_CPUID_TRAP = 0x1 ) const ( -- cgit v1.2.3 From 8b8aad91d581ee5f600f5ec0b7fb407b36d07db1 Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Thu, 19 Jul 2018 15:48:08 -0700 Subject: kernel: mutations on creds now require a copy. PiperOrigin-RevId: 205315612 Change-Id: I9a0a1e32c8abfb7467a38743b82449cc92830316 --- pkg/sentry/kernel/fasync/fasync.go | 2 +- pkg/sentry/kernel/task.go | 4 +++- pkg/sentry/kernel/task_identity.go | 16 +++++++++++++--- pkg/sentry/syscalls/linux/sys_file.go | 4 ++-- 4 files changed, 19 insertions(+), 7 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/fasync/fasync.go b/pkg/sentry/kernel/fasync/fasync.go index 028d6766f..15218fb5a 100644 --- a/pkg/sentry/kernel/fasync/fasync.go +++ b/pkg/sentry/kernel/fasync/fasync.go @@ -35,7 +35,7 @@ func New() fs.FileAsync { type FileAsync struct { mu sync.Mutex e waiter.Entry - requester auth.Credentials + requester *auth.Credentials // Only one of the following is allowed to be non-nil. recipientPG *kernel.ProcessGroup diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 7763050a5..7f6735320 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -334,7 +334,9 @@ type Task struct { // creds is the task's credentials. // - // creds is protected by mu. + // creds is protected by mu, however the value itself is immutable and + // can only be changed by a copy. After reading the pointer, access + // will proceed outside the scope of mu. creds *auth.Credentials // utsns is the task's UTS namespace. diff --git a/pkg/sentry/kernel/task_identity.go b/pkg/sentry/kernel/task_identity.go index a51fa9d7e..b0921b2eb 100644 --- a/pkg/sentry/kernel/task_identity.go +++ b/pkg/sentry/kernel/task_identity.go @@ -20,11 +20,13 @@ import ( "gvisor.googlesource.com/gvisor/pkg/syserror" ) -// Credentials returns t's credentials by value. -func (t *Task) Credentials() auth.Credentials { +// Credentials returns t's credentials. +// +// This value must be considered immutable. +func (t *Task) Credentials() *auth.Credentials { t.mu.Lock() defer t.mu.Unlock() - return *t.creds // Copy out with lock held. + return t.creds } // UserNamespace returns the user namespace associated with the task. @@ -162,6 +164,7 @@ func (t *Task) SetRESUID(r, e, s auth.UID) error { func (t *Task) setKUIDsUncheckedLocked(newR, newE, newS auth.KUID) { root := t.creds.UserNamespace.MapToKUID(auth.RootUID) oldR, oldE, oldS := t.creds.RealKUID, t.creds.EffectiveKUID, t.creds.SavedKUID + t.creds = t.creds.Fork() // See doc for creds. t.creds.RealKUID, t.creds.EffectiveKUID, t.creds.SavedKUID = newR, newE, newS // "1. If one or more of the real, effective or saved set user IDs was @@ -297,6 +300,7 @@ func (t *Task) SetRESGID(r, e, s auth.GID) error { func (t *Task) setKGIDsUncheckedLocked(newR, newE, newS auth.KGID) { oldE := t.creds.EffectiveKGID + t.creds = t.creds.Fork() // See doc for creds. t.creds.RealKGID, t.creds.EffectiveKGID, t.creds.SavedKGID = newR, newE, newS // Not documented, but compare Linux's kernel/cred.c:commit_creds(). @@ -321,6 +325,7 @@ func (t *Task) SetExtraGIDs(gids []auth.GID) error { } kgids[i] = kgid } + t.creds = t.creds.Fork() // See doc for creds. t.creds.ExtraKGIDs = kgids return nil } @@ -352,6 +357,7 @@ func (t *Task) SetCapabilitySets(permitted, inheritable, effective auth.Capabili if inheritable & ^(t.creds.InheritableCaps|t.creds.BoundingCaps) != 0 { return syserror.EPERM } + t.creds = t.creds.Fork() // See doc for creds. t.creds.PermittedCaps = permitted t.creds.InheritableCaps = inheritable t.creds.EffectiveCaps = effective @@ -384,6 +390,7 @@ func (t *Task) SetUserNamespace(ns *auth.UserNamespace) error { return syserror.EPERM } + t.creds = t.creds.Fork() // See doc for creds. t.creds.UserNamespace = ns // "The child process created by clone(2) with the CLONE_NEWUSER flag // starts out with a complete set of capabilities in the new user @@ -407,6 +414,7 @@ func (t *Task) SetUserNamespace(ns *auth.UserNamespace) error { func (t *Task) SetKeepCaps(k bool) { t.mu.Lock() defer t.mu.Unlock() + t.creds = t.creds.Fork() // See doc for creds. t.creds.KeepCaps = k } @@ -491,6 +499,8 @@ func (t *Task) updateCredsForExecLocked() { } } + t.creds = t.creds.Fork() // See doc for creds. + // Now we enter poorly-documented, somewhat confusing territory. (The // accompanying comment in Linux's security/commoncap.c:cap_bprm_set_creds // is not very helpful.) My reading of it is: diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 490649f87..66e6fd9d4 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -415,14 +415,14 @@ func Creat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall // accessContext should only be used for access(2). type accessContext struct { context.Context - creds auth.Credentials + creds *auth.Credentials } // Value implements context.Context. func (ac accessContext) Value(key interface{}) interface{} { switch key { case auth.CtxCredentials: - return &ac.creds + return ac.creds default: return ac.Context.Value(key) } -- cgit v1.2.3 From be7fcbc5582fe831b5ec63f773d867d7591e27a1 Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Fri, 27 Jul 2018 10:16:27 -0700 Subject: stateify: support explicit annotation mode; convert refs and stack packages. We have been unnecessarily creating too many savable types implicitly. PiperOrigin-RevId: 206334201 Change-Id: Idc5a3a14bfb7ee125c4f2bb2b1c53164e46f29a8 --- pkg/abi/BUILD | 3 +- pkg/abi/linux/BUILD | 3 +- pkg/amutex/BUILD | 2 +- pkg/atomicbitops/BUILD | 2 +- pkg/binary/BUILD | 2 +- pkg/bits/BUILD | 2 +- pkg/bpf/BUILD | 3 +- pkg/compressio/BUILD | 2 +- pkg/control/client/BUILD | 2 +- pkg/control/server/BUILD | 2 +- pkg/cpuid/BUILD | 3 +- pkg/dhcp/BUILD | 2 +- pkg/eventchannel/BUILD | 2 +- pkg/fd/BUILD | 2 +- pkg/gate/BUILD | 2 +- pkg/hashio/BUILD | 2 +- pkg/ilist/BUILD | 3 +- pkg/linewriter/BUILD | 2 +- pkg/log/BUILD | 2 +- pkg/metric/BUILD | 2 +- pkg/p9/BUILD | 2 +- pkg/p9/p9test/BUILD | 2 +- pkg/rand/BUILD | 2 +- pkg/refs/BUILD | 20 +------- pkg/refs/refcounter.go | 4 ++ pkg/refs/refcounter_state.go | 1 + pkg/seccomp/BUILD | 3 +- pkg/secio/BUILD | 2 +- pkg/segment/test/BUILD | 2 +- pkg/sentry/arch/BUILD | 3 +- pkg/sentry/context/BUILD | 2 +- pkg/sentry/context/contexttest/BUILD | 3 +- pkg/sentry/control/BUILD | 2 +- pkg/sentry/device/BUILD | 2 +- pkg/sentry/fs/BUILD | 3 +- pkg/sentry/fs/anon/BUILD | 2 +- pkg/sentry/fs/ashmem/BUILD | 3 +- pkg/sentry/fs/binder/BUILD | 3 +- pkg/sentry/fs/dev/BUILD | 3 +- pkg/sentry/fs/fdpipe/BUILD | 3 +- pkg/sentry/fs/filetest/BUILD | 3 +- pkg/sentry/fs/fsutil/BUILD | 3 +- pkg/sentry/fs/gofer/BUILD | 3 +- pkg/sentry/fs/host/BUILD | 3 +- pkg/sentry/fs/lock/BUILD | 3 +- pkg/sentry/fs/proc/BUILD | 3 +- pkg/sentry/fs/proc/device/BUILD | 2 +- pkg/sentry/fs/proc/seqfile/BUILD | 3 +- pkg/sentry/fs/ramfs/BUILD | 3 +- pkg/sentry/fs/ramfs/test/BUILD | 3 +- pkg/sentry/fs/sys/BUILD | 3 +- pkg/sentry/fs/timerfd/BUILD | 3 +- pkg/sentry/fs/tmpfs/BUILD | 3 +- pkg/sentry/fs/tty/BUILD | 3 +- pkg/sentry/hostcpu/BUILD | 2 +- pkg/sentry/inet/BUILD | 4 +- pkg/sentry/kernel/BUILD | 3 +- pkg/sentry/kernel/auth/BUILD | 3 +- pkg/sentry/kernel/epoll/BUILD | 3 +- pkg/sentry/kernel/eventfd/BUILD | 3 +- pkg/sentry/kernel/fasync/BUILD | 2 +- pkg/sentry/kernel/futex/BUILD | 3 +- pkg/sentry/kernel/kdefs/BUILD | 2 +- pkg/sentry/kernel/memevent/BUILD | 2 +- pkg/sentry/kernel/pipe/BUILD | 3 +- pkg/sentry/kernel/sched/BUILD | 2 +- pkg/sentry/kernel/semaphore/BUILD | 3 +- pkg/sentry/kernel/shm/BUILD | 3 +- pkg/sentry/kernel/time/BUILD | 3 +- pkg/sentry/limits/BUILD | 3 +- pkg/sentry/loader/BUILD | 4 +- pkg/sentry/memmap/BUILD | 3 +- pkg/sentry/memutil/BUILD | 2 +- pkg/sentry/mm/BUILD | 3 +- pkg/sentry/platform/BUILD | 3 +- pkg/sentry/platform/filemem/BUILD | 3 +- pkg/sentry/platform/interrupt/BUILD | 2 +- pkg/sentry/platform/kvm/BUILD | 2 +- pkg/sentry/platform/kvm/testutil/BUILD | 2 +- pkg/sentry/platform/procid/BUILD | 2 +- pkg/sentry/platform/ptrace/BUILD | 2 +- pkg/sentry/platform/ring0/BUILD | 2 +- pkg/sentry/platform/ring0/pagetables/BUILD | 2 +- pkg/sentry/platform/safecopy/BUILD | 2 +- pkg/sentry/safemem/BUILD | 2 +- pkg/sentry/sighandling/BUILD | 2 +- pkg/sentry/socket/BUILD | 3 +- pkg/sentry/socket/control/BUILD | 3 +- pkg/sentry/socket/epsocket/BUILD | 3 +- pkg/sentry/socket/hostinet/BUILD | 3 +- pkg/sentry/socket/netlink/BUILD | 3 +- pkg/sentry/socket/netlink/port/BUILD | 3 +- pkg/sentry/socket/netlink/route/BUILD | 3 +- pkg/sentry/socket/rpcinet/BUILD | 2 +- pkg/sentry/socket/rpcinet/conn/BUILD | 2 +- pkg/sentry/socket/rpcinet/notifier/BUILD | 2 +- pkg/sentry/socket/unix/BUILD | 3 +- pkg/sentry/state/BUILD | 2 +- pkg/sentry/strace/BUILD | 2 +- pkg/sentry/syscalls/BUILD | 2 +- pkg/sentry/syscalls/linux/BUILD | 3 +- pkg/sentry/time/BUILD | 2 +- pkg/sentry/uniqueid/BUILD | 2 +- pkg/sentry/usage/BUILD | 3 +- pkg/sentry/usermem/BUILD | 3 +- pkg/sentry/watchdog/BUILD | 2 +- pkg/sleep/BUILD | 2 +- pkg/state/BUILD | 2 +- pkg/state/statefile/BUILD | 2 +- pkg/sync/BUILD | 2 +- pkg/sync/seqatomictest/BUILD | 2 +- pkg/syserr/BUILD | 2 +- pkg/syserror/BUILD | 2 +- pkg/tcpip/BUILD | 3 +- pkg/tcpip/adapters/gonet/BUILD | 2 +- pkg/tcpip/buffer/BUILD | 3 +- pkg/tcpip/checker/BUILD | 2 +- pkg/tcpip/header/BUILD | 3 +- pkg/tcpip/link/channel/BUILD | 2 +- pkg/tcpip/link/fdbased/BUILD | 2 +- pkg/tcpip/link/loopback/BUILD | 2 +- pkg/tcpip/link/rawfile/BUILD | 2 +- pkg/tcpip/link/sharedmem/BUILD | 2 +- pkg/tcpip/link/sharedmem/pipe/BUILD | 2 +- pkg/tcpip/link/sharedmem/queue/BUILD | 2 +- pkg/tcpip/link/sniffer/BUILD | 2 +- pkg/tcpip/link/tun/BUILD | 2 +- pkg/tcpip/link/waitable/BUILD | 2 +- pkg/tcpip/network/BUILD | 2 +- pkg/tcpip/network/arp/BUILD | 2 +- pkg/tcpip/network/fragmentation/BUILD | 3 +- pkg/tcpip/network/hash/BUILD | 2 +- pkg/tcpip/network/ipv4/BUILD | 2 +- pkg/tcpip/network/ipv6/BUILD | 2 +- pkg/tcpip/ports/BUILD | 2 +- pkg/tcpip/seqnum/BUILD | 3 +- pkg/tcpip/stack/BUILD | 15 +----- pkg/tcpip/stack/registration.go | 2 + pkg/tcpip/transport/ping/BUILD | 3 +- pkg/tcpip/transport/queue/BUILD | 3 +- pkg/tcpip/transport/tcp/BUILD | 3 +- pkg/tcpip/transport/tcp/testing/context/BUILD | 2 +- pkg/tcpip/transport/tcpconntrack/BUILD | 2 +- pkg/tcpip/transport/udp/BUILD | 3 +- pkg/tcpip/transport/unix/BUILD | 3 +- pkg/tmutex/BUILD | 2 +- pkg/unet/BUILD | 2 +- pkg/urpc/BUILD | 2 +- pkg/waiter/BUILD | 3 +- pkg/waiter/fdnotifier/BUILD | 2 +- tools/go_stateify/defs.bzl | 58 ++++++++++++++++++----- tools/go_stateify/main.go | 66 +++++++++++++++++++++------ 152 files changed, 255 insertions(+), 267 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/BUILD b/pkg/abi/BUILD index 4d507161f..f1e6bac67 100644 --- a/pkg/abi/BUILD +++ b/pkg/abi/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "abi_state", diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index ae7e4378c..38b4829c9 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -4,8 +4,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "linux_state", diff --git a/pkg/amutex/BUILD b/pkg/amutex/BUILD index 442096319..84e6b79a5 100644 --- a/pkg/amutex/BUILD +++ b/pkg/amutex/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "amutex", diff --git a/pkg/atomicbitops/BUILD b/pkg/atomicbitops/BUILD index f20a9f855..a8dd17825 100644 --- a/pkg/atomicbitops/BUILD +++ b/pkg/atomicbitops/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "atomicbitops", diff --git a/pkg/binary/BUILD b/pkg/binary/BUILD index 16f08b13f..586d05634 100644 --- a/pkg/binary/BUILD +++ b/pkg/binary/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "binary", diff --git a/pkg/bits/BUILD b/pkg/bits/BUILD index 9897e5dc3..8c943b615 100644 --- a/pkg/bits/BUILD +++ b/pkg/bits/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") go_library( diff --git a/pkg/bpf/BUILD b/pkg/bpf/BUILD index d4f12f13a..403270049 100644 --- a/pkg/bpf/BUILD +++ b/pkg/bpf/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "bpf_state", diff --git a/pkg/compressio/BUILD b/pkg/compressio/BUILD index 721b2d983..d70f982c1 100644 --- a/pkg/compressio/BUILD +++ b/pkg/compressio/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "compressio", diff --git a/pkg/control/client/BUILD b/pkg/control/client/BUILD index 9e1c058e4..d58cd1b71 100644 --- a/pkg/control/client/BUILD +++ b/pkg/control/client/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "client", diff --git a/pkg/control/server/BUILD b/pkg/control/server/BUILD index 2d0fdd8b8..c3f74a532 100644 --- a/pkg/control/server/BUILD +++ b/pkg/control/server/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "server", diff --git a/pkg/cpuid/BUILD b/pkg/cpuid/BUILD index a503b7ae8..9a0ca1b33 100644 --- a/pkg/cpuid/BUILD +++ b/pkg/cpuid/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "cpuid_state", diff --git a/pkg/dhcp/BUILD b/pkg/dhcp/BUILD index f56969ad8..bd9f592b4 100644 --- a/pkg/dhcp/BUILD +++ b/pkg/dhcp/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "dhcp", diff --git a/pkg/eventchannel/BUILD b/pkg/eventchannel/BUILD index ea0c587be..ac2ea869d 100644 --- a/pkg/eventchannel/BUILD +++ b/pkg/eventchannel/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 +load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("@io_bazel_rules_go//go:def.bzl", "go_library") go_library( name = "eventchannel", diff --git a/pkg/fd/BUILD b/pkg/fd/BUILD index e69d83d06..435b6fa34 100644 --- a/pkg/fd/BUILD +++ b/pkg/fd/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "fd", diff --git a/pkg/gate/BUILD b/pkg/gate/BUILD index 0b8b01da8..872eff531 100644 --- a/pkg/gate/BUILD +++ b/pkg/gate/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "gate", diff --git a/pkg/hashio/BUILD b/pkg/hashio/BUILD index aaa58b58f..5736e2e73 100644 --- a/pkg/hashio/BUILD +++ b/pkg/hashio/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "hashio", diff --git a/pkg/ilist/BUILD b/pkg/ilist/BUILD index 16a738e89..e32f26ffa 100644 --- a/pkg/ilist/BUILD +++ b/pkg/ilist/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "list_state", diff --git a/pkg/linewriter/BUILD b/pkg/linewriter/BUILD index 4a96c6f1d..6c3795432 100644 --- a/pkg/linewriter/BUILD +++ b/pkg/linewriter/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "linewriter", diff --git a/pkg/log/BUILD b/pkg/log/BUILD index 2530cfd18..fc9281079 100644 --- a/pkg/log/BUILD +++ b/pkg/log/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "log", diff --git a/pkg/metric/BUILD b/pkg/metric/BUILD index e3f50d528..c0cd40c7b 100644 --- a/pkg/metric/BUILD +++ b/pkg/metric/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") go_library( name = "metric", diff --git a/pkg/p9/BUILD b/pkg/p9/BUILD index f348ff2e9..1cf5c6458 100644 --- a/pkg/p9/BUILD +++ b/pkg/p9/BUILD @@ -1,4 +1,4 @@ -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") package( default_visibility = ["//visibility:public"], diff --git a/pkg/p9/p9test/BUILD b/pkg/p9/p9test/BUILD index 339c86089..d6f428e11 100644 --- a/pkg/p9/p9test/BUILD +++ b/pkg/p9/p9test/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_test( name = "p9test_test", diff --git a/pkg/rand/BUILD b/pkg/rand/BUILD index 2bb59f895..12e6cf25a 100644 --- a/pkg/rand/BUILD +++ b/pkg/rand/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "rand", diff --git a/pkg/refs/BUILD b/pkg/refs/BUILD index 4b7c9345d..3ea877ccf 100644 --- a/pkg/refs/BUILD +++ b/pkg/refs/BUILD @@ -1,32 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") - -go_stateify( - name = "refs_state", - srcs = [ - "refcounter.go", - "refcounter_state.go", - ], - out = "refs_state.go", - package = "refs", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "refs", srcs = [ "refcounter.go", "refcounter_state.go", - "refs_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/refs", visibility = ["//:sandbox"], - deps = [ - "//pkg/ilist", - "//pkg/log", - "//pkg/state", - ], + deps = ["//pkg/ilist"], ) go_test( diff --git a/pkg/refs/refcounter.go b/pkg/refs/refcounter.go index 3162001e1..0d44c2499 100644 --- a/pkg/refs/refcounter.go +++ b/pkg/refs/refcounter.go @@ -58,6 +58,8 @@ type WeakRefUser interface { } // WeakRef is a weak reference. +// +// +stateify savable type WeakRef struct { ilist.Entry `state:"nosave"` @@ -177,6 +179,8 @@ func (w *WeakRef) zap() { // // N.B. To allow the zero-object to be initialized, the count is offset by // 1, that is, when refCount is n, there are really n+1 references. +// +// +stateify savable type AtomicRefCount struct { // refCount is composed of two fields: // diff --git a/pkg/refs/refcounter_state.go b/pkg/refs/refcounter_state.go index 1be67f951..093eae785 100644 --- a/pkg/refs/refcounter_state.go +++ b/pkg/refs/refcounter_state.go @@ -14,6 +14,7 @@ package refs +// +stateify savable type savedReference struct { obj interface{} } diff --git a/pkg/seccomp/BUILD b/pkg/seccomp/BUILD index cadd24505..b3e2f0b38 100644 --- a/pkg/seccomp/BUILD +++ b/pkg/seccomp/BUILD @@ -1,6 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_binary", "go_embed_data", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("@io_bazel_rules_go//go:def.bzl", "go_binary", "go_embed_data") go_binary( name = "victim", diff --git a/pkg/secio/BUILD b/pkg/secio/BUILD index 9a28d2c1f..0ed38c64a 100644 --- a/pkg/secio/BUILD +++ b/pkg/secio/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "secio", diff --git a/pkg/segment/test/BUILD b/pkg/segment/test/BUILD index 9d398d71a..bdf53e24e 100644 --- a/pkg/segment/test/BUILD +++ b/pkg/segment/test/BUILD @@ -1,4 +1,4 @@ -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") package( default_visibility = ["//visibility:private"], diff --git a/pkg/sentry/arch/BUILD b/pkg/sentry/arch/BUILD index a88f57ac7..0a2a35400 100644 --- a/pkg/sentry/arch/BUILD +++ b/pkg/sentry/arch/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "arch_state", diff --git a/pkg/sentry/context/BUILD b/pkg/sentry/context/BUILD index ff39f94ba..2a7a6df23 100644 --- a/pkg/sentry/context/BUILD +++ b/pkg/sentry/context/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "context", diff --git a/pkg/sentry/context/contexttest/BUILD b/pkg/sentry/context/contexttest/BUILD index 5977344de..591b11a4d 100644 --- a/pkg/sentry/context/contexttest/BUILD +++ b/pkg/sentry/context/contexttest/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "contexttest_state", diff --git a/pkg/sentry/control/BUILD b/pkg/sentry/control/BUILD index 6169891f7..fbdde0721 100644 --- a/pkg/sentry/control/BUILD +++ b/pkg/sentry/control/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "control", diff --git a/pkg/sentry/device/BUILD b/pkg/sentry/device/BUILD index 1a8b461ba..69c99b0b3 100644 --- a/pkg/sentry/device/BUILD +++ b/pkg/sentry/device/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "device", diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 9b7264753..e3c9a9b70 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "fs_state", diff --git a/pkg/sentry/fs/anon/BUILD b/pkg/sentry/fs/anon/BUILD index 6b18aee47..ff4ab850a 100644 --- a/pkg/sentry/fs/anon/BUILD +++ b/pkg/sentry/fs/anon/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "anon", diff --git a/pkg/sentry/fs/ashmem/BUILD b/pkg/sentry/fs/ashmem/BUILD index e20e22a0f..9f166799a 100644 --- a/pkg/sentry/fs/ashmem/BUILD +++ b/pkg/sentry/fs/ashmem/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_stateify( diff --git a/pkg/sentry/fs/binder/BUILD b/pkg/sentry/fs/binder/BUILD index 15f91699f..ec3928baf 100644 --- a/pkg/sentry/fs/binder/BUILD +++ b/pkg/sentry/fs/binder/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "binder_state", diff --git a/pkg/sentry/fs/dev/BUILD b/pkg/sentry/fs/dev/BUILD index d33a19c2f..ea41615fd 100644 --- a/pkg/sentry/fs/dev/BUILD +++ b/pkg/sentry/fs/dev/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "dev_state", diff --git a/pkg/sentry/fs/fdpipe/BUILD b/pkg/sentry/fs/fdpipe/BUILD index 9e1f65d3e..4fcb06f1f 100644 --- a/pkg/sentry/fs/fdpipe/BUILD +++ b/pkg/sentry/fs/fdpipe/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "pipe_state", diff --git a/pkg/sentry/fs/filetest/BUILD b/pkg/sentry/fs/filetest/BUILD index 51a390d77..f481c57fb 100644 --- a/pkg/sentry/fs/filetest/BUILD +++ b/pkg/sentry/fs/filetest/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "filetest_state", diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index 4fa6395f7..6eea64298 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "fsutil_state", diff --git a/pkg/sentry/fs/gofer/BUILD b/pkg/sentry/fs/gofer/BUILD index e6f659c53..1277379e7 100644 --- a/pkg/sentry/fs/gofer/BUILD +++ b/pkg/sentry/fs/gofer/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "gofer_state", diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 97b64daed..23ec66f50 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "host_state", diff --git a/pkg/sentry/fs/lock/BUILD b/pkg/sentry/fs/lock/BUILD index c15dde800..2607d7ed3 100644 --- a/pkg/sentry/fs/lock/BUILD +++ b/pkg/sentry/fs/lock/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "lock_state", diff --git a/pkg/sentry/fs/proc/BUILD b/pkg/sentry/fs/proc/BUILD index 21b5fc0c3..870df47b2 100644 --- a/pkg/sentry/fs/proc/BUILD +++ b/pkg/sentry/fs/proc/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "proc_state", diff --git a/pkg/sentry/fs/proc/device/BUILD b/pkg/sentry/fs/proc/device/BUILD index b62062bd7..34582f275 100644 --- a/pkg/sentry/fs/proc/device/BUILD +++ b/pkg/sentry/fs/proc/device/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "device", diff --git a/pkg/sentry/fs/proc/seqfile/BUILD b/pkg/sentry/fs/proc/seqfile/BUILD index 48dd25e5b..c84f7e20d 100644 --- a/pkg/sentry/fs/proc/seqfile/BUILD +++ b/pkg/sentry/fs/proc/seqfile/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "seqfile_state", diff --git a/pkg/sentry/fs/ramfs/BUILD b/pkg/sentry/fs/ramfs/BUILD index 663a1aeb9..d84f2c624 100644 --- a/pkg/sentry/fs/ramfs/BUILD +++ b/pkg/sentry/fs/ramfs/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "ramfs_state", diff --git a/pkg/sentry/fs/ramfs/test/BUILD b/pkg/sentry/fs/ramfs/test/BUILD index 074b0f5ad..57fee45e2 100644 --- a/pkg/sentry/fs/ramfs/test/BUILD +++ b/pkg/sentry/fs/ramfs/test/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "test_state", diff --git a/pkg/sentry/fs/sys/BUILD b/pkg/sentry/fs/sys/BUILD index 0ae2cbac8..095ff1f25 100644 --- a/pkg/sentry/fs/sys/BUILD +++ b/pkg/sentry/fs/sys/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "sys_state", diff --git a/pkg/sentry/fs/timerfd/BUILD b/pkg/sentry/fs/timerfd/BUILD index 7fddc29f4..8b1b7872e 100644 --- a/pkg/sentry/fs/timerfd/BUILD +++ b/pkg/sentry/fs/timerfd/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "timerfd_state", diff --git a/pkg/sentry/fs/tmpfs/BUILD b/pkg/sentry/fs/tmpfs/BUILD index be4e695d3..473ab4296 100644 --- a/pkg/sentry/fs/tmpfs/BUILD +++ b/pkg/sentry/fs/tmpfs/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "tmpfs_state", diff --git a/pkg/sentry/fs/tty/BUILD b/pkg/sentry/fs/tty/BUILD index fce327dfe..363897b2c 100644 --- a/pkg/sentry/fs/tty/BUILD +++ b/pkg/sentry/fs/tty/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "tty_state", diff --git a/pkg/sentry/hostcpu/BUILD b/pkg/sentry/hostcpu/BUILD index 9457618d8..f362d15c8 100644 --- a/pkg/sentry/hostcpu/BUILD +++ b/pkg/sentry/hostcpu/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "hostcpu", diff --git a/pkg/sentry/inet/BUILD b/pkg/sentry/inet/BUILD index 1150ced57..eaf8f15b2 100644 --- a/pkg/sentry/inet/BUILD +++ b/pkg/sentry/inet/BUILD @@ -1,11 +1,9 @@ -load("@io_bazel_rules_go//go:def.bzl", "go_library") - package( default_visibility = ["//:sandbox"], licenses = ["notice"], # Apache 2.0 ) -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "inet_state", diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 07568b47c..c4a7dacb2 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "kernel_state", diff --git a/pkg/sentry/kernel/auth/BUILD b/pkg/sentry/kernel/auth/BUILD index 7f0680b88..5b7b30557 100644 --- a/pkg/sentry/kernel/auth/BUILD +++ b/pkg/sentry/kernel/auth/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "auth_state", diff --git a/pkg/sentry/kernel/epoll/BUILD b/pkg/sentry/kernel/epoll/BUILD index 04651d961..7d491efbc 100644 --- a/pkg/sentry/kernel/epoll/BUILD +++ b/pkg/sentry/kernel/epoll/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "epoll_autogen_state", diff --git a/pkg/sentry/kernel/eventfd/BUILD b/pkg/sentry/kernel/eventfd/BUILD index 561ced852..7ec179bd8 100644 --- a/pkg/sentry/kernel/eventfd/BUILD +++ b/pkg/sentry/kernel/eventfd/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "eventfd_state", diff --git a/pkg/sentry/kernel/fasync/BUILD b/pkg/sentry/kernel/fasync/BUILD index 8d06e1182..17749c0de 100644 --- a/pkg/sentry/kernel/fasync/BUILD +++ b/pkg/sentry/kernel/fasync/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "fasync", diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index de9897c58..a97a43549 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_template_instance( name = "waiter_list", diff --git a/pkg/sentry/kernel/kdefs/BUILD b/pkg/sentry/kernel/kdefs/BUILD index b6c00042a..fe6fa2260 100644 --- a/pkg/sentry/kernel/kdefs/BUILD +++ b/pkg/sentry/kernel/kdefs/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "kdefs", diff --git a/pkg/sentry/kernel/memevent/BUILD b/pkg/sentry/kernel/memevent/BUILD index c7779e1d5..66899910c 100644 --- a/pkg/sentry/kernel/memevent/BUILD +++ b/pkg/sentry/kernel/memevent/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 +load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("@io_bazel_rules_go//go:def.bzl", "go_library") go_library( name = "memevent", diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD index ca9825f9d..4600d19bd 100644 --- a/pkg/sentry/kernel/pipe/BUILD +++ b/pkg/sentry/kernel/pipe/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "pipe_state", diff --git a/pkg/sentry/kernel/sched/BUILD b/pkg/sentry/kernel/sched/BUILD index b533c51c4..125792f39 100644 --- a/pkg/sentry/kernel/sched/BUILD +++ b/pkg/sentry/kernel/sched/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "sched", diff --git a/pkg/sentry/kernel/semaphore/BUILD b/pkg/sentry/kernel/semaphore/BUILD index 1656ad126..969145fe1 100644 --- a/pkg/sentry/kernel/semaphore/BUILD +++ b/pkg/sentry/kernel/semaphore/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_template_instance( name = "waiter_list", diff --git a/pkg/sentry/kernel/shm/BUILD b/pkg/sentry/kernel/shm/BUILD index 182cc1c76..0f88eb0ac 100644 --- a/pkg/sentry/kernel/shm/BUILD +++ b/pkg/sentry/kernel/shm/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "shm_state", diff --git a/pkg/sentry/kernel/time/BUILD b/pkg/sentry/kernel/time/BUILD index 84f31b2dc..b3ed42aa4 100644 --- a/pkg/sentry/kernel/time/BUILD +++ b/pkg/sentry/kernel/time/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "time_state", diff --git a/pkg/sentry/limits/BUILD b/pkg/sentry/limits/BUILD index 06c3e72b0..3ce41cacc 100644 --- a/pkg/sentry/limits/BUILD +++ b/pkg/sentry/limits/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "limits_state", diff --git a/pkg/sentry/loader/BUILD b/pkg/sentry/loader/BUILD index 01a0ec426..e63052c6d 100644 --- a/pkg/sentry/loader/BUILD +++ b/pkg/sentry/loader/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_embed_data", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("@io_bazel_rules_go//go:def.bzl", "go_embed_data") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_embed_data( name = "vdso_bin", diff --git a/pkg/sentry/memmap/BUILD b/pkg/sentry/memmap/BUILD index 7525fea45..2e367e189 100644 --- a/pkg/sentry/memmap/BUILD +++ b/pkg/sentry/memmap/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "memmap_state", diff --git a/pkg/sentry/memutil/BUILD b/pkg/sentry/memutil/BUILD index a387a0c9f..341b30b98 100644 --- a/pkg/sentry/memutil/BUILD +++ b/pkg/sentry/memutil/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "memutil", diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index 258389bb2..3f396986a 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "mm_state", diff --git a/pkg/sentry/platform/BUILD b/pkg/sentry/platform/BUILD index d5be81f8d..15a7fbbc3 100644 --- a/pkg/sentry/platform/BUILD +++ b/pkg/sentry/platform/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "platform_state", diff --git a/pkg/sentry/platform/filemem/BUILD b/pkg/sentry/platform/filemem/BUILD index 3c4d5b0b6..dadba1d38 100644 --- a/pkg/sentry/platform/filemem/BUILD +++ b/pkg/sentry/platform/filemem/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "filemem_autogen_state", diff --git a/pkg/sentry/platform/interrupt/BUILD b/pkg/sentry/platform/interrupt/BUILD index 33dde2a31..35121321a 100644 --- a/pkg/sentry/platform/interrupt/BUILD +++ b/pkg/sentry/platform/interrupt/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "interrupt", diff --git a/pkg/sentry/platform/kvm/BUILD b/pkg/sentry/platform/kvm/BUILD index 673393fad..4ef9e20d7 100644 --- a/pkg/sentry/platform/kvm/BUILD +++ b/pkg/sentry/platform/kvm/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/sentry/platform/kvm/testutil/BUILD b/pkg/sentry/platform/kvm/testutil/BUILD index 8533a8d89..e779e3893 100644 --- a/pkg/sentry/platform/kvm/testutil/BUILD +++ b/pkg/sentry/platform/kvm/testutil/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "testutil", diff --git a/pkg/sentry/platform/procid/BUILD b/pkg/sentry/platform/procid/BUILD index 5db4f6261..ba68d48f4 100644 --- a/pkg/sentry/platform/procid/BUILD +++ b/pkg/sentry/platform/procid/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "procid", diff --git a/pkg/sentry/platform/ptrace/BUILD b/pkg/sentry/platform/ptrace/BUILD index 16b0b3c69..ceee895dc 100644 --- a/pkg/sentry/platform/ptrace/BUILD +++ b/pkg/sentry/platform/ptrace/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "ptrace", diff --git a/pkg/sentry/platform/ring0/BUILD b/pkg/sentry/platform/ring0/BUILD index 2df232a64..2485eb2eb 100644 --- a/pkg/sentry/platform/ring0/BUILD +++ b/pkg/sentry/platform/ring0/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") go_template( diff --git a/pkg/sentry/platform/ring0/pagetables/BUILD b/pkg/sentry/platform/ring0/pagetables/BUILD index 023e298a0..7a86e2234 100644 --- a/pkg/sentry/platform/ring0/pagetables/BUILD +++ b/pkg/sentry/platform/ring0/pagetables/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") go_template( diff --git a/pkg/sentry/platform/safecopy/BUILD b/pkg/sentry/platform/safecopy/BUILD index 8b9f29403..7dcf6e561 100644 --- a/pkg/sentry/platform/safecopy/BUILD +++ b/pkg/sentry/platform/safecopy/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "safecopy", diff --git a/pkg/sentry/safemem/BUILD b/pkg/sentry/safemem/BUILD index dc4cfce41..e96509ce1 100644 --- a/pkg/sentry/safemem/BUILD +++ b/pkg/sentry/safemem/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "safemem", diff --git a/pkg/sentry/sighandling/BUILD b/pkg/sentry/sighandling/BUILD index daaad7c90..f480f0735 100644 --- a/pkg/sentry/sighandling/BUILD +++ b/pkg/sentry/sighandling/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "sighandling", diff --git a/pkg/sentry/socket/BUILD b/pkg/sentry/socket/BUILD index 5500a676e..929787aa0 100644 --- a/pkg/sentry/socket/BUILD +++ b/pkg/sentry/socket/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "socket_state", diff --git a/pkg/sentry/socket/control/BUILD b/pkg/sentry/socket/control/BUILD index 25de2f655..faf2b4c27 100644 --- a/pkg/sentry/socket/control/BUILD +++ b/pkg/sentry/socket/control/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "control_state", diff --git a/pkg/sentry/socket/epsocket/BUILD b/pkg/sentry/socket/epsocket/BUILD index 8430886cb..7ad5e88c5 100644 --- a/pkg/sentry/socket/epsocket/BUILD +++ b/pkg/sentry/socket/epsocket/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "epsocket_state", diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD index 60ec265ba..227ca3926 100644 --- a/pkg/sentry/socket/hostinet/BUILD +++ b/pkg/sentry/socket/hostinet/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "hostinet_state", diff --git a/pkg/sentry/socket/netlink/BUILD b/pkg/sentry/socket/netlink/BUILD index 9df3ab17c..b23a243f7 100644 --- a/pkg/sentry/socket/netlink/BUILD +++ b/pkg/sentry/socket/netlink/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "netlink_state", diff --git a/pkg/sentry/socket/netlink/port/BUILD b/pkg/sentry/socket/netlink/port/BUILD index 7340b95c9..ba6f686e4 100644 --- a/pkg/sentry/socket/netlink/port/BUILD +++ b/pkg/sentry/socket/netlink/port/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "port_state", diff --git a/pkg/sentry/socket/netlink/route/BUILD b/pkg/sentry/socket/netlink/route/BUILD index ff3f7b7a4..726469fc9 100644 --- a/pkg/sentry/socket/netlink/route/BUILD +++ b/pkg/sentry/socket/netlink/route/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "route_state", diff --git a/pkg/sentry/socket/rpcinet/BUILD b/pkg/sentry/socket/rpcinet/BUILD index 8973453f9..288199779 100644 --- a/pkg/sentry/socket/rpcinet/BUILD +++ b/pkg/sentry/socket/rpcinet/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 +load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("@io_bazel_rules_go//go:def.bzl", "go_library") go_library( name = "rpcinet", diff --git a/pkg/sentry/socket/rpcinet/conn/BUILD b/pkg/sentry/socket/rpcinet/conn/BUILD index 4923dee4b..c51ca14b1 100644 --- a/pkg/sentry/socket/rpcinet/conn/BUILD +++ b/pkg/sentry/socket/rpcinet/conn/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # BSD -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "conn", diff --git a/pkg/sentry/socket/rpcinet/notifier/BUILD b/pkg/sentry/socket/rpcinet/notifier/BUILD index 6f3b06a05..2ae902b3f 100644 --- a/pkg/sentry/socket/rpcinet/notifier/BUILD +++ b/pkg/sentry/socket/rpcinet/notifier/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # BSD -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "notifier", diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index 1ec6eb7ed..7d04d6b6b 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "unix_state", diff --git a/pkg/sentry/state/BUILD b/pkg/sentry/state/BUILD index 9bd98f445..a57a8298e 100644 --- a/pkg/sentry/state/BUILD +++ b/pkg/sentry/state/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "state", diff --git a/pkg/sentry/strace/BUILD b/pkg/sentry/strace/BUILD index c5946a564..e1c8db67a 100644 --- a/pkg/sentry/strace/BUILD +++ b/pkg/sentry/strace/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 +load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("@io_bazel_rules_go//go:def.bzl", "go_library") go_library( name = "strace", diff --git a/pkg/sentry/syscalls/BUILD b/pkg/sentry/syscalls/BUILD index d667b42c8..22a757095 100644 --- a/pkg/sentry/syscalls/BUILD +++ b/pkg/sentry/syscalls/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 +load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("@io_bazel_rules_go//go:def.bzl", "go_library") go_library( name = "syscalls", diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index d3f3cc459..574621ad2 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "linux_state", diff --git a/pkg/sentry/time/BUILD b/pkg/sentry/time/BUILD index cbcd699d5..9452787fb 100644 --- a/pkg/sentry/time/BUILD +++ b/pkg/sentry/time/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/sentry/uniqueid/BUILD b/pkg/sentry/uniqueid/BUILD index c8ab03c3d..8eba3609e 100644 --- a/pkg/sentry/uniqueid/BUILD +++ b/pkg/sentry/uniqueid/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "uniqueid", diff --git a/pkg/sentry/usage/BUILD b/pkg/sentry/usage/BUILD index a0fe0aa07..edee44d96 100644 --- a/pkg/sentry/usage/BUILD +++ b/pkg/sentry/usage/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "usage_state", diff --git a/pkg/sentry/usermem/BUILD b/pkg/sentry/usermem/BUILD index 36c0760dd..9dd1cd2b5 100644 --- a/pkg/sentry/usermem/BUILD +++ b/pkg/sentry/usermem/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "usermem_state", diff --git a/pkg/sentry/watchdog/BUILD b/pkg/sentry/watchdog/BUILD index 28fae4490..13bc33eb1 100644 --- a/pkg/sentry/watchdog/BUILD +++ b/pkg/sentry/watchdog/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "watchdog", diff --git a/pkg/sleep/BUILD b/pkg/sleep/BUILD index f2b69b225..05e4ca540 100644 --- a/pkg/sleep/BUILD +++ b/pkg/sleep/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "sleep", diff --git a/pkg/state/BUILD b/pkg/state/BUILD index bb6415d9b..012b0484e 100644 --- a/pkg/state/BUILD +++ b/pkg/state/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/state/statefile/BUILD b/pkg/state/statefile/BUILD index df2c6a578..16abe1930 100644 --- a/pkg/state/statefile/BUILD +++ b/pkg/state/statefile/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "statefile", diff --git a/pkg/sync/BUILD b/pkg/sync/BUILD index 1fc0c25b5..3959fea36 100644 --- a/pkg/sync/BUILD +++ b/pkg/sync/BUILD @@ -1,4 +1,4 @@ -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") package( default_visibility = ["//:sandbox"], diff --git a/pkg/sync/seqatomictest/BUILD b/pkg/sync/seqatomictest/BUILD index 9d6ee2dfb..07b4f85ab 100644 --- a/pkg/sync/seqatomictest/BUILD +++ b/pkg/sync/seqatomictest/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/syserr/BUILD b/pkg/syserr/BUILD index e5ce48412..c0850f3d9 100644 --- a/pkg/syserr/BUILD +++ b/pkg/syserr/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "syserr", diff --git a/pkg/syserror/BUILD b/pkg/syserror/BUILD index 68ddec786..e050c2043 100644 --- a/pkg/syserror/BUILD +++ b/pkg/syserror/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "syserror", diff --git a/pkg/tcpip/BUILD b/pkg/tcpip/BUILD index 186a0d3bf..391d801d0 100644 --- a/pkg/tcpip/BUILD +++ b/pkg/tcpip/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "tcpip_state", diff --git a/pkg/tcpip/adapters/gonet/BUILD b/pkg/tcpip/adapters/gonet/BUILD index 97da46776..bf618831a 100644 --- a/pkg/tcpip/adapters/gonet/BUILD +++ b/pkg/tcpip/adapters/gonet/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "gonet", diff --git a/pkg/tcpip/buffer/BUILD b/pkg/tcpip/buffer/BUILD index 08adf18cd..efeb6a448 100644 --- a/pkg/tcpip/buffer/BUILD +++ b/pkg/tcpip/buffer/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "buffer_state", diff --git a/pkg/tcpip/checker/BUILD b/pkg/tcpip/checker/BUILD index 5447cfbf4..e8a524918 100644 --- a/pkg/tcpip/checker/BUILD +++ b/pkg/tcpip/checker/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "checker", diff --git a/pkg/tcpip/header/BUILD b/pkg/tcpip/header/BUILD index 859c2a106..3aa2cfb24 100644 --- a/pkg/tcpip/header/BUILD +++ b/pkg/tcpip/header/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "tcp_header_state", diff --git a/pkg/tcpip/link/channel/BUILD b/pkg/tcpip/link/channel/BUILD index f2f0c8b6f..9a6f49c45 100644 --- a/pkg/tcpip/link/channel/BUILD +++ b/pkg/tcpip/link/channel/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "channel", diff --git a/pkg/tcpip/link/fdbased/BUILD b/pkg/tcpip/link/fdbased/BUILD index aca3b14ca..6e75e9f47 100644 --- a/pkg/tcpip/link/fdbased/BUILD +++ b/pkg/tcpip/link/fdbased/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "fdbased", diff --git a/pkg/tcpip/link/loopback/BUILD b/pkg/tcpip/link/loopback/BUILD index 9714e93db..cc4247ffd 100644 --- a/pkg/tcpip/link/loopback/BUILD +++ b/pkg/tcpip/link/loopback/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "loopback", diff --git a/pkg/tcpip/link/rawfile/BUILD b/pkg/tcpip/link/rawfile/BUILD index 4b30c7c1c..10b35a37e 100644 --- a/pkg/tcpip/link/rawfile/BUILD +++ b/pkg/tcpip/link/rawfile/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "rawfile", diff --git a/pkg/tcpip/link/sharedmem/BUILD b/pkg/tcpip/link/sharedmem/BUILD index 1bd79a3f4..5390257c5 100644 --- a/pkg/tcpip/link/sharedmem/BUILD +++ b/pkg/tcpip/link/sharedmem/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "sharedmem", diff --git a/pkg/tcpip/link/sharedmem/pipe/BUILD b/pkg/tcpip/link/sharedmem/pipe/BUILD index e6c658071..ff798ae6f 100644 --- a/pkg/tcpip/link/sharedmem/pipe/BUILD +++ b/pkg/tcpip/link/sharedmem/pipe/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "pipe", diff --git a/pkg/tcpip/link/sharedmem/queue/BUILD b/pkg/tcpip/link/sharedmem/queue/BUILD index 80cedade1..c4a7879c4 100644 --- a/pkg/tcpip/link/sharedmem/queue/BUILD +++ b/pkg/tcpip/link/sharedmem/queue/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "queue", diff --git a/pkg/tcpip/link/sniffer/BUILD b/pkg/tcpip/link/sniffer/BUILD index d14f150d1..1e844f949 100644 --- a/pkg/tcpip/link/sniffer/BUILD +++ b/pkg/tcpip/link/sniffer/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "sniffer", diff --git a/pkg/tcpip/link/tun/BUILD b/pkg/tcpip/link/tun/BUILD index 21da7d57e..a8bb03661 100644 --- a/pkg/tcpip/link/tun/BUILD +++ b/pkg/tcpip/link/tun/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "tun", diff --git a/pkg/tcpip/link/waitable/BUILD b/pkg/tcpip/link/waitable/BUILD index 3b513383a..7582df32e 100644 --- a/pkg/tcpip/link/waitable/BUILD +++ b/pkg/tcpip/link/waitable/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "waitable", diff --git a/pkg/tcpip/network/BUILD b/pkg/tcpip/network/BUILD index 963857f51..9a26b46c4 100644 --- a/pkg/tcpip/network/BUILD +++ b/pkg/tcpip/network/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_test") +load("//tools/go_stateify:defs.bzl", "go_test") go_test( name = "ip_test", diff --git a/pkg/tcpip/network/arp/BUILD b/pkg/tcpip/network/arp/BUILD index 689f66d6e..44f2b66e5 100644 --- a/pkg/tcpip/network/arp/BUILD +++ b/pkg/tcpip/network/arp/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "arp", diff --git a/pkg/tcpip/network/fragmentation/BUILD b/pkg/tcpip/network/fragmentation/BUILD index a173f87fb..ac97ebe43 100644 --- a/pkg/tcpip/network/fragmentation/BUILD +++ b/pkg/tcpip/network/fragmentation/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "fragmentation_state", diff --git a/pkg/tcpip/network/hash/BUILD b/pkg/tcpip/network/hash/BUILD index e1b5f26c4..1c22c52fc 100644 --- a/pkg/tcpip/network/hash/BUILD +++ b/pkg/tcpip/network/hash/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "hash", diff --git a/pkg/tcpip/network/ipv4/BUILD b/pkg/tcpip/network/ipv4/BUILD index ae42b662f..19314e9bd 100644 --- a/pkg/tcpip/network/ipv4/BUILD +++ b/pkg/tcpip/network/ipv4/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "ipv4", diff --git a/pkg/tcpip/network/ipv6/BUILD b/pkg/tcpip/network/ipv6/BUILD index d008ac7fb..1c3eccae0 100644 --- a/pkg/tcpip/network/ipv6/BUILD +++ b/pkg/tcpip/network/ipv6/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "ipv6", diff --git a/pkg/tcpip/ports/BUILD b/pkg/tcpip/ports/BUILD index 710c283f7..3c3374275 100644 --- a/pkg/tcpip/ports/BUILD +++ b/pkg/tcpip/ports/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "ports", diff --git a/pkg/tcpip/seqnum/BUILD b/pkg/tcpip/seqnum/BUILD index 6d28dbc3f..a75869dac 100644 --- a/pkg/tcpip/seqnum/BUILD +++ b/pkg/tcpip/seqnum/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "seqnum_state", diff --git a/pkg/tcpip/stack/BUILD b/pkg/tcpip/stack/BUILD index 6d201d0a2..5e7355135 100644 --- a/pkg/tcpip/stack/BUILD +++ b/pkg/tcpip/stack/BUILD @@ -1,17 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") - -go_stateify( - name = "stack_state", - srcs = [ - "registration.go", - "stack.go", - ], - out = "stack_state.go", - package = "stack", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "stack", @@ -22,7 +11,6 @@ go_library( "route.go", "stack.go", "stack_global_state.go", - "stack_state.go", "transport_demuxer.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/stack", @@ -32,7 +20,6 @@ go_library( deps = [ "//pkg/ilist", "//pkg/sleep", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/stack/registration.go b/pkg/tcpip/stack/registration.go index e9550a062..c66f925a8 100644 --- a/pkg/tcpip/stack/registration.go +++ b/pkg/tcpip/stack/registration.go @@ -31,6 +31,8 @@ type NetworkEndpointID struct { } // TransportEndpointID is the identifier of a transport layer protocol endpoint. +// +// +stateify savable type TransportEndpointID struct { // LocalPort is the local port associated with the endpoint. LocalPort uint16 diff --git a/pkg/tcpip/transport/ping/BUILD b/pkg/tcpip/transport/ping/BUILD index 1febbf7f5..28e3e1700 100644 --- a/pkg/tcpip/transport/ping/BUILD +++ b/pkg/tcpip/transport/ping/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "ping_state", diff --git a/pkg/tcpip/transport/queue/BUILD b/pkg/tcpip/transport/queue/BUILD index 7e8ee1f66..fb878ad36 100644 --- a/pkg/tcpip/transport/queue/BUILD +++ b/pkg/tcpip/transport/queue/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "queue_state", diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index 53623787d..6a7153e4d 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "tcp_state", diff --git a/pkg/tcpip/transport/tcp/testing/context/BUILD b/pkg/tcpip/transport/tcp/testing/context/BUILD index 3caa38bcb..7a95594ef 100644 --- a/pkg/tcpip/transport/tcp/testing/context/BUILD +++ b/pkg/tcpip/transport/tcp/testing/context/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "context", diff --git a/pkg/tcpip/transport/tcpconntrack/BUILD b/pkg/tcpip/transport/tcpconntrack/BUILD index 3d748528e..46da3e6f1 100644 --- a/pkg/tcpip/transport/tcpconntrack/BUILD +++ b/pkg/tcpip/transport/tcpconntrack/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tcpconntrack", diff --git a/pkg/tcpip/transport/udp/BUILD b/pkg/tcpip/transport/udp/BUILD index 4f7a47973..790dd55a3 100644 --- a/pkg/tcpip/transport/udp/BUILD +++ b/pkg/tcpip/transport/udp/BUILD @@ -1,8 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "udp_state", diff --git a/pkg/tcpip/transport/unix/BUILD b/pkg/tcpip/transport/unix/BUILD index d58f06544..676f2cf92 100644 --- a/pkg/tcpip/transport/unix/BUILD +++ b/pkg/tcpip/transport/unix/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_stateify( name = "unix_state", diff --git a/pkg/tmutex/BUILD b/pkg/tmutex/BUILD index d9a2c5ae5..d18338fff 100644 --- a/pkg/tmutex/BUILD +++ b/pkg/tmutex/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tmutex", diff --git a/pkg/unet/BUILD b/pkg/unet/BUILD index e8e40315a..acdfd7cb6 100644 --- a/pkg/unet/BUILD +++ b/pkg/unet/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "unet", diff --git a/pkg/urpc/BUILD b/pkg/urpc/BUILD index b29b25637..d32c57d1a 100644 --- a/pkg/urpc/BUILD +++ b/pkg/urpc/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "urpc", diff --git a/pkg/waiter/BUILD b/pkg/waiter/BUILD index 032ec3237..8256acdb4 100644 --- a/pkg/waiter/BUILD +++ b/pkg/waiter/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -load("//tools/go_stateify:defs.bzl", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_stateify( name = "waiter_state", diff --git a/pkg/waiter/fdnotifier/BUILD b/pkg/waiter/fdnotifier/BUILD index d5b5ee82d..4e582755d 100644 --- a/pkg/waiter/fdnotifier/BUILD +++ b/pkg/waiter/fdnotifier/BUILD @@ -1,6 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "fdnotifier", diff --git a/tools/go_stateify/defs.bzl b/tools/go_stateify/defs.bzl index 60a9895ff..2b2582b7a 100644 --- a/tools/go_stateify/defs.bzl +++ b/tools/go_stateify/defs.bzl @@ -22,6 +22,8 @@ go_library( ) """ +load("@io_bazel_rules_go//go:def.bzl", _go_library = "go_library", _go_test = "go_test") + def _go_stateify_impl(ctx): """Implementation for the stateify tool.""" output = ctx.outputs.out @@ -33,6 +35,8 @@ def _go_stateify_impl(ctx): args += ["-statepkg=%s" % ctx.attr._statepkg] if ctx.attr.imports: args += ["-imports=%s" % ",".join(ctx.attr.imports)] + if ctx.attr.explicit: + args += ["-explicit=true"] args += ["--"] for src in ctx.attr.srcs: args += [f.path for f in src.files] @@ -45,17 +49,15 @@ def _go_stateify_impl(ctx): executable = ctx.executable._tool, ) -""" -Generates save and restore logic from a set of Go files. - - -Args: - name: the name of the rule. - srcs: the input source files. These files should include all structs in the package that need to be saved. - imports: an optional list of extra non-aliased, Go-style absolute import paths. - out: the name of the generated file output. This must not conflict with any other files and must be added to the srcs of the relevant go_library. - package: the package name for the input sources. -""" +# Generates save and restore logic from a set of Go files. +# +# Args: +# name: the name of the rule. +# srcs: the input source files. These files should include all structs in the package that need to be saved. +# imports: an optional list of extra non-aliased, Go-style absolute import paths. +# out: the name of the generated file output. This must not conflict with any other files and must be added to the srcs of the relevant go_library. +# package: the package name for the input sources. +# explicit: only generate for types explicitly annotated as savable. go_stateify = rule( implementation = _go_stateify_impl, attrs = { @@ -63,7 +65,41 @@ go_stateify = rule( "imports": attr.string_list(mandatory = False), "package": attr.string(mandatory = True), "out": attr.output(mandatory = True), + "explicit": attr.bool(default = False), "_tool": attr.label(executable = True, cfg = "host", default = Label("//tools/go_stateify:stateify")), "_statepkg": attr.string(default = "gvisor.googlesource.com/gvisor/pkg/state"), }, ) + +def go_library(name, srcs, deps = [], imports = [], **kwargs): + """wraps the standard go_library and does stateification.""" + if "encode_unsafe.go" not in srcs and (name + "_state_autogen.go") not in srcs: + # Only do stateification for non-state packages without manual autogen. + go_stateify( + name = name + "_state_autogen", + srcs = [src for src in srcs if src.endswith(".go")], + imports = imports, + package = name, + out = name + "_state_autogen.go", + explicit = True, + ) + all_srcs = srcs + [name + "_state_autogen.go"] + if "//pkg/state" not in deps: + all_deps = deps + ["//pkg/state"] + else: + all_deps = deps + else: + all_deps = deps + all_srcs = srcs + _go_library( + name = name, + srcs = all_srcs, + deps = all_deps, + **kwargs + ) + +def go_test(**kwargs): + """Wraps the standard go_test.""" + _go_test( + **kwargs + ) diff --git a/tools/go_stateify/main.go b/tools/go_stateify/main.go index 6c3583c62..231c6d80b 100644 --- a/tools/go_stateify/main.go +++ b/tools/go_stateify/main.go @@ -25,6 +25,7 @@ import ( "os" "reflect" "strings" + "sync" ) var ( @@ -32,6 +33,7 @@ var ( imports = flag.String("imports", "", "extra imports for the output file") output = flag.String("output", "", "output file") statePkg = flag.String("statepkg", "", "state import package; defaults to empty") + explicit = flag.Bool("explicit", false, "only generate for types explicitly tagged '// +stateify savable'") ) // resolveTypeName returns a qualified type name. @@ -224,16 +226,24 @@ func main() { // Emit the package name. fmt.Fprint(outputFile, "// automatically generated by stateify.\n\n") fmt.Fprintf(outputFile, "package %s\n\n", *pkg) - fmt.Fprint(outputFile, "import (\n") - if *statePkg != "" { - fmt.Fprintf(outputFile, " \"%s\"\n", *statePkg) - } - if *imports != "" { - for _, i := range strings.Split(*imports, ",") { - fmt.Fprintf(outputFile, " \"%s\"\n", i) - } + + // Emit the imports lazily. + var once sync.Once + maybeEmitImports := func() { + once.Do(func() { + // Emit the imports. + fmt.Fprint(outputFile, "import (\n") + if *statePkg != "" { + fmt.Fprintf(outputFile, " \"%s\"\n", *statePkg) + } + if *imports != "" { + for _, i := range strings.Split(*imports, ",") { + fmt.Fprintf(outputFile, " \"%s\"\n", i) + } + } + fmt.Fprint(outputFile, ")\n\n") + }) } - fmt.Fprint(outputFile, ")\n\n") files := make([]*ast.File, 0, len(flag.Args())) @@ -241,7 +251,7 @@ func main() { for _, filename := range flag.Args() { // Parse the file. fset := token.NewFileSet() - f, err := parser.ParseFile(fset, filename, nil, 0) + f, err := parser.ParseFile(fset, filename, nil, parser.ParseComments) if err != nil { // Not a valid input file? fmt.Fprintf(os.Stderr, "Input %q can't be parsed: %v\n", filename, err) @@ -308,6 +318,26 @@ func main() { continue } + if *explicit { + // In explicit mode, only generate code for + // types explicitly marked + // "// +stateify savable" in one of the + // proceeding comment lines. + if d.Doc == nil { + continue + } + savable := false + for _, l := range d.Doc.List { + if l.Text == "// +stateify savable" { + savable = true + break + } + } + if !savable { + continue + } + } + for _, gs := range d.Specs { ts := gs.(*ast.TypeSpec) switch ts.Type.(type) { @@ -315,6 +345,8 @@ func main() { // Don't register. break case *ast.StructType: + maybeEmitImports() + ss := ts.Type.(*ast.StructType) // Define beforeSave if a definition was not found. This @@ -360,6 +392,8 @@ func main() { // Add to our registration. emitRegister(ts.Name.Name) case *ast.Ident, *ast.SelectorExpr, *ast.ArrayType: + maybeEmitImports() + _, val := resolveTypeName(ts.Name.Name, ts.Type) // Dispatch directly. @@ -377,10 +411,12 @@ func main() { } } - // Emit the init() function. - fmt.Fprintf(outputFile, "func init() {\n") - for _, ic := range initCalls { - fmt.Fprintf(outputFile, " %s\n", ic) + if len(initCalls) > 0 { + // Emit the init() function. + fmt.Fprintf(outputFile, "func init() {\n") + for _, ic := range initCalls { + fmt.Fprintf(outputFile, " %s\n", ic) + } + fmt.Fprintf(outputFile, "}\n") } - fmt.Fprintf(outputFile, "}\n") } -- cgit v1.2.3 From a7a0167716d71895919021692b15bd000f63b24f Mon Sep 17 00:00:00 2001 From: Andrei Vagin Date: Tue, 31 Jul 2018 11:18:02 -0700 Subject: proc: show file flags in fdinfo Currently, there is an attempt to print FD flags, but they are not decoded into a number, so we see something like this: /criu # cat /proc/self/fdinfo/0 flags: {%!o(bool=000false)} Actually, fdinfo has to contain file flags. Change-Id: Idcbb7db908067447eb9ae6f2c3cfb861f2be1a97 PiperOrigin-RevId: 206794498 --- pkg/sentry/fs/flags.go | 39 +++++++++++++++++++++++++++++++ pkg/sentry/fs/proc/fds.go | 21 +++++++++-------- pkg/sentry/kernel/fd_map.go | 16 +++++++++++++ pkg/sentry/syscalls/linux/flags.go | 43 ----------------------------------- pkg/sentry/syscalls/linux/sys_file.go | 4 ++-- 5 files changed, 69 insertions(+), 54 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/flags.go b/pkg/sentry/fs/flags.go index 810452584..da0ff58af 100644 --- a/pkg/sentry/fs/flags.go +++ b/pkg/sentry/fs/flags.go @@ -14,6 +14,10 @@ package fs +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" +) + // FileFlags encodes file flags. type FileFlags struct { // Direct indicates that I/O should be done directly. @@ -78,3 +82,38 @@ func (f FileFlags) Settable() SettableFileFlags { Async: f.Async, } } + +// ToLinux converts a FileFlags object to a Linux representation. +func (f FileFlags) ToLinux() (mask uint) { + if f.Direct { + mask |= linux.O_DIRECT + } + if f.NonBlocking { + mask |= linux.O_NONBLOCK + } + if f.Sync { + mask |= linux.O_SYNC + } + if f.Append { + mask |= linux.O_APPEND + } + if f.Directory { + mask |= linux.O_DIRECTORY + } + if f.Async { + mask |= linux.O_ASYNC + } + if f.LargeFile { + mask |= linux.O_LARGEFILE + } + + switch { + case f.Read && f.Write: + mask |= linux.O_RDWR + case f.Write: + mask |= linux.O_WRONLY + case f.Read: + mask |= linux.O_RDONLY + } + return +} diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index 2eca9ac31..194a9c12a 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -40,16 +40,16 @@ func walkDescriptors(t *kernel.Task, p string, toInode func(*fs.File, kernel.FDF } var file *fs.File - var flags kernel.FDFlags + var fdFlags kernel.FDFlags t.WithMuLocked(func(t *kernel.Task) { if fdm := t.FDMap(); fdm != nil { - file, flags = fdm.GetDescriptor(kdefs.FD(n)) + file, fdFlags = fdm.GetDescriptor(kdefs.FD(n)) } }) if file == nil { return nil, syserror.ENOENT } - return toInode(file, flags), nil + return toInode(file, fdFlags), nil } // readDescriptors reads fds in the task starting at offset, and calls the @@ -200,17 +200,20 @@ func (f *fdDir) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset type fdInfo struct { ramfs.File - flags kernel.FDFlags + flags fs.FileFlags + fdFlags kernel.FDFlags } // newFdInfo returns a new fdInfo based on an existing file. -func newFdInfo(t *kernel.Task, _ *fs.File, flags kernel.FDFlags, msrc *fs.MountSource) *fs.Inode { - fdi := &fdInfo{flags: flags} +func newFdInfo(t *kernel.Task, file *fs.File, fdFlags kernel.FDFlags, msrc *fs.MountSource) *fs.Inode { + fdi := &fdInfo{flags: file.Flags(), fdFlags: fdFlags} fdi.InitFile(t, fs.RootOwner, fs.FilePermissions{User: fs.PermMask{Read: true}}) // TODO: Get pos, locks, and other data. For now we only // have flags. // See https://www.kernel.org/doc/Documentation/filesystems/proc.txt - fdi.Append([]byte(fmt.Sprintf("flags: %08o\n", flags))) + + flags := file.Flags().ToLinux() | fdFlags.ToLinuxFileFlags() + fdi.Append([]byte(fmt.Sprintf("flags:\t0%o\n", flags))) return newFile(fdi, msrc, fs.SpecialFile, t) } @@ -241,8 +244,8 @@ func newFdInfoDir(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { // Lookup loads an fd in /proc/TID/fdinfo into a Dirent. func (fdid *fdInfoDir) Lookup(ctx context.Context, dir *fs.Inode, p string) (*fs.Dirent, error) { - n, err := walkDescriptors(fdid.t, p, func(file *fs.File, flags kernel.FDFlags) *fs.Inode { - return newFdInfo(fdid.t, file, flags, dir.MountSource) + n, err := walkDescriptors(fdid.t, p, func(file *fs.File, fdFlags kernel.FDFlags) *fs.Inode { + return newFdInfo(fdid.t, file, fdFlags, dir.MountSource) }) if err != nil { return nil, err diff --git a/pkg/sentry/kernel/fd_map.go b/pkg/sentry/kernel/fd_map.go index ef73125fd..299506330 100644 --- a/pkg/sentry/kernel/fd_map.go +++ b/pkg/sentry/kernel/fd_map.go @@ -51,6 +51,22 @@ type FDFlags struct { CloseOnExec bool } +// ToLinuxFileFlags converts a kernel.FDFlags object to a Linux file flags representation. +func (f FDFlags) ToLinuxFileFlags() (mask uint) { + if f.CloseOnExec { + mask |= linux.O_CLOEXEC + } + return +} + +// ToLinuxFDFlags converts a kernel.FDFlags object to a Linux descriptor flags representation. +func (f FDFlags) ToLinuxFDFlags() (mask uint) { + if f.CloseOnExec { + mask |= linux.FD_CLOEXEC + } + return +} + // descriptor holds the details about a file descriptor, namely a pointer the // file itself and the descriptor flags. type descriptor struct { diff --git a/pkg/sentry/syscalls/linux/flags.go b/pkg/sentry/syscalls/linux/flags.go index b2e173f3e..e8db3e0de 100644 --- a/pkg/sentry/syscalls/linux/flags.go +++ b/pkg/sentry/syscalls/linux/flags.go @@ -17,7 +17,6 @@ package linux import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" ) // flagsToPermissions returns a Permissions object from Linux flags. @@ -35,48 +34,6 @@ func flagsToPermissions(mask uint) (p fs.PermMask) { return } -// fdFlagsToLinux converts a kernel.FDFlags object to a Linux representation. -func fdFlagsToLinux(flags kernel.FDFlags) (mask uint) { - if flags.CloseOnExec { - mask |= linux.FD_CLOEXEC - } - return -} - -// flagsToLinux converts a FileFlags object to a Linux representation. -func flagsToLinux(flags fs.FileFlags) (mask uint) { - if flags.Direct { - mask |= linux.O_DIRECT - } - if flags.NonBlocking { - mask |= linux.O_NONBLOCK - } - if flags.Sync { - mask |= linux.O_SYNC - } - if flags.Append { - mask |= linux.O_APPEND - } - if flags.Directory { - mask |= linux.O_DIRECTORY - } - if flags.Async { - mask |= linux.O_ASYNC - } - if flags.LargeFile { - mask |= linux.O_LARGEFILE - } - switch { - case flags.Read && flags.Write: - mask |= linux.O_RDWR - case flags.Write: - mask |= linux.O_WRONLY - case flags.Read: - mask |= linux.O_RDONLY - } - return -} - // linuxToFlags converts linux file flags to a FileFlags object. func linuxToFlags(mask uint) (flags fs.FileFlags) { return fs.FileFlags{ diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 2f28fbea6..692f11ed7 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -809,14 +809,14 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall } return uintptr(fd), nil, nil case linux.F_GETFD: - return uintptr(fdFlagsToLinux(flags)), nil, nil + return uintptr(flags.ToLinuxFDFlags()), nil, nil case linux.F_SETFD: flags := args[2].Uint() t.FDMap().SetFlags(fd, kernel.FDFlags{ CloseOnExec: flags&syscall.FD_CLOEXEC != 0, }) case linux.F_GETFL: - return uintptr(flagsToLinux(file.Flags())), nil, nil + return uintptr(file.Flags().ToLinux()), nil, nil case linux.F_SETFL: flags := uint(args[2].Uint()) file.SetFlags(linuxToSettableFlags(flags)) -- cgit v1.2.3 From b9e1cf8404ce1263176643dee1a1cc835c9d1448 Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Wed, 1 Aug 2018 15:42:07 -0700 Subject: stateify: convert all packages to use explicit mode. PiperOrigin-RevId: 207007153 Change-Id: Ifedf1cc3758dc18be16647a4ece9c840c1c636c9 --- pkg/abi/BUILD | 13 +----- pkg/abi/linux/BUILD | 16 +------ pkg/abi/linux/bpf.go | 2 + pkg/abi/linux/tty.go | 2 + pkg/bpf/BUILD | 17 +------- pkg/bpf/interpreter.go | 2 + pkg/cpuid/BUILD | 15 +------ pkg/cpuid/cpuid.go | 2 + pkg/ilist/BUILD | 15 +------ pkg/ilist/list.go | 4 ++ pkg/segment/range.go | 2 + pkg/segment/set.go | 5 +++ pkg/sentry/arch/BUILD | 18 +------- pkg/sentry/arch/arch.go | 2 + pkg/sentry/arch/arch_amd64.go | 2 + pkg/sentry/arch/arch_state_x86.go | 1 + pkg/sentry/arch/arch_x86.go | 2 + pkg/sentry/arch/auxv.go | 2 + pkg/sentry/arch/signal_amd64.go | 6 +++ pkg/sentry/context/contexttest/BUILD | 17 +------- pkg/sentry/fs/BUILD | 36 +--------------- pkg/sentry/fs/ashmem/BUILD | 17 +------- pkg/sentry/fs/ashmem/area.go | 8 ++-- pkg/sentry/fs/ashmem/device.go | 22 +++++----- pkg/sentry/fs/ashmem/pin_board.go | 2 + pkg/sentry/fs/attr.go | 12 ++++++ pkg/sentry/fs/binder/BUILD | 13 +----- pkg/sentry/fs/binder/binder.go | 26 +++++++----- pkg/sentry/fs/dentry.go | 4 ++ pkg/sentry/fs/dev/BUILD | 20 +-------- pkg/sentry/fs/dev/dev.go | 2 + pkg/sentry/fs/dev/fs.go | 2 + pkg/sentry/fs/dev/full.go | 2 + pkg/sentry/fs/dev/null.go | 3 ++ pkg/sentry/fs/dev/random.go | 1 + pkg/sentry/fs/dirent.go | 2 + pkg/sentry/fs/dirent_cache.go | 2 + pkg/sentry/fs/fdpipe/BUILD | 31 +------------- pkg/sentry/fs/fdpipe/pipe.go | 2 + pkg/sentry/fs/file.go | 2 + pkg/sentry/fs/file_overlay.go | 4 ++ pkg/sentry/fs/filesystems.go | 2 + pkg/sentry/fs/filetest/BUILD | 18 +------- pkg/sentry/fs/flags.go | 2 + pkg/sentry/fs/fsutil/BUILD | 20 +-------- pkg/sentry/fs/fsutil/dirty_set.go | 2 + pkg/sentry/fs/fsutil/handle.go | 2 + pkg/sentry/fs/fsutil/host_file_mapper.go | 2 + pkg/sentry/fs/fsutil/inode.go | 6 +++ pkg/sentry/fs/fsutil/inode_cached.go | 2 + pkg/sentry/fs/gofer/BUILD | 23 +--------- pkg/sentry/fs/gofer/file.go | 2 + pkg/sentry/fs/gofer/fs.go | 2 + pkg/sentry/fs/gofer/inode.go | 4 ++ pkg/sentry/fs/gofer/session.go | 3 ++ pkg/sentry/fs/host/BUILD | 27 +----------- pkg/sentry/fs/host/descriptor.go | 2 + pkg/sentry/fs/host/file.go | 2 + pkg/sentry/fs/host/fs.go | 6 ++- pkg/sentry/fs/host/inode.go | 4 ++ pkg/sentry/fs/inode.go | 4 ++ pkg/sentry/fs/inode_inotify.go | 2 + pkg/sentry/fs/inotify.go | 2 + pkg/sentry/fs/inotify_event.go | 2 + pkg/sentry/fs/inotify_watch.go | 2 + pkg/sentry/fs/lock/BUILD | 15 +------ pkg/sentry/fs/lock/lock.go | 6 ++- pkg/sentry/fs/mount.go | 4 ++ pkg/sentry/fs/mount_overlay.go | 4 ++ pkg/sentry/fs/mounts.go | 2 + pkg/sentry/fs/overlay.go | 2 + pkg/sentry/fs/proc/BUILD | 34 +-------------- pkg/sentry/fs/proc/cpuinfo.go | 2 + pkg/sentry/fs/proc/exec_args.go | 2 + pkg/sentry/fs/proc/fds.go | 6 +++ pkg/sentry/fs/proc/file.go | 1 + pkg/sentry/fs/proc/filesystems.go | 2 + pkg/sentry/fs/proc/fs.go | 2 + pkg/sentry/fs/proc/loadavg.go | 2 + pkg/sentry/fs/proc/meminfo.go | 2 + pkg/sentry/fs/proc/mounts.go | 4 ++ pkg/sentry/fs/proc/proc.go | 4 ++ pkg/sentry/fs/proc/seqfile/BUILD | 30 ++----------- pkg/sentry/fs/proc/seqfile/seqfile.go | 4 ++ pkg/sentry/fs/proc/stat.go | 2 + pkg/sentry/fs/proc/sys.go | 5 +++ pkg/sentry/fs/proc/sys_net.go | 2 + pkg/sentry/fs/proc/task.go | 20 +++++++++ pkg/sentry/fs/proc/uid_gid_map.go | 3 ++ pkg/sentry/fs/proc/uptime.go | 2 + pkg/sentry/fs/proc/version.go | 2 + pkg/sentry/fs/ramfs/BUILD | 21 +-------- pkg/sentry/fs/ramfs/dir.go | 2 + pkg/sentry/fs/ramfs/ramfs.go | 2 + pkg/sentry/fs/ramfs/socket.go | 2 + pkg/sentry/fs/ramfs/symlink.go | 2 + pkg/sentry/fs/ramfs/test/BUILD | 18 +------- pkg/sentry/fs/sys/BUILD | 14 +----- pkg/sentry/fs/sys/fs.go | 2 + pkg/sentry/fs/sys/sys.go | 5 ++- pkg/sentry/fs/timerfd/BUILD | 18 +------- pkg/sentry/fs/timerfd/timerfd.go | 4 +- pkg/sentry/fs/tmpfs/BUILD | 17 +------- pkg/sentry/fs/tmpfs/file_regular.go | 2 + pkg/sentry/fs/tmpfs/fs.go | 2 + pkg/sentry/fs/tmpfs/inode_file.go | 2 + pkg/sentry/fs/tmpfs/tmpfs.go | 8 ++++ pkg/sentry/fs/tty/BUILD | 20 +-------- pkg/sentry/fs/tty/dir.go | 18 +++++--- pkg/sentry/fs/tty/fs.go | 4 ++ pkg/sentry/fs/tty/inode.go | 2 + pkg/sentry/fs/tty/line_discipline.go | 6 +++ pkg/sentry/fs/tty/master.go | 4 ++ pkg/sentry/fs/tty/queue.go | 4 +- pkg/sentry/fs/tty/slave.go | 4 ++ pkg/sentry/fs/tty/terminal.go | 2 + pkg/sentry/inet/BUILD | 15 +------ pkg/sentry/inet/inet.go | 2 + pkg/sentry/kernel/BUILD | 59 +++----------------------- pkg/sentry/kernel/abstract_socket_namespace.go | 3 ++ pkg/sentry/kernel/auth/BUILD | 17 +------- pkg/sentry/kernel/auth/credentials.go | 2 + pkg/sentry/kernel/auth/id_map.go | 2 + pkg/sentry/kernel/auth/user_namespace.go | 2 + pkg/sentry/kernel/epoll/BUILD | 15 +------ pkg/sentry/kernel/epoll/epoll.go | 8 +++- pkg/sentry/kernel/eventfd/BUILD | 18 +------- pkg/sentry/kernel/eventfd/eventfd.go | 4 +- pkg/sentry/kernel/fd_map.go | 6 +++ pkg/sentry/kernel/fs_context.go | 2 + pkg/sentry/kernel/futex/BUILD | 18 +------- pkg/sentry/kernel/futex/futex.go | 2 + pkg/sentry/kernel/ipc_namespace.go | 2 + pkg/sentry/kernel/kernel.go | 4 +- pkg/sentry/kernel/pending_signals.go | 5 +++ pkg/sentry/kernel/pipe/BUILD | 20 +-------- pkg/sentry/kernel/pipe/buffers.go | 2 + pkg/sentry/kernel/pipe/node.go | 2 + pkg/sentry/kernel/pipe/pipe.go | 2 + pkg/sentry/kernel/pipe/reader.go | 2 + pkg/sentry/kernel/pipe/reader_writer.go | 2 + pkg/sentry/kernel/pipe/writer.go | 2 + pkg/sentry/kernel/ptrace.go | 4 ++ pkg/sentry/kernel/rseq.go | 2 + pkg/sentry/kernel/semaphore/BUILD | 15 +------ pkg/sentry/kernel/semaphore/semaphore.go | 8 ++++ pkg/sentry/kernel/sessions.go | 4 ++ pkg/sentry/kernel/shm/BUILD | 13 +----- pkg/sentry/kernel/shm/shm.go | 4 ++ pkg/sentry/kernel/signal_handlers.go | 2 + pkg/sentry/kernel/syscalls.go | 2 + pkg/sentry/kernel/syslog.go | 2 + pkg/sentry/kernel/task.go | 2 + pkg/sentry/kernel/task_clone.go | 4 ++ pkg/sentry/kernel/task_context.go | 2 + pkg/sentry/kernel/task_exec.go | 4 ++ pkg/sentry/kernel/task_exit.go | 6 +++ pkg/sentry/kernel/task_resources.go | 2 + pkg/sentry/kernel/task_run.go | 2 + pkg/sentry/kernel/task_sched.go | 2 + pkg/sentry/kernel/task_signals.go | 5 +++ pkg/sentry/kernel/task_syscall.go | 4 ++ pkg/sentry/kernel/thread_group.go | 2 + pkg/sentry/kernel/threads.go | 8 ++++ pkg/sentry/kernel/time/BUILD | 14 +----- pkg/sentry/kernel/time/time.go | 6 +++ pkg/sentry/kernel/timekeeper.go | 2 + pkg/sentry/kernel/timer.go | 8 ++++ pkg/sentry/kernel/uts_namespace.go | 2 + pkg/sentry/kernel/vdso.go | 2 + pkg/sentry/limits/BUILD | 13 +----- pkg/sentry/limits/limits.go | 4 ++ pkg/sentry/loader/BUILD | 15 +------ pkg/sentry/loader/vdso.go | 2 + pkg/sentry/loader/vdso_state.go | 1 + pkg/sentry/memmap/BUILD | 15 +------ pkg/sentry/memmap/mapping_set.go | 2 + pkg/sentry/mm/BUILD | 21 +-------- pkg/sentry/mm/aio_context.go | 8 ++++ pkg/sentry/mm/mm.go | 7 +++ pkg/sentry/mm/special_mappable.go | 2 + pkg/sentry/platform/BUILD | 13 +----- pkg/sentry/platform/filemem/BUILD | 14 +----- pkg/sentry/platform/filemem/filemem.go | 2 + pkg/sentry/socket/BUILD | 17 +------- pkg/sentry/socket/control/BUILD | 23 +++------- pkg/sentry/socket/control/control.go | 4 ++ pkg/sentry/socket/epsocket/BUILD | 16 +------ pkg/sentry/socket/epsocket/epsocket.go | 2 + pkg/sentry/socket/epsocket/stack.go | 2 + pkg/sentry/socket/hostinet/BUILD | 15 +------ pkg/sentry/socket/netlink/BUILD | 13 +----- pkg/sentry/socket/netlink/port/BUILD | 15 +------ pkg/sentry/socket/netlink/port/port.go | 2 + pkg/sentry/socket/netlink/route/BUILD | 17 +------- pkg/sentry/socket/netlink/route/protocol.go | 2 + pkg/sentry/socket/netlink/socket.go | 2 + pkg/sentry/socket/socket.go | 2 + pkg/sentry/socket/unix/BUILD | 13 +----- pkg/sentry/socket/unix/unix.go | 2 + pkg/sentry/syscalls/linux/BUILD | 20 +-------- pkg/sentry/syscalls/linux/sys_aio.go | 2 + pkg/sentry/syscalls/linux/sys_futex.go | 2 + pkg/sentry/syscalls/linux/sys_poll.go | 2 + pkg/sentry/syscalls/linux/sys_time.go | 2 + pkg/sentry/usage/BUILD | 17 +------- pkg/sentry/usage/cpu.go | 2 + pkg/sentry/usage/io.go | 2 + pkg/sentry/usermem/BUILD | 16 +------ pkg/sentry/usermem/access_type.go | 2 + pkg/sentry/usermem/addr.go | 2 + pkg/tcpip/BUILD | 17 +------- pkg/tcpip/buffer/BUILD | 13 +----- pkg/tcpip/buffer/view.go | 2 + pkg/tcpip/header/BUILD | 13 +----- pkg/tcpip/header/tcp.go | 4 ++ pkg/tcpip/network/fragmentation/BUILD | 11 +---- pkg/tcpip/seqnum/BUILD | 17 +------- pkg/tcpip/tcpip.go | 4 ++ pkg/tcpip/transport/ping/BUILD | 17 +------- pkg/tcpip/transport/ping/endpoint.go | 1 + pkg/tcpip/transport/queue/BUILD | 17 +------- pkg/tcpip/transport/queue/queue.go | 2 + pkg/tcpip/transport/tcp/BUILD | 25 +---------- pkg/tcpip/transport/tcp/endpoint.go | 4 ++ pkg/tcpip/transport/tcp/rcv.go | 2 + pkg/tcpip/transport/tcp/reno.go | 2 + pkg/tcpip/transport/tcp/segment.go | 2 + pkg/tcpip/transport/tcp/segment_queue.go | 2 + pkg/tcpip/transport/tcp/snd.go | 4 ++ pkg/tcpip/transport/tcp/snd_state.go | 1 + pkg/tcpip/transport/udp/BUILD | 17 +------- pkg/tcpip/transport/udp/endpoint.go | 3 ++ pkg/tcpip/transport/unix/BUILD | 16 +------ pkg/tcpip/transport/unix/connectioned.go | 2 + pkg/tcpip/transport/unix/connectionless.go | 2 + pkg/tcpip/transport/unix/unix.go | 11 +++++ pkg/waiter/BUILD | 21 ++------- pkg/waiter/waiter.go | 2 + 239 files changed, 662 insertions(+), 1108 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/BUILD b/pkg/abi/BUILD index f1e6bac67..c014d2c4b 100644 --- a/pkg/abi/BUILD +++ b/pkg/abi/BUILD @@ -1,24 +1,13 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "abi_state", - srcs = [ - "abi.go", - ], - out = "abi_state.go", - package = "abi", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "abi", srcs = [ "abi.go", - "abi_state.go", "flag.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/abi", visibility = ["//:sandbox"], - deps = ["//pkg/state"], ) diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index 38b4829c9..ac4ceefbc 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -4,19 +4,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "linux_state", - srcs = [ - "binder.go", - "bpf.go", - "time.go", - "tty.go", - ], - out = "linux_state.go", - package = "linux", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "linux", @@ -41,7 +29,6 @@ go_library( "ipc.go", "limits.go", "linux.go", - "linux_state.go", "mm.go", "netdevice.go", "netlink.go", @@ -67,6 +54,5 @@ go_library( "//pkg/abi", "//pkg/binary", "//pkg/bits", - "//pkg/state", ], ) diff --git a/pkg/abi/linux/bpf.go b/pkg/abi/linux/bpf.go index f597ef4f5..80e5b1af1 100644 --- a/pkg/abi/linux/bpf.go +++ b/pkg/abi/linux/bpf.go @@ -15,6 +15,8 @@ package linux // BPFInstruction is a raw BPF virtual machine instruction. +// +// +stateify savable type BPFInstruction struct { // OpCode is the operation to execute. OpCode uint16 diff --git a/pkg/abi/linux/tty.go b/pkg/abi/linux/tty.go index 84b6ccc87..b640f7627 100644 --- a/pkg/abi/linux/tty.go +++ b/pkg/abi/linux/tty.go @@ -38,6 +38,8 @@ type Termios struct { // KernelTermios is struct ktermios/struct termios2, defined in // uapi/asm-generic/termbits.h. +// +// +stateify savable type KernelTermios struct { InputFlags uint32 OutputFlags uint32 diff --git a/pkg/bpf/BUILD b/pkg/bpf/BUILD index 403270049..564df3af5 100644 --- a/pkg/bpf/BUILD +++ b/pkg/bpf/BUILD @@ -1,21 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "bpf_state", - srcs = [ - "interpreter.go", - ], - out = "bpf_state.go", - package = "bpf", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "bpf", srcs = [ "bpf.go", - "bpf_state.go", "decoder.go", "input_bytes.go", "interpreter.go", @@ -23,10 +13,7 @@ go_library( ], importpath = "gvisor.googlesource.com/gvisor/pkg/bpf", visibility = ["//visibility:public"], - deps = [ - "//pkg/abi/linux", - "//pkg/state", - ], + deps = ["//pkg/abi/linux"], ) go_test( diff --git a/pkg/bpf/interpreter.go b/pkg/bpf/interpreter.go index b7dee86a8..111ada9d1 100644 --- a/pkg/bpf/interpreter.go +++ b/pkg/bpf/interpreter.go @@ -88,6 +88,8 @@ func (e Error) Error() string { } // Program is a BPF program that has been validated for consistency. +// +// +stateify savable type Program struct { instructions []linux.BPFInstruction } diff --git a/pkg/cpuid/BUILD b/pkg/cpuid/BUILD index 9a0ca1b33..46fc4703b 100644 --- a/pkg/cpuid/BUILD +++ b/pkg/cpuid/BUILD @@ -1,27 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "cpuid_state", - srcs = ["cpuid.go"], - out = "cpuid_state.go", - package = "cpuid", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "cpuid", srcs = [ "cpu_amd64.s", "cpuid.go", - "cpuid_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/cpuid", visibility = ["//:sandbox"], - deps = [ - "//pkg/log", - "//pkg/state", - ], + deps = ["//pkg/log"], ) go_test( diff --git a/pkg/cpuid/cpuid.go b/pkg/cpuid/cpuid.go index b486ab037..e91e34dc7 100644 --- a/pkg/cpuid/cpuid.go +++ b/pkg/cpuid/cpuid.go @@ -409,6 +409,8 @@ func (f Feature) flagString(cpuinfoOnly bool) string { } // FeatureSet is a set of Features for a cpu. +// +// +stateify savable type FeatureSet struct { // Set is the set of features that are enabled in this FeatureSet. Set map[Feature]bool diff --git a/pkg/ilist/BUILD b/pkg/ilist/BUILD index e32f26ffa..b26a39132 100644 --- a/pkg/ilist/BUILD +++ b/pkg/ilist/BUILD @@ -1,28 +1,15 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "list_state", - srcs = [ - "interface_list.go", - ], - out = "interface_list_state.go", - package = "ilist", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "ilist", srcs = [ "interface_list.go", - "interface_list_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/ilist", visibility = ["//visibility:public"], - deps = [ - "//pkg/state", - ], ) go_template_instance( diff --git a/pkg/ilist/list.go b/pkg/ilist/list.go index 5efb6c072..a88b82196 100644 --- a/pkg/ilist/list.go +++ b/pkg/ilist/list.go @@ -36,6 +36,8 @@ type Linker interface { // for e := l.Front(); e != nil; e = e.Next() { // // do something with e. // } +// +// +stateify savable type List struct { head Linker tail Linker @@ -155,6 +157,8 @@ func (l *List) Remove(e Linker) { // Entry is a default implementation of Linker. Users can add anonymous fields // of this type to their structs to make them automatically implement the // methods needed by List. +// +// +stateify savable type Entry struct { next Linker prev Linker diff --git a/pkg/segment/range.go b/pkg/segment/range.go index 5ff30d489..34c067265 100644 --- a/pkg/segment/range.go +++ b/pkg/segment/range.go @@ -18,6 +18,8 @@ package segment type T uint64 // A Range represents a contiguous range of T. +// +// +stateify savable type Range struct { // Start is the inclusive start of the range. Start T diff --git a/pkg/segment/set.go b/pkg/segment/set.go index 6eed1d930..cffec2a2c 100644 --- a/pkg/segment/set.go +++ b/pkg/segment/set.go @@ -88,6 +88,8 @@ const ( // A Set is a mapping of segments with non-overlapping Range keys. The zero // value for a Set is an empty set. Set values are not safely movable nor // copyable. Set is thread-compatible. +// +// +stateify savable type Set struct { root node `state:".(*SegmentDataSlices)"` } @@ -596,6 +598,7 @@ func (s *Set) ApplyContiguous(r Range, fn func(seg Iterator)) GapIterator { } } +// +stateify savable type node struct { // An internal binary tree node looks like: // @@ -1317,6 +1320,8 @@ func (n *node) writeDebugString(buf *bytes.Buffer, prefix string) { // SegmentDataSlices represents segments from a set as slices of start, end, and // values. SegmentDataSlices is primarily used as an intermediate representation // for save/restore and the layout here is optimized for that. +// +// +stateify savable type SegmentDataSlices struct { Start []Key End []Key diff --git a/pkg/sentry/arch/BUILD b/pkg/sentry/arch/BUILD index 0a2a35400..314b3e962 100644 --- a/pkg/sentry/arch/BUILD +++ b/pkg/sentry/arch/BUILD @@ -1,21 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "arch_state", - srcs = [ - "arch.go", - "arch_amd64.go", - "arch_state_x86.go", - "arch_x86.go", - "auxv.go", - "signal_amd64.go", - ], - out = "arch_state.go", - package = "arch", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "arch", @@ -24,7 +10,6 @@ go_library( "arch.go", "arch_amd64.go", "arch_amd64.s", - "arch_state.go", "arch_state_x86.go", "arch_x86.go", "auxv.go", @@ -46,7 +31,6 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/limits", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go index 0189e958d..21cb84502 100644 --- a/pkg/sentry/arch/arch.go +++ b/pkg/sentry/arch/arch.go @@ -254,6 +254,8 @@ const ( // MemoryManager. // // Note that "highest address" below is always exclusive. +// +// +stateify savable type MmapLayout struct { // MinAddr is the lowest mappable address. MinAddr usermem.Addr diff --git a/pkg/sentry/arch/arch_amd64.go b/pkg/sentry/arch/arch_amd64.go index 23526fe8e..f1e408af9 100644 --- a/pkg/sentry/arch/arch_amd64.go +++ b/pkg/sentry/arch/arch_amd64.go @@ -95,6 +95,8 @@ const ( ) // context64 represents an AMD64 context. +// +// +stateify savable type context64 struct { State sigFPState []x86FPState // fpstate to be restored on sigreturn. diff --git a/pkg/sentry/arch/arch_state_x86.go b/pkg/sentry/arch/arch_state_x86.go index cb38d098a..e9c23a06b 100644 --- a/pkg/sentry/arch/arch_state_x86.go +++ b/pkg/sentry/arch/arch_state_x86.go @@ -56,6 +56,7 @@ func (s *State) afterLoad() { copy(s.x86FPState, old) } +// +stateify savable type syscallPtraceRegs struct { R15 uint64 R14 uint64 diff --git a/pkg/sentry/arch/arch_x86.go b/pkg/sentry/arch/arch_x86.go index 5cc4f8377..b35eec53c 100644 --- a/pkg/sentry/arch/arch_x86.go +++ b/pkg/sentry/arch/arch_x86.go @@ -153,6 +153,8 @@ func NewFloatingPointData() *FloatingPointData { // State contains the common architecture bits for X86 (the build tag of this // file ensures it's only built on x86). +// +// +stateify savable type State struct { // The system registers. Regs syscall.PtraceRegs `state:".(syscallPtraceRegs)"` diff --git a/pkg/sentry/arch/auxv.go b/pkg/sentry/arch/auxv.go index 70e0e35b7..81cfb4a01 100644 --- a/pkg/sentry/arch/auxv.go +++ b/pkg/sentry/arch/auxv.go @@ -19,6 +19,8 @@ import ( ) // An AuxEntry represents an entry in an ELF auxiliary vector. +// +// +stateify savable type AuxEntry struct { Key uint64 Value usermem.Addr diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index c1d743f38..e81717e8b 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -28,6 +28,8 @@ import ( // SignalAct represents the action that should be taken when a signal is // delivered, and is equivalent to struct sigaction on 64-bit x86. +// +// +stateify savable type SignalAct struct { Handler uint64 Flags uint64 @@ -47,6 +49,8 @@ func (s *SignalAct) DeserializeTo(other *SignalAct) { // SignalStack represents information about a user stack, and is equivalent to // stack_t on 64-bit x86. +// +// +stateify savable type SignalStack struct { Addr uint64 Flags uint32 @@ -66,6 +70,8 @@ func (s *SignalStack) DeserializeTo(other *SignalStack) { // SignalInfo represents information about a signal being delivered, and is // equivalent to struct siginfo on 64-bit x86. +// +// +stateify savable type SignalInfo struct { Signo int32 // Signal number Errno int32 // Errno value diff --git a/pkg/sentry/context/contexttest/BUILD b/pkg/sentry/context/contexttest/BUILD index 591b11a4d..01bb40b04 100644 --- a/pkg/sentry/context/contexttest/BUILD +++ b/pkg/sentry/context/contexttest/BUILD @@ -1,23 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "contexttest_state", - srcs = [ - "contexttest.go", - ], - out = "contexttest_state.go", - package = "contexttest", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "contexttest", testonly = 1, - srcs = [ - "contexttest.go", - "contexttest_state.go", - ], + srcs = ["contexttest.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest", visibility = ["//pkg/sentry:internal"], deps = [ @@ -28,6 +16,5 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/platform/ptrace", "//pkg/sentry/uniqueid", - "//pkg/state", ], ) diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index e3c9a9b70..18cd5ae8e 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -1,40 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "fs_state", - srcs = [ - "attr.go", - "dentry.go", - "dirent.go", - "dirent_cache.go", - "dirent_list.go", - "dirent_state.go", - "file.go", - "file_overlay.go", - "file_state.go", - "filesystems.go", - "flags.go", - "inode.go", - "inode_inotify.go", - "inode_operations.go", - "inode_overlay.go", - "inotify.go", - "inotify_event.go", - "inotify_watch.go", - "mock.go", - "mount.go", - "mount_overlay.go", - "mount_state.go", - "mounts.go", - "overlay.go", - "path.go", - ], - out = "fs_state.go", - package = "fs", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "fs", @@ -54,7 +21,6 @@ go_library( "filesystems.go", "flags.go", "fs.go", - "fs_state.go", "inode.go", "inode_inotify.go", "inode_operations.go", diff --git a/pkg/sentry/fs/ashmem/BUILD b/pkg/sentry/fs/ashmem/BUILD index 9f166799a..dc893d22f 100644 --- a/pkg/sentry/fs/ashmem/BUILD +++ b/pkg/sentry/fs/ashmem/BUILD @@ -1,26 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -go_stateify( - name = "ashmem_state", - srcs = [ - "area.go", - "device.go", - "pin_board.go", - "uint64_range.go", - "uint64_set.go", - ], - out = "ashmem_state.go", - package = "ashmem", -) - go_library( name = "ashmem", srcs = [ "area.go", - "ashmem_state.go", "device.go", "pin_board.go", "uint64_range.go", @@ -41,7 +27,6 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", ], diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index e4f76f0d0..bfd7f2762 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -39,10 +39,12 @@ const ( ) // Area implements fs.FileOperations. +// +// +stateify savable type Area struct { - fsutil.NoFsync - fsutil.DeprecatedFileOperations - fsutil.NotDirReaddir + fsutil.NoFsync `state:"nosave"` + fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.NotDirReaddir `state:"nosave"` ad *Device diff --git a/pkg/sentry/fs/ashmem/device.go b/pkg/sentry/fs/ashmem/device.go index c5b51d4a7..d0986fa11 100644 --- a/pkg/sentry/fs/ashmem/device.go +++ b/pkg/sentry/fs/ashmem/device.go @@ -27,17 +27,19 @@ import ( ) // Device implements fs.InodeOperations. +// +// +stateify savable type Device struct { - fsutil.DeprecatedFileOperations - fsutil.InodeNoExtendedAttributes - fsutil.InodeNotDirectory - fsutil.InodeNotRenameable - fsutil.InodeNotSocket - fsutil.InodeNotSymlink - fsutil.NoFsync - fsutil.NoMappable - fsutil.NoopWriteOut - fsutil.NotDirReaddir + fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotRenameable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.NoFsync `state:"nosave"` + fsutil.NoMappable `state:"nosave"` + fsutil.NoopWriteOut `state:"nosave"` + fsutil.NotDirReaddir `state:"nosave"` mu sync.Mutex `state:"nosave"` unstable fs.UnstableAttr diff --git a/pkg/sentry/fs/ashmem/pin_board.go b/pkg/sentry/fs/ashmem/pin_board.go index c7fb3822c..ecba395a0 100644 --- a/pkg/sentry/fs/ashmem/pin_board.go +++ b/pkg/sentry/fs/ashmem/pin_board.go @@ -56,6 +56,8 @@ func (setFunctions) Split(Range, noValue, uint64) (noValue, noValue) { // segment.Set is used for implementation where segments represent // ranges of pinned bytes, while gaps represent ranges of unpinned // bytes. All ranges are page-aligned. +// +// +stateify savable type PinBoard struct { Set } diff --git a/pkg/sentry/fs/attr.go b/pkg/sentry/fs/attr.go index 56a2ad6f7..4178f18b2 100644 --- a/pkg/sentry/fs/attr.go +++ b/pkg/sentry/fs/attr.go @@ -91,6 +91,8 @@ func (n InodeType) String() string { // StableAttr contains Inode attributes that will be stable throughout the // lifetime of the Inode. +// +// +stateify savable type StableAttr struct { // Type is the InodeType of a InodeOperations. Type InodeType @@ -150,6 +152,8 @@ func IsCharDevice(s StableAttr) bool { // UnstableAttr contains Inode attributes that may change over the lifetime // of the Inode. +// +// +stateify savable type UnstableAttr struct { // Size is the file size in bytes. Size int64 @@ -186,6 +190,8 @@ func WithCurrentTime(ctx context.Context, u UnstableAttr) UnstableAttr { } // AttrMask contains fields to mask StableAttr and UnstableAttr. +// +// +stateify savable type AttrMask struct { Type bool DeviceID bool @@ -227,6 +233,8 @@ func (a AttrMask) Union(b AttrMask) AttrMask { } // PermMask are file access permissions. +// +// +stateify savable type PermMask struct { // Read indicates reading is permitted. Read bool @@ -280,6 +288,8 @@ func (p PermMask) SupersetOf(other PermMask) bool { // FilePermissions represents the permissions of a file, with // Read/Write/Execute bits for user, group, and other. +// +// +stateify savable type FilePermissions struct { User PermMask Group PermMask @@ -370,6 +380,8 @@ func (f FilePermissions) AnyRead() bool { } // FileOwner represents ownership of a file. +// +// +stateify savable type FileOwner struct { UID auth.KUID GID auth.KGID diff --git a/pkg/sentry/fs/binder/BUILD b/pkg/sentry/fs/binder/BUILD index ec3928baf..a077b91d2 100644 --- a/pkg/sentry/fs/binder/BUILD +++ b/pkg/sentry/fs/binder/BUILD @@ -1,25 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "binder_state", - srcs = ["binder.go"], - out = "binder_state.go", - package = "binder", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "binder", srcs = [ "binder.go", - "binder_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/binder", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/log", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", @@ -30,8 +21,6 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", - "//pkg/tcpip/transport/unix", ], ) diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index 3f87b6b08..502a262dd 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -40,15 +40,17 @@ const ( ) // Device implements fs.InodeOperations. +// +// +stateify savable type Device struct { - fsutil.InodeNoExtendedAttributes - fsutil.InodeNotDirectory - fsutil.InodeNotRenameable - fsutil.InodeNotSocket - fsutil.InodeNotSymlink - fsutil.NoMappable - fsutil.NoopWriteOut - fsutil.DeprecatedFileOperations + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotRenameable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.NoMappable `state:"nosave"` + fsutil.NoopWriteOut `state:"nosave"` + fsutil.DeprecatedFileOperations `state:"nosave"` // mu protects unstable. mu sync.Mutex `state:"nosave"` @@ -186,10 +188,12 @@ func (bd *Device) StatFS(context.Context) (fs.Info, error) { } // Proc implements fs.FileOperations and fs.IoctlGetter. +// +// +stateify savable type Proc struct { - fsutil.NoFsync - fsutil.DeprecatedFileOperations - fsutil.NotDirReaddir + fsutil.NoFsync `state:"nosave"` + fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.NotDirReaddir `state:"nosave"` bd *Device task *kernel.Task diff --git a/pkg/sentry/fs/dentry.go b/pkg/sentry/fs/dentry.go index d42e8da81..b347468ff 100644 --- a/pkg/sentry/fs/dentry.go +++ b/pkg/sentry/fs/dentry.go @@ -21,6 +21,8 @@ import ( ) // DentAttr is the metadata of a directory entry. It is a subset of StableAttr. +// +// +stateify savable type DentAttr struct { // Type is the InodeType of an Inode. Type InodeType @@ -154,6 +156,8 @@ func GenericReaddir(ctx *DirCtx, s *SortedDentryMap) (int, error) { } // SortedDentryMap is a sorted map of names and fs.DentAttr entries. +// +// +stateify savable type SortedDentryMap struct { // names is always kept in sorted-order. names []string diff --git a/pkg/sentry/fs/dev/BUILD b/pkg/sentry/fs/dev/BUILD index ea41615fd..fc069bb5f 100644 --- a/pkg/sentry/fs/dev/BUILD +++ b/pkg/sentry/fs/dev/BUILD @@ -1,25 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "dev_state", - srcs = [ - "dev.go", - "fs.go", - "full.go", - "null.go", - "random.go", - ], - out = "dev_state.go", - package = "dev", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "dev", srcs = [ "dev.go", - "dev_state.go", "device.go", "fs.go", "full.go", @@ -30,8 +16,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/amutex", - "//pkg/log", "//pkg/rand", "//pkg/sentry/context", "//pkg/sentry/device", @@ -45,9 +29,7 @@ go_library( "//pkg/sentry/mm", "//pkg/sentry/platform", "//pkg/sentry/safemem", - "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/fs/dev/dev.go b/pkg/sentry/fs/dev/dev.go index 36c61bfc2..3f4f2a40a 100644 --- a/pkg/sentry/fs/dev/dev.go +++ b/pkg/sentry/fs/dev/dev.go @@ -27,6 +27,8 @@ import ( ) // Dev is the root node. +// +// +stateify savable type Dev struct { ramfs.Dir } diff --git a/pkg/sentry/fs/dev/fs.go b/pkg/sentry/fs/dev/fs.go index 3c79f3782..2ae49be4e 100644 --- a/pkg/sentry/fs/dev/fs.go +++ b/pkg/sentry/fs/dev/fs.go @@ -29,6 +29,8 @@ const binderEnabledKey = "binder_enabled" const ashmemEnabledKey = "ashmem_enabled" // filesystem is a devtmpfs. +// +// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index e13eb6c03..492b8eb3a 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -26,6 +26,8 @@ import ( ) // fullDevice is used to implement /dev/full. +// +// +stateify savable type fullDevice struct { ramfs.Entry } diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 66b8ba967..2977c8670 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -29,6 +29,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) +// +stateify savable type nullDevice struct { ramfs.Entry } @@ -54,6 +55,7 @@ func (n *nullDevice) Truncate(context.Context, *fs.Inode, int64) error { return nil } +// +stateify savable type zeroDevice struct { nullDevice } @@ -80,6 +82,7 @@ func (zd *zeroDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.F }), nil } +// +stateify savable type zeroFileOperations struct { fs.FileOperations } diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index 33a045a05..47b76218f 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -24,6 +24,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) +// +stateify savable type randomDevice struct { ramfs.Entry } diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go index f9bf2fba6..4658d044f 100644 --- a/pkg/sentry/fs/dirent.go +++ b/pkg/sentry/fs/dirent.go @@ -81,6 +81,8 @@ var renameMu sync.RWMutex // // Dirents currently do not attempt to free entries that lack application references under // memory pressure. +// +// +stateify savable type Dirent struct { // AtomicRefCount is our reference count. refs.AtomicRefCount diff --git a/pkg/sentry/fs/dirent_cache.go b/pkg/sentry/fs/dirent_cache.go index e786e4f65..c680e4828 100644 --- a/pkg/sentry/fs/dirent_cache.go +++ b/pkg/sentry/fs/dirent_cache.go @@ -25,6 +25,8 @@ import ( // // A nil DirentCache corresponds to a cache with size 0. All methods can be // called, but nothing is actually cached. +// +// +stateify savable type DirentCache struct { // Maximum size of the cache. This must be saved manually, to handle the case // when cache is nil. diff --git a/pkg/sentry/fs/fdpipe/BUILD b/pkg/sentry/fs/fdpipe/BUILD index 4fcb06f1f..ffe4204bc 100644 --- a/pkg/sentry/fs/fdpipe/BUILD +++ b/pkg/sentry/fs/fdpipe/BUILD @@ -1,54 +1,27 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "pipe_state", - srcs = [ - "pipe.go", - "pipe_state.go", - ], - out = "pipe_autogen_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/sentry/fs"], - package = "fdpipe", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "fdpipe", srcs = [ "pipe.go", - "pipe_autogen_state.go", "pipe_opener.go", "pipe_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fdpipe", + imports = ["gvisor.googlesource.com/gvisor/pkg/sentry/fs"], visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/abi/linux", - "//pkg/amutex", "//pkg/fd", "//pkg/log", - "//pkg/metric", - "//pkg/p9", - "//pkg/refs", "//pkg/secio", "//pkg/sentry/context", - "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", - "//pkg/sentry/fs/lock", - "//pkg/sentry/kernel/auth", - "//pkg/sentry/kernel/time", - "//pkg/sentry/memmap", - "//pkg/sentry/platform", "//pkg/sentry/safemem", - "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", - "//pkg/tcpip", - "//pkg/tcpip/transport/unix", - "//pkg/unet", "//pkg/waiter", "//pkg/waiter/fdnotifier", ], diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index 7b318e35f..2e34604e6 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -34,6 +34,8 @@ import ( ) // pipeOperations are the fs.FileOperations of a host pipe. +// +// +stateify savable type pipeOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 6d93ef760..8e535a618 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -47,6 +47,8 @@ const FileMaxOffset = math.MaxInt64 // and write(2). // // FIXME: Split synchronization from cancellation. +// +// +stateify savable type File struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index 36b2cf75e..113962368 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -60,6 +60,8 @@ func overlayFile(ctx context.Context, inode *Inode, flags FileFlags) (*File, err } // overlayFileOperations implements FileOperations for a file in an overlay. +// +// +stateify savable type overlayFileOperations struct { // upperMu protects upper below. In contrast lower is stable. upperMu sync.Mutex `state:"nosave"` @@ -375,6 +377,8 @@ func readdirOne(ctx context.Context, d *Dirent) (map[string]DentAttr, error) { // overlayMappingIdentity wraps a MappingIdentity, and also holds a reference // on a file during its lifetime. +// +// +stateify savable type overlayMappingIdentity struct { refs.AtomicRefCount id memmap.MappingIdentity diff --git a/pkg/sentry/fs/filesystems.go b/pkg/sentry/fs/filesystems.go index 200e792f4..5a1e7a270 100644 --- a/pkg/sentry/fs/filesystems.go +++ b/pkg/sentry/fs/filesystems.go @@ -125,6 +125,8 @@ func GetFilesystems() []Filesystem { } // MountSourceFlags represents all mount option flags as a struct. +// +// +stateify savable type MountSourceFlags struct { // ReadOnly corresponds to mount(2)'s "MS_RDONLY" and indicates that // the filesystem should be mounted read-only. diff --git a/pkg/sentry/fs/filetest/BUILD b/pkg/sentry/fs/filetest/BUILD index f481c57fb..d137fee4c 100644 --- a/pkg/sentry/fs/filetest/BUILD +++ b/pkg/sentry/fs/filetest/BUILD @@ -1,34 +1,20 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "filetest_state", - srcs = [ - "filetest.go", - ], - out = "filetest_state.go", - package = "filetest", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "filetest", testonly = 1, - srcs = [ - "filetest.go", - "filetest_state.go", - ], + srcs = ["filetest.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/filetest", visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/flags.go b/pkg/sentry/fs/flags.go index da0ff58af..1aa271560 100644 --- a/pkg/sentry/fs/flags.go +++ b/pkg/sentry/fs/flags.go @@ -19,6 +19,8 @@ import ( ) // FileFlags encodes file flags. +// +// +stateify savable type FileFlags struct { // Direct indicates that I/O should be done directly. Direct bool diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index 6eea64298..3512bae6f 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -1,24 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "fsutil_state", - srcs = [ - "dirty_set_impl.go", - "file.go", - "file_range_set_impl.go", - "frame_ref_set_impl.go", - "handle.go", - "host_file_mapper.go", - "host_file_mapper_state.go", - "inode.go", - "inode_cached.go", - ], - out = "fsutil_state.go", - package = "fsutil", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "dirty_set_impl", @@ -84,7 +67,6 @@ go_library( "frame_ref_set.go", "frame_ref_set_impl.go", "fsutil.go", - "fsutil_state.go", "handle.go", "host_file_mapper.go", "host_file_mapper_state.go", diff --git a/pkg/sentry/fs/fsutil/dirty_set.go b/pkg/sentry/fs/fsutil/dirty_set.go index 9c6c98542..8e31e48fd 100644 --- a/pkg/sentry/fs/fsutil/dirty_set.go +++ b/pkg/sentry/fs/fsutil/dirty_set.go @@ -32,6 +32,8 @@ import ( // DirtyInfo is the value type of DirtySet, and represents information about a // Mappable offset that is dirty (the cached data for that offset is newer than // its source). +// +// +stateify savable type DirtyInfo struct { // Keep is true if the represented offset is concurrently writable, such // that writing the data for that offset back to the source does not diff --git a/pkg/sentry/fs/fsutil/handle.go b/pkg/sentry/fs/fsutil/handle.go index 149c0f84a..e7efd3c0f 100644 --- a/pkg/sentry/fs/fsutil/handle.go +++ b/pkg/sentry/fs/fsutil/handle.go @@ -27,6 +27,8 @@ import ( // // FIXME: Remove Handle entirely in favor of individual fs.File // implementations using simple generic utilities. +// +// +stateify savable type Handle struct { NoopRelease `state:"nosave"` NoIoctl `state:"nosave"` diff --git a/pkg/sentry/fs/fsutil/host_file_mapper.go b/pkg/sentry/fs/fsutil/host_file_mapper.go index d0a27fc1c..9c1e2f76f 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper.go @@ -29,6 +29,8 @@ import ( // HostFileMapper caches mappings of an arbitrary host file descriptor. It is // used by implementations of memmap.Mappable that represent a host file // descriptor. +// +// +stateify savable type HostFileMapper struct { // HostFile conceptually breaks the file into pieces called chunks, of // size and alignment chunkSize, and caches mappings of the file on a chunk diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index e1ad07df2..177396fdc 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -31,6 +31,8 @@ func NewSimpleInodeOperations(i InodeSimpleAttributes) fs.InodeOperations { } // simpleInodeOperations is a simple implementation of Inode. +// +// +stateify savable type simpleInodeOperations struct { DeprecatedFileOperations `state:"nosave"` InodeNotDirectory `state:"nosave"` @@ -48,6 +50,8 @@ type simpleInodeOperations struct { // InodeSimpleAttributes implements a subset of the Inode interface. It provides // read-only access to attributes. +// +// +stateify savable type InodeSimpleAttributes struct { // FSType is the filesystem type reported by StatFS. FSType uint64 @@ -110,6 +114,8 @@ func (*InodeSimpleAttributes) Truncate(context.Context, *fs.Inode, int64) error // // Users need not initialize Xattrs to non-nil (it will be initialized // when the first extended attribute is set. +// +// +stateify savable type InMemoryAttributes struct { Unstable fs.UnstableAttr Xattrs map[string][]byte diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index cba642a8f..0a320e2d8 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -55,6 +55,8 @@ import ( // // Implementations of InodeOperations.WriteOut must call Sync to write out // in-memory modifications of data and metadata to the CachedFileObject. +// +// +stateify savable type CachingInodeOperations struct { // backingFile is a handle to a cached file object. backingFile CachedFileObject diff --git a/pkg/sentry/fs/gofer/BUILD b/pkg/sentry/fs/gofer/BUILD index 1277379e7..cb17339c9 100644 --- a/pkg/sentry/fs/gofer/BUILD +++ b/pkg/sentry/fs/gofer/BUILD @@ -1,21 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "gofer_state", - srcs = [ - "file.go", - "file_state.go", - "fs.go", - "inode.go", - "inode_state.go", - "session.go", - "session_state.go", - ], - out = "gofer_state.go", - package = "gofer", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "gofer", @@ -27,7 +12,6 @@ go_library( "file.go", "file_state.go", "fs.go", - "gofer_state.go", "handles.go", "inode.go", "inode_state.go", @@ -41,7 +25,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/amutex", "//pkg/fd", "//pkg/log", "//pkg/metric", @@ -54,15 +37,11 @@ go_library( "//pkg/sentry/fs/fdpipe", "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/host", - "//pkg/sentry/fs/lock", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", - "//pkg/sentry/platform", "//pkg/sentry/safemem", - "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index 039618808..46a6bbd5d 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -33,6 +33,8 @@ import ( var openedWX = metric.MustCreateNewUint64Metric("/gofer/opened_write_execute_file", true /* sync */, "Number of times a writable+executable file was opened from a gofer.") // fileOperations implements fs.FileOperations for a remote file system. +// +// +stateify savable type fileOperations struct { fsutil.NoIoctl `state:"nosave"` waiter.AlwaysReady `state:"nosave"` diff --git a/pkg/sentry/fs/gofer/fs.go b/pkg/sentry/fs/gofer/fs.go index dd5d43c47..3ae93f059 100644 --- a/pkg/sentry/fs/gofer/fs.go +++ b/pkg/sentry/fs/gofer/fs.go @@ -83,6 +83,8 @@ var ( ) // filesystem is a 9p client. +// +// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go index df584c382..7fc8f77b0 100644 --- a/pkg/sentry/fs/gofer/inode.go +++ b/pkg/sentry/fs/gofer/inode.go @@ -35,6 +35,8 @@ import ( ) // inodeOperations implements fs.InodeOperations. +// +// +stateify savable type inodeOperations struct { fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` @@ -68,6 +70,8 @@ type inodeOperations struct { // circular load dependency between it and inodeOperations). Even with // lazy loading, this approach defines the dependencies between objects // and the expected load behavior more concretely. +// +// +stateify savable type inodeFileState struct { // s is common file system state for Gofers. s *session `state:"wait"` diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index b6841526a..648a11435 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -27,6 +27,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/unet" ) +// +stateify savable type endpointMap struct { mu sync.RWMutex `state:"nosave"` // TODO: Make map with private unix sockets savable. @@ -63,6 +64,8 @@ func (e *endpointMap) get(key device.MultiDeviceKey) unix.BoundEndpoint { } // session holds state for each 9p session established during sys_mount. +// +// +stateify savable type session struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 23ec66f50..29c79284a 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -1,23 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "host_state", - srcs = [ - "control.go", - "descriptor.go", - "descriptor_state.go", - "file.go", - "fs.go", - "inode.go", - "inode_state.go", - "socket.go", - "socket_state.go", - ], - out = "host_state.go", - package = "host", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "host", @@ -28,7 +11,6 @@ go_library( "device.go", "file.go", "fs.go", - "host_state.go", "inode.go", "inode_state.go", "ioctl_unsafe.go", @@ -42,7 +24,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/amutex", "//pkg/fd", "//pkg/log", "//pkg/refs", @@ -52,20 +33,14 @@ go_library( "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", - "//pkg/sentry/fs/lock", "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", - "//pkg/sentry/platform", "//pkg/sentry/safemem", - "//pkg/sentry/socket", "//pkg/sentry/socket/control", "//pkg/sentry/socket/unix", - "//pkg/sentry/uniqueid", - "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/link/rawfile", diff --git a/pkg/sentry/fs/host/descriptor.go b/pkg/sentry/fs/host/descriptor.go index 613bd06e8..3aee4d11c 100644 --- a/pkg/sentry/fs/host/descriptor.go +++ b/pkg/sentry/fs/host/descriptor.go @@ -25,6 +25,8 @@ import ( ) // descriptor wraps a host fd. +// +// +stateify savable type descriptor struct { // donated is true if the host fd was donated by another process. donated bool diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index bdf844337..f9bef6d93 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -37,6 +37,8 @@ import ( ) // fileOperations implements fs.FileOperations for a host file descriptor. +// +// +stateify savable type fileOperations struct { fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/host/fs.go b/pkg/sentry/fs/host/fs.go index 974700636..e46ae433c 100644 --- a/pkg/sentry/fs/host/fs.go +++ b/pkg/sentry/fs/host/fs.go @@ -51,6 +51,8 @@ const maxTraversals = 10 // to lock down the configurations. This filesystem should only be mounted at root. // // Think twice before exposing this to applications. +// +// +stateify savable type Filesystem struct { // whitelist is a set of host paths to whitelist. paths []string @@ -266,8 +268,10 @@ func newMountSource(ctx context.Context, root string, mounter fs.FileOwner, file } // superOperations implements fs.MountSourceOperations. +// +// +stateify savable type superOperations struct { - fs.SimpleMountSourceOperations `state:"nosave"` + fs.SimpleMountSourceOperations // root is the path of the mount point. All inode mappings // are relative to this root. diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index 226bc5164..761ccde33 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -34,6 +34,8 @@ import ( // inodeOperations implements fs.InodeOperations for an fs.Inodes backed // by a host file descriptor. +// +// +stateify savable type inodeOperations struct { fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` @@ -65,6 +67,8 @@ type inodeOperations struct { // circular load dependency between it and inodeOperations). Even with // lazy loading, this approach defines the dependencies between objects // and the expected load behavior more concretely. +// +// +stateify savable type inodeFileState struct { // Common file system state. mops *superOperations `state:"wait"` diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index 6c8e6f188..d0dbce5dd 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -28,6 +28,8 @@ import ( // Inode is a file system object that can be simultaneously referenced by different // components of the VFS (Dirent, fs.File, etc). +// +// +stateify savable type Inode struct { // AtomicRefCount is our reference count. refs.AtomicRefCount @@ -58,6 +60,8 @@ type Inode struct { // Note that in Linux fcntl(2) and flock(2) locks are _not_ cooperative, because race and // deadlock conditions make merging them prohibitive. We do the same and keep them oblivious // to each other but provide a "context" as a convenient container. +// +// +stateify savable type LockCtx struct { // Posix is a set of POSIX-style regional advisory locks, see fcntl(2). Posix lock.Locks diff --git a/pkg/sentry/fs/inode_inotify.go b/pkg/sentry/fs/inode_inotify.go index 358bbecdf..683140afe 100644 --- a/pkg/sentry/fs/inode_inotify.go +++ b/pkg/sentry/fs/inode_inotify.go @@ -20,6 +20,8 @@ import ( ) // Watches is the collection of inotify watches on an inode. +// +// +stateify savable type Watches struct { // mu protects the fields below. mu sync.RWMutex `state:"nosave"` diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 6f5e8ce5e..2aabdded8 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -34,6 +34,8 @@ import ( // // Lock ordering: // Inotify.mu -> Inode.Watches.mu -> Watch.mu -> Inotify.evMu +// +// +stateify savable type Inotify struct { // Unique identifier for this inotify instance. We don't just reuse the // inotify fd because fds can be duped. These should not be exposed to the diff --git a/pkg/sentry/fs/inotify_event.go b/pkg/sentry/fs/inotify_event.go index 217915ba4..e9b5e0f56 100644 --- a/pkg/sentry/fs/inotify_event.go +++ b/pkg/sentry/fs/inotify_event.go @@ -28,6 +28,8 @@ import ( const inotifyEventBaseSize = 16 // Event represents a struct inotify_event from linux. +// +// +stateify savable type Event struct { ilist.Entry diff --git a/pkg/sentry/fs/inotify_watch.go b/pkg/sentry/fs/inotify_watch.go index 8904ef544..3e1959e83 100644 --- a/pkg/sentry/fs/inotify_watch.go +++ b/pkg/sentry/fs/inotify_watch.go @@ -27,6 +27,8 @@ import ( // holding an extra ref on each dirent known (by inotify) to point to the // inode. These are known as pins. For a full discussion, see // fs/g3doc/inotify.md. +// +// +stateify savable type Watch struct { // Inotify instance which owns this watch. owner *Inotify diff --git a/pkg/sentry/fs/lock/BUILD b/pkg/sentry/fs/lock/BUILD index 2607d7ed3..3159ff1da 100644 --- a/pkg/sentry/fs/lock/BUILD +++ b/pkg/sentry/fs/lock/BUILD @@ -1,18 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "lock_state", - srcs = [ - "lock.go", - "lock_range.go", - "lock_set.go", - ], - out = "lock_state.go", - package = "lock", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "lock_range", @@ -49,13 +38,11 @@ go_library( "lock_range.go", "lock_set.go", "lock_set_functions.go", - "lock_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/lock", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/log", - "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/lock/lock.go b/pkg/sentry/fs/lock/lock.go index 24d54c989..e9b376eb6 100644 --- a/pkg/sentry/fs/lock/lock.go +++ b/pkg/sentry/fs/lock/lock.go @@ -88,6 +88,8 @@ const LockEOF = math.MaxUint64 // // A Lock may be downgraded from a write lock to a read lock only if // the write lock's uid is the same as the read lock. +// +// +stateify savable type Lock struct { // Readers are the set of read lock holders identified by UniqueID. // If len(Readers) > 0 then HasWriter must be false. @@ -103,6 +105,8 @@ type Lock struct { } // Locks is a thread-safe wrapper around a LockSet. +// +// +stateify savable type Locks struct { // mu protects locks below. mu sync.Mutex `state:"nosave"` @@ -111,7 +115,7 @@ type Locks struct { locks LockSet // blockedQueue is the queue of waiters that are waiting on a lock. - blockedQueue waiter.Queue + blockedQueue waiter.Queue `state:"zerovalue"` } // Blocker is the interface used for blocking locks. Passing a nil Blocker diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index eb1897174..4ede767f9 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -101,6 +101,8 @@ func (i InodeMappings) String() string { // (e.g. cannot be mounted at different locations). // // TODO: Move mount-specific information out of MountSource. +// +// +stateify savable type MountSource struct { refs.AtomicRefCount @@ -260,6 +262,8 @@ func NewNonCachingMountSource(filesystem Filesystem, flags MountSourceFlags) *Mo } // SimpleMountSourceOperations implements MountSourceOperations. +// +// +stateify savable type SimpleMountSourceOperations struct { keep bool } diff --git a/pkg/sentry/fs/mount_overlay.go b/pkg/sentry/fs/mount_overlay.go index 1be81e3a1..d135e8a37 100644 --- a/pkg/sentry/fs/mount_overlay.go +++ b/pkg/sentry/fs/mount_overlay.go @@ -18,6 +18,8 @@ import "gvisor.googlesource.com/gvisor/pkg/sentry/context" // overlayMountSourceOperations implements MountSourceOperations for an overlay // mount point. +// +// +stateify savable type overlayMountSourceOperations struct { upper *MountSource lower *MountSource @@ -72,6 +74,8 @@ func (o *overlayMountSourceOperations) Destroy() { } // type overlayFilesystem is the filesystem for overlay mounts. +// +// +stateify savable type overlayFilesystem struct{} // Name implements Filesystem.Name. diff --git a/pkg/sentry/fs/mounts.go b/pkg/sentry/fs/mounts.go index 87da4ee0e..144d3427d 100644 --- a/pkg/sentry/fs/mounts.go +++ b/pkg/sentry/fs/mounts.go @@ -32,6 +32,8 @@ import ( const DefaultTraversalLimit = 10 // MountNamespace defines a collection of mounts. +// +// +stateify savable type MountNamespace struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/overlay.go b/pkg/sentry/fs/overlay.go index 7357d6401..af13dc8c7 100644 --- a/pkg/sentry/fs/overlay.go +++ b/pkg/sentry/fs/overlay.go @@ -145,6 +145,8 @@ func newOverlayInode(ctx context.Context, o *overlayEntry, msrc *MountSource) *I } // overlayEntry is the overlay metadata of an Inode. It implements Mappable. +// +// +stateify savable type overlayEntry struct { // lowerExists is true if an Inode exists for this file in the lower // filesystem. If lowerExists is true, then the overlay must create diff --git a/pkg/sentry/fs/proc/BUILD b/pkg/sentry/fs/proc/BUILD index 870df47b2..2d9f07f2f 100644 --- a/pkg/sentry/fs/proc/BUILD +++ b/pkg/sentry/fs/proc/BUILD @@ -1,32 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "proc_state", - srcs = [ - "cpuinfo.go", - "exec_args.go", - "fds.go", - "file.go", - "filesystems.go", - "fs.go", - "loadavg.go", - "meminfo.go", - "mounts.go", - "net.go", - "proc.go", - "stat.go", - "sys.go", - "sys_net.go", - "task.go", - "uid_gid_map.go", - "uptime.go", - "version.go", - ], - out = "proc_state.go", - package = "proc", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "proc", @@ -42,7 +16,6 @@ go_library( "mounts.go", "net.go", "proc.go", - "proc_state.go", "rpcinet_proc.go", "stat.go", "sys.go", @@ -56,9 +29,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/amutex", - "//pkg/log", - "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/proc/device", @@ -73,8 +43,6 @@ go_library( "//pkg/sentry/socket/rpcinet", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", - "//pkg/syserr", "//pkg/syserror", ], ) diff --git a/pkg/sentry/fs/proc/cpuinfo.go b/pkg/sentry/fs/proc/cpuinfo.go index f80aaa5b1..4dfec03a4 100644 --- a/pkg/sentry/fs/proc/cpuinfo.go +++ b/pkg/sentry/fs/proc/cpuinfo.go @@ -27,6 +27,8 @@ import ( // cpuinfo is a file describing the CPU capabilities. // // Presently cpuinfo never changes, so it doesn't need to be a SeqFile. +// +// +stateify savable type cpuinfo struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/exec_args.go b/pkg/sentry/fs/proc/exec_args.go index 0e1523bf1..a69cbaa0e 100644 --- a/pkg/sentry/fs/proc/exec_args.go +++ b/pkg/sentry/fs/proc/exec_args.go @@ -37,6 +37,8 @@ const ( // execArgFile is a file containing the exec args (either cmdline or environ) // for a given task. +// +// +stateify savable type execArgFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index 194a9c12a..cca8f874c 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -138,6 +138,8 @@ func (f *fd) Close() error { } // fdDir implements /proc/TID/fd. +// +// +stateify savable type fdDir struct { ramfs.Dir @@ -197,6 +199,8 @@ func (f *fdDir) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset } // fdInfo is a single file in /proc/TID/fdinfo/. +// +// +stateify savable type fdInfo struct { ramfs.File @@ -229,6 +233,8 @@ func (*fdInfo) Truncate(ctx context.Context, inode *fs.Inode, size int64) error // fdInfoDir implements /proc/TID/fdinfo. It embeds an fdDir, but overrides // Lookup and Readdir. +// +// +stateify savable type fdInfoDir struct { ramfs.Dir diff --git a/pkg/sentry/fs/proc/file.go b/pkg/sentry/fs/proc/file.go index 9a433cdf8..4b2d08e75 100644 --- a/pkg/sentry/fs/proc/file.go +++ b/pkg/sentry/fs/proc/file.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) +// +stateify savable type file struct { fs.InodeOperations diff --git a/pkg/sentry/fs/proc/filesystems.go b/pkg/sentry/fs/proc/filesystems.go index 37db9cf9c..49b92fd8a 100644 --- a/pkg/sentry/fs/proc/filesystems.go +++ b/pkg/sentry/fs/proc/filesystems.go @@ -24,6 +24,8 @@ import ( ) // filesystemsData backs /proc/filesystems. +// +// +stateify savable type filesystemsData struct{} // NeedsUpdate returns true on the first generation. The set of registered file diff --git a/pkg/sentry/fs/proc/fs.go b/pkg/sentry/fs/proc/fs.go index 3aadd6ac4..061824b8c 100644 --- a/pkg/sentry/fs/proc/fs.go +++ b/pkg/sentry/fs/proc/fs.go @@ -22,6 +22,8 @@ import ( ) // filesystem is a procfs. +// +// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/proc/loadavg.go b/pkg/sentry/fs/proc/loadavg.go index 7583b6ccd..6fac251d2 100644 --- a/pkg/sentry/fs/proc/loadavg.go +++ b/pkg/sentry/fs/proc/loadavg.go @@ -23,6 +23,8 @@ import ( ) // loadavgData backs /proc/loadavg. +// +// +stateify savable type loadavgData struct{} // NeedsUpdate implements seqfile.SeqSource.NeedsUpdate. diff --git a/pkg/sentry/fs/proc/meminfo.go b/pkg/sentry/fs/proc/meminfo.go index 49cb0faed..53dfd59ef 100644 --- a/pkg/sentry/fs/proc/meminfo.go +++ b/pkg/sentry/fs/proc/meminfo.go @@ -26,6 +26,8 @@ import ( ) // meminfoData backs /proc/meminfo. +// +// +stateify savable type meminfoData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/proc/mounts.go b/pkg/sentry/fs/proc/mounts.go index 108432f4e..2b8167c28 100644 --- a/pkg/sentry/fs/proc/mounts.go +++ b/pkg/sentry/fs/proc/mounts.go @@ -71,6 +71,8 @@ func forEachMountSource(t *kernel.Task, fn func(string, *fs.MountSource)) { } // mountInfoFile is used to implement /proc/[pid]/mountinfo. +// +// +stateify savable type mountInfoFile struct { t *kernel.Task } @@ -152,6 +154,8 @@ func (mif *mountInfoFile) ReadSeqFileData(ctx context.Context, handle seqfile.Se } // mountsFile is used to implement /proc/[pid]/mountinfo. +// +// +stateify savable type mountsFile struct { t *kernel.Task } diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index b2a8d639c..07029a7bb 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -33,6 +33,8 @@ import ( ) // proc is a root proc node. +// +// +stateify savable type proc struct { ramfs.Dir @@ -47,6 +49,8 @@ type proc struct { // stubProcFSFile is a file type that can be used to return file contents // which are constant. This file is not writable and will always have mode // 0444. +// +// +stateify savable type stubProcFSFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/seqfile/BUILD b/pkg/sentry/fs/proc/seqfile/BUILD index c84f7e20d..53c475652 100644 --- a/pkg/sentry/fs/proc/seqfile/BUILD +++ b/pkg/sentry/fs/proc/seqfile/BUILD @@ -1,22 +1,10 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "seqfile_state", - srcs = [ - "seqfile.go", - ], - out = "seqfile_state.go", - package = "seqfile", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "seqfile", - srcs = [ - "seqfile.go", - "seqfile_state.go", - ], + srcs = ["seqfile.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile", visibility = ["//pkg/sentry:internal"], deps = [ @@ -26,26 +14,16 @@ go_library( "//pkg/sentry/fs/ramfs", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", ], ) -go_stateify( - name = "seqfile_test_state", - srcs = ["seqfile_test.go"], - out = "seqfile_test_state.go", - package = "seqfile", -) - go_test( name = "seqfile_test", size = "small", - srcs = [ - "seqfile_test.go", - "seqfile_test_state.go", - ], + srcs = ["seqfile_test.go"], embed = [":seqfile"], deps = [ + "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs/test", diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index c08565f8a..51cae5e37 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -30,6 +30,8 @@ import ( type SeqHandle interface{} // SeqData holds the data for one unit in the file. +// +// +stateify savable type SeqData struct { // The data to be returned to the user. Buf []byte @@ -82,6 +84,8 @@ func (s *SeqGenerationCounter) IsCurrent(generation int64) bool { } // SeqFile is used to provide dynamic files that can be ordered by record. +// +// +stateify savable type SeqFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/stat.go b/pkg/sentry/fs/proc/stat.go index 284f3e52b..bf7650211 100644 --- a/pkg/sentry/fs/proc/stat.go +++ b/pkg/sentry/fs/proc/stat.go @@ -25,6 +25,8 @@ import ( ) // statData backs /proc/stat. +// +// +stateify savable type statData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index aab891c53..a2d36ca23 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -28,6 +28,8 @@ import ( ) // hostname is a file containing the system hostname. +// +// +stateify savable type hostname struct { ramfs.Entry } @@ -52,6 +54,8 @@ func (p *proc) newHostname(ctx context.Context, msrc *fs.MountSource) *fs.Inode } // mmapMinAddrData backs /proc/sys/vm/mmap_min_addr. +// +// +stateify savable type mmapMinAddrData struct { k *kernel.Kernel } @@ -74,6 +78,7 @@ func (d *mmapMinAddrData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHand }, 0 } +// +stateify savable type overcommitMemory struct{} func (*overcommitMemory) NeedsUpdate(generation int64) bool { diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index f3a5043f8..beb25be20 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -33,6 +33,7 @@ const ( tcpWMem ) +// +stateify savable type tcpMem struct { ramfs.Entry s inet.Stack @@ -100,6 +101,7 @@ func (m *tcpMem) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, return n, cperr } +// +stateify savable type tcpSack struct { ramfs.Entry s inet.Stack diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index efc635946..748ca4320 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -52,6 +52,8 @@ func getTaskMM(t *kernel.Task) (*mm.MemoryManager, error) { } // taskDir represents a task-level directory. +// +// +stateify savable type taskDir struct { ramfs.Dir @@ -92,6 +94,8 @@ func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace } // subtasks represents a /proc/TID/task directory. +// +// +stateify savable type subtasks struct { ramfs.Dir @@ -167,6 +171,8 @@ func (s *subtasks) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, off } // exe is an fs.InodeOperations symlink for the /proc/PID/exe file. +// +// +stateify savable type exe struct { ramfs.Symlink @@ -226,6 +232,8 @@ func (e *exe) Readlink(ctx context.Context, inode *fs.Inode) (string, error) { // namespaceFile represents a file in the namespacefs, such as the files in // /proc//ns. +// +// +stateify savable type namespaceFile struct { ramfs.Symlink @@ -274,6 +282,8 @@ func newNamespaceDir(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { } // mapsData implements seqfile.SeqSource for /proc/[pid]/maps. +// +// +stateify savable type mapsData struct { t *kernel.Task } @@ -311,6 +321,7 @@ func (md *mapsData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([ return []seqfile.SeqData{}, 0 } +// +stateify savable type taskStatData struct { t *kernel.Task @@ -391,6 +402,8 @@ func (s *taskStatData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) } // statmData implements seqfile.SeqSource for /proc/[pid]/statm. +// +// +stateify savable type statmData struct { t *kernel.Task } @@ -425,6 +438,8 @@ func (s *statmData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([ } // statusData implements seqfile.SeqSource for /proc/[pid]/status. +// +// +stateify savable type statusData struct { t *kernel.Task pidns *kernel.PIDNamespace @@ -490,6 +505,7 @@ type ioUsage interface { IOUsage() *usage.IO } +// +stateify savable type ioData struct { ioUsage } @@ -530,6 +546,8 @@ func (i *ioData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([]se // On Linux, /proc/[pid]/comm is writable, and writing to the comm file changes // the thread name. We don't implement this yet as there are no known users of // this feature. +// +// +stateify savable type comm struct { ramfs.Entry @@ -559,6 +577,8 @@ func (c *comm) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, off } // auxvec is a file containing the auxiliary vector for a task. +// +// +stateify savable type auxvec struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index 85acb5163..9811d9c9d 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -29,6 +29,8 @@ import ( // An idMapSeqSource is a seqfile.SeqSource that returns UID or GID mappings // from a task's user namespace. +// +// +stateify savable type idMapSeqSource struct { t *kernel.Task gids bool @@ -70,6 +72,7 @@ type idMapSeqHandle struct { value int } +// +stateify savable type idMapSeqFile struct { seqfile.SeqFile } diff --git a/pkg/sentry/fs/proc/uptime.go b/pkg/sentry/fs/proc/uptime.go index 4679d5821..f3a9b81df 100644 --- a/pkg/sentry/fs/proc/uptime.go +++ b/pkg/sentry/fs/proc/uptime.go @@ -27,6 +27,8 @@ import ( ) // uptime is a file containing the system uptime. +// +// +stateify savable type uptime struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/version.go b/pkg/sentry/fs/proc/version.go index c0f2e87e3..00f6a2afd 100644 --- a/pkg/sentry/fs/proc/version.go +++ b/pkg/sentry/fs/proc/version.go @@ -23,6 +23,8 @@ import ( ) // versionData backs /proc/version. +// +// +stateify savable type versionData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/ramfs/BUILD b/pkg/sentry/fs/ramfs/BUILD index d84f2c624..5230157fe 100644 --- a/pkg/sentry/fs/ramfs/BUILD +++ b/pkg/sentry/fs/ramfs/BUILD @@ -1,19 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "ramfs_state", - srcs = [ - "dir.go", - "file.go", - "ramfs.go", - "socket.go", - "symlink.go", - ], - out = "ramfs_state.go", - package = "ramfs", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "ramfs", @@ -21,7 +8,6 @@ go_library( "dir.go", "file.go", "ramfs.go", - "ramfs_state.go", "socket.go", "symlink.go", "tree.go", @@ -29,12 +15,8 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs", visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/amutex", - "//pkg/log", - "//pkg/refs", "//pkg/secio", "//pkg/sentry/context", - "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", @@ -42,7 +24,6 @@ go_library( "//pkg/sentry/memmap", "//pkg/sentry/safemem", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index 19d5612ed..04432f28c 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -44,6 +44,8 @@ type CreateOps struct { } // Dir represents a single directory in the filesystem. +// +// +stateify savable type Dir struct { Entry diff --git a/pkg/sentry/fs/ramfs/ramfs.go b/pkg/sentry/fs/ramfs/ramfs.go index d6cfaf753..13e72e775 100644 --- a/pkg/sentry/fs/ramfs/ramfs.go +++ b/pkg/sentry/fs/ramfs/ramfs.go @@ -60,6 +60,8 @@ var ( // Entry represents common internal state for file and directory nodes. // This may be used by other packages to easily create ramfs files. +// +// +stateify savable type Entry struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoMappable `state:"nosave"` diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index b0c79325f..93427a1ff 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -21,6 +21,8 @@ import ( ) // Socket represents a socket. +// +// +stateify savable type Socket struct { Entry diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index 9bbf78619..1c54d9991 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -22,6 +22,8 @@ import ( ) // Symlink represents a symlink. +// +// +stateify savable type Symlink struct { Entry diff --git a/pkg/sentry/fs/ramfs/test/BUILD b/pkg/sentry/fs/ramfs/test/BUILD index 57fee45e2..187eac49d 100644 --- a/pkg/sentry/fs/ramfs/test/BUILD +++ b/pkg/sentry/fs/ramfs/test/BUILD @@ -1,30 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "test_state", - srcs = [ - "test.go", - ], - out = "test_state.go", - package = "test", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "test", testonly = 1, - srcs = [ - "test.go", - "test_state.go", - ], + srcs = ["test.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/sentry/context", - "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs", - "//pkg/state", ], ) diff --git a/pkg/sentry/fs/sys/BUILD b/pkg/sentry/fs/sys/BUILD index 095ff1f25..bc24e980e 100644 --- a/pkg/sentry/fs/sys/BUILD +++ b/pkg/sentry/fs/sys/BUILD @@ -1,16 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "sys_state", - srcs = [ - "fs.go", - "sys.go", - ], - out = "sys_state.go", - package = "sys", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "sys", @@ -18,7 +8,6 @@ go_library( "device.go", "fs.go", "sys.go", - "sys_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/sys", visibility = ["//pkg/sentry:internal"], @@ -28,6 +17,5 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs", "//pkg/sentry/usermem", - "//pkg/state", ], ) diff --git a/pkg/sentry/fs/sys/fs.go b/pkg/sentry/fs/sys/fs.go index c6d5f7fd8..625525540 100644 --- a/pkg/sentry/fs/sys/fs.go +++ b/pkg/sentry/fs/sys/fs.go @@ -20,6 +20,8 @@ import ( ) // filesystem is a sysfs. +// +// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/sys/sys.go b/pkg/sentry/fs/sys/sys.go index ccf56f644..b9b2fb4a1 100644 --- a/pkg/sentry/fs/sys/sys.go +++ b/pkg/sentry/fs/sys/sys.go @@ -22,12 +22,13 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -type Dir struct { +// +stateify savable +type dir struct { ramfs.Dir } func newDir(ctx context.Context, msrc *fs.MountSource, contents map[string]*fs.Inode) *fs.Inode { - d := &Dir{} + d := &dir{} d.InitDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) return fs.NewInode(d, msrc, fs.StableAttr{ DeviceID: sysfsDevice.DeviceID(), diff --git a/pkg/sentry/fs/timerfd/BUILD b/pkg/sentry/fs/timerfd/BUILD index 8b1b7872e..ffdd7e0dc 100644 --- a/pkg/sentry/fs/timerfd/BUILD +++ b/pkg/sentry/fs/timerfd/BUILD @@ -1,33 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "timerfd_state", - srcs = [ - "timerfd.go", - ], - out = "timerfd_state.go", - package = "timerfd", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "timerfd", - srcs = [ - "timerfd.go", - "timerfd_state.go", - ], + srcs = ["timerfd.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/timerfd", visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/fs/timerfd/timerfd.go b/pkg/sentry/fs/timerfd/timerfd.go index ae58f6fd7..767db95a0 100644 --- a/pkg/sentry/fs/timerfd/timerfd.go +++ b/pkg/sentry/fs/timerfd/timerfd.go @@ -30,6 +30,8 @@ import ( ) // TimerOperations implements fs.FileOperations for timerfds. +// +// +stateify savable type TimerOperations struct { fsutil.ZeroSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` @@ -38,7 +40,7 @@ type TimerOperations struct { fsutil.NoMMap `state:"nosave"` fsutil.NoIoctl `state:"nosave"` - events waiter.Queue `state:"nosave"` + events waiter.Queue `state:"zerovalue"` timer *ktime.Timer // val is the number of timer expirations since the last successful call to diff --git a/pkg/sentry/fs/tmpfs/BUILD b/pkg/sentry/fs/tmpfs/BUILD index 473ab4296..cfe11ab02 100644 --- a/pkg/sentry/fs/tmpfs/BUILD +++ b/pkg/sentry/fs/tmpfs/BUILD @@ -1,18 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tmpfs_state", - srcs = [ - "file_regular.go", - "fs.go", - "inode_file.go", - "tmpfs.go", - ], - out = "tmpfs_state.go", - package = "tmpfs", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tmpfs", @@ -22,13 +10,11 @@ go_library( "fs.go", "inode_file.go", "tmpfs.go", - "tmpfs_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tmpfs", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/log", "//pkg/sentry/context", "//pkg/sentry/device", "//pkg/sentry/fs", @@ -41,7 +27,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/tcpip/transport/unix", "//pkg/waiter", ], diff --git a/pkg/sentry/fs/tmpfs/file_regular.go b/pkg/sentry/fs/tmpfs/file_regular.go index 9811d90bc..342688f81 100644 --- a/pkg/sentry/fs/tmpfs/file_regular.go +++ b/pkg/sentry/fs/tmpfs/file_regular.go @@ -25,6 +25,8 @@ import ( // regularFileOperations implements fs.FileOperations for a regular // tmpfs file. +// +// +stateify savable type regularFileOperations struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go index 5bd9ade52..ca620e65e 100644 --- a/pkg/sentry/fs/tmpfs/fs.go +++ b/pkg/sentry/fs/tmpfs/fs.go @@ -47,6 +47,8 @@ const ( var modeRegexp = regexp.MustCompile("0[0-7][0-7][0-7]") // Filesystem is a tmpfs. +// +// +stateify savable type Filesystem struct{} func init() { diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 4e803c9ff..1e4fe47d2 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -43,6 +43,8 @@ import ( // include an InvalidatorRegion associated with that reference. When the // referenced portion of the file is removed (with Truncate), the associated // InvalidatorRegion is invalidated. +// +// +stateify savable type fileInodeOperations struct { fsutil.DeprecatedFileOperations `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 1cc7ae491..10cb5451d 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -49,6 +49,8 @@ func rename(ctx context.Context, oldParent *fs.Inode, oldName string, newParent } // Dir is a directory. +// +// +stateify savable type Dir struct { ramfs.Dir @@ -122,6 +124,8 @@ func (*Dir) StatFS(context.Context) (fs.Info, error) { } // Symlink is a symlink. +// +// +stateify savable type Symlink struct { ramfs.Symlink } @@ -149,6 +153,8 @@ func (s *Symlink) StatFS(context.Context) (fs.Info, error) { } // Socket is a socket. +// +// +stateify savable type Socket struct { ramfs.Socket } @@ -176,6 +182,8 @@ func (s *Socket) StatFS(context.Context) (fs.Info, error) { } // Fifo is a tmpfs named pipe. +// +// +stateify savable type Fifo struct { ramfs.Entry } diff --git a/pkg/sentry/fs/tty/BUILD b/pkg/sentry/fs/tty/BUILD index 363897b2c..3c446eef4 100644 --- a/pkg/sentry/fs/tty/BUILD +++ b/pkg/sentry/fs/tty/BUILD @@ -1,22 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tty_state", - srcs = [ - "dir.go", - "fs.go", - "inode.go", - "line_discipline.go", - "master.go", - "queue.go", - "slave.go", - "terminal.go", - ], - out = "tty_state.go", - package = "tty", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tty", @@ -29,7 +13,6 @@ go_library( "queue.go", "slave.go", "terminal.go", - "tty_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tty", visibility = ["//pkg/sentry:internal"], @@ -44,7 +27,6 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index 2c5b2aed6..c91091db4 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -49,14 +49,16 @@ import ( // corresponding Dirents hold on their parent (this directory). // // dirInodeOperations implements fs.InodeOperations. +// +// +stateify savable type dirInodeOperations struct { - fsutil.DeprecatedFileOperations - fsutil.InodeNotSocket - fsutil.InodeNotRenameable - fsutil.InodeNotSymlink - fsutil.InodeNoExtendedAttributes - fsutil.NoMappable - fsutil.NoopWriteOut + fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotRenameable `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.NoMappable `state:"nosave"` + fsutil.NoopWriteOut `state:"nosave"` // msrc is the super block this directory is on. // @@ -348,6 +350,8 @@ func (d *dirInodeOperations) masterClose(t *Terminal) { // // This is nearly identical to fsutil.DirFileOperations, except that it takes // df.di.mu in IterateDir. +// +// +stateify savable type dirFileOperations struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/tty/fs.go b/pkg/sentry/fs/tty/fs.go index dbaffe95e..e28635607 100644 --- a/pkg/sentry/fs/tty/fs.go +++ b/pkg/sentry/fs/tty/fs.go @@ -28,6 +28,8 @@ var ptsDevice = device.NewAnonDevice() // // This devpts is always in the new "multi-instance" mode. i.e., it contains a // ptmx device tied to this mount. +// +// +stateify savable type filesystem struct{} func init() { @@ -69,6 +71,8 @@ func (f *filesystem) Mount(ctx context.Context, device string, flags fs.MountSou } // superOperations implements fs.MountSourceOperations, preventing caching. +// +// +stateify savable type superOperations struct{} // Revalidate implements fs.DirentOperations.Revalidate. diff --git a/pkg/sentry/fs/tty/inode.go b/pkg/sentry/fs/tty/inode.go index 04b9a7727..c0fa2b407 100644 --- a/pkg/sentry/fs/tty/inode.go +++ b/pkg/sentry/fs/tty/inode.go @@ -31,6 +31,8 @@ import ( // // * fs.InodeOperations.Release // * fs.InodeOperations.GetFile +// +// +stateify savable type inodeOperations struct { fsutil.DeprecatedFileOperations `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` diff --git a/pkg/sentry/fs/tty/line_discipline.go b/pkg/sentry/fs/tty/line_discipline.go index f094635f5..d243ee40e 100644 --- a/pkg/sentry/fs/tty/line_discipline.go +++ b/pkg/sentry/fs/tty/line_discipline.go @@ -72,6 +72,8 @@ const ( // termiosMu // inQueue.mu // outQueue.mu +// +// +stateify savable type lineDiscipline struct { // inQueue is the input queue of the terminal. inQueue queue @@ -183,6 +185,8 @@ type transformer interface { // outputQueueTransformer implements transformer. It performs line discipline // transformations on the output queue. +// +// +stateify savable type outputQueueTransformer struct{} // transform does output processing for one end of the pty. See @@ -254,6 +258,8 @@ func (*outputQueueTransformer) transform(l *lineDiscipline, q *queue, buf []byte // inputQueueTransformer implements transformer. It performs line discipline // transformations on the input queue. +// +// +stateify savable type inputQueueTransformer struct{} // transform does input processing for one end of the pty. Characters read are diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index 74cdbe874..c7198e218 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -27,6 +27,8 @@ import ( // masterInodeOperations are the fs.InodeOperations for the master end of the // Terminal (ptmx file). +// +// +stateify savable type masterInodeOperations struct { inodeOperations @@ -96,6 +98,8 @@ func (mi *masterInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flag } // masterFileOperations are the fs.FileOperations for the master end of a terminal. +// +// +stateify savable type masterFileOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/tty/queue.go b/pkg/sentry/fs/tty/queue.go index 026d5e077..42c105abc 100644 --- a/pkg/sentry/fs/tty/queue.go +++ b/pkg/sentry/fs/tty/queue.go @@ -32,11 +32,13 @@ import ( // processed (i.e. undergo termios transformations) as they are added to the // read buffer. The read buffer is readable when its length is nonzero and // readable is true. +// +// +stateify savable type queue struct { // mu protects everything in queue. mu sync.Mutex `state:"nosave"` - waiter.Queue `state:"nosave"` + waiter.Queue `state:"zerovalue"` // readBuf is buffer of data ready to be read when readable is true. // This data has been processed. diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index f5eec726e..1c562b172 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -27,6 +27,8 @@ import ( // slaveInodeOperations are the fs.InodeOperations for the slave end of the // Terminal (pts file). +// +// +stateify savable type slaveInodeOperations struct { inodeOperations @@ -86,6 +88,8 @@ func (si *slaveInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags } // slaveFileOperations are the fs.FileOperations for the slave end of a terminal. +// +// +stateify savable type slaveFileOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/tty/terminal.go b/pkg/sentry/fs/tty/terminal.go index fa5b00409..3cb135124 100644 --- a/pkg/sentry/fs/tty/terminal.go +++ b/pkg/sentry/fs/tty/terminal.go @@ -21,6 +21,8 @@ import ( ) // Terminal is a pseudoterminal. +// +// +stateify savable type Terminal struct { refs.AtomicRefCount diff --git a/pkg/sentry/inet/BUILD b/pkg/sentry/inet/BUILD index eaf8f15b2..159c50efb 100644 --- a/pkg/sentry/inet/BUILD +++ b/pkg/sentry/inet/BUILD @@ -3,26 +3,15 @@ package( licenses = ["notice"], # Apache 2.0 ) -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "inet_state", - srcs = ["inet.go"], - out = "inet_state.go", - package = "inet", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "inet", srcs = [ "context.go", "inet.go", - "inet_state.go", "test_stack.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/inet", - deps = [ - "//pkg/sentry/context", - "//pkg/state", - ], + deps = ["//pkg/sentry/context"], ) diff --git a/pkg/sentry/inet/inet.go b/pkg/sentry/inet/inet.go index e4b326993..e54a61196 100644 --- a/pkg/sentry/inet/inet.go +++ b/pkg/sentry/inet/inet.go @@ -87,6 +87,8 @@ type InterfaceAddr struct { } // TCPBufferSize contains settings controlling TCP buffer sizing. +// +// +stateify savable type TCPBufferSize struct { // Min is the minimum size. Min int diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index c4a7dacb2..0ebacefa6 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -1,58 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "kernel_state", - srcs = [ - "abstract_socket_namespace.go", - "fd_map.go", - "fs_context.go", - "ipc_namespace.go", - "kernel.go", - "kernel_state.go", - "pending_signals.go", - "pending_signals_state.go", - "process_group_list.go", - "ptrace.go", - "rseq.go", - "session_list.go", - "sessions.go", - "signal.go", - "signal_handlers.go", - "syscalls.go", - "syscalls_state.go", - "syslog.go", - "task.go", - "task_clone.go", - "task_context.go", - "task_exec.go", - "task_exit.go", - "task_list.go", - "task_resources.go", - "task_run.go", - "task_sched.go", - "task_signals.go", - "task_start.go", - "task_syscall.go", - "thread_group.go", - "threads.go", - "timekeeper.go", - "timekeeper_state.go", - "timer.go", - "uts_namespace.go", - "vdso.go", - "version.go", - ], - out = "kernel_autogen_state.go", - imports = [ - "gvisor.googlesource.com/gvisor/pkg/sentry/arch", - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", - "gvisor.googlesource.com/gvisor/pkg/tcpip", - ], - package = "kernel", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "pending_signals_list", @@ -118,7 +67,6 @@ go_library( "fs_context.go", "ipc_namespace.go", "kernel.go", - "kernel_autogen_state.go", "kernel_state.go", "pending_signals.go", "pending_signals_list.go", @@ -164,6 +112,11 @@ go_library( "version.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel", + imports = [ + "gvisor.googlesource.com/gvisor/pkg/sentry/arch", + # "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", + "gvisor.googlesource.com/gvisor/pkg/tcpip", + ], visibility = ["//:sandbox"], deps = [ "//pkg/abi", diff --git a/pkg/sentry/kernel/abstract_socket_namespace.go b/pkg/sentry/kernel/abstract_socket_namespace.go index 014c4a3bf..d6d1d341d 100644 --- a/pkg/sentry/kernel/abstract_socket_namespace.go +++ b/pkg/sentry/kernel/abstract_socket_namespace.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) +// +stateify savable type abstractEndpoint struct { ep unix.BoundEndpoint wr *refs.WeakRef @@ -39,6 +40,8 @@ func (e *abstractEndpoint) WeakRefGone() { } // AbstractSocketNamespace is used to implement the Linux abstract socket functionality. +// +// +stateify savable type AbstractSocketNamespace struct { mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/auth/BUILD b/pkg/sentry/kernel/auth/BUILD index 5b7b30557..a81085372 100644 --- a/pkg/sentry/kernel/auth/BUILD +++ b/pkg/sentry/kernel/auth/BUILD @@ -1,20 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "auth_state", - srcs = [ - "credentials.go", - "id.go", - "id_map_range.go", - "id_map_set.go", - "user_namespace.go", - ], - out = "auth_state.go", - package = "auth", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_template_instance( name = "id_map_range", @@ -48,7 +35,6 @@ go_library( name = "auth", srcs = [ "auth.go", - "auth_state.go", "capability_set.go", "context.go", "credentials.go", @@ -66,7 +52,6 @@ go_library( "//pkg/bits", "//pkg/log", "//pkg/sentry/context", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go index f6fb05285..f18f7dac9 100644 --- a/pkg/sentry/kernel/auth/credentials.go +++ b/pkg/sentry/kernel/auth/credentials.go @@ -21,6 +21,8 @@ import ( // Credentials contains information required to authorize privileged operations // in a user namespace. +// +// +stateify savable type Credentials struct { // Real/effective/saved user/group IDs in the root user namespace. None of // these should ever be NoID. diff --git a/pkg/sentry/kernel/auth/id_map.go b/pkg/sentry/kernel/auth/id_map.go index 6adb33530..bd0090e0f 100644 --- a/pkg/sentry/kernel/auth/id_map.go +++ b/pkg/sentry/kernel/auth/id_map.go @@ -77,6 +77,8 @@ func (ns *UserNamespace) allIDsMapped(m *idMapSet, start, end uint32) bool { // An IDMapEntry represents a mapping from a range of contiguous IDs in a user // namespace to an equally-sized range of contiguous IDs in the namespace's // parent. +// +// +stateify savable type IDMapEntry struct { // FirstID is the first ID in the range in the namespace. FirstID uint32 diff --git a/pkg/sentry/kernel/auth/user_namespace.go b/pkg/sentry/kernel/auth/user_namespace.go index 0980aeadf..d359f3f31 100644 --- a/pkg/sentry/kernel/auth/user_namespace.go +++ b/pkg/sentry/kernel/auth/user_namespace.go @@ -23,6 +23,8 @@ import ( // A UserNamespace represents a user namespace. See user_namespaces(7) for // details. +// +// +stateify savable type UserNamespace struct { // parent is this namespace's parent. If this is the root namespace, parent // is nil. The parent pointer is immutable. diff --git a/pkg/sentry/kernel/epoll/BUILD b/pkg/sentry/kernel/epoll/BUILD index 7d491efbc..5e8b36ed6 100644 --- a/pkg/sentry/kernel/epoll/BUILD +++ b/pkg/sentry/kernel/epoll/BUILD @@ -1,22 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "epoll_autogen_state", - srcs = [ - "epoll.go", - "epoll_state.go", - ], - out = "epoll_autogen_state.go", - package = "epoll", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "epoll", srcs = [ "epoll.go", - "epoll_autogen_state.go", "epoll_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll", @@ -29,9 +18,7 @@ go_library( "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/kdefs", - "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index b572fcd7e..d87e64a1c 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -58,6 +58,8 @@ const ( // potentially be reassigned. We also cannot use just the file pointer because // it is possible to have multiple entries for the same file object as long as // they are created with different FDs (i.e., the FDs point to the same file). +// +// +stateify savable type FileIdentifier struct { File *fs.File Fd kdefs.FD @@ -65,6 +67,8 @@ type FileIdentifier struct { // pollEntry holds all the state associated with an event poll entry, that is, // a file being observed by an event poll object. +// +// +stateify savable type pollEntry struct { ilist.Entry file *refs.WeakRef `state:"manual"` @@ -92,6 +96,8 @@ func (p *pollEntry) WeakRefGone() { // EventPoll holds all the state associated with an event poll object, that is, // collection of files to observe and their current state. +// +// +stateify savable type EventPoll struct { fsutil.PipeSeek `state:"zerovalue"` fsutil.NotDirReaddir `state:"zerovalue"` @@ -102,7 +108,7 @@ type EventPoll struct { // Wait queue is used to notify interested parties when the event poll // object itself becomes readable or writable. - waiter.Queue + waiter.Queue `state:"zerovalue"` // files is the map of all the files currently being observed, it is // protected by mu. diff --git a/pkg/sentry/kernel/eventfd/BUILD b/pkg/sentry/kernel/eventfd/BUILD index 7ec179bd8..cc1120b4f 100644 --- a/pkg/sentry/kernel/eventfd/BUILD +++ b/pkg/sentry/kernel/eventfd/BUILD @@ -1,33 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "eventfd_state", - srcs = [ - "eventfd.go", - ], - out = "eventfd_state.go", - package = "eventfd", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "eventfd", - srcs = [ - "eventfd.go", - "eventfd_state.go", - ], + srcs = ["eventfd.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/eventfd", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", "//pkg/waiter/fdnotifier", diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index bd50bd9fe..a4ada0e78 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -35,6 +35,8 @@ import ( // EventOperations represents an event with the semantics of Linux's file-based event // notification (eventfd). Eventfds are usually internal to the Sentry but in certain // situations they may be converted into a host-backed eventfd. +// +// +stateify savable type EventOperations struct { fsutil.NoopRelease `state:"nosave"` fsutil.PipeSeek `state:"nosave"` @@ -49,7 +51,7 @@ type EventOperations struct { // Queue is used to notify interested parties when the event object // becomes readable or writable. - wq waiter.Queue `state:"nosave"` + wq waiter.Queue `state:"zerovalue"` // val is the current value of the event counter. val uint64 diff --git a/pkg/sentry/kernel/fd_map.go b/pkg/sentry/kernel/fd_map.go index 299506330..d5d4aaacb 100644 --- a/pkg/sentry/kernel/fd_map.go +++ b/pkg/sentry/kernel/fd_map.go @@ -46,6 +46,8 @@ func (f FDs) Less(i, j int) bool { } // FDFlags define flags for an individual descriptor. +// +// +stateify savable type FDFlags struct { // CloseOnExec indicates the descriptor should be closed on exec. CloseOnExec bool @@ -69,12 +71,16 @@ func (f FDFlags) ToLinuxFDFlags() (mask uint) { // descriptor holds the details about a file descriptor, namely a pointer the // file itself and the descriptor flags. +// +// +stateify savable type descriptor struct { file *fs.File flags FDFlags } // FDMap is used to manage File references and flags. +// +// +stateify savable type FDMap struct { refs.AtomicRefCount k *Kernel diff --git a/pkg/sentry/kernel/fs_context.go b/pkg/sentry/kernel/fs_context.go index dbc097696..f3f05e8f5 100644 --- a/pkg/sentry/kernel/fs_context.go +++ b/pkg/sentry/kernel/fs_context.go @@ -25,6 +25,8 @@ import ( // FSContext contains filesystem context. // // This includes umask and working directory. +// +// +stateify savable type FSContext struct { refs.AtomicRefCount diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index a97a43549..b44a26974 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "waiter_list", @@ -14,29 +14,15 @@ go_template_instance( }, ) -go_stateify( - name = "futex_state", - srcs = [ - "futex.go", - "waiter_list.go", - ], - out = "futex_state.go", - package = "futex", -) - go_library( name = "futex", srcs = [ "futex.go", - "futex_state.go", "waiter_list.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex", visibility = ["//pkg/sentry:internal"], - deps = [ - "//pkg/state", - "//pkg/syserror", - ], + deps = ["//pkg/syserror"], ) go_test( diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index 15e3e5e2c..4a1f2a0ef 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -196,6 +196,8 @@ func bucketIndexForAddr(addr uintptr) uintptr { } // Manager holds futex state for a single virtual address space. +// +// +stateify savable type Manager struct { buckets [bucketCount]bucket `state:"zerovalue"` } diff --git a/pkg/sentry/kernel/ipc_namespace.go b/pkg/sentry/kernel/ipc_namespace.go index a86bda77b..5eef49f59 100644 --- a/pkg/sentry/kernel/ipc_namespace.go +++ b/pkg/sentry/kernel/ipc_namespace.go @@ -21,6 +21,8 @@ import ( ) // IPCNamespace represents an IPC namespace. +// +// +stateify savable type IPCNamespace struct { // User namespace which owns this IPC namespace. Immutable. userNS *auth.UserNamespace diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 64439cd9d..419a1d473 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -62,6 +62,8 @@ import ( // Kernel represents an emulated Linux kernel. It must be initialized by calling // Init() or LoadFrom(). +// +// +stateify savable type Kernel struct { // extMu serializes external changes to the Kernel with calls to // Kernel.SaveTo. (Kernel.SaveTo requires that the state of the Kernel @@ -158,7 +160,7 @@ type Kernel struct { // exitErr is the error causing the sandbox to exit, if any. It is // protected by extMu. - exitErr error + exitErr error `state:"nosave"` // danglingEndpoints is used to save / restore tcpip.DanglingEndpoints. danglingEndpoints struct{} `state:".([]tcpip.Endpoint)"` diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go index 5dc0f266c..06be5a7e1 100644 --- a/pkg/sentry/kernel/pending_signals.go +++ b/pkg/sentry/kernel/pending_signals.go @@ -38,6 +38,8 @@ const ( // pendingSignals holds a collection of pending signals. The zero value of // pendingSignals is a valid empty collection. pendingSignals is thread-unsafe; // users must provide synchronization. +// +// +stateify savable type pendingSignals struct { // signals contains all pending signals. // @@ -52,11 +54,14 @@ type pendingSignals struct { } // pendingSignalQueue holds a pendingSignalList for a single signal number. +// +// +stateify savable type pendingSignalQueue struct { pendingSignalList length int } +// +stateify savable type pendingSignal struct { // pendingSignalEntry links into a pendingSignalList. pendingSignalEntry diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD index 4600d19bd..19b23c6d2 100644 --- a/pkg/sentry/kernel/pipe/BUILD +++ b/pkg/sentry/kernel/pipe/BUILD @@ -1,20 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "pipe_state", - srcs = [ - "buffers.go", - "node.go", - "pipe.go", - "reader.go", - "reader_writer.go", - "writer.go", - ], - out = "pipe_state.go", - package = "pipe", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "pipe", @@ -23,7 +9,6 @@ go_library( "device.go", "node.go", "pipe.go", - "pipe_state.go", "reader.go", "reader_writer.go", "writer.go", @@ -34,15 +19,12 @@ go_library( "//pkg/abi/linux", "//pkg/amutex", "//pkg/ilist", - "//pkg/log", - "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/kernel/pipe/buffers.go b/pkg/sentry/kernel/pipe/buffers.go index f300537c5..a82e45c3f 100644 --- a/pkg/sentry/kernel/pipe/buffers.go +++ b/pkg/sentry/kernel/pipe/buffers.go @@ -20,6 +20,8 @@ import ( // Buffer encapsulates a queueable byte buffer that can // easily be truncated. It is designed only for use with pipes. +// +// +stateify savable type Buffer struct { ilist.Entry data []byte diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index e418cf174..23d692da1 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -24,6 +24,8 @@ import ( ) // inodeOperations wraps fs.InodeOperations operations with common pipe opening semantics. +// +// +stateify savable type inodeOperations struct { fs.InodeOperations diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index 9a21df5b4..ced2559a7 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -41,6 +41,8 @@ const DefaultPipeSize = 65536 // Pipe is an encapsulation of a platform-independent pipe. // It manages a buffered byte queue shared between a reader/writer // pair. +// +// +stateify savable type Pipe struct { waiter.Queue `state:"nosave"` diff --git a/pkg/sentry/kernel/pipe/reader.go b/pkg/sentry/kernel/pipe/reader.go index 40d5e4943..1fa5e9a32 100644 --- a/pkg/sentry/kernel/pipe/reader.go +++ b/pkg/sentry/kernel/pipe/reader.go @@ -20,6 +20,8 @@ import ( // Reader satisfies the fs.FileOperations interface for read-only pipes. // Reader should be used with !fs.FileFlags.Write to reject writes. +// +// +stateify savable type Reader struct { ReaderWriter } diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index dc642a3a6..82607367b 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -31,6 +31,8 @@ import ( // read and write requests. This should only be used directly for named pipes. // pipe(2) and pipe2(2) only support unidirectional pipes and should use // either pipe.Reader or pipe.Writer. +// +// +stateify savable type ReaderWriter struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/kernel/pipe/writer.go b/pkg/sentry/kernel/pipe/writer.go index fd13008ac..d93324b53 100644 --- a/pkg/sentry/kernel/pipe/writer.go +++ b/pkg/sentry/kernel/pipe/writer.go @@ -20,6 +20,8 @@ import ( // Writer satisfies the fs.FileOperations interface for write-only pipes. // Writer should be used with !fs.FileFlags.Read to reject reads. +// +// +stateify savable type Writer struct { ReaderWriter } diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index f1c2c4bf0..e9e69004d 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -25,6 +25,8 @@ import ( // ptraceOptions are the subset of options controlling a task's ptrace behavior // that are set by ptrace(PTRACE_SETOPTIONS). +// +// +stateify savable type ptraceOptions struct { // ExitKill is true if the tracee should be sent SIGKILL when the tracer // exits. @@ -185,6 +187,8 @@ func (t *Task) hasTracer() bool { } // ptraceStop is a TaskStop placed on tasks in a ptrace-stop. +// +// +stateify savable type ptraceStop struct { // If frozen is true, the stopped task's tracer is currently operating on // it, so Task.Kill should not remove the stop. diff --git a/pkg/sentry/kernel/rseq.go b/pkg/sentry/kernel/rseq.go index 635372993..1f3de58e3 100644 --- a/pkg/sentry/kernel/rseq.go +++ b/pkg/sentry/kernel/rseq.go @@ -23,6 +23,8 @@ import ( // Restartable sequences, as described in https://lwn.net/Articles/650333/. // RSEQCriticalRegion describes a restartable sequence critical region. +// +// +stateify savable type RSEQCriticalRegion struct { // When a task in this thread group has its CPU preempted (as defined by // platform.ErrContextCPUPreempted) or has a signal delivered to an diff --git a/pkg/sentry/kernel/semaphore/BUILD b/pkg/sentry/kernel/semaphore/BUILD index 969145fe1..e7fa44e2c 100644 --- a/pkg/sentry/kernel/semaphore/BUILD +++ b/pkg/sentry/kernel/semaphore/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "waiter_list", @@ -14,21 +14,10 @@ go_template_instance( }, ) -go_stateify( - name = "semaphore_state", - srcs = [ - "semaphore.go", - "waiter_list.go", - ], - out = "semaphore_autogen_state.go", - package = "semaphore", -) - go_library( name = "semaphore", srcs = [ "semaphore.go", - "semaphore_autogen_state.go", "waiter_list.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/semaphore", @@ -40,8 +29,6 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", - "//pkg/state", - "//pkg/state/statefile", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index a1ee83ce5..aa07946cf 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -42,6 +42,8 @@ const ( ) // Registry maintains a set of semaphores that can be found by key or ID. +// +// +stateify savable type Registry struct { // userNS owning the ipc name this registry belongs to. Immutable. userNS *auth.UserNamespace @@ -52,6 +54,8 @@ type Registry struct { } // Set represents a set of semaphores that can be operated atomically. +// +// +stateify savable type Set struct { // registry owning this sem set. Immutable. registry *Registry @@ -79,6 +83,8 @@ type Set struct { } // sem represents a single semanphore from a set. +// +// +stateify savable type sem struct { value int16 waiters waiterList `state:"zerovalue"` @@ -86,6 +92,8 @@ type sem struct { // waiter represents a caller that is waiting for the semaphore value to // become positive or zero. +// +// +stateify savable type waiter struct { waiterEntry diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index fa4c7b8f6..cf4e18805 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -27,6 +27,8 @@ type SessionID ThreadID type ProcessGroupID ThreadID // Session contains a leader threadgroup and a list of ProcessGroups. +// +// +stateify savable type Session struct { refs refs.AtomicRefCount @@ -76,6 +78,8 @@ func (s *Session) decRef() { } // ProcessGroup contains an originator threadgroup and a parent Session. +// +// +stateify savable type ProcessGroup struct { refs refs.AtomicRefCount // not exported. diff --git a/pkg/sentry/kernel/shm/BUILD b/pkg/sentry/kernel/shm/BUILD index 0f88eb0ac..40e641355 100644 --- a/pkg/sentry/kernel/shm/BUILD +++ b/pkg/sentry/kernel/shm/BUILD @@ -1,22 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "shm_state", - srcs = [ - "shm.go", - ], - out = "shm_autogen_state.go", - package = "shm", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "shm", srcs = [ "device.go", "shm.go", - "shm_autogen_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/shm", visibility = ["//pkg/sentry:internal"], @@ -33,7 +23,6 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 7217e8103..1ac444094 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -72,6 +72,8 @@ const ( // Registry tracks all shared memory segments in an IPC namespace. The registry // provides the mechanisms for creating and finding segments, and reporting // global shm parameters. +// +// +stateify savable type Registry struct { // userNS owns the IPC namespace this registry belong to. Immutable. userNS *auth.UserNamespace @@ -288,6 +290,8 @@ func (r *Registry) remove(s *Shm) { // shmctl(SHM_RMID). // // Shm implements memmap.Mappable and memmap.MappingIdentity. +// +// +stateify savable type Shm struct { // AtomicRefCount tracks the number of references to this segment from // maps. A segment always holds a reference to itself, until it's marked for diff --git a/pkg/sentry/kernel/signal_handlers.go b/pkg/sentry/kernel/signal_handlers.go index 21ba4ee70..3649f5e4d 100644 --- a/pkg/sentry/kernel/signal_handlers.go +++ b/pkg/sentry/kernel/signal_handlers.go @@ -22,6 +22,8 @@ import ( ) // SignalHandlers holds information about signal actions. +// +// +stateify savable type SignalHandlers struct { // mu protects actions, as well as the signal state of all tasks and thread // groups using this SignalHandlers object. (See comment on diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go index e20fa3eb6..4c7811b6c 100644 --- a/pkg/sentry/kernel/syscalls.go +++ b/pkg/sentry/kernel/syscalls.go @@ -176,6 +176,8 @@ type Stracer interface { // SyscallTable is a lookup table of system calls. Critically, a SyscallTable // is *immutable*. In order to make supporting suspend and resume sane, they // must be uniquely registered and may not change during operation. +// +// +stateify savable type SyscallTable struct { // OS is the operating system that this syscall table implements. OS abi.OS `state:"wait"` diff --git a/pkg/sentry/kernel/syslog.go b/pkg/sentry/kernel/syslog.go index 31541749e..125312b6a 100644 --- a/pkg/sentry/kernel/syslog.go +++ b/pkg/sentry/kernel/syslog.go @@ -23,6 +23,8 @@ import ( // syslog represents a sentry-global kernel log. // // Currently, it contains only fun messages for a dmesg easter egg. +// +// +stateify savable type syslog struct { // mu protects the below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 7f6735320..ae9b3d175 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -52,6 +52,8 @@ import ( // All fields that are "exclusive to the task goroutine" can only be accessed // by the task goroutine while it is running. The task goroutine does not // require synchronization to read or write these fields. +// +// +stateify savable type Task struct { taskNode diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index a61283267..38f7826e2 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -349,6 +349,7 @@ func (t *Task) unstopVforkParent() { } } +// +stateify savable type runSyscallAfterPtraceEventClone struct { vforkChild *Task @@ -366,6 +367,7 @@ func (r *runSyscallAfterPtraceEventClone) execute(t *Task) taskRunState { return (*runSyscallExit)(nil) } +// +stateify savable type runSyscallAfterVforkStop struct { // childTID has the same meaning as // runSyscallAfterPtraceEventClone.vforkChildTID. @@ -471,6 +473,8 @@ func (t *Task) Unshare(opts *SharingOptions) error { // current MM. (Normally, CLONE_VFORK is used in conjunction with CLONE_VM, so // that the child and parent share mappings until the child execve()s into a // new process image or exits.) +// +// +stateify savable type vforkStop struct{} // StopIgnoresKill implements TaskStop.Killable. diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index 5c563ba08..9a59cbd33 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -35,6 +35,8 @@ var ErrNoSyscalls = errors.New("no syscall table found") type Auxmap map[string]interface{} // TaskContext is the subset of a task's data that is provided by the loader. +// +// +stateify savable type TaskContext struct { // Name is the thread name set by the prctl(PR_SET_NAME) system call. Name string diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index 2285847a2..385299b24 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -73,6 +73,8 @@ import ( // execStop is a TaskStop that a task sets on itself when it wants to execve // and is waiting for the other tasks in its thread group to exit first. +// +// +stateify savable type execStop struct{} // Killable implements TaskStop.Killable. @@ -119,6 +121,8 @@ func (t *Task) Execve(newTC *TaskContext) (*SyscallControl, error) { // The runSyscallAfterExecStop state continues execve(2) after all siblings of // a thread in the execve syscall have exited. +// +// +stateify savable type runSyscallAfterExecStop struct { tc *TaskContext } diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index d6604f37b..b16844e91 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -38,6 +38,8 @@ import ( // An ExitStatus is a value communicated from an exiting task or thread group // to the party that reaps it. +// +// +stateify savable type ExitStatus struct { // Code is the numeric value passed to the call to exit or exit_group that // caused the exit. If the exit was not caused by such a call, Code is 0. @@ -222,6 +224,8 @@ func (t *Task) advanceExitStateLocked(oldExit, newExit TaskExitState) { } // runExit is the entry point into the task exit path. +// +// +stateify savable type runExit struct{} func (*runExit) execute(t *Task) taskRunState { @@ -229,6 +233,7 @@ func (*runExit) execute(t *Task) taskRunState { return (*runExitMain)(nil) } +// +stateify savable type runExitMain struct{} func (*runExitMain) execute(t *Task) taskRunState { @@ -531,6 +536,7 @@ func (t *Task) reparentLocked(parent *Task) { // tracer (if one exists) and reaps the leader immediately. In Linux, this is // in fs/exec.c:de_thread(); in the sentry, this is in Task.promoteLocked(). +// +stateify savable type runExitNotify struct{} func (*runExitNotify) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_resources.go b/pkg/sentry/kernel/task_resources.go index 4ca25664a..0832bf989 100644 --- a/pkg/sentry/kernel/task_resources.go +++ b/pkg/sentry/kernel/task_resources.go @@ -21,6 +21,8 @@ import ( // TaskResources is the subset of a task's data provided by its creator that is // not provided by the loader. +// +// +stateify savable type TaskResources struct { // SignalMask is the set of signals whose delivery is currently blocked. // diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index a03fa6ac0..8dd0ef6ea 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -131,6 +131,8 @@ func (t *Task) doStop() { // The runApp state checks for interrupts before executing untrusted // application code. +// +// +stateify savable type runApp struct{} func (*runApp) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_sched.go b/pkg/sentry/kernel/task_sched.go index b50139077..49141ab74 100644 --- a/pkg/sentry/kernel/task_sched.go +++ b/pkg/sentry/kernel/task_sched.go @@ -65,6 +65,8 @@ const ( // TaskGoroutineSchedInfo contains task goroutine scheduling state which must // be read and updated atomically. +// +// +stateify savable type TaskGoroutineSchedInfo struct { // Timestamp was the value of Kernel.cpuClock when this // TaskGoroutineSchedInfo was last updated. diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 91f6c0874..62ec530be 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -748,6 +748,8 @@ func (t *Task) CopyInSignalStack(addr usermem.Addr) (arch.SignalStack, error) { // groupStop is a TaskStop placed on tasks that have received a stop signal // (SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU). (The term "group-stop" originates from // the ptrace man page.) +// +// +stateify savable type groupStop struct{} // Killable implements TaskStop.Killable. @@ -881,6 +883,8 @@ func (t *Task) signalStop(target *Task, code int32, status int32) { } // The runInterrupt state handles conditions indicated by interrupts. +// +// +stateify savable type runInterrupt struct{} func (*runInterrupt) execute(t *Task) taskRunState { @@ -1020,6 +1024,7 @@ func (*runInterrupt) execute(t *Task) taskRunState { return (*runApp)(nil) } +// +stateify savable type runInterruptAfterSignalDeliveryStop struct{} func (*runInterruptAfterSignalDeliveryStop) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index 79f4ff60c..3b9652504 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -241,6 +241,7 @@ func (t *Task) doSyscallEnter(sysno uintptr, args arch.SyscallArguments) taskRun return t.doSyscallInvoke(sysno, args) } +// +stateify savable type runSyscallAfterSyscallEnterStop struct{} func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState { @@ -260,6 +261,7 @@ func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState { return t.doSyscallInvoke(sysno, args) } +// +stateify savable type runSyscallAfterSysemuStop struct{} func (*runSyscallAfterSysemuStop) execute(t *Task) taskRunState { @@ -294,6 +296,7 @@ func (t *Task) doSyscallInvoke(sysno uintptr, args arch.SyscallArguments) taskRu return (*runSyscallExit)(nil).execute(t) } +// +stateify savable type runSyscallReinvoke struct{} func (*runSyscallReinvoke) execute(t *Task) taskRunState { @@ -310,6 +313,7 @@ func (*runSyscallReinvoke) execute(t *Task) taskRunState { return t.doSyscallInvoke(sysno, args) } +// +stateify savable type runSyscallExit struct{} func (*runSyscallExit) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index 8fffd3446..441b8a822 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -28,6 +28,8 @@ import ( // groups" are usually called "processes" in userspace documentation.) // // ThreadGroup is a superset of Linux's struct signal_struct. +// +// +stateify savable type ThreadGroup struct { threadGroupNode diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index 440da9dad..844213c35 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -50,6 +50,8 @@ func (tid ThreadID) String() string { const InitTID ThreadID = 1 // A TaskSet comprises all tasks in a system. +// +// +stateify savable type TaskSet struct { // mu protects all relationships betweens tasks and thread groups in the // TaskSet. (mu is approximately equivalent to Linux's tasklist_lock.) @@ -110,6 +112,8 @@ func (ts *TaskSet) forEachThreadGroupLocked(f func(tg *ThreadGroup)) { // // N.B. A task is said to be visible in a PID namespace if the PID namespace // contains a thread ID that maps to that task. +// +// +stateify savable type PIDNamespace struct { // owner is the TaskSet that this PID namespace belongs to. The owner // pointer is immutable. @@ -263,6 +267,8 @@ func (ns *PIDNamespace) UserNamespace() *auth.UserNamespace { // (threadGroupNode is an anonymous field in ThreadGroup; this is to expose // threadGroupEntry's methods on ThreadGroup to make it implement // threadGroupLinker.) +// +// +stateify savable type threadGroupNode struct { // pidns is the PID namespace containing the thread group and all of its // member tasks. The pidns pointer is immutable. @@ -382,6 +388,8 @@ func (tg *ThreadGroup) ID() ThreadID { // A taskNode defines the relationship between a task and the rest of the // system. The comments on threadGroupNode also apply to taskNode. +// +// +stateify savable type taskNode struct { // tg is the thread group that this task belongs to. The tg pointer is // immutable. diff --git a/pkg/sentry/kernel/time/BUILD b/pkg/sentry/kernel/time/BUILD index b3ed42aa4..5d8db2273 100644 --- a/pkg/sentry/kernel/time/BUILD +++ b/pkg/sentry/kernel/time/BUILD @@ -1,30 +1,18 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "time_state", - srcs = [ - "time.go", - ], - out = "time_state.go", - package = "time", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "time", srcs = [ "context.go", "time.go", - "time_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/log", "//pkg/sentry/context", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go index c223c2f19..6eadd2878 100644 --- a/pkg/sentry/kernel/time/time.go +++ b/pkg/sentry/kernel/time/time.go @@ -42,6 +42,8 @@ const ( // // Time may represent time with respect to any clock and may not have any // meaning in the real world. +// +// +stateify savable type Time struct { ns int64 } @@ -286,6 +288,8 @@ type TimerListener interface { } // Setting contains user-controlled mutable Timer properties. +// +// +stateify savable type Setting struct { // Enabled is true if the timer is running. Enabled bool @@ -371,6 +375,8 @@ func (s Setting) advancedTo(now Time) (Setting, uint64) { // // Timers should be created using NewTimer and must be cleaned up by calling // Timer.Destroy when no longer used. +// +// +stateify savable type Timer struct { // clock is the time source. clock is immutable. clock Clock diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go index 4de8ac13b..df5dbe128 100644 --- a/pkg/sentry/kernel/timekeeper.go +++ b/pkg/sentry/kernel/timekeeper.go @@ -25,6 +25,8 @@ import ( ) // Timekeeper manages all of the kernel clocks. +// +// +stateify savable type Timekeeper struct { // clocks are the clock sources. // diff --git a/pkg/sentry/kernel/timer.go b/pkg/sentry/kernel/timer.go index 03a3310be..534d03d0f 100644 --- a/pkg/sentry/kernel/timer.go +++ b/pkg/sentry/kernel/timer.go @@ -26,6 +26,8 @@ import ( // timekeeperClock is a ktime.Clock that reads time from a // kernel.Timekeeper-managed clock. +// +// +stateify savable type timekeeperClock struct { tk *Timekeeper c sentrytime.ClockID @@ -49,6 +51,8 @@ func (tc *timekeeperClock) Now() ktime.Time { // tgClock is a ktime.Clock that measures the time a thread group has spent // executing. +// +// +stateify savable type tgClock struct { tg *ThreadGroup @@ -155,6 +159,8 @@ func (tc *taskClock) Now() ktime.Time { } // signalNotifier is a ktime.Listener that sends signals to a ThreadGroup. +// +// +stateify savable type signalNotifier struct { tg *ThreadGroup signal linux.Signal @@ -179,6 +185,8 @@ func (s *signalNotifier) Notify(exp uint64) { func (s *signalNotifier) Destroy() {} // TimerManager is a collection of supported process cpu timers. +// +// +stateify savable type TimerManager struct { // Clocks used to drive thread group execution time timers. virtClock *tgClock diff --git a/pkg/sentry/kernel/uts_namespace.go b/pkg/sentry/kernel/uts_namespace.go index 58e9b4d1b..7e0fe0d21 100644 --- a/pkg/sentry/kernel/uts_namespace.go +++ b/pkg/sentry/kernel/uts_namespace.go @@ -22,6 +22,8 @@ import ( // UTSNamespace represents a UTS namespace, a holder of two system identifiers: // the hostname and domain name. +// +// +stateify savable type UTSNamespace struct { // mu protects all fields below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/vdso.go b/pkg/sentry/kernel/vdso.go index 0bacbea49..971e8bc59 100644 --- a/pkg/sentry/kernel/vdso.go +++ b/pkg/sentry/kernel/vdso.go @@ -52,6 +52,8 @@ type vdsoParams struct { // Everything in the struct is 8 bytes for easy alignment. // // It must be kept in sync with params in vdso/vdso_time.cc. +// +// +stateify savable type VDSOParamPage struct { // The parameter page is fr, allocated from platform.Memory(). platform platform.Platform diff --git a/pkg/sentry/limits/BUILD b/pkg/sentry/limits/BUILD index 3ce41cacc..90f4395d4 100644 --- a/pkg/sentry/limits/BUILD +++ b/pkg/sentry/limits/BUILD @@ -1,22 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "limits_state", - srcs = [ - "limits.go", - ], - out = "limits_state.go", - package = "limits", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "limits", srcs = [ "context.go", "limits.go", - "limits_state.go", "linux.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/limits", @@ -24,7 +14,6 @@ go_library( deps = [ "//pkg/abi/linux", "//pkg/sentry/context", - "//pkg/state", ], ) diff --git a/pkg/sentry/limits/limits.go b/pkg/sentry/limits/limits.go index 4230ba958..02c8b60e3 100644 --- a/pkg/sentry/limits/limits.go +++ b/pkg/sentry/limits/limits.go @@ -47,6 +47,8 @@ const ( const Infinity = ^uint64(0) // Limit specifies a system limit. +// +// +stateify savable type Limit struct { // Cur specifies the current limit. Cur uint64 @@ -55,6 +57,8 @@ type Limit struct { } // LimitSet represents the Limits that correspond to each LimitType. +// +// +stateify savable type LimitSet struct { mu sync.Mutex `state:"nosave"` data map[LimitType]Limit diff --git a/pkg/sentry/loader/BUILD b/pkg/sentry/loader/BUILD index e63052c6d..0beb4561b 100644 --- a/pkg/sentry/loader/BUILD +++ b/pkg/sentry/loader/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("@io_bazel_rules_go//go:def.bzl", "go_embed_data") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library") go_embed_data( name = "vdso_bin", @@ -10,23 +10,12 @@ go_embed_data( var = "vdsoBin", ) -go_stateify( - name = "loader_state", - srcs = [ - "vdso.go", - "vdso_state.go", - ], - out = "loader_state.go", - package = "loader", -) - go_library( name = "loader", srcs = [ "elf.go", "interpreter.go", "loader.go", - "loader_state.go", "vdso.go", "vdso_state.go", ":vdso_bin", @@ -40,7 +29,6 @@ go_library( "//pkg/cpuid", "//pkg/log", "//pkg/rand", - "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", @@ -55,7 +43,6 @@ go_library( "//pkg/sentry/uniqueid", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index 2e8693f8e..a06e27ac9 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -193,6 +193,8 @@ func validateVDSO(ctx context.Context, f *fs.File, size uint64) (elfInfo, error) // // NOTE: to support multiple architectures or operating systems, this // would need to contain a VDSO for each. +// +// +stateify savable type VDSO struct { // ParamPage is the VDSO parameter page. This page should be updated to // inform the VDSO for timekeeping data. diff --git a/pkg/sentry/loader/vdso_state.go b/pkg/sentry/loader/vdso_state.go index 92004ad9e..dc71e1c2d 100644 --- a/pkg/sentry/loader/vdso_state.go +++ b/pkg/sentry/loader/vdso_state.go @@ -18,6 +18,7 @@ import ( "debug/elf" ) +// +stateify savable type elfProgHeader struct { Type elf.ProgType Flags elf.ProgFlag diff --git a/pkg/sentry/memmap/BUILD b/pkg/sentry/memmap/BUILD index 2e367e189..c9e0b95a0 100644 --- a/pkg/sentry/memmap/BUILD +++ b/pkg/sentry/memmap/BUILD @@ -1,18 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "memmap_state", - srcs = [ - "mappable_range.go", - "mapping_set.go", - "mapping_set_impl.go", - ], - out = "memmap_state.go", - package = "memmap", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "mappable_range", @@ -46,7 +35,6 @@ go_library( "mapping_set.go", "mapping_set_impl.go", "memmap.go", - "memmap_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/memmap", visibility = ["//pkg/sentry:internal"], @@ -56,7 +44,6 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/platform", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/memmap/mapping_set.go b/pkg/sentry/memmap/mapping_set.go index 0cd42ffbf..c9483905d 100644 --- a/pkg/sentry/memmap/mapping_set.go +++ b/pkg/sentry/memmap/mapping_set.go @@ -35,6 +35,8 @@ import ( type MappingsOfRange map[MappingOfRange]struct{} // MappingOfRange represents a mapping of a MappableRange. +// +// +stateify savable type MappingOfRange struct { MappingSpace MappingSpace AddrRange usermem.AddrRange diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index 3f396986a..bbdfae247 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -1,24 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "mm_state", - srcs = [ - "aio_context.go", - "aio_context_state.go", - "file_refcount_set.go", - "io_list.go", - "mm.go", - "pma_set.go", - "save_restore.go", - "special_mappable.go", - "vma_set.go", - ], - out = "mm_state.go", - package = "mm", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "file_refcount_set", @@ -101,7 +84,6 @@ go_library( "lifecycle.go", "metadata.go", "mm.go", - "mm_state.go", "pma.go", "pma_set.go", "proc_pid_maps.go", @@ -131,7 +113,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/sync", "//pkg/syserror", "//pkg/tcpip/buffer", diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index 992bde5a5..b42156d45 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -28,6 +28,8 @@ import ( ) // aioManager creates and manages asynchronous I/O contexts. +// +// +stateify savable type aioManager struct { // mu protects below. mu sync.Mutex `state:"nosave"` @@ -89,12 +91,16 @@ func (a *aioManager) lookupAIOContext(id uint64) (*AIOContext, bool) { } // ioResult is a completed I/O operation. +// +// +stateify savable type ioResult struct { data interface{} ioEntry } // AIOContext is a single asynchronous I/O context. +// +// +stateify savable type AIOContext struct { // done is the notification channel used for all requests. done chan struct{} `state:"nosave"` @@ -190,6 +196,8 @@ func (ctx *AIOContext) WaitChannel() (chan struct{}, bool) { // aioMappable implements memmap.MappingIdentity and memmap.Mappable for AIO // ring buffers. +// +// +stateify savable type aioMappable struct { refs.AtomicRefCount diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index ce8097b7f..3299ae164 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -46,6 +46,8 @@ import ( ) // MemoryManager implements a virtual address space. +// +// +stateify savable type MemoryManager struct { // p is the platform. // @@ -207,6 +209,8 @@ type MemoryManager struct { } // vma represents a virtual memory area. +// +// +stateify savable type vma struct { // mappable is the virtual memory object mapped by this vma. If mappable is // nil, the vma represents a private anonymous mapping. @@ -346,6 +350,8 @@ func (v *vma) loadRealPerms(b int) { } // pma represents a platform mapping area. +// +// +stateify savable type pma struct { // file is the file mapped by this pma. Only pmas for which file == // platform.Platform.Memory() may be saved. pmas hold a reference to the @@ -380,6 +386,7 @@ type pma struct { internalMappings safemem.BlockSeq `state:"nosave"` } +// +stateify savable type privateRefs struct { mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index 9d3614034..aa2f87107 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -28,6 +28,8 @@ import ( // semantics similar to Linux's mm/mmap.c:_install_special_mapping(), except // that SpecialMappable takes ownership of the memory that it represents // (_install_special_mapping() does not.) +// +// +stateify savable type SpecialMappable struct { refs.AtomicRefCount diff --git a/pkg/sentry/platform/BUILD b/pkg/sentry/platform/BUILD index 15a7fbbc3..af9ba5394 100644 --- a/pkg/sentry/platform/BUILD +++ b/pkg/sentry/platform/BUILD @@ -1,16 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "platform_state", - srcs = [ - "file_range.go", - ], - out = "platform_state.go", - package = "platform", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_template_instance( name = "file_range", @@ -30,7 +21,6 @@ go_library( "file_range.go", "mmap_min_addr.go", "platform.go", - "platform_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/platform", visibility = ["//pkg/sentry:internal"], @@ -44,7 +34,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/platform/filemem/BUILD b/pkg/sentry/platform/filemem/BUILD index dadba1d38..2a5982763 100644 --- a/pkg/sentry/platform/filemem/BUILD +++ b/pkg/sentry/platform/filemem/BUILD @@ -1,18 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "filemem_autogen_state", - srcs = [ - "filemem.go", - "filemem_state.go", - "usage_set.go", - ], - out = "filemem_autogen_state.go", - package = "filemem", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "usage_set", @@ -38,7 +27,6 @@ go_library( name = "filemem", srcs = [ "filemem.go", - "filemem_autogen_state.go", "filemem_state.go", "filemem_unsafe.go", "usage_set.go", diff --git a/pkg/sentry/platform/filemem/filemem.go b/pkg/sentry/platform/filemem/filemem.go index 870274ae1..feb020ef8 100644 --- a/pkg/sentry/platform/filemem/filemem.go +++ b/pkg/sentry/platform/filemem/filemem.go @@ -155,6 +155,8 @@ type FileMem struct { } // usage tracks usage information. +// +// +stateify savable type usageInfo struct { // kind is the usage kind. kind usage.MemoryKind diff --git a/pkg/sentry/socket/BUILD b/pkg/sentry/socket/BUILD index 929787aa0..a320fca0b 100644 --- a/pkg/sentry/socket/BUILD +++ b/pkg/sentry/socket/BUILD @@ -1,22 +1,10 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "socket_state", - srcs = [ - "socket.go", - ], - out = "socket_state_autogen.go", - package = "socket", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "socket", - srcs = [ - "socket.go", - "socket_state_autogen.go", - ], + srcs = ["socket.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket", visibility = ["//pkg/sentry:internal"], deps = [ @@ -29,7 +17,6 @@ go_library( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/tcpip", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/socket/control/BUILD b/pkg/sentry/socket/control/BUILD index faf2b4c27..c4874fdfb 100644 --- a/pkg/sentry/socket/control/BUILD +++ b/pkg/sentry/socket/control/BUILD @@ -1,26 +1,14 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "control_state", - srcs = [ - "control.go", - ], - out = "control_state.go", - imports = [ - "gvisor.googlesource.com/gvisor/pkg/sentry/fs", - ], - package = "control", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "control", - srcs = [ - "control.go", - "control_state.go", - ], + srcs = ["control.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control", + imports = [ + "gvisor.googlesource.com/gvisor/pkg/sentry/fs", + ], visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", @@ -31,7 +19,6 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/kdefs", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", ], diff --git a/pkg/sentry/socket/control/control.go b/pkg/sentry/socket/control/control.go index 17ecdd11c..c31182e69 100644 --- a/pkg/sentry/socket/control/control.go +++ b/pkg/sentry/socket/control/control.go @@ -51,6 +51,8 @@ type SCMRights interface { // RightsFiles represents a SCM_RIGHTS socket control message. A reference is // maintained for each fs.File and is release either when an FD is created or // when the Release method is called. +// +// +stateify savable type RightsFiles []*fs.File // NewSCMRights creates a new SCM_RIGHTS socket control message representation @@ -128,6 +130,8 @@ func PackRights(t *kernel.Task, rights SCMRights, cloexec bool, buf []byte) []by } // scmCredentials represents an SCM_CREDENTIALS socket control message. +// +// +stateify savable type scmCredentials struct { t *kernel.Task kuid auth.KUID diff --git a/pkg/sentry/socket/epsocket/BUILD b/pkg/sentry/socket/epsocket/BUILD index 7ad5e88c5..49af8db85 100644 --- a/pkg/sentry/socket/epsocket/BUILD +++ b/pkg/sentry/socket/epsocket/BUILD @@ -1,24 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "epsocket_state", - srcs = [ - "epsocket.go", - "save_restore.go", - "stack.go", - ], - out = "epsocket_state.go", - package = "epsocket", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "epsocket", srcs = [ "device.go", "epsocket.go", - "epsocket_state.go", "provider.go", "save_restore.go", "stack.go", @@ -31,7 +19,6 @@ go_library( "//pkg/abi/linux", "//pkg/binary", "//pkg/log", - "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/device", @@ -44,7 +31,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index a2927e1b9..f969a1d7c 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -95,6 +95,8 @@ type commonEndpoint interface { // SocketOperations encapsulates all the state needed to represent a network stack // endpoint in the kernel context. +// +// +stateify savable type SocketOperations struct { socket.ReceiveTimeout fsutil.PipeSeek `state:"nosave"` diff --git a/pkg/sentry/socket/epsocket/stack.go b/pkg/sentry/socket/epsocket/stack.go index ec1d96ccb..12b4b4767 100644 --- a/pkg/sentry/socket/epsocket/stack.go +++ b/pkg/sentry/socket/epsocket/stack.go @@ -26,6 +26,8 @@ import ( ) // Stack implements inet.Stack for netstack/tcpip/stack.Stack. +// +// +stateify savable type Stack struct { Stack *stack.Stack `state:"manual"` } diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD index 227ca3926..d623718b3 100644 --- a/pkg/sentry/socket/hostinet/BUILD +++ b/pkg/sentry/socket/hostinet/BUILD @@ -1,24 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "hostinet_state", - srcs = [ - "save_restore.go", - "socket.go", - "stack.go", - ], - out = "hostinet_autogen_state.go", - package = "hostinet", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "hostinet", srcs = [ "device.go", "hostinet.go", - "hostinet_autogen_state.go", "save_restore.go", "socket.go", "socket_unsafe.go", @@ -42,7 +30,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/socket/netlink/BUILD b/pkg/sentry/socket/netlink/BUILD index b23a243f7..b852165f7 100644 --- a/pkg/sentry/socket/netlink/BUILD +++ b/pkg/sentry/socket/netlink/BUILD @@ -1,21 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "netlink_state", - srcs = [ - "socket.go", - ], - out = "netlink_state.go", - package = "netlink", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "netlink", srcs = [ "message.go", - "netlink_state.go", "provider.go", "socket.go", ], @@ -36,7 +26,6 @@ go_library( "//pkg/sentry/socket/netlink/port", "//pkg/sentry/socket/unix", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/netlink/port/BUILD b/pkg/sentry/socket/netlink/port/BUILD index ba6f686e4..3a7dbc5ed 100644 --- a/pkg/sentry/socket/netlink/port/BUILD +++ b/pkg/sentry/socket/netlink/port/BUILD @@ -1,23 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "port_state", - srcs = ["port.go"], - out = "port_state.go", - package = "port", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "port", - srcs = [ - "port.go", - "port_state.go", - ], + srcs = ["port.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/port", visibility = ["//pkg/sentry:internal"], - deps = ["//pkg/state"], ) go_test( diff --git a/pkg/sentry/socket/netlink/port/port.go b/pkg/sentry/socket/netlink/port/port.go index 4ccf0b84c..1c5d4c3a5 100644 --- a/pkg/sentry/socket/netlink/port/port.go +++ b/pkg/sentry/socket/netlink/port/port.go @@ -32,6 +32,8 @@ import ( const maxPorts = 10000 // Manager allocates netlink port IDs. +// +// +stateify savable type Manager struct { // mu protects the fields below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/socket/netlink/route/BUILD b/pkg/sentry/socket/netlink/route/BUILD index 726469fc9..e1bcfe252 100644 --- a/pkg/sentry/socket/netlink/route/BUILD +++ b/pkg/sentry/socket/netlink/route/BUILD @@ -1,32 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "route_state", - srcs = ["protocol.go"], - out = "route_state.go", - package = "route", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "route", - srcs = [ - "protocol.go", - "route_state.go", - ], + srcs = ["protocol.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/route", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", "//pkg/sentry/context", - "//pkg/sentry/fs", "//pkg/sentry/inet", "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/socket/netlink", - "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", ], ) diff --git a/pkg/sentry/socket/netlink/route/protocol.go b/pkg/sentry/socket/netlink/route/protocol.go index e8030c518..55a76e916 100644 --- a/pkg/sentry/socket/netlink/route/protocol.go +++ b/pkg/sentry/socket/netlink/route/protocol.go @@ -43,6 +43,8 @@ func typeKind(typ uint16) commandKind { } // Protocol implements netlink.Protocol. +// +// +stateify savable type Protocol struct{} var _ netlink.Protocol = (*Protocol)(nil) diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index 0b8f528d0..e15d1546c 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -51,6 +51,8 @@ var netlinkSocketDevice = device.NewAnonDevice() // to/from the kernel. // // Socket implements socket.Socket. +// +// +stateify savable type Socket struct { socket.ReceiveTimeout fsutil.PipeSeek `state:"nosave"` diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index bd4858a34..54fe64595 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -195,6 +195,8 @@ func NewDirent(ctx context.Context, d *device.Device) *fs.Dirent { // // Care must be taken when copying ReceiveTimeout as it contains atomic // variables. +// +// +stateify savable type ReceiveTimeout struct { // ns is length of the timeout in nanoseconds. // diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index 7d04d6b6b..9fe681e9a 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -1,15 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "unix_state", - srcs = [ - "unix.go", - ], - out = "unix_state.go", - package = "unix", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "unix", @@ -17,7 +8,6 @@ go_library( "device.go", "io.go", "unix.go", - "unix_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix", visibility = ["//pkg/sentry:internal"], @@ -37,7 +27,6 @@ go_library( "//pkg/sentry/socket/control", "//pkg/sentry/socket/epsocket", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 27bacbbc3..5b6411f97 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -42,6 +42,8 @@ import ( // SocketOperations is a Unix socket. It is similar to an epsocket, except it is backed // by a unix.Endpoint instead of a tcpip.Endpoint. +// +// +stateify savable type SocketOperations struct { refs.AtomicRefCount socket.ReceiveTimeout diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index 574621ad2..e4450a093 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -1,18 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "linux_state", - srcs = [ - "sys_aio.go", - "sys_futex.go", - "sys_poll.go", - "sys_time.go", - ], - out = "linux_state.go", - package = "linux", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "linux", @@ -20,7 +8,6 @@ go_library( "error.go", "flags.go", "linux64.go", - "linux_state.go", "sigset.go", "sys_aio.go", "sys_capability.go", @@ -66,7 +53,6 @@ go_library( "//pkg/abi/linux", "//pkg/binary", "//pkg/bpf", - "//pkg/eventchannel", "//pkg/log", "//pkg/metric", "//pkg/rand", @@ -74,7 +60,6 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", - "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/lock", "//pkg/sentry/fs/timerfd", "//pkg/sentry/kernel", @@ -85,7 +70,6 @@ go_library( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/pipe", "//pkg/sentry/kernel/sched", - "//pkg/sentry/kernel/semaphore", "//pkg/sentry/kernel/shm", "//pkg/sentry/kernel/time", "//pkg/sentry/limits", @@ -97,8 +81,6 @@ go_library( "//pkg/sentry/syscalls", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", - "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/syscalls/linux/sys_aio.go b/pkg/sentry/syscalls/linux/sys_aio.go index fc3397081..54e4afa9e 100644 --- a/pkg/sentry/syscalls/linux/sys_aio.go +++ b/pkg/sentry/syscalls/linux/sys_aio.go @@ -69,6 +69,8 @@ type ioCallback struct { } // ioEvent describes an I/O result. +// +// +stateify savable type ioEvent struct { Data uint64 Obj uint64 diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index 57762d058..1a0e1f5fb 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -132,6 +132,8 @@ func (f futexChecker) Op(addr uintptr, opIn uint32) (bool, error) { // futexWaitRestartBlock encapsulates the state required to restart futex(2) // via restart_syscall(2). +// +// +stateify savable type futexWaitRestartBlock struct { duration time.Duration diff --git a/pkg/sentry/syscalls/linux/sys_poll.go b/pkg/sentry/syscalls/linux/sys_poll.go index d4dbfd285..b9bdefadb 100644 --- a/pkg/sentry/syscalls/linux/sys_poll.go +++ b/pkg/sentry/syscalls/linux/sys_poll.go @@ -274,6 +274,8 @@ func copyOutTimevalRemaining(t *kernel.Task, startNs ktime.Time, timeout time.Du // pollRestartBlock encapsulates the state required to restart poll(2) via // restart_syscall(2). +// +// +stateify savable type pollRestartBlock struct { pfdAddr usermem.Addr nfds uint diff --git a/pkg/sentry/syscalls/linux/sys_time.go b/pkg/sentry/syscalls/linux/sys_time.go index dcee694b2..8e6683444 100644 --- a/pkg/sentry/syscalls/linux/sys_time.go +++ b/pkg/sentry/syscalls/linux/sys_time.go @@ -168,6 +168,8 @@ func Time(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallC // clockNanosleepRestartBlock encapsulates the state required to restart // clock_nanosleep(2) via restart_syscall(2). +// +// +stateify savable type clockNanosleepRestartBlock struct { c ktime.Clock duration time.Duration diff --git a/pkg/sentry/usage/BUILD b/pkg/sentry/usage/BUILD index edee44d96..868dfd400 100644 --- a/pkg/sentry/usage/BUILD +++ b/pkg/sentry/usage/BUILD @@ -1,17 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "usage_state", - srcs = [ - "cpu.go", - "io.go", - "memory.go", - ], - out = "usage_state.go", - package = "usage", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "usage", @@ -21,7 +10,6 @@ go_library( "memory.go", "memory_unsafe.go", "usage.go", - "usage_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/usage", visibility = [ @@ -29,9 +17,6 @@ go_library( ], deps = [ "//pkg/bits", - "//pkg/log", "//pkg/sentry/memutil", - "//pkg/state", - "@org_golang_x_sys//unix:go_default_library", ], ) diff --git a/pkg/sentry/usage/cpu.go b/pkg/sentry/usage/cpu.go index 1c2cc90e1..ed7b04b9e 100644 --- a/pkg/sentry/usage/cpu.go +++ b/pkg/sentry/usage/cpu.go @@ -20,6 +20,8 @@ import ( // CPUStats contains the subset of struct rusage fields that relate to CPU // scheduling. +// +// +stateify savable type CPUStats struct { // UserTime is the amount of time spent executing application code. UserTime time.Duration diff --git a/pkg/sentry/usage/io.go b/pkg/sentry/usage/io.go index a05053c32..49faa507d 100644 --- a/pkg/sentry/usage/io.go +++ b/pkg/sentry/usage/io.go @@ -19,6 +19,8 @@ import ( ) // IO contains I/O-related statistics. +// +// +stateify savable type IO struct { // CharsRead is the number of bytes read by read syscalls. CharsRead uint64 diff --git a/pkg/sentry/usermem/BUILD b/pkg/sentry/usermem/BUILD index 9dd1cd2b5..69ba919e0 100644 --- a/pkg/sentry/usermem/BUILD +++ b/pkg/sentry/usermem/BUILD @@ -1,19 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "usermem_state", - srcs = [ - "access_type.go", - "addr.go", - "addr_range.go", - "addr_range_seq_unsafe.go", - ], - out = "usermem_state.go", - package = "usermem", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "addr_range", @@ -36,7 +24,6 @@ go_library( "bytes_io.go", "bytes_io_unsafe.go", "usermem.go", - "usermem_state.go", "usermem_x86.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/usermem", @@ -47,7 +34,6 @@ go_library( "//pkg/log", "//pkg/sentry/context", "//pkg/sentry/safemem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/buffer", ], diff --git a/pkg/sentry/usermem/access_type.go b/pkg/sentry/usermem/access_type.go index 7eabecf30..75346d854 100644 --- a/pkg/sentry/usermem/access_type.go +++ b/pkg/sentry/usermem/access_type.go @@ -20,6 +20,8 @@ import ( // AccessType specifies memory access types. This is used for // setting mapping permissions, as well as communicating faults. +// +// +stateify savable type AccessType struct { // Read is read access. Read bool diff --git a/pkg/sentry/usermem/addr.go b/pkg/sentry/usermem/addr.go index d175fdc74..fc94bee80 100644 --- a/pkg/sentry/usermem/addr.go +++ b/pkg/sentry/usermem/addr.go @@ -19,6 +19,8 @@ import ( ) // Addr represents a generic virtual address. +// +// +stateify savable type Addr uintptr // AddLength adds the given length to start and returns the result. ok is true diff --git a/pkg/tcpip/BUILD b/pkg/tcpip/BUILD index 391d801d0..5153bd3b4 100644 --- a/pkg/tcpip/BUILD +++ b/pkg/tcpip/BUILD @@ -1,26 +1,13 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tcpip_state", - srcs = [ - "tcpip.go", - ], - out = "tcpip_state.go", - package = "tcpip", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tcpip", - srcs = [ - "tcpip.go", - "tcpip_state.go", - ], + srcs = ["tcpip.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip", visibility = ["//visibility:public"], deps = [ - "//pkg/state", "//pkg/tcpip/buffer", "//pkg/waiter", ], diff --git a/pkg/tcpip/buffer/BUILD b/pkg/tcpip/buffer/BUILD index efeb6a448..11a725423 100644 --- a/pkg/tcpip/buffer/BUILD +++ b/pkg/tcpip/buffer/BUILD @@ -1,26 +1,15 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "buffer_state", - srcs = [ - "view.go", - ], - out = "buffer_state.go", - package = "buffer", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "buffer", srcs = [ - "buffer_state.go", "prependable.go", "view.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer", visibility = ["//visibility:public"], - deps = ["//pkg/state"], ) go_test( diff --git a/pkg/tcpip/buffer/view.go b/pkg/tcpip/buffer/view.go index a5774a327..bbb4e1d24 100644 --- a/pkg/tcpip/buffer/view.go +++ b/pkg/tcpip/buffer/view.go @@ -54,6 +54,8 @@ func (v *View) ToVectorisedView(views [1]View) VectorisedView { // VectorisedView is a vectorised version of View using non contigous memory. // It supports all the convenience methods supported by View. +// +// +stateify savable type VectorisedView struct { views []View size int diff --git a/pkg/tcpip/header/BUILD b/pkg/tcpip/header/BUILD index 3aa2cfb24..8f22ba3a5 100644 --- a/pkg/tcpip/header/BUILD +++ b/pkg/tcpip/header/BUILD @@ -1,15 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tcp_header_state", - srcs = [ - "tcp.go", - ], - out = "tcp_header_state.go", - package = "header", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "header", @@ -25,13 +16,11 @@ go_library( "ipv6.go", "ipv6_fragment.go", "tcp.go", - "tcp_header_state.go", "udp.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/header", visibility = ["//visibility:public"], deps = [ - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/seqnum", ], diff --git a/pkg/tcpip/header/tcp.go b/pkg/tcpip/header/tcp.go index a95d282b0..6689a6dc5 100644 --- a/pkg/tcpip/header/tcp.go +++ b/pkg/tcpip/header/tcp.go @@ -120,6 +120,8 @@ type TCPSynOptions struct { } // SACKBlock represents a single contiguous SACK block. +// +// +stateify savable type SACKBlock struct { // Start indicates the lowest sequence number in the block. Start seqnum.Value @@ -131,6 +133,8 @@ type SACKBlock struct { // TCPOptions are used to parse and cache the TCP segment options for a non // syn/syn-ack segment. +// +// +stateify savable type TCPOptions struct { // TS is true if the TimeStamp option is enabled. TS bool diff --git a/pkg/tcpip/network/fragmentation/BUILD b/pkg/tcpip/network/fragmentation/BUILD index ac97ebe43..83b4d253f 100644 --- a/pkg/tcpip/network/fragmentation/BUILD +++ b/pkg/tcpip/network/fragmentation/BUILD @@ -1,14 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "fragmentation_state", - srcs = ["reassembler_list.go"], - out = "fragmentation_state.go", - package = "fragmentation", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "reassembler_list", @@ -26,7 +19,6 @@ go_library( srcs = [ "frag_heap.go", "fragmentation.go", - "fragmentation_state.go", "reassembler.go", "reassembler_list.go", ], @@ -34,7 +26,6 @@ go_library( visibility = ["//:sandbox"], deps = [ "//pkg/log", - "//pkg/state", "//pkg/tcpip/buffer", ], ) diff --git a/pkg/tcpip/seqnum/BUILD b/pkg/tcpip/seqnum/BUILD index a75869dac..c5c889239 100644 --- a/pkg/tcpip/seqnum/BUILD +++ b/pkg/tcpip/seqnum/BUILD @@ -1,25 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "seqnum_state", - srcs = [ - "seqnum.go", - ], - out = "seqnum_state.go", - package = "seqnum", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "seqnum", - srcs = [ - "seqnum.go", - "seqnum_state.go", - ], + srcs = ["seqnum.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/seqnum", visibility = [ "//visibility:public", ], - deps = ["//pkg/state"], ) diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index eb1e4645d..af0aec85c 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -213,6 +213,8 @@ const ( // FullAddress represents a full transport node address, as required by the // Connect() and Bind() methods. +// +// +stateify savable type FullAddress struct { // NIC is the ID of the NIC this address refers to. // @@ -256,6 +258,8 @@ func (s SlicePayload) Size() int { } // A ControlMessages contains socket control messages for IP sockets. +// +// +stateify savable type ControlMessages struct { // HasTimestamp indicates whether Timestamp is valid/set. HasTimestamp bool diff --git a/pkg/tcpip/transport/ping/BUILD b/pkg/tcpip/transport/ping/BUILD index 28e3e1700..117532fea 100644 --- a/pkg/tcpip/transport/ping/BUILD +++ b/pkg/tcpip/transport/ping/BUILD @@ -1,19 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "ping_state", - srcs = [ - "endpoint.go", - "endpoint_state.go", - "ping_packet_list.go", - ], - out = "ping_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], - package = "ping", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_template_instance( name = "ping_packet_list", @@ -32,14 +20,13 @@ go_library( "endpoint.go", "endpoint_state.go", "ping_packet_list.go", - "ping_state.go", "protocol.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/ping", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/sleep", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/ping/endpoint.go b/pkg/tcpip/transport/ping/endpoint.go index f15e44b61..a22684de9 100644 --- a/pkg/tcpip/transport/ping/endpoint.go +++ b/pkg/tcpip/transport/ping/endpoint.go @@ -26,6 +26,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) +// +stateify savable type pingPacket struct { pingPacketEntry senderAddress tcpip.FullAddress diff --git a/pkg/tcpip/transport/queue/BUILD b/pkg/tcpip/transport/queue/BUILD index fb878ad36..6dcec312e 100644 --- a/pkg/tcpip/transport/queue/BUILD +++ b/pkg/tcpip/transport/queue/BUILD @@ -1,27 +1,14 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "queue_state", - srcs = [ - "queue.go", - ], - out = "queue_state.go", - package = "queue", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "queue", - srcs = [ - "queue.go", - "queue_state.go", - ], + srcs = ["queue.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/queue", visibility = ["//:sandbox"], deps = [ "//pkg/ilist", - "//pkg/state", "//pkg/tcpip", "//pkg/waiter", ], diff --git a/pkg/tcpip/transport/queue/queue.go b/pkg/tcpip/transport/queue/queue.go index 6a17441ae..eb9ee8a3f 100644 --- a/pkg/tcpip/transport/queue/queue.go +++ b/pkg/tcpip/transport/queue/queue.go @@ -33,6 +33,8 @@ type Entry interface { } // Queue is a buffer queue. +// +// +stateify savable type Queue struct { ReaderQueue *waiter.Queue WriterQueue *waiter.Queue diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index 6a7153e4d..9ebae6cc7 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -1,27 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tcp_state", - srcs = [ - "endpoint.go", - "endpoint_state.go", - "rcv.go", - "reno.go", - "segment.go", - "segment_heap.go", - "segment_queue.go", - "segment_state.go", - "snd.go", - "snd_state.go", - "tcp_segment_list.go", - ], - out = "tcp_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], - package = "tcp", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "tcp_segment_list", @@ -53,15 +33,14 @@ go_library( "snd.go", "snd_state.go", "tcp_segment_list.go", - "tcp_state.go", "timer.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/tcp", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/rand", "//pkg/sleep", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index 5b8a1e20f..de1883d84 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -54,6 +54,8 @@ const ( ) // SACKInfo holds TCP SACK related information for a given endpoint. +// +// +stateify savable type SACKInfo struct { // Blocks is the maximum number of SACK blocks we track // per endpoint. @@ -69,6 +71,8 @@ type SACKInfo struct { // have concurrent goroutines make calls into the endpoint, they are properly // synchronized. The protocol implementation, however, runs in a single // goroutine. +// +// +stateify savable type endpoint struct { // workMu is used to arbitrate which goroutine may perform protocol // work. Only the main protocol goroutine is expected to call Lock() on diff --git a/pkg/tcpip/transport/tcp/rcv.go b/pkg/tcpip/transport/tcp/rcv.go index b22a00ce1..92ef9c6f7 100644 --- a/pkg/tcpip/transport/tcp/rcv.go +++ b/pkg/tcpip/transport/tcp/rcv.go @@ -22,6 +22,8 @@ import ( // receiver holds the state necessary to receive TCP segments and turn them // into a stream of bytes. +// +// +stateify savable type receiver struct { ep *endpoint diff --git a/pkg/tcpip/transport/tcp/reno.go b/pkg/tcpip/transport/tcp/reno.go index 60f170a27..03ae8d747 100644 --- a/pkg/tcpip/transport/tcp/reno.go +++ b/pkg/tcpip/transport/tcp/reno.go @@ -16,6 +16,8 @@ package tcp // renoState stores the variables related to TCP New Reno congestion // control algorithm. +// +// +stateify savable type renoState struct { s *sender } diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index 40928ba2c..8dccea2ba 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -36,6 +36,8 @@ const ( // segment represents a TCP segment. It holds the payload and parsed TCP segment // information, and can be added to intrusive lists. // segment is mostly immutable, the only field allowed to change is viewToDeliver. +// +// +stateify savable type segment struct { segmentEntry refCnt int32 diff --git a/pkg/tcpip/transport/tcp/segment_queue.go b/pkg/tcpip/transport/tcp/segment_queue.go index 2ddcf5f10..6a2d7bc0b 100644 --- a/pkg/tcpip/transport/tcp/segment_queue.go +++ b/pkg/tcpip/transport/tcp/segment_queue.go @@ -21,6 +21,8 @@ import ( ) // segmentQueue is a bounded, thread-safe queue of TCP segments. +// +// +stateify savable type segmentQueue struct { mu sync.Mutex `state:"nosave"` list segmentList `state:"wait"` diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index e38686e1b..376e81846 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -54,6 +54,8 @@ type congestionControl interface { } // sender holds the state necessary to send TCP segments. +// +// +stateify savable type sender struct { ep *endpoint @@ -133,6 +135,8 @@ type sender struct { } // fastRecovery holds information related to fast recovery from a packet loss. +// +// +stateify savable type fastRecovery struct { // active whether the endpoint is in fast recovery. The following fields // are only meaningful when active is true. diff --git a/pkg/tcpip/transport/tcp/snd_state.go b/pkg/tcpip/transport/tcp/snd_state.go index 33c8867f4..d536839af 100644 --- a/pkg/tcpip/transport/tcp/snd_state.go +++ b/pkg/tcpip/transport/tcp/snd_state.go @@ -18,6 +18,7 @@ import ( "time" ) +// +stateify savable type unixTime struct { second int64 nano int64 diff --git a/pkg/tcpip/transport/udp/BUILD b/pkg/tcpip/transport/udp/BUILD index 790dd55a3..1a3a62d3d 100644 --- a/pkg/tcpip/transport/udp/BUILD +++ b/pkg/tcpip/transport/udp/BUILD @@ -1,19 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "udp_state", - srcs = [ - "endpoint.go", - "endpoint_state.go", - "udp_packet_list.go", - ], - out = "udp_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], - package = "udp", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "udp_packet_list", @@ -33,13 +21,12 @@ go_library( "endpoint_state.go", "protocol.go", "udp_packet_list.go", - "udp_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/udp", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/sleep", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 2a32c3a87..03fb76f92 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -25,6 +25,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) +// +stateify savable type udpPacket struct { udpPacketEntry senderAddress tcpip.FullAddress @@ -49,6 +50,8 @@ const ( // between users of the endpoint and the protocol implementation; it is legal to // have concurrent goroutines make calls into the endpoint, they are properly // synchronized. +// +// +stateify savable type endpoint struct { // The following fields are initialized at creation time and do not // change throughout the lifetime of the endpoint. diff --git a/pkg/tcpip/transport/unix/BUILD b/pkg/tcpip/transport/unix/BUILD index 676f2cf92..dae0bd079 100644 --- a/pkg/tcpip/transport/unix/BUILD +++ b/pkg/tcpip/transport/unix/BUILD @@ -1,17 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "unix_state", - srcs = [ - "connectioned.go", - "connectionless.go", - "unix.go", - ], - out = "unix_state.go", - package = "unix", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "unix", @@ -20,14 +9,11 @@ go_library( "connectioned_state.go", "connectionless.go", "unix.go", - "unix_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix", visibility = ["//:sandbox"], deps = [ "//pkg/ilist", - "//pkg/log", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/transport/queue", diff --git a/pkg/tcpip/transport/unix/connectioned.go b/pkg/tcpip/transport/unix/connectioned.go index 0e63186b2..dd7c03cf1 100644 --- a/pkg/tcpip/transport/unix/connectioned.go +++ b/pkg/tcpip/transport/unix/connectioned.go @@ -85,6 +85,8 @@ type ConnectingEndpoint interface { // path != "" && acceptedChan != nil => bound and listening. // // Only one of these will be true at any moment. +// +// +stateify savable type connectionedEndpoint struct { baseEndpoint diff --git a/pkg/tcpip/transport/unix/connectionless.go b/pkg/tcpip/transport/unix/connectionless.go index 3276ddcd0..2a6ec8b4b 100644 --- a/pkg/tcpip/transport/unix/connectionless.go +++ b/pkg/tcpip/transport/unix/connectionless.go @@ -25,6 +25,8 @@ import ( // // Specifically, this means datagram unix sockets not created with // socketpair(2). +// +// +stateify savable type connectionlessEndpoint struct { baseEndpoint } diff --git a/pkg/tcpip/transport/unix/unix.go b/pkg/tcpip/transport/unix/unix.go index 190a1ccdb..8e4af3139 100644 --- a/pkg/tcpip/transport/unix/unix.go +++ b/pkg/tcpip/transport/unix/unix.go @@ -60,6 +60,8 @@ type CredentialsControlMessage interface { } // A ControlMessages represents a collection of socket control messages. +// +// +stateify savable type ControlMessages struct { // Rights is a control message containing FDs. Rights RightsControlMessage @@ -235,6 +237,8 @@ type BoundEndpoint interface { } // message represents a message passed over a Unix domain socket. +// +// +stateify savable type message struct { ilist.Entry @@ -306,6 +310,8 @@ type Receiver interface { } // queueReceiver implements Receiver for datagram sockets. +// +// +stateify savable type queueReceiver struct { readQueue *queue.Queue } @@ -369,6 +375,8 @@ func (q *queueReceiver) RecvMaxQueueSize() int64 { func (*queueReceiver) Release() {} // streamQueueReceiver implements Receiver for stream sockets. +// +// +stateify savable type streamQueueReceiver struct { queueReceiver @@ -579,6 +587,7 @@ type ConnectedEndpoint interface { Release() } +// +stateify savable type connectedEndpoint struct { // endpoint represents the subset of the Endpoint functionality needed by // the connectedEndpoint. It is implemented by both connectionedEndpoint @@ -671,6 +680,8 @@ func (*connectedEndpoint) Release() {} // unix domain socket Endpoint implementations. // // Not to be used on its own. +// +// +stateify savable type baseEndpoint struct { *waiter.Queue diff --git a/pkg/waiter/BUILD b/pkg/waiter/BUILD index 8256acdb4..5e611c54f 100644 --- a/pkg/waiter/BUILD +++ b/pkg/waiter/BUILD @@ -1,28 +1,13 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "waiter_state", - srcs = [ - "waiter.go", - ], - out = "waiter_state.go", - package = "waiter", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "waiter", - srcs = [ - "waiter.go", - "waiter_state.go", - ], + srcs = ["waiter.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/waiter", visibility = ["//visibility:public"], - deps = [ - "//pkg/ilist", - "//pkg/state", - ], + deps = ["//pkg/ilist"], ) go_test( diff --git a/pkg/waiter/waiter.go b/pkg/waiter/waiter.go index 9b189bb9e..9825880ca 100644 --- a/pkg/waiter/waiter.go +++ b/pkg/waiter/waiter.go @@ -157,6 +157,8 @@ func NewChannelEntry(c chan struct{}) (Entry, chan struct{}) { // notifiers can notify them when events happen. // // The zero value for waiter.Queue is an empty queue ready for use. +// +// +stateify savable type Queue struct { list ilist.List `state:"zerovalue"` mu sync.RWMutex `state:"nosave"` -- cgit v1.2.3 From 60add78980737a7330100d98bf6a214892dee3c0 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Wed, 1 Aug 2018 19:56:12 -0700 Subject: Automated rollback of changelist 207007153 PiperOrigin-RevId: 207037226 Change-Id: I8b5f1a056d4f3eab17846f2e0193bb737ecb5428 --- pkg/abi/BUILD | 13 +++++- pkg/abi/linux/BUILD | 16 ++++++- pkg/abi/linux/bpf.go | 2 - pkg/abi/linux/tty.go | 2 - pkg/bpf/BUILD | 17 +++++++- pkg/bpf/interpreter.go | 2 - pkg/cpuid/BUILD | 15 ++++++- pkg/cpuid/cpuid.go | 2 - pkg/ilist/BUILD | 15 ++++++- pkg/ilist/list.go | 4 -- pkg/segment/range.go | 2 - pkg/segment/set.go | 5 --- pkg/sentry/arch/BUILD | 18 +++++++- pkg/sentry/arch/arch.go | 2 - pkg/sentry/arch/arch_amd64.go | 2 - pkg/sentry/arch/arch_state_x86.go | 1 - pkg/sentry/arch/arch_x86.go | 2 - pkg/sentry/arch/auxv.go | 2 - pkg/sentry/arch/signal_amd64.go | 6 --- pkg/sentry/context/contexttest/BUILD | 17 +++++++- pkg/sentry/fs/BUILD | 36 +++++++++++++++- pkg/sentry/fs/ashmem/BUILD | 17 +++++++- pkg/sentry/fs/ashmem/area.go | 8 ++-- pkg/sentry/fs/ashmem/device.go | 22 +++++----- pkg/sentry/fs/ashmem/pin_board.go | 2 - pkg/sentry/fs/attr.go | 12 ------ pkg/sentry/fs/binder/BUILD | 13 +++++- pkg/sentry/fs/binder/binder.go | 26 +++++------- pkg/sentry/fs/dentry.go | 4 -- pkg/sentry/fs/dev/BUILD | 20 ++++++++- pkg/sentry/fs/dev/dev.go | 2 - pkg/sentry/fs/dev/fs.go | 2 - pkg/sentry/fs/dev/full.go | 2 - pkg/sentry/fs/dev/null.go | 3 -- pkg/sentry/fs/dev/random.go | 1 - pkg/sentry/fs/dirent.go | 2 - pkg/sentry/fs/dirent_cache.go | 2 - pkg/sentry/fs/fdpipe/BUILD | 31 +++++++++++++- pkg/sentry/fs/fdpipe/pipe.go | 2 - pkg/sentry/fs/file.go | 2 - pkg/sentry/fs/file_overlay.go | 4 -- pkg/sentry/fs/filesystems.go | 2 - pkg/sentry/fs/filetest/BUILD | 18 +++++++- pkg/sentry/fs/flags.go | 2 - pkg/sentry/fs/fsutil/BUILD | 20 ++++++++- pkg/sentry/fs/fsutil/dirty_set.go | 2 - pkg/sentry/fs/fsutil/handle.go | 2 - pkg/sentry/fs/fsutil/host_file_mapper.go | 2 - pkg/sentry/fs/fsutil/inode.go | 6 --- pkg/sentry/fs/fsutil/inode_cached.go | 2 - pkg/sentry/fs/gofer/BUILD | 23 +++++++++- pkg/sentry/fs/gofer/file.go | 2 - pkg/sentry/fs/gofer/fs.go | 2 - pkg/sentry/fs/gofer/inode.go | 4 -- pkg/sentry/fs/gofer/session.go | 3 -- pkg/sentry/fs/host/BUILD | 27 +++++++++++- pkg/sentry/fs/host/descriptor.go | 2 - pkg/sentry/fs/host/file.go | 2 - pkg/sentry/fs/host/fs.go | 6 +-- pkg/sentry/fs/host/inode.go | 4 -- pkg/sentry/fs/inode.go | 4 -- pkg/sentry/fs/inode_inotify.go | 2 - pkg/sentry/fs/inotify.go | 2 - pkg/sentry/fs/inotify_event.go | 2 - pkg/sentry/fs/inotify_watch.go | 2 - pkg/sentry/fs/lock/BUILD | 15 ++++++- pkg/sentry/fs/lock/lock.go | 6 +-- pkg/sentry/fs/mount.go | 4 -- pkg/sentry/fs/mount_overlay.go | 4 -- pkg/sentry/fs/mounts.go | 2 - pkg/sentry/fs/overlay.go | 2 - pkg/sentry/fs/proc/BUILD | 34 ++++++++++++++- pkg/sentry/fs/proc/cpuinfo.go | 2 - pkg/sentry/fs/proc/exec_args.go | 2 - pkg/sentry/fs/proc/fds.go | 6 --- pkg/sentry/fs/proc/file.go | 1 - pkg/sentry/fs/proc/filesystems.go | 2 - pkg/sentry/fs/proc/fs.go | 2 - pkg/sentry/fs/proc/loadavg.go | 2 - pkg/sentry/fs/proc/meminfo.go | 2 - pkg/sentry/fs/proc/mounts.go | 4 -- pkg/sentry/fs/proc/proc.go | 4 -- pkg/sentry/fs/proc/seqfile/BUILD | 30 +++++++++++-- pkg/sentry/fs/proc/seqfile/seqfile.go | 4 -- pkg/sentry/fs/proc/stat.go | 2 - pkg/sentry/fs/proc/sys.go | 5 --- pkg/sentry/fs/proc/sys_net.go | 2 - pkg/sentry/fs/proc/task.go | 20 --------- pkg/sentry/fs/proc/uid_gid_map.go | 3 -- pkg/sentry/fs/proc/uptime.go | 2 - pkg/sentry/fs/proc/version.go | 2 - pkg/sentry/fs/ramfs/BUILD | 21 ++++++++- pkg/sentry/fs/ramfs/dir.go | 2 - pkg/sentry/fs/ramfs/ramfs.go | 2 - pkg/sentry/fs/ramfs/socket.go | 2 - pkg/sentry/fs/ramfs/symlink.go | 2 - pkg/sentry/fs/ramfs/test/BUILD | 18 +++++++- pkg/sentry/fs/sys/BUILD | 14 +++++- pkg/sentry/fs/sys/fs.go | 2 - pkg/sentry/fs/sys/sys.go | 5 +-- pkg/sentry/fs/timerfd/BUILD | 18 +++++++- pkg/sentry/fs/timerfd/timerfd.go | 4 +- pkg/sentry/fs/tmpfs/BUILD | 17 +++++++- pkg/sentry/fs/tmpfs/file_regular.go | 2 - pkg/sentry/fs/tmpfs/fs.go | 2 - pkg/sentry/fs/tmpfs/inode_file.go | 2 - pkg/sentry/fs/tmpfs/tmpfs.go | 8 ---- pkg/sentry/fs/tty/BUILD | 20 ++++++++- pkg/sentry/fs/tty/dir.go | 18 +++----- pkg/sentry/fs/tty/fs.go | 4 -- pkg/sentry/fs/tty/inode.go | 2 - pkg/sentry/fs/tty/line_discipline.go | 6 --- pkg/sentry/fs/tty/master.go | 4 -- pkg/sentry/fs/tty/queue.go | 4 +- pkg/sentry/fs/tty/slave.go | 4 -- pkg/sentry/fs/tty/terminal.go | 2 - pkg/sentry/inet/BUILD | 15 ++++++- pkg/sentry/inet/inet.go | 2 - pkg/sentry/kernel/BUILD | 59 +++++++++++++++++++++++--- pkg/sentry/kernel/abstract_socket_namespace.go | 3 -- pkg/sentry/kernel/auth/BUILD | 17 +++++++- pkg/sentry/kernel/auth/credentials.go | 2 - pkg/sentry/kernel/auth/id_map.go | 2 - pkg/sentry/kernel/auth/user_namespace.go | 2 - pkg/sentry/kernel/epoll/BUILD | 15 ++++++- pkg/sentry/kernel/epoll/epoll.go | 8 +--- pkg/sentry/kernel/eventfd/BUILD | 18 +++++++- pkg/sentry/kernel/eventfd/eventfd.go | 4 +- pkg/sentry/kernel/fd_map.go | 6 --- pkg/sentry/kernel/fs_context.go | 2 - pkg/sentry/kernel/futex/BUILD | 18 +++++++- pkg/sentry/kernel/futex/futex.go | 2 - pkg/sentry/kernel/ipc_namespace.go | 2 - pkg/sentry/kernel/kernel.go | 4 +- pkg/sentry/kernel/pending_signals.go | 5 --- pkg/sentry/kernel/pipe/BUILD | 20 ++++++++- pkg/sentry/kernel/pipe/buffers.go | 2 - pkg/sentry/kernel/pipe/node.go | 2 - pkg/sentry/kernel/pipe/pipe.go | 2 - pkg/sentry/kernel/pipe/reader.go | 2 - pkg/sentry/kernel/pipe/reader_writer.go | 2 - pkg/sentry/kernel/pipe/writer.go | 2 - pkg/sentry/kernel/ptrace.go | 4 -- pkg/sentry/kernel/rseq.go | 2 - pkg/sentry/kernel/semaphore/BUILD | 15 ++++++- pkg/sentry/kernel/semaphore/semaphore.go | 8 ---- pkg/sentry/kernel/sessions.go | 4 -- pkg/sentry/kernel/shm/BUILD | 13 +++++- pkg/sentry/kernel/shm/shm.go | 4 -- pkg/sentry/kernel/signal_handlers.go | 2 - pkg/sentry/kernel/syscalls.go | 2 - pkg/sentry/kernel/syslog.go | 2 - pkg/sentry/kernel/task.go | 2 - pkg/sentry/kernel/task_clone.go | 4 -- pkg/sentry/kernel/task_context.go | 2 - pkg/sentry/kernel/task_exec.go | 4 -- pkg/sentry/kernel/task_exit.go | 6 --- pkg/sentry/kernel/task_resources.go | 2 - pkg/sentry/kernel/task_run.go | 2 - pkg/sentry/kernel/task_sched.go | 2 - pkg/sentry/kernel/task_signals.go | 5 --- pkg/sentry/kernel/task_syscall.go | 4 -- pkg/sentry/kernel/thread_group.go | 2 - pkg/sentry/kernel/threads.go | 8 ---- pkg/sentry/kernel/time/BUILD | 14 +++++- pkg/sentry/kernel/time/time.go | 6 --- pkg/sentry/kernel/timekeeper.go | 2 - pkg/sentry/kernel/timer.go | 8 ---- pkg/sentry/kernel/uts_namespace.go | 2 - pkg/sentry/kernel/vdso.go | 2 - pkg/sentry/limits/BUILD | 13 +++++- pkg/sentry/limits/limits.go | 4 -- pkg/sentry/loader/BUILD | 15 ++++++- pkg/sentry/loader/vdso.go | 2 - pkg/sentry/loader/vdso_state.go | 1 - pkg/sentry/memmap/BUILD | 15 ++++++- pkg/sentry/memmap/mapping_set.go | 2 - pkg/sentry/mm/BUILD | 21 ++++++++- pkg/sentry/mm/aio_context.go | 8 ---- pkg/sentry/mm/mm.go | 7 --- pkg/sentry/mm/special_mappable.go | 2 - pkg/sentry/platform/BUILD | 13 +++++- pkg/sentry/platform/filemem/BUILD | 14 +++++- pkg/sentry/platform/filemem/filemem.go | 2 - pkg/sentry/socket/BUILD | 17 +++++++- pkg/sentry/socket/control/BUILD | 23 +++++++--- pkg/sentry/socket/control/control.go | 4 -- pkg/sentry/socket/epsocket/BUILD | 16 ++++++- pkg/sentry/socket/epsocket/epsocket.go | 2 - pkg/sentry/socket/epsocket/stack.go | 2 - pkg/sentry/socket/hostinet/BUILD | 15 ++++++- pkg/sentry/socket/netlink/BUILD | 13 +++++- pkg/sentry/socket/netlink/port/BUILD | 15 ++++++- pkg/sentry/socket/netlink/port/port.go | 2 - pkg/sentry/socket/netlink/route/BUILD | 17 +++++++- pkg/sentry/socket/netlink/route/protocol.go | 2 - pkg/sentry/socket/netlink/socket.go | 2 - pkg/sentry/socket/socket.go | 2 - pkg/sentry/socket/unix/BUILD | 13 +++++- pkg/sentry/socket/unix/unix.go | 2 - pkg/sentry/syscalls/linux/BUILD | 20 ++++++++- pkg/sentry/syscalls/linux/sys_aio.go | 2 - pkg/sentry/syscalls/linux/sys_futex.go | 2 - pkg/sentry/syscalls/linux/sys_poll.go | 2 - pkg/sentry/syscalls/linux/sys_time.go | 2 - pkg/sentry/usage/BUILD | 17 +++++++- pkg/sentry/usage/cpu.go | 2 - pkg/sentry/usage/io.go | 2 - pkg/sentry/usermem/BUILD | 16 ++++++- pkg/sentry/usermem/access_type.go | 2 - pkg/sentry/usermem/addr.go | 2 - pkg/tcpip/BUILD | 17 +++++++- pkg/tcpip/buffer/BUILD | 13 +++++- pkg/tcpip/buffer/view.go | 2 - pkg/tcpip/header/BUILD | 13 +++++- pkg/tcpip/header/tcp.go | 4 -- pkg/tcpip/network/fragmentation/BUILD | 11 ++++- pkg/tcpip/seqnum/BUILD | 17 +++++++- pkg/tcpip/tcpip.go | 4 -- pkg/tcpip/transport/ping/BUILD | 17 +++++++- pkg/tcpip/transport/ping/endpoint.go | 1 - pkg/tcpip/transport/queue/BUILD | 17 +++++++- pkg/tcpip/transport/queue/queue.go | 2 - pkg/tcpip/transport/tcp/BUILD | 25 ++++++++++- pkg/tcpip/transport/tcp/endpoint.go | 4 -- pkg/tcpip/transport/tcp/rcv.go | 2 - pkg/tcpip/transport/tcp/reno.go | 2 - pkg/tcpip/transport/tcp/segment.go | 2 - pkg/tcpip/transport/tcp/segment_queue.go | 2 - pkg/tcpip/transport/tcp/snd.go | 4 -- pkg/tcpip/transport/tcp/snd_state.go | 1 - pkg/tcpip/transport/udp/BUILD | 17 +++++++- pkg/tcpip/transport/udp/endpoint.go | 3 -- pkg/tcpip/transport/unix/BUILD | 16 ++++++- pkg/tcpip/transport/unix/connectioned.go | 2 - pkg/tcpip/transport/unix/connectionless.go | 2 - pkg/tcpip/transport/unix/unix.go | 11 ----- pkg/waiter/BUILD | 21 +++++++-- pkg/waiter/waiter.go | 2 - 239 files changed, 1108 insertions(+), 662 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/BUILD b/pkg/abi/BUILD index c014d2c4b..f1e6bac67 100644 --- a/pkg/abi/BUILD +++ b/pkg/abi/BUILD @@ -1,13 +1,24 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "abi_state", + srcs = [ + "abi.go", + ], + out = "abi_state.go", + package = "abi", +) go_library( name = "abi", srcs = [ "abi.go", + "abi_state.go", "flag.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/abi", visibility = ["//:sandbox"], + deps = ["//pkg/state"], ) diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index ac4ceefbc..38b4829c9 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -4,7 +4,19 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "linux_state", + srcs = [ + "binder.go", + "bpf.go", + "time.go", + "tty.go", + ], + out = "linux_state.go", + package = "linux", +) go_library( name = "linux", @@ -29,6 +41,7 @@ go_library( "ipc.go", "limits.go", "linux.go", + "linux_state.go", "mm.go", "netdevice.go", "netlink.go", @@ -54,5 +67,6 @@ go_library( "//pkg/abi", "//pkg/binary", "//pkg/bits", + "//pkg/state", ], ) diff --git a/pkg/abi/linux/bpf.go b/pkg/abi/linux/bpf.go index 80e5b1af1..f597ef4f5 100644 --- a/pkg/abi/linux/bpf.go +++ b/pkg/abi/linux/bpf.go @@ -15,8 +15,6 @@ package linux // BPFInstruction is a raw BPF virtual machine instruction. -// -// +stateify savable type BPFInstruction struct { // OpCode is the operation to execute. OpCode uint16 diff --git a/pkg/abi/linux/tty.go b/pkg/abi/linux/tty.go index b640f7627..84b6ccc87 100644 --- a/pkg/abi/linux/tty.go +++ b/pkg/abi/linux/tty.go @@ -38,8 +38,6 @@ type Termios struct { // KernelTermios is struct ktermios/struct termios2, defined in // uapi/asm-generic/termbits.h. -// -// +stateify savable type KernelTermios struct { InputFlags uint32 OutputFlags uint32 diff --git a/pkg/bpf/BUILD b/pkg/bpf/BUILD index 564df3af5..403270049 100644 --- a/pkg/bpf/BUILD +++ b/pkg/bpf/BUILD @@ -1,11 +1,21 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "bpf_state", + srcs = [ + "interpreter.go", + ], + out = "bpf_state.go", + package = "bpf", +) go_library( name = "bpf", srcs = [ "bpf.go", + "bpf_state.go", "decoder.go", "input_bytes.go", "interpreter.go", @@ -13,7 +23,10 @@ go_library( ], importpath = "gvisor.googlesource.com/gvisor/pkg/bpf", visibility = ["//visibility:public"], - deps = ["//pkg/abi/linux"], + deps = [ + "//pkg/abi/linux", + "//pkg/state", + ], ) go_test( diff --git a/pkg/bpf/interpreter.go b/pkg/bpf/interpreter.go index 111ada9d1..b7dee86a8 100644 --- a/pkg/bpf/interpreter.go +++ b/pkg/bpf/interpreter.go @@ -88,8 +88,6 @@ func (e Error) Error() string { } // Program is a BPF program that has been validated for consistency. -// -// +stateify savable type Program struct { instructions []linux.BPFInstruction } diff --git a/pkg/cpuid/BUILD b/pkg/cpuid/BUILD index 46fc4703b..9a0ca1b33 100644 --- a/pkg/cpuid/BUILD +++ b/pkg/cpuid/BUILD @@ -1,16 +1,27 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "cpuid_state", + srcs = ["cpuid.go"], + out = "cpuid_state.go", + package = "cpuid", +) go_library( name = "cpuid", srcs = [ "cpu_amd64.s", "cpuid.go", + "cpuid_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/cpuid", visibility = ["//:sandbox"], - deps = ["//pkg/log"], + deps = [ + "//pkg/log", + "//pkg/state", + ], ) go_test( diff --git a/pkg/cpuid/cpuid.go b/pkg/cpuid/cpuid.go index e91e34dc7..b486ab037 100644 --- a/pkg/cpuid/cpuid.go +++ b/pkg/cpuid/cpuid.go @@ -409,8 +409,6 @@ func (f Feature) flagString(cpuinfoOnly bool) string { } // FeatureSet is a set of Features for a cpu. -// -// +stateify savable type FeatureSet struct { // Set is the set of features that are enabled in this FeatureSet. Set map[Feature]bool diff --git a/pkg/ilist/BUILD b/pkg/ilist/BUILD index b26a39132..e32f26ffa 100644 --- a/pkg/ilist/BUILD +++ b/pkg/ilist/BUILD @@ -1,15 +1,28 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "list_state", + srcs = [ + "interface_list.go", + ], + out = "interface_list_state.go", + package = "ilist", +) go_library( name = "ilist", srcs = [ "interface_list.go", + "interface_list_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/ilist", visibility = ["//visibility:public"], + deps = [ + "//pkg/state", + ], ) go_template_instance( diff --git a/pkg/ilist/list.go b/pkg/ilist/list.go index a88b82196..5efb6c072 100644 --- a/pkg/ilist/list.go +++ b/pkg/ilist/list.go @@ -36,8 +36,6 @@ type Linker interface { // for e := l.Front(); e != nil; e = e.Next() { // // do something with e. // } -// -// +stateify savable type List struct { head Linker tail Linker @@ -157,8 +155,6 @@ func (l *List) Remove(e Linker) { // Entry is a default implementation of Linker. Users can add anonymous fields // of this type to their structs to make them automatically implement the // methods needed by List. -// -// +stateify savable type Entry struct { next Linker prev Linker diff --git a/pkg/segment/range.go b/pkg/segment/range.go index 34c067265..5ff30d489 100644 --- a/pkg/segment/range.go +++ b/pkg/segment/range.go @@ -18,8 +18,6 @@ package segment type T uint64 // A Range represents a contiguous range of T. -// -// +stateify savable type Range struct { // Start is the inclusive start of the range. Start T diff --git a/pkg/segment/set.go b/pkg/segment/set.go index cffec2a2c..6eed1d930 100644 --- a/pkg/segment/set.go +++ b/pkg/segment/set.go @@ -88,8 +88,6 @@ const ( // A Set is a mapping of segments with non-overlapping Range keys. The zero // value for a Set is an empty set. Set values are not safely movable nor // copyable. Set is thread-compatible. -// -// +stateify savable type Set struct { root node `state:".(*SegmentDataSlices)"` } @@ -598,7 +596,6 @@ func (s *Set) ApplyContiguous(r Range, fn func(seg Iterator)) GapIterator { } } -// +stateify savable type node struct { // An internal binary tree node looks like: // @@ -1320,8 +1317,6 @@ func (n *node) writeDebugString(buf *bytes.Buffer, prefix string) { // SegmentDataSlices represents segments from a set as slices of start, end, and // values. SegmentDataSlices is primarily used as an intermediate representation // for save/restore and the layout here is optimized for that. -// -// +stateify savable type SegmentDataSlices struct { Start []Key End []Key diff --git a/pkg/sentry/arch/BUILD b/pkg/sentry/arch/BUILD index 314b3e962..0a2a35400 100644 --- a/pkg/sentry/arch/BUILD +++ b/pkg/sentry/arch/BUILD @@ -1,7 +1,21 @@ package(licenses = ["notice"]) # Apache 2.0 load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "arch_state", + srcs = [ + "arch.go", + "arch_amd64.go", + "arch_state_x86.go", + "arch_x86.go", + "auxv.go", + "signal_amd64.go", + ], + out = "arch_state.go", + package = "arch", +) go_library( name = "arch", @@ -10,6 +24,7 @@ go_library( "arch.go", "arch_amd64.go", "arch_amd64.s", + "arch_state.go", "arch_state_x86.go", "arch_x86.go", "auxv.go", @@ -31,6 +46,7 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/limits", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go index 21cb84502..0189e958d 100644 --- a/pkg/sentry/arch/arch.go +++ b/pkg/sentry/arch/arch.go @@ -254,8 +254,6 @@ const ( // MemoryManager. // // Note that "highest address" below is always exclusive. -// -// +stateify savable type MmapLayout struct { // MinAddr is the lowest mappable address. MinAddr usermem.Addr diff --git a/pkg/sentry/arch/arch_amd64.go b/pkg/sentry/arch/arch_amd64.go index f1e408af9..23526fe8e 100644 --- a/pkg/sentry/arch/arch_amd64.go +++ b/pkg/sentry/arch/arch_amd64.go @@ -95,8 +95,6 @@ const ( ) // context64 represents an AMD64 context. -// -// +stateify savable type context64 struct { State sigFPState []x86FPState // fpstate to be restored on sigreturn. diff --git a/pkg/sentry/arch/arch_state_x86.go b/pkg/sentry/arch/arch_state_x86.go index e9c23a06b..cb38d098a 100644 --- a/pkg/sentry/arch/arch_state_x86.go +++ b/pkg/sentry/arch/arch_state_x86.go @@ -56,7 +56,6 @@ func (s *State) afterLoad() { copy(s.x86FPState, old) } -// +stateify savable type syscallPtraceRegs struct { R15 uint64 R14 uint64 diff --git a/pkg/sentry/arch/arch_x86.go b/pkg/sentry/arch/arch_x86.go index b35eec53c..5cc4f8377 100644 --- a/pkg/sentry/arch/arch_x86.go +++ b/pkg/sentry/arch/arch_x86.go @@ -153,8 +153,6 @@ func NewFloatingPointData() *FloatingPointData { // State contains the common architecture bits for X86 (the build tag of this // file ensures it's only built on x86). -// -// +stateify savable type State struct { // The system registers. Regs syscall.PtraceRegs `state:".(syscallPtraceRegs)"` diff --git a/pkg/sentry/arch/auxv.go b/pkg/sentry/arch/auxv.go index 81cfb4a01..70e0e35b7 100644 --- a/pkg/sentry/arch/auxv.go +++ b/pkg/sentry/arch/auxv.go @@ -19,8 +19,6 @@ import ( ) // An AuxEntry represents an entry in an ELF auxiliary vector. -// -// +stateify savable type AuxEntry struct { Key uint64 Value usermem.Addr diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index e81717e8b..c1d743f38 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -28,8 +28,6 @@ import ( // SignalAct represents the action that should be taken when a signal is // delivered, and is equivalent to struct sigaction on 64-bit x86. -// -// +stateify savable type SignalAct struct { Handler uint64 Flags uint64 @@ -49,8 +47,6 @@ func (s *SignalAct) DeserializeTo(other *SignalAct) { // SignalStack represents information about a user stack, and is equivalent to // stack_t on 64-bit x86. -// -// +stateify savable type SignalStack struct { Addr uint64 Flags uint32 @@ -70,8 +66,6 @@ func (s *SignalStack) DeserializeTo(other *SignalStack) { // SignalInfo represents information about a signal being delivered, and is // equivalent to struct siginfo on 64-bit x86. -// -// +stateify savable type SignalInfo struct { Signo int32 // Signal number Errno int32 // Errno value diff --git a/pkg/sentry/context/contexttest/BUILD b/pkg/sentry/context/contexttest/BUILD index 01bb40b04..591b11a4d 100644 --- a/pkg/sentry/context/contexttest/BUILD +++ b/pkg/sentry/context/contexttest/BUILD @@ -1,11 +1,23 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "contexttest_state", + srcs = [ + "contexttest.go", + ], + out = "contexttest_state.go", + package = "contexttest", +) go_library( name = "contexttest", testonly = 1, - srcs = ["contexttest.go"], + srcs = [ + "contexttest.go", + "contexttest_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest", visibility = ["//pkg/sentry:internal"], deps = [ @@ -16,5 +28,6 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/platform/ptrace", "//pkg/sentry/uniqueid", + "//pkg/state", ], ) diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 18cd5ae8e..e3c9a9b70 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -1,7 +1,40 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "fs_state", + srcs = [ + "attr.go", + "dentry.go", + "dirent.go", + "dirent_cache.go", + "dirent_list.go", + "dirent_state.go", + "file.go", + "file_overlay.go", + "file_state.go", + "filesystems.go", + "flags.go", + "inode.go", + "inode_inotify.go", + "inode_operations.go", + "inode_overlay.go", + "inotify.go", + "inotify_event.go", + "inotify_watch.go", + "mock.go", + "mount.go", + "mount_overlay.go", + "mount_state.go", + "mounts.go", + "overlay.go", + "path.go", + ], + out = "fs_state.go", + package = "fs", +) go_library( name = "fs", @@ -21,6 +54,7 @@ go_library( "filesystems.go", "flags.go", "fs.go", + "fs_state.go", "inode.go", "inode_inotify.go", "inode_operations.go", diff --git a/pkg/sentry/fs/ashmem/BUILD b/pkg/sentry/fs/ashmem/BUILD index dc893d22f..9f166799a 100644 --- a/pkg/sentry/fs/ashmem/BUILD +++ b/pkg/sentry/fs/ashmem/BUILD @@ -1,12 +1,26 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") +go_stateify( + name = "ashmem_state", + srcs = [ + "area.go", + "device.go", + "pin_board.go", + "uint64_range.go", + "uint64_set.go", + ], + out = "ashmem_state.go", + package = "ashmem", +) + go_library( name = "ashmem", srcs = [ "area.go", + "ashmem_state.go", "device.go", "pin_board.go", "uint64_range.go", @@ -27,6 +41,7 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", ], diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index bfd7f2762..e4f76f0d0 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -39,12 +39,10 @@ const ( ) // Area implements fs.FileOperations. -// -// +stateify savable type Area struct { - fsutil.NoFsync `state:"nosave"` - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` + fsutil.NoFsync + fsutil.DeprecatedFileOperations + fsutil.NotDirReaddir ad *Device diff --git a/pkg/sentry/fs/ashmem/device.go b/pkg/sentry/fs/ashmem/device.go index d0986fa11..c5b51d4a7 100644 --- a/pkg/sentry/fs/ashmem/device.go +++ b/pkg/sentry/fs/ashmem/device.go @@ -27,19 +27,17 @@ import ( ) // Device implements fs.InodeOperations. -// -// +stateify savable type Device struct { - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.InodeNoExtendedAttributes `state:"nosave"` - fsutil.InodeNotDirectory `state:"nosave"` - fsutil.InodeNotRenameable `state:"nosave"` - fsutil.InodeNotSocket `state:"nosave"` - fsutil.InodeNotSymlink `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoMappable `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` + fsutil.DeprecatedFileOperations + fsutil.InodeNoExtendedAttributes + fsutil.InodeNotDirectory + fsutil.InodeNotRenameable + fsutil.InodeNotSocket + fsutil.InodeNotSymlink + fsutil.NoFsync + fsutil.NoMappable + fsutil.NoopWriteOut + fsutil.NotDirReaddir mu sync.Mutex `state:"nosave"` unstable fs.UnstableAttr diff --git a/pkg/sentry/fs/ashmem/pin_board.go b/pkg/sentry/fs/ashmem/pin_board.go index ecba395a0..c7fb3822c 100644 --- a/pkg/sentry/fs/ashmem/pin_board.go +++ b/pkg/sentry/fs/ashmem/pin_board.go @@ -56,8 +56,6 @@ func (setFunctions) Split(Range, noValue, uint64) (noValue, noValue) { // segment.Set is used for implementation where segments represent // ranges of pinned bytes, while gaps represent ranges of unpinned // bytes. All ranges are page-aligned. -// -// +stateify savable type PinBoard struct { Set } diff --git a/pkg/sentry/fs/attr.go b/pkg/sentry/fs/attr.go index 4178f18b2..56a2ad6f7 100644 --- a/pkg/sentry/fs/attr.go +++ b/pkg/sentry/fs/attr.go @@ -91,8 +91,6 @@ func (n InodeType) String() string { // StableAttr contains Inode attributes that will be stable throughout the // lifetime of the Inode. -// -// +stateify savable type StableAttr struct { // Type is the InodeType of a InodeOperations. Type InodeType @@ -152,8 +150,6 @@ func IsCharDevice(s StableAttr) bool { // UnstableAttr contains Inode attributes that may change over the lifetime // of the Inode. -// -// +stateify savable type UnstableAttr struct { // Size is the file size in bytes. Size int64 @@ -190,8 +186,6 @@ func WithCurrentTime(ctx context.Context, u UnstableAttr) UnstableAttr { } // AttrMask contains fields to mask StableAttr and UnstableAttr. -// -// +stateify savable type AttrMask struct { Type bool DeviceID bool @@ -233,8 +227,6 @@ func (a AttrMask) Union(b AttrMask) AttrMask { } // PermMask are file access permissions. -// -// +stateify savable type PermMask struct { // Read indicates reading is permitted. Read bool @@ -288,8 +280,6 @@ func (p PermMask) SupersetOf(other PermMask) bool { // FilePermissions represents the permissions of a file, with // Read/Write/Execute bits for user, group, and other. -// -// +stateify savable type FilePermissions struct { User PermMask Group PermMask @@ -380,8 +370,6 @@ func (f FilePermissions) AnyRead() bool { } // FileOwner represents ownership of a file. -// -// +stateify savable type FileOwner struct { UID auth.KUID GID auth.KGID diff --git a/pkg/sentry/fs/binder/BUILD b/pkg/sentry/fs/binder/BUILD index a077b91d2..ec3928baf 100644 --- a/pkg/sentry/fs/binder/BUILD +++ b/pkg/sentry/fs/binder/BUILD @@ -1,16 +1,25 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "binder_state", + srcs = ["binder.go"], + out = "binder_state.go", + package = "binder", +) go_library( name = "binder", srcs = [ "binder.go", + "binder_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/binder", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/log", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", @@ -21,6 +30,8 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", + "//pkg/tcpip/transport/unix", ], ) diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index 502a262dd..3f87b6b08 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -40,17 +40,15 @@ const ( ) // Device implements fs.InodeOperations. -// -// +stateify savable type Device struct { - fsutil.InodeNoExtendedAttributes `state:"nosave"` - fsutil.InodeNotDirectory `state:"nosave"` - fsutil.InodeNotRenameable `state:"nosave"` - fsutil.InodeNotSocket `state:"nosave"` - fsutil.InodeNotSymlink `state:"nosave"` - fsutil.NoMappable `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` - fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.InodeNoExtendedAttributes + fsutil.InodeNotDirectory + fsutil.InodeNotRenameable + fsutil.InodeNotSocket + fsutil.InodeNotSymlink + fsutil.NoMappable + fsutil.NoopWriteOut + fsutil.DeprecatedFileOperations // mu protects unstable. mu sync.Mutex `state:"nosave"` @@ -188,12 +186,10 @@ func (bd *Device) StatFS(context.Context) (fs.Info, error) { } // Proc implements fs.FileOperations and fs.IoctlGetter. -// -// +stateify savable type Proc struct { - fsutil.NoFsync `state:"nosave"` - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` + fsutil.NoFsync + fsutil.DeprecatedFileOperations + fsutil.NotDirReaddir bd *Device task *kernel.Task diff --git a/pkg/sentry/fs/dentry.go b/pkg/sentry/fs/dentry.go index b347468ff..d42e8da81 100644 --- a/pkg/sentry/fs/dentry.go +++ b/pkg/sentry/fs/dentry.go @@ -21,8 +21,6 @@ import ( ) // DentAttr is the metadata of a directory entry. It is a subset of StableAttr. -// -// +stateify savable type DentAttr struct { // Type is the InodeType of an Inode. Type InodeType @@ -156,8 +154,6 @@ func GenericReaddir(ctx *DirCtx, s *SortedDentryMap) (int, error) { } // SortedDentryMap is a sorted map of names and fs.DentAttr entries. -// -// +stateify savable type SortedDentryMap struct { // names is always kept in sorted-order. names []string diff --git a/pkg/sentry/fs/dev/BUILD b/pkg/sentry/fs/dev/BUILD index fc069bb5f..ea41615fd 100644 --- a/pkg/sentry/fs/dev/BUILD +++ b/pkg/sentry/fs/dev/BUILD @@ -1,11 +1,25 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "dev_state", + srcs = [ + "dev.go", + "fs.go", + "full.go", + "null.go", + "random.go", + ], + out = "dev_state.go", + package = "dev", +) go_library( name = "dev", srcs = [ "dev.go", + "dev_state.go", "device.go", "fs.go", "full.go", @@ -16,6 +30,8 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/amutex", + "//pkg/log", "//pkg/rand", "//pkg/sentry/context", "//pkg/sentry/device", @@ -29,7 +45,9 @@ go_library( "//pkg/sentry/mm", "//pkg/sentry/platform", "//pkg/sentry/safemem", + "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/fs/dev/dev.go b/pkg/sentry/fs/dev/dev.go index 3f4f2a40a..36c61bfc2 100644 --- a/pkg/sentry/fs/dev/dev.go +++ b/pkg/sentry/fs/dev/dev.go @@ -27,8 +27,6 @@ import ( ) // Dev is the root node. -// -// +stateify savable type Dev struct { ramfs.Dir } diff --git a/pkg/sentry/fs/dev/fs.go b/pkg/sentry/fs/dev/fs.go index 2ae49be4e..3c79f3782 100644 --- a/pkg/sentry/fs/dev/fs.go +++ b/pkg/sentry/fs/dev/fs.go @@ -29,8 +29,6 @@ const binderEnabledKey = "binder_enabled" const ashmemEnabledKey = "ashmem_enabled" // filesystem is a devtmpfs. -// -// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index 492b8eb3a..e13eb6c03 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -26,8 +26,6 @@ import ( ) // fullDevice is used to implement /dev/full. -// -// +stateify savable type fullDevice struct { ramfs.Entry } diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 2977c8670..66b8ba967 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -29,7 +29,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// +stateify savable type nullDevice struct { ramfs.Entry } @@ -55,7 +54,6 @@ func (n *nullDevice) Truncate(context.Context, *fs.Inode, int64) error { return nil } -// +stateify savable type zeroDevice struct { nullDevice } @@ -82,7 +80,6 @@ func (zd *zeroDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.F }), nil } -// +stateify savable type zeroFileOperations struct { fs.FileOperations } diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index 47b76218f..33a045a05 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -24,7 +24,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// +stateify savable type randomDevice struct { ramfs.Entry } diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go index 4658d044f..f9bf2fba6 100644 --- a/pkg/sentry/fs/dirent.go +++ b/pkg/sentry/fs/dirent.go @@ -81,8 +81,6 @@ var renameMu sync.RWMutex // // Dirents currently do not attempt to free entries that lack application references under // memory pressure. -// -// +stateify savable type Dirent struct { // AtomicRefCount is our reference count. refs.AtomicRefCount diff --git a/pkg/sentry/fs/dirent_cache.go b/pkg/sentry/fs/dirent_cache.go index c680e4828..e786e4f65 100644 --- a/pkg/sentry/fs/dirent_cache.go +++ b/pkg/sentry/fs/dirent_cache.go @@ -25,8 +25,6 @@ import ( // // A nil DirentCache corresponds to a cache with size 0. All methods can be // called, but nothing is actually cached. -// -// +stateify savable type DirentCache struct { // Maximum size of the cache. This must be saved manually, to handle the case // when cache is nil. diff --git a/pkg/sentry/fs/fdpipe/BUILD b/pkg/sentry/fs/fdpipe/BUILD index ffe4204bc..4fcb06f1f 100644 --- a/pkg/sentry/fs/fdpipe/BUILD +++ b/pkg/sentry/fs/fdpipe/BUILD @@ -1,27 +1,54 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "pipe_state", + srcs = [ + "pipe.go", + "pipe_state.go", + ], + out = "pipe_autogen_state.go", + imports = ["gvisor.googlesource.com/gvisor/pkg/sentry/fs"], + package = "fdpipe", +) go_library( name = "fdpipe", srcs = [ "pipe.go", + "pipe_autogen_state.go", "pipe_opener.go", "pipe_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fdpipe", - imports = ["gvisor.googlesource.com/gvisor/pkg/sentry/fs"], visibility = ["//pkg/sentry:internal"], deps = [ + "//pkg/abi/linux", + "//pkg/amutex", "//pkg/fd", "//pkg/log", + "//pkg/metric", + "//pkg/p9", + "//pkg/refs", "//pkg/secio", "//pkg/sentry/context", + "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", + "//pkg/sentry/fs/lock", + "//pkg/sentry/kernel/auth", + "//pkg/sentry/kernel/time", + "//pkg/sentry/memmap", + "//pkg/sentry/platform", "//pkg/sentry/safemem", + "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", + "//pkg/tcpip", + "//pkg/tcpip/transport/unix", + "//pkg/unet", "//pkg/waiter", "//pkg/waiter/fdnotifier", ], diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index 2e34604e6..7b318e35f 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -34,8 +34,6 @@ import ( ) // pipeOperations are the fs.FileOperations of a host pipe. -// -// +stateify savable type pipeOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 8e535a618..6d93ef760 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -47,8 +47,6 @@ const FileMaxOffset = math.MaxInt64 // and write(2). // // FIXME: Split synchronization from cancellation. -// -// +stateify savable type File struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index 113962368..36b2cf75e 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -60,8 +60,6 @@ func overlayFile(ctx context.Context, inode *Inode, flags FileFlags) (*File, err } // overlayFileOperations implements FileOperations for a file in an overlay. -// -// +stateify savable type overlayFileOperations struct { // upperMu protects upper below. In contrast lower is stable. upperMu sync.Mutex `state:"nosave"` @@ -377,8 +375,6 @@ func readdirOne(ctx context.Context, d *Dirent) (map[string]DentAttr, error) { // overlayMappingIdentity wraps a MappingIdentity, and also holds a reference // on a file during its lifetime. -// -// +stateify savable type overlayMappingIdentity struct { refs.AtomicRefCount id memmap.MappingIdentity diff --git a/pkg/sentry/fs/filesystems.go b/pkg/sentry/fs/filesystems.go index 5a1e7a270..200e792f4 100644 --- a/pkg/sentry/fs/filesystems.go +++ b/pkg/sentry/fs/filesystems.go @@ -125,8 +125,6 @@ func GetFilesystems() []Filesystem { } // MountSourceFlags represents all mount option flags as a struct. -// -// +stateify savable type MountSourceFlags struct { // ReadOnly corresponds to mount(2)'s "MS_RDONLY" and indicates that // the filesystem should be mounted read-only. diff --git a/pkg/sentry/fs/filetest/BUILD b/pkg/sentry/fs/filetest/BUILD index d137fee4c..f481c57fb 100644 --- a/pkg/sentry/fs/filetest/BUILD +++ b/pkg/sentry/fs/filetest/BUILD @@ -1,20 +1,34 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "filetest_state", + srcs = [ + "filetest.go", + ], + out = "filetest_state.go", + package = "filetest", +) go_library( name = "filetest", testonly = 1, - srcs = ["filetest.go"], + srcs = [ + "filetest.go", + "filetest_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/filetest", visibility = ["//pkg/sentry:internal"], deps = [ + "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/flags.go b/pkg/sentry/fs/flags.go index 1aa271560..da0ff58af 100644 --- a/pkg/sentry/fs/flags.go +++ b/pkg/sentry/fs/flags.go @@ -19,8 +19,6 @@ import ( ) // FileFlags encodes file flags. -// -// +stateify savable type FileFlags struct { // Direct indicates that I/O should be done directly. Direct bool diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index 3512bae6f..6eea64298 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -1,7 +1,24 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "fsutil_state", + srcs = [ + "dirty_set_impl.go", + "file.go", + "file_range_set_impl.go", + "frame_ref_set_impl.go", + "handle.go", + "host_file_mapper.go", + "host_file_mapper_state.go", + "inode.go", + "inode_cached.go", + ], + out = "fsutil_state.go", + package = "fsutil", +) go_template_instance( name = "dirty_set_impl", @@ -67,6 +84,7 @@ go_library( "frame_ref_set.go", "frame_ref_set_impl.go", "fsutil.go", + "fsutil_state.go", "handle.go", "host_file_mapper.go", "host_file_mapper_state.go", diff --git a/pkg/sentry/fs/fsutil/dirty_set.go b/pkg/sentry/fs/fsutil/dirty_set.go index 8e31e48fd..9c6c98542 100644 --- a/pkg/sentry/fs/fsutil/dirty_set.go +++ b/pkg/sentry/fs/fsutil/dirty_set.go @@ -32,8 +32,6 @@ import ( // DirtyInfo is the value type of DirtySet, and represents information about a // Mappable offset that is dirty (the cached data for that offset is newer than // its source). -// -// +stateify savable type DirtyInfo struct { // Keep is true if the represented offset is concurrently writable, such // that writing the data for that offset back to the source does not diff --git a/pkg/sentry/fs/fsutil/handle.go b/pkg/sentry/fs/fsutil/handle.go index e7efd3c0f..149c0f84a 100644 --- a/pkg/sentry/fs/fsutil/handle.go +++ b/pkg/sentry/fs/fsutil/handle.go @@ -27,8 +27,6 @@ import ( // // FIXME: Remove Handle entirely in favor of individual fs.File // implementations using simple generic utilities. -// -// +stateify savable type Handle struct { NoopRelease `state:"nosave"` NoIoctl `state:"nosave"` diff --git a/pkg/sentry/fs/fsutil/host_file_mapper.go b/pkg/sentry/fs/fsutil/host_file_mapper.go index 9c1e2f76f..d0a27fc1c 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper.go @@ -29,8 +29,6 @@ import ( // HostFileMapper caches mappings of an arbitrary host file descriptor. It is // used by implementations of memmap.Mappable that represent a host file // descriptor. -// -// +stateify savable type HostFileMapper struct { // HostFile conceptually breaks the file into pieces called chunks, of // size and alignment chunkSize, and caches mappings of the file on a chunk diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index 177396fdc..e1ad07df2 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -31,8 +31,6 @@ func NewSimpleInodeOperations(i InodeSimpleAttributes) fs.InodeOperations { } // simpleInodeOperations is a simple implementation of Inode. -// -// +stateify savable type simpleInodeOperations struct { DeprecatedFileOperations `state:"nosave"` InodeNotDirectory `state:"nosave"` @@ -50,8 +48,6 @@ type simpleInodeOperations struct { // InodeSimpleAttributes implements a subset of the Inode interface. It provides // read-only access to attributes. -// -// +stateify savable type InodeSimpleAttributes struct { // FSType is the filesystem type reported by StatFS. FSType uint64 @@ -114,8 +110,6 @@ func (*InodeSimpleAttributes) Truncate(context.Context, *fs.Inode, int64) error // // Users need not initialize Xattrs to non-nil (it will be initialized // when the first extended attribute is set. -// -// +stateify savable type InMemoryAttributes struct { Unstable fs.UnstableAttr Xattrs map[string][]byte diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index 0a320e2d8..cba642a8f 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -55,8 +55,6 @@ import ( // // Implementations of InodeOperations.WriteOut must call Sync to write out // in-memory modifications of data and metadata to the CachedFileObject. -// -// +stateify savable type CachingInodeOperations struct { // backingFile is a handle to a cached file object. backingFile CachedFileObject diff --git a/pkg/sentry/fs/gofer/BUILD b/pkg/sentry/fs/gofer/BUILD index cb17339c9..1277379e7 100644 --- a/pkg/sentry/fs/gofer/BUILD +++ b/pkg/sentry/fs/gofer/BUILD @@ -1,6 +1,21 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "gofer_state", + srcs = [ + "file.go", + "file_state.go", + "fs.go", + "inode.go", + "inode_state.go", + "session.go", + "session_state.go", + ], + out = "gofer_state.go", + package = "gofer", +) go_library( name = "gofer", @@ -12,6 +27,7 @@ go_library( "file.go", "file_state.go", "fs.go", + "gofer_state.go", "handles.go", "inode.go", "inode_state.go", @@ -25,6 +41,7 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/amutex", "//pkg/fd", "//pkg/log", "//pkg/metric", @@ -37,11 +54,15 @@ go_library( "//pkg/sentry/fs/fdpipe", "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/host", + "//pkg/sentry/fs/lock", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", + "//pkg/sentry/platform", "//pkg/sentry/safemem", + "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index 46a6bbd5d..039618808 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -33,8 +33,6 @@ import ( var openedWX = metric.MustCreateNewUint64Metric("/gofer/opened_write_execute_file", true /* sync */, "Number of times a writable+executable file was opened from a gofer.") // fileOperations implements fs.FileOperations for a remote file system. -// -// +stateify savable type fileOperations struct { fsutil.NoIoctl `state:"nosave"` waiter.AlwaysReady `state:"nosave"` diff --git a/pkg/sentry/fs/gofer/fs.go b/pkg/sentry/fs/gofer/fs.go index 3ae93f059..dd5d43c47 100644 --- a/pkg/sentry/fs/gofer/fs.go +++ b/pkg/sentry/fs/gofer/fs.go @@ -83,8 +83,6 @@ var ( ) // filesystem is a 9p client. -// -// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go index 7fc8f77b0..df584c382 100644 --- a/pkg/sentry/fs/gofer/inode.go +++ b/pkg/sentry/fs/gofer/inode.go @@ -35,8 +35,6 @@ import ( ) // inodeOperations implements fs.InodeOperations. -// -// +stateify savable type inodeOperations struct { fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` @@ -70,8 +68,6 @@ type inodeOperations struct { // circular load dependency between it and inodeOperations). Even with // lazy loading, this approach defines the dependencies between objects // and the expected load behavior more concretely. -// -// +stateify savable type inodeFileState struct { // s is common file system state for Gofers. s *session `state:"wait"` diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index 648a11435..b6841526a 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -27,7 +27,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/unet" ) -// +stateify savable type endpointMap struct { mu sync.RWMutex `state:"nosave"` // TODO: Make map with private unix sockets savable. @@ -64,8 +63,6 @@ func (e *endpointMap) get(key device.MultiDeviceKey) unix.BoundEndpoint { } // session holds state for each 9p session established during sys_mount. -// -// +stateify savable type session struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 29c79284a..23ec66f50 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -1,6 +1,23 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "host_state", + srcs = [ + "control.go", + "descriptor.go", + "descriptor_state.go", + "file.go", + "fs.go", + "inode.go", + "inode_state.go", + "socket.go", + "socket_state.go", + ], + out = "host_state.go", + package = "host", +) go_library( name = "host", @@ -11,6 +28,7 @@ go_library( "device.go", "file.go", "fs.go", + "host_state.go", "inode.go", "inode_state.go", "ioctl_unsafe.go", @@ -24,6 +42,7 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/amutex", "//pkg/fd", "//pkg/log", "//pkg/refs", @@ -33,14 +52,20 @@ go_library( "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", + "//pkg/sentry/fs/lock", "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", + "//pkg/sentry/platform", "//pkg/sentry/safemem", + "//pkg/sentry/socket", "//pkg/sentry/socket/control", "//pkg/sentry/socket/unix", + "//pkg/sentry/uniqueid", + "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/link/rawfile", diff --git a/pkg/sentry/fs/host/descriptor.go b/pkg/sentry/fs/host/descriptor.go index 3aee4d11c..613bd06e8 100644 --- a/pkg/sentry/fs/host/descriptor.go +++ b/pkg/sentry/fs/host/descriptor.go @@ -25,8 +25,6 @@ import ( ) // descriptor wraps a host fd. -// -// +stateify savable type descriptor struct { // donated is true if the host fd was donated by another process. donated bool diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index f9bef6d93..bdf844337 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -37,8 +37,6 @@ import ( ) // fileOperations implements fs.FileOperations for a host file descriptor. -// -// +stateify savable type fileOperations struct { fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/host/fs.go b/pkg/sentry/fs/host/fs.go index e46ae433c..974700636 100644 --- a/pkg/sentry/fs/host/fs.go +++ b/pkg/sentry/fs/host/fs.go @@ -51,8 +51,6 @@ const maxTraversals = 10 // to lock down the configurations. This filesystem should only be mounted at root. // // Think twice before exposing this to applications. -// -// +stateify savable type Filesystem struct { // whitelist is a set of host paths to whitelist. paths []string @@ -268,10 +266,8 @@ func newMountSource(ctx context.Context, root string, mounter fs.FileOwner, file } // superOperations implements fs.MountSourceOperations. -// -// +stateify savable type superOperations struct { - fs.SimpleMountSourceOperations + fs.SimpleMountSourceOperations `state:"nosave"` // root is the path of the mount point. All inode mappings // are relative to this root. diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index 761ccde33..226bc5164 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -34,8 +34,6 @@ import ( // inodeOperations implements fs.InodeOperations for an fs.Inodes backed // by a host file descriptor. -// -// +stateify savable type inodeOperations struct { fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` @@ -67,8 +65,6 @@ type inodeOperations struct { // circular load dependency between it and inodeOperations). Even with // lazy loading, this approach defines the dependencies between objects // and the expected load behavior more concretely. -// -// +stateify savable type inodeFileState struct { // Common file system state. mops *superOperations `state:"wait"` diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index d0dbce5dd..6c8e6f188 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -28,8 +28,6 @@ import ( // Inode is a file system object that can be simultaneously referenced by different // components of the VFS (Dirent, fs.File, etc). -// -// +stateify savable type Inode struct { // AtomicRefCount is our reference count. refs.AtomicRefCount @@ -60,8 +58,6 @@ type Inode struct { // Note that in Linux fcntl(2) and flock(2) locks are _not_ cooperative, because race and // deadlock conditions make merging them prohibitive. We do the same and keep them oblivious // to each other but provide a "context" as a convenient container. -// -// +stateify savable type LockCtx struct { // Posix is a set of POSIX-style regional advisory locks, see fcntl(2). Posix lock.Locks diff --git a/pkg/sentry/fs/inode_inotify.go b/pkg/sentry/fs/inode_inotify.go index 683140afe..358bbecdf 100644 --- a/pkg/sentry/fs/inode_inotify.go +++ b/pkg/sentry/fs/inode_inotify.go @@ -20,8 +20,6 @@ import ( ) // Watches is the collection of inotify watches on an inode. -// -// +stateify savable type Watches struct { // mu protects the fields below. mu sync.RWMutex `state:"nosave"` diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 2aabdded8..6f5e8ce5e 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -34,8 +34,6 @@ import ( // // Lock ordering: // Inotify.mu -> Inode.Watches.mu -> Watch.mu -> Inotify.evMu -// -// +stateify savable type Inotify struct { // Unique identifier for this inotify instance. We don't just reuse the // inotify fd because fds can be duped. These should not be exposed to the diff --git a/pkg/sentry/fs/inotify_event.go b/pkg/sentry/fs/inotify_event.go index e9b5e0f56..217915ba4 100644 --- a/pkg/sentry/fs/inotify_event.go +++ b/pkg/sentry/fs/inotify_event.go @@ -28,8 +28,6 @@ import ( const inotifyEventBaseSize = 16 // Event represents a struct inotify_event from linux. -// -// +stateify savable type Event struct { ilist.Entry diff --git a/pkg/sentry/fs/inotify_watch.go b/pkg/sentry/fs/inotify_watch.go index 3e1959e83..8904ef544 100644 --- a/pkg/sentry/fs/inotify_watch.go +++ b/pkg/sentry/fs/inotify_watch.go @@ -27,8 +27,6 @@ import ( // holding an extra ref on each dirent known (by inotify) to point to the // inode. These are known as pins. For a full discussion, see // fs/g3doc/inotify.md. -// -// +stateify savable type Watch struct { // Inotify instance which owns this watch. owner *Inotify diff --git a/pkg/sentry/fs/lock/BUILD b/pkg/sentry/fs/lock/BUILD index 3159ff1da..2607d7ed3 100644 --- a/pkg/sentry/fs/lock/BUILD +++ b/pkg/sentry/fs/lock/BUILD @@ -1,7 +1,18 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "lock_state", + srcs = [ + "lock.go", + "lock_range.go", + "lock_set.go", + ], + out = "lock_state.go", + package = "lock", +) go_template_instance( name = "lock_range", @@ -38,11 +49,13 @@ go_library( "lock_range.go", "lock_set.go", "lock_set_functions.go", + "lock_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/lock", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/log", + "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/lock/lock.go b/pkg/sentry/fs/lock/lock.go index e9b376eb6..24d54c989 100644 --- a/pkg/sentry/fs/lock/lock.go +++ b/pkg/sentry/fs/lock/lock.go @@ -88,8 +88,6 @@ const LockEOF = math.MaxUint64 // // A Lock may be downgraded from a write lock to a read lock only if // the write lock's uid is the same as the read lock. -// -// +stateify savable type Lock struct { // Readers are the set of read lock holders identified by UniqueID. // If len(Readers) > 0 then HasWriter must be false. @@ -105,8 +103,6 @@ type Lock struct { } // Locks is a thread-safe wrapper around a LockSet. -// -// +stateify savable type Locks struct { // mu protects locks below. mu sync.Mutex `state:"nosave"` @@ -115,7 +111,7 @@ type Locks struct { locks LockSet // blockedQueue is the queue of waiters that are waiting on a lock. - blockedQueue waiter.Queue `state:"zerovalue"` + blockedQueue waiter.Queue } // Blocker is the interface used for blocking locks. Passing a nil Blocker diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index 4ede767f9..eb1897174 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -101,8 +101,6 @@ func (i InodeMappings) String() string { // (e.g. cannot be mounted at different locations). // // TODO: Move mount-specific information out of MountSource. -// -// +stateify savable type MountSource struct { refs.AtomicRefCount @@ -262,8 +260,6 @@ func NewNonCachingMountSource(filesystem Filesystem, flags MountSourceFlags) *Mo } // SimpleMountSourceOperations implements MountSourceOperations. -// -// +stateify savable type SimpleMountSourceOperations struct { keep bool } diff --git a/pkg/sentry/fs/mount_overlay.go b/pkg/sentry/fs/mount_overlay.go index d135e8a37..1be81e3a1 100644 --- a/pkg/sentry/fs/mount_overlay.go +++ b/pkg/sentry/fs/mount_overlay.go @@ -18,8 +18,6 @@ import "gvisor.googlesource.com/gvisor/pkg/sentry/context" // overlayMountSourceOperations implements MountSourceOperations for an overlay // mount point. -// -// +stateify savable type overlayMountSourceOperations struct { upper *MountSource lower *MountSource @@ -74,8 +72,6 @@ func (o *overlayMountSourceOperations) Destroy() { } // type overlayFilesystem is the filesystem for overlay mounts. -// -// +stateify savable type overlayFilesystem struct{} // Name implements Filesystem.Name. diff --git a/pkg/sentry/fs/mounts.go b/pkg/sentry/fs/mounts.go index 144d3427d..87da4ee0e 100644 --- a/pkg/sentry/fs/mounts.go +++ b/pkg/sentry/fs/mounts.go @@ -32,8 +32,6 @@ import ( const DefaultTraversalLimit = 10 // MountNamespace defines a collection of mounts. -// -// +stateify savable type MountNamespace struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/overlay.go b/pkg/sentry/fs/overlay.go index af13dc8c7..7357d6401 100644 --- a/pkg/sentry/fs/overlay.go +++ b/pkg/sentry/fs/overlay.go @@ -145,8 +145,6 @@ func newOverlayInode(ctx context.Context, o *overlayEntry, msrc *MountSource) *I } // overlayEntry is the overlay metadata of an Inode. It implements Mappable. -// -// +stateify savable type overlayEntry struct { // lowerExists is true if an Inode exists for this file in the lower // filesystem. If lowerExists is true, then the overlay must create diff --git a/pkg/sentry/fs/proc/BUILD b/pkg/sentry/fs/proc/BUILD index 2d9f07f2f..870df47b2 100644 --- a/pkg/sentry/fs/proc/BUILD +++ b/pkg/sentry/fs/proc/BUILD @@ -1,6 +1,32 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "proc_state", + srcs = [ + "cpuinfo.go", + "exec_args.go", + "fds.go", + "file.go", + "filesystems.go", + "fs.go", + "loadavg.go", + "meminfo.go", + "mounts.go", + "net.go", + "proc.go", + "stat.go", + "sys.go", + "sys_net.go", + "task.go", + "uid_gid_map.go", + "uptime.go", + "version.go", + ], + out = "proc_state.go", + package = "proc", +) go_library( name = "proc", @@ -16,6 +42,7 @@ go_library( "mounts.go", "net.go", "proc.go", + "proc_state.go", "rpcinet_proc.go", "stat.go", "sys.go", @@ -29,6 +56,9 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/amutex", + "//pkg/log", + "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/proc/device", @@ -43,6 +73,8 @@ go_library( "//pkg/sentry/socket/rpcinet", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", + "//pkg/syserr", "//pkg/syserror", ], ) diff --git a/pkg/sentry/fs/proc/cpuinfo.go b/pkg/sentry/fs/proc/cpuinfo.go index 4dfec03a4..f80aaa5b1 100644 --- a/pkg/sentry/fs/proc/cpuinfo.go +++ b/pkg/sentry/fs/proc/cpuinfo.go @@ -27,8 +27,6 @@ import ( // cpuinfo is a file describing the CPU capabilities. // // Presently cpuinfo never changes, so it doesn't need to be a SeqFile. -// -// +stateify savable type cpuinfo struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/exec_args.go b/pkg/sentry/fs/proc/exec_args.go index a69cbaa0e..0e1523bf1 100644 --- a/pkg/sentry/fs/proc/exec_args.go +++ b/pkg/sentry/fs/proc/exec_args.go @@ -37,8 +37,6 @@ const ( // execArgFile is a file containing the exec args (either cmdline or environ) // for a given task. -// -// +stateify savable type execArgFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index cca8f874c..194a9c12a 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -138,8 +138,6 @@ func (f *fd) Close() error { } // fdDir implements /proc/TID/fd. -// -// +stateify savable type fdDir struct { ramfs.Dir @@ -199,8 +197,6 @@ func (f *fdDir) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset } // fdInfo is a single file in /proc/TID/fdinfo/. -// -// +stateify savable type fdInfo struct { ramfs.File @@ -233,8 +229,6 @@ func (*fdInfo) Truncate(ctx context.Context, inode *fs.Inode, size int64) error // fdInfoDir implements /proc/TID/fdinfo. It embeds an fdDir, but overrides // Lookup and Readdir. -// -// +stateify savable type fdInfoDir struct { ramfs.Dir diff --git a/pkg/sentry/fs/proc/file.go b/pkg/sentry/fs/proc/file.go index 4b2d08e75..9a433cdf8 100644 --- a/pkg/sentry/fs/proc/file.go +++ b/pkg/sentry/fs/proc/file.go @@ -22,7 +22,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// +stateify savable type file struct { fs.InodeOperations diff --git a/pkg/sentry/fs/proc/filesystems.go b/pkg/sentry/fs/proc/filesystems.go index 49b92fd8a..37db9cf9c 100644 --- a/pkg/sentry/fs/proc/filesystems.go +++ b/pkg/sentry/fs/proc/filesystems.go @@ -24,8 +24,6 @@ import ( ) // filesystemsData backs /proc/filesystems. -// -// +stateify savable type filesystemsData struct{} // NeedsUpdate returns true on the first generation. The set of registered file diff --git a/pkg/sentry/fs/proc/fs.go b/pkg/sentry/fs/proc/fs.go index 061824b8c..3aadd6ac4 100644 --- a/pkg/sentry/fs/proc/fs.go +++ b/pkg/sentry/fs/proc/fs.go @@ -22,8 +22,6 @@ import ( ) // filesystem is a procfs. -// -// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/proc/loadavg.go b/pkg/sentry/fs/proc/loadavg.go index 6fac251d2..7583b6ccd 100644 --- a/pkg/sentry/fs/proc/loadavg.go +++ b/pkg/sentry/fs/proc/loadavg.go @@ -23,8 +23,6 @@ import ( ) // loadavgData backs /proc/loadavg. -// -// +stateify savable type loadavgData struct{} // NeedsUpdate implements seqfile.SeqSource.NeedsUpdate. diff --git a/pkg/sentry/fs/proc/meminfo.go b/pkg/sentry/fs/proc/meminfo.go index 53dfd59ef..49cb0faed 100644 --- a/pkg/sentry/fs/proc/meminfo.go +++ b/pkg/sentry/fs/proc/meminfo.go @@ -26,8 +26,6 @@ import ( ) // meminfoData backs /proc/meminfo. -// -// +stateify savable type meminfoData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/proc/mounts.go b/pkg/sentry/fs/proc/mounts.go index 2b8167c28..108432f4e 100644 --- a/pkg/sentry/fs/proc/mounts.go +++ b/pkg/sentry/fs/proc/mounts.go @@ -71,8 +71,6 @@ func forEachMountSource(t *kernel.Task, fn func(string, *fs.MountSource)) { } // mountInfoFile is used to implement /proc/[pid]/mountinfo. -// -// +stateify savable type mountInfoFile struct { t *kernel.Task } @@ -154,8 +152,6 @@ func (mif *mountInfoFile) ReadSeqFileData(ctx context.Context, handle seqfile.Se } // mountsFile is used to implement /proc/[pid]/mountinfo. -// -// +stateify savable type mountsFile struct { t *kernel.Task } diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index 07029a7bb..b2a8d639c 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -33,8 +33,6 @@ import ( ) // proc is a root proc node. -// -// +stateify savable type proc struct { ramfs.Dir @@ -49,8 +47,6 @@ type proc struct { // stubProcFSFile is a file type that can be used to return file contents // which are constant. This file is not writable and will always have mode // 0444. -// -// +stateify savable type stubProcFSFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/seqfile/BUILD b/pkg/sentry/fs/proc/seqfile/BUILD index 53c475652..c84f7e20d 100644 --- a/pkg/sentry/fs/proc/seqfile/BUILD +++ b/pkg/sentry/fs/proc/seqfile/BUILD @@ -1,10 +1,22 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "seqfile_state", + srcs = [ + "seqfile.go", + ], + out = "seqfile_state.go", + package = "seqfile", +) go_library( name = "seqfile", - srcs = ["seqfile.go"], + srcs = [ + "seqfile.go", + "seqfile_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile", visibility = ["//pkg/sentry:internal"], deps = [ @@ -14,16 +26,26 @@ go_library( "//pkg/sentry/fs/ramfs", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", + "//pkg/state", ], ) +go_stateify( + name = "seqfile_test_state", + srcs = ["seqfile_test.go"], + out = "seqfile_test_state.go", + package = "seqfile", +) + go_test( name = "seqfile_test", size = "small", - srcs = ["seqfile_test.go"], + srcs = [ + "seqfile_test.go", + "seqfile_test_state.go", + ], embed = [":seqfile"], deps = [ - "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs/test", diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index 51cae5e37..c08565f8a 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -30,8 +30,6 @@ import ( type SeqHandle interface{} // SeqData holds the data for one unit in the file. -// -// +stateify savable type SeqData struct { // The data to be returned to the user. Buf []byte @@ -84,8 +82,6 @@ func (s *SeqGenerationCounter) IsCurrent(generation int64) bool { } // SeqFile is used to provide dynamic files that can be ordered by record. -// -// +stateify savable type SeqFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/stat.go b/pkg/sentry/fs/proc/stat.go index bf7650211..284f3e52b 100644 --- a/pkg/sentry/fs/proc/stat.go +++ b/pkg/sentry/fs/proc/stat.go @@ -25,8 +25,6 @@ import ( ) // statData backs /proc/stat. -// -// +stateify savable type statData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index a2d36ca23..aab891c53 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -28,8 +28,6 @@ import ( ) // hostname is a file containing the system hostname. -// -// +stateify savable type hostname struct { ramfs.Entry } @@ -54,8 +52,6 @@ func (p *proc) newHostname(ctx context.Context, msrc *fs.MountSource) *fs.Inode } // mmapMinAddrData backs /proc/sys/vm/mmap_min_addr. -// -// +stateify savable type mmapMinAddrData struct { k *kernel.Kernel } @@ -78,7 +74,6 @@ func (d *mmapMinAddrData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHand }, 0 } -// +stateify savable type overcommitMemory struct{} func (*overcommitMemory) NeedsUpdate(generation int64) bool { diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index beb25be20..f3a5043f8 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -33,7 +33,6 @@ const ( tcpWMem ) -// +stateify savable type tcpMem struct { ramfs.Entry s inet.Stack @@ -101,7 +100,6 @@ func (m *tcpMem) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, return n, cperr } -// +stateify savable type tcpSack struct { ramfs.Entry s inet.Stack diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 748ca4320..efc635946 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -52,8 +52,6 @@ func getTaskMM(t *kernel.Task) (*mm.MemoryManager, error) { } // taskDir represents a task-level directory. -// -// +stateify savable type taskDir struct { ramfs.Dir @@ -94,8 +92,6 @@ func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace } // subtasks represents a /proc/TID/task directory. -// -// +stateify savable type subtasks struct { ramfs.Dir @@ -171,8 +167,6 @@ func (s *subtasks) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, off } // exe is an fs.InodeOperations symlink for the /proc/PID/exe file. -// -// +stateify savable type exe struct { ramfs.Symlink @@ -232,8 +226,6 @@ func (e *exe) Readlink(ctx context.Context, inode *fs.Inode) (string, error) { // namespaceFile represents a file in the namespacefs, such as the files in // /proc//ns. -// -// +stateify savable type namespaceFile struct { ramfs.Symlink @@ -282,8 +274,6 @@ func newNamespaceDir(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { } // mapsData implements seqfile.SeqSource for /proc/[pid]/maps. -// -// +stateify savable type mapsData struct { t *kernel.Task } @@ -321,7 +311,6 @@ func (md *mapsData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([ return []seqfile.SeqData{}, 0 } -// +stateify savable type taskStatData struct { t *kernel.Task @@ -402,8 +391,6 @@ func (s *taskStatData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) } // statmData implements seqfile.SeqSource for /proc/[pid]/statm. -// -// +stateify savable type statmData struct { t *kernel.Task } @@ -438,8 +425,6 @@ func (s *statmData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([ } // statusData implements seqfile.SeqSource for /proc/[pid]/status. -// -// +stateify savable type statusData struct { t *kernel.Task pidns *kernel.PIDNamespace @@ -505,7 +490,6 @@ type ioUsage interface { IOUsage() *usage.IO } -// +stateify savable type ioData struct { ioUsage } @@ -546,8 +530,6 @@ func (i *ioData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([]se // On Linux, /proc/[pid]/comm is writable, and writing to the comm file changes // the thread name. We don't implement this yet as there are no known users of // this feature. -// -// +stateify savable type comm struct { ramfs.Entry @@ -577,8 +559,6 @@ func (c *comm) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, off } // auxvec is a file containing the auxiliary vector for a task. -// -// +stateify savable type auxvec struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index 9811d9c9d..85acb5163 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -29,8 +29,6 @@ import ( // An idMapSeqSource is a seqfile.SeqSource that returns UID or GID mappings // from a task's user namespace. -// -// +stateify savable type idMapSeqSource struct { t *kernel.Task gids bool @@ -72,7 +70,6 @@ type idMapSeqHandle struct { value int } -// +stateify savable type idMapSeqFile struct { seqfile.SeqFile } diff --git a/pkg/sentry/fs/proc/uptime.go b/pkg/sentry/fs/proc/uptime.go index f3a9b81df..4679d5821 100644 --- a/pkg/sentry/fs/proc/uptime.go +++ b/pkg/sentry/fs/proc/uptime.go @@ -27,8 +27,6 @@ import ( ) // uptime is a file containing the system uptime. -// -// +stateify savable type uptime struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/version.go b/pkg/sentry/fs/proc/version.go index 00f6a2afd..c0f2e87e3 100644 --- a/pkg/sentry/fs/proc/version.go +++ b/pkg/sentry/fs/proc/version.go @@ -23,8 +23,6 @@ import ( ) // versionData backs /proc/version. -// -// +stateify savable type versionData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/ramfs/BUILD b/pkg/sentry/fs/ramfs/BUILD index 5230157fe..d84f2c624 100644 --- a/pkg/sentry/fs/ramfs/BUILD +++ b/pkg/sentry/fs/ramfs/BUILD @@ -1,6 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "ramfs_state", + srcs = [ + "dir.go", + "file.go", + "ramfs.go", + "socket.go", + "symlink.go", + ], + out = "ramfs_state.go", + package = "ramfs", +) go_library( name = "ramfs", @@ -8,6 +21,7 @@ go_library( "dir.go", "file.go", "ramfs.go", + "ramfs_state.go", "socket.go", "symlink.go", "tree.go", @@ -15,8 +29,12 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs", visibility = ["//pkg/sentry:internal"], deps = [ + "//pkg/amutex", + "//pkg/log", + "//pkg/refs", "//pkg/secio", "//pkg/sentry/context", + "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", @@ -24,6 +42,7 @@ go_library( "//pkg/sentry/memmap", "//pkg/sentry/safemem", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index 04432f28c..19d5612ed 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -44,8 +44,6 @@ type CreateOps struct { } // Dir represents a single directory in the filesystem. -// -// +stateify savable type Dir struct { Entry diff --git a/pkg/sentry/fs/ramfs/ramfs.go b/pkg/sentry/fs/ramfs/ramfs.go index 13e72e775..d6cfaf753 100644 --- a/pkg/sentry/fs/ramfs/ramfs.go +++ b/pkg/sentry/fs/ramfs/ramfs.go @@ -60,8 +60,6 @@ var ( // Entry represents common internal state for file and directory nodes. // This may be used by other packages to easily create ramfs files. -// -// +stateify savable type Entry struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoMappable `state:"nosave"` diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index 93427a1ff..b0c79325f 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -21,8 +21,6 @@ import ( ) // Socket represents a socket. -// -// +stateify savable type Socket struct { Entry diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index 1c54d9991..9bbf78619 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -22,8 +22,6 @@ import ( ) // Symlink represents a symlink. -// -// +stateify savable type Symlink struct { Entry diff --git a/pkg/sentry/fs/ramfs/test/BUILD b/pkg/sentry/fs/ramfs/test/BUILD index 187eac49d..57fee45e2 100644 --- a/pkg/sentry/fs/ramfs/test/BUILD +++ b/pkg/sentry/fs/ramfs/test/BUILD @@ -1,16 +1,30 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "test_state", + srcs = [ + "test.go", + ], + out = "test_state.go", + package = "test", +) go_library( name = "test", testonly = 1, - srcs = ["test.go"], + srcs = [ + "test.go", + "test_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/sentry/context", + "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs", + "//pkg/state", ], ) diff --git a/pkg/sentry/fs/sys/BUILD b/pkg/sentry/fs/sys/BUILD index bc24e980e..095ff1f25 100644 --- a/pkg/sentry/fs/sys/BUILD +++ b/pkg/sentry/fs/sys/BUILD @@ -1,6 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "sys_state", + srcs = [ + "fs.go", + "sys.go", + ], + out = "sys_state.go", + package = "sys", +) go_library( name = "sys", @@ -8,6 +18,7 @@ go_library( "device.go", "fs.go", "sys.go", + "sys_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/sys", visibility = ["//pkg/sentry:internal"], @@ -17,5 +28,6 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs", "//pkg/sentry/usermem", + "//pkg/state", ], ) diff --git a/pkg/sentry/fs/sys/fs.go b/pkg/sentry/fs/sys/fs.go index 625525540..c6d5f7fd8 100644 --- a/pkg/sentry/fs/sys/fs.go +++ b/pkg/sentry/fs/sys/fs.go @@ -20,8 +20,6 @@ import ( ) // filesystem is a sysfs. -// -// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/sys/sys.go b/pkg/sentry/fs/sys/sys.go index b9b2fb4a1..ccf56f644 100644 --- a/pkg/sentry/fs/sys/sys.go +++ b/pkg/sentry/fs/sys/sys.go @@ -22,13 +22,12 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// +stateify savable -type dir struct { +type Dir struct { ramfs.Dir } func newDir(ctx context.Context, msrc *fs.MountSource, contents map[string]*fs.Inode) *fs.Inode { - d := &dir{} + d := &Dir{} d.InitDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) return fs.NewInode(d, msrc, fs.StableAttr{ DeviceID: sysfsDevice.DeviceID(), diff --git a/pkg/sentry/fs/timerfd/BUILD b/pkg/sentry/fs/timerfd/BUILD index ffdd7e0dc..8b1b7872e 100644 --- a/pkg/sentry/fs/timerfd/BUILD +++ b/pkg/sentry/fs/timerfd/BUILD @@ -1,19 +1,33 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "timerfd_state", + srcs = [ + "timerfd.go", + ], + out = "timerfd_state.go", + package = "timerfd", +) go_library( name = "timerfd", - srcs = ["timerfd.go"], + srcs = [ + "timerfd.go", + "timerfd_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/timerfd", visibility = ["//pkg/sentry:internal"], deps = [ + "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/fs/timerfd/timerfd.go b/pkg/sentry/fs/timerfd/timerfd.go index 767db95a0..ae58f6fd7 100644 --- a/pkg/sentry/fs/timerfd/timerfd.go +++ b/pkg/sentry/fs/timerfd/timerfd.go @@ -30,8 +30,6 @@ import ( ) // TimerOperations implements fs.FileOperations for timerfds. -// -// +stateify savable type TimerOperations struct { fsutil.ZeroSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` @@ -40,7 +38,7 @@ type TimerOperations struct { fsutil.NoMMap `state:"nosave"` fsutil.NoIoctl `state:"nosave"` - events waiter.Queue `state:"zerovalue"` + events waiter.Queue `state:"nosave"` timer *ktime.Timer // val is the number of timer expirations since the last successful call to diff --git a/pkg/sentry/fs/tmpfs/BUILD b/pkg/sentry/fs/tmpfs/BUILD index cfe11ab02..473ab4296 100644 --- a/pkg/sentry/fs/tmpfs/BUILD +++ b/pkg/sentry/fs/tmpfs/BUILD @@ -1,6 +1,18 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "tmpfs_state", + srcs = [ + "file_regular.go", + "fs.go", + "inode_file.go", + "tmpfs.go", + ], + out = "tmpfs_state.go", + package = "tmpfs", +) go_library( name = "tmpfs", @@ -10,11 +22,13 @@ go_library( "fs.go", "inode_file.go", "tmpfs.go", + "tmpfs_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tmpfs", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/log", "//pkg/sentry/context", "//pkg/sentry/device", "//pkg/sentry/fs", @@ -27,6 +41,7 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/tcpip/transport/unix", "//pkg/waiter", ], diff --git a/pkg/sentry/fs/tmpfs/file_regular.go b/pkg/sentry/fs/tmpfs/file_regular.go index 342688f81..9811d90bc 100644 --- a/pkg/sentry/fs/tmpfs/file_regular.go +++ b/pkg/sentry/fs/tmpfs/file_regular.go @@ -25,8 +25,6 @@ import ( // regularFileOperations implements fs.FileOperations for a regular // tmpfs file. -// -// +stateify savable type regularFileOperations struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go index ca620e65e..5bd9ade52 100644 --- a/pkg/sentry/fs/tmpfs/fs.go +++ b/pkg/sentry/fs/tmpfs/fs.go @@ -47,8 +47,6 @@ const ( var modeRegexp = regexp.MustCompile("0[0-7][0-7][0-7]") // Filesystem is a tmpfs. -// -// +stateify savable type Filesystem struct{} func init() { diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 1e4fe47d2..4e803c9ff 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -43,8 +43,6 @@ import ( // include an InvalidatorRegion associated with that reference. When the // referenced portion of the file is removed (with Truncate), the associated // InvalidatorRegion is invalidated. -// -// +stateify savable type fileInodeOperations struct { fsutil.DeprecatedFileOperations `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 10cb5451d..1cc7ae491 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -49,8 +49,6 @@ func rename(ctx context.Context, oldParent *fs.Inode, oldName string, newParent } // Dir is a directory. -// -// +stateify savable type Dir struct { ramfs.Dir @@ -124,8 +122,6 @@ func (*Dir) StatFS(context.Context) (fs.Info, error) { } // Symlink is a symlink. -// -// +stateify savable type Symlink struct { ramfs.Symlink } @@ -153,8 +149,6 @@ func (s *Symlink) StatFS(context.Context) (fs.Info, error) { } // Socket is a socket. -// -// +stateify savable type Socket struct { ramfs.Socket } @@ -182,8 +176,6 @@ func (s *Socket) StatFS(context.Context) (fs.Info, error) { } // Fifo is a tmpfs named pipe. -// -// +stateify savable type Fifo struct { ramfs.Entry } diff --git a/pkg/sentry/fs/tty/BUILD b/pkg/sentry/fs/tty/BUILD index 3c446eef4..363897b2c 100644 --- a/pkg/sentry/fs/tty/BUILD +++ b/pkg/sentry/fs/tty/BUILD @@ -1,6 +1,22 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "tty_state", + srcs = [ + "dir.go", + "fs.go", + "inode.go", + "line_discipline.go", + "master.go", + "queue.go", + "slave.go", + "terminal.go", + ], + out = "tty_state.go", + package = "tty", +) go_library( name = "tty", @@ -13,6 +29,7 @@ go_library( "queue.go", "slave.go", "terminal.go", + "tty_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tty", visibility = ["//pkg/sentry:internal"], @@ -27,6 +44,7 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index c91091db4..2c5b2aed6 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -49,16 +49,14 @@ import ( // corresponding Dirents hold on their parent (this directory). // // dirInodeOperations implements fs.InodeOperations. -// -// +stateify savable type dirInodeOperations struct { - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.InodeNotSocket `state:"nosave"` - fsutil.InodeNotRenameable `state:"nosave"` - fsutil.InodeNotSymlink `state:"nosave"` - fsutil.InodeNoExtendedAttributes `state:"nosave"` - fsutil.NoMappable `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` + fsutil.DeprecatedFileOperations + fsutil.InodeNotSocket + fsutil.InodeNotRenameable + fsutil.InodeNotSymlink + fsutil.InodeNoExtendedAttributes + fsutil.NoMappable + fsutil.NoopWriteOut // msrc is the super block this directory is on. // @@ -350,8 +348,6 @@ func (d *dirInodeOperations) masterClose(t *Terminal) { // // This is nearly identical to fsutil.DirFileOperations, except that it takes // df.di.mu in IterateDir. -// -// +stateify savable type dirFileOperations struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/tty/fs.go b/pkg/sentry/fs/tty/fs.go index e28635607..dbaffe95e 100644 --- a/pkg/sentry/fs/tty/fs.go +++ b/pkg/sentry/fs/tty/fs.go @@ -28,8 +28,6 @@ var ptsDevice = device.NewAnonDevice() // // This devpts is always in the new "multi-instance" mode. i.e., it contains a // ptmx device tied to this mount. -// -// +stateify savable type filesystem struct{} func init() { @@ -71,8 +69,6 @@ func (f *filesystem) Mount(ctx context.Context, device string, flags fs.MountSou } // superOperations implements fs.MountSourceOperations, preventing caching. -// -// +stateify savable type superOperations struct{} // Revalidate implements fs.DirentOperations.Revalidate. diff --git a/pkg/sentry/fs/tty/inode.go b/pkg/sentry/fs/tty/inode.go index c0fa2b407..04b9a7727 100644 --- a/pkg/sentry/fs/tty/inode.go +++ b/pkg/sentry/fs/tty/inode.go @@ -31,8 +31,6 @@ import ( // // * fs.InodeOperations.Release // * fs.InodeOperations.GetFile -// -// +stateify savable type inodeOperations struct { fsutil.DeprecatedFileOperations `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` diff --git a/pkg/sentry/fs/tty/line_discipline.go b/pkg/sentry/fs/tty/line_discipline.go index d243ee40e..f094635f5 100644 --- a/pkg/sentry/fs/tty/line_discipline.go +++ b/pkg/sentry/fs/tty/line_discipline.go @@ -72,8 +72,6 @@ const ( // termiosMu // inQueue.mu // outQueue.mu -// -// +stateify savable type lineDiscipline struct { // inQueue is the input queue of the terminal. inQueue queue @@ -185,8 +183,6 @@ type transformer interface { // outputQueueTransformer implements transformer. It performs line discipline // transformations on the output queue. -// -// +stateify savable type outputQueueTransformer struct{} // transform does output processing for one end of the pty. See @@ -258,8 +254,6 @@ func (*outputQueueTransformer) transform(l *lineDiscipline, q *queue, buf []byte // inputQueueTransformer implements transformer. It performs line discipline // transformations on the input queue. -// -// +stateify savable type inputQueueTransformer struct{} // transform does input processing for one end of the pty. Characters read are diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index c7198e218..74cdbe874 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -27,8 +27,6 @@ import ( // masterInodeOperations are the fs.InodeOperations for the master end of the // Terminal (ptmx file). -// -// +stateify savable type masterInodeOperations struct { inodeOperations @@ -98,8 +96,6 @@ func (mi *masterInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flag } // masterFileOperations are the fs.FileOperations for the master end of a terminal. -// -// +stateify savable type masterFileOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/tty/queue.go b/pkg/sentry/fs/tty/queue.go index 42c105abc..026d5e077 100644 --- a/pkg/sentry/fs/tty/queue.go +++ b/pkg/sentry/fs/tty/queue.go @@ -32,13 +32,11 @@ import ( // processed (i.e. undergo termios transformations) as they are added to the // read buffer. The read buffer is readable when its length is nonzero and // readable is true. -// -// +stateify savable type queue struct { // mu protects everything in queue. mu sync.Mutex `state:"nosave"` - waiter.Queue `state:"zerovalue"` + waiter.Queue `state:"nosave"` // readBuf is buffer of data ready to be read when readable is true. // This data has been processed. diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index 1c562b172..f5eec726e 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -27,8 +27,6 @@ import ( // slaveInodeOperations are the fs.InodeOperations for the slave end of the // Terminal (pts file). -// -// +stateify savable type slaveInodeOperations struct { inodeOperations @@ -88,8 +86,6 @@ func (si *slaveInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags } // slaveFileOperations are the fs.FileOperations for the slave end of a terminal. -// -// +stateify savable type slaveFileOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/tty/terminal.go b/pkg/sentry/fs/tty/terminal.go index 3cb135124..fa5b00409 100644 --- a/pkg/sentry/fs/tty/terminal.go +++ b/pkg/sentry/fs/tty/terminal.go @@ -21,8 +21,6 @@ import ( ) // Terminal is a pseudoterminal. -// -// +stateify savable type Terminal struct { refs.AtomicRefCount diff --git a/pkg/sentry/inet/BUILD b/pkg/sentry/inet/BUILD index 159c50efb..eaf8f15b2 100644 --- a/pkg/sentry/inet/BUILD +++ b/pkg/sentry/inet/BUILD @@ -3,15 +3,26 @@ package( licenses = ["notice"], # Apache 2.0 ) -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "inet_state", + srcs = ["inet.go"], + out = "inet_state.go", + package = "inet", +) go_library( name = "inet", srcs = [ "context.go", "inet.go", + "inet_state.go", "test_stack.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/inet", - deps = ["//pkg/sentry/context"], + deps = [ + "//pkg/sentry/context", + "//pkg/state", + ], ) diff --git a/pkg/sentry/inet/inet.go b/pkg/sentry/inet/inet.go index e54a61196..e4b326993 100644 --- a/pkg/sentry/inet/inet.go +++ b/pkg/sentry/inet/inet.go @@ -87,8 +87,6 @@ type InterfaceAddr struct { } // TCPBufferSize contains settings controlling TCP buffer sizing. -// -// +stateify savable type TCPBufferSize struct { // Min is the minimum size. Min int diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 0ebacefa6..c4a7dacb2 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -1,7 +1,58 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "kernel_state", + srcs = [ + "abstract_socket_namespace.go", + "fd_map.go", + "fs_context.go", + "ipc_namespace.go", + "kernel.go", + "kernel_state.go", + "pending_signals.go", + "pending_signals_state.go", + "process_group_list.go", + "ptrace.go", + "rseq.go", + "session_list.go", + "sessions.go", + "signal.go", + "signal_handlers.go", + "syscalls.go", + "syscalls_state.go", + "syslog.go", + "task.go", + "task_clone.go", + "task_context.go", + "task_exec.go", + "task_exit.go", + "task_list.go", + "task_resources.go", + "task_run.go", + "task_sched.go", + "task_signals.go", + "task_start.go", + "task_syscall.go", + "thread_group.go", + "threads.go", + "timekeeper.go", + "timekeeper_state.go", + "timer.go", + "uts_namespace.go", + "vdso.go", + "version.go", + ], + out = "kernel_autogen_state.go", + imports = [ + "gvisor.googlesource.com/gvisor/pkg/sentry/arch", + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", + "gvisor.googlesource.com/gvisor/pkg/tcpip", + ], + package = "kernel", +) go_template_instance( name = "pending_signals_list", @@ -67,6 +118,7 @@ go_library( "fs_context.go", "ipc_namespace.go", "kernel.go", + "kernel_autogen_state.go", "kernel_state.go", "pending_signals.go", "pending_signals_list.go", @@ -112,11 +164,6 @@ go_library( "version.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel", - imports = [ - "gvisor.googlesource.com/gvisor/pkg/sentry/arch", - # "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", - "gvisor.googlesource.com/gvisor/pkg/tcpip", - ], visibility = ["//:sandbox"], deps = [ "//pkg/abi", diff --git a/pkg/sentry/kernel/abstract_socket_namespace.go b/pkg/sentry/kernel/abstract_socket_namespace.go index d6d1d341d..014c4a3bf 100644 --- a/pkg/sentry/kernel/abstract_socket_namespace.go +++ b/pkg/sentry/kernel/abstract_socket_namespace.go @@ -22,7 +22,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) -// +stateify savable type abstractEndpoint struct { ep unix.BoundEndpoint wr *refs.WeakRef @@ -40,8 +39,6 @@ func (e *abstractEndpoint) WeakRefGone() { } // AbstractSocketNamespace is used to implement the Linux abstract socket functionality. -// -// +stateify savable type AbstractSocketNamespace struct { mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/auth/BUILD b/pkg/sentry/kernel/auth/BUILD index a81085372..5b7b30557 100644 --- a/pkg/sentry/kernel/auth/BUILD +++ b/pkg/sentry/kernel/auth/BUILD @@ -1,7 +1,20 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "auth_state", + srcs = [ + "credentials.go", + "id.go", + "id_map_range.go", + "id_map_set.go", + "user_namespace.go", + ], + out = "auth_state.go", + package = "auth", +) go_template_instance( name = "id_map_range", @@ -35,6 +48,7 @@ go_library( name = "auth", srcs = [ "auth.go", + "auth_state.go", "capability_set.go", "context.go", "credentials.go", @@ -52,6 +66,7 @@ go_library( "//pkg/bits", "//pkg/log", "//pkg/sentry/context", + "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go index f18f7dac9..f6fb05285 100644 --- a/pkg/sentry/kernel/auth/credentials.go +++ b/pkg/sentry/kernel/auth/credentials.go @@ -21,8 +21,6 @@ import ( // Credentials contains information required to authorize privileged operations // in a user namespace. -// -// +stateify savable type Credentials struct { // Real/effective/saved user/group IDs in the root user namespace. None of // these should ever be NoID. diff --git a/pkg/sentry/kernel/auth/id_map.go b/pkg/sentry/kernel/auth/id_map.go index bd0090e0f..6adb33530 100644 --- a/pkg/sentry/kernel/auth/id_map.go +++ b/pkg/sentry/kernel/auth/id_map.go @@ -77,8 +77,6 @@ func (ns *UserNamespace) allIDsMapped(m *idMapSet, start, end uint32) bool { // An IDMapEntry represents a mapping from a range of contiguous IDs in a user // namespace to an equally-sized range of contiguous IDs in the namespace's // parent. -// -// +stateify savable type IDMapEntry struct { // FirstID is the first ID in the range in the namespace. FirstID uint32 diff --git a/pkg/sentry/kernel/auth/user_namespace.go b/pkg/sentry/kernel/auth/user_namespace.go index d359f3f31..0980aeadf 100644 --- a/pkg/sentry/kernel/auth/user_namespace.go +++ b/pkg/sentry/kernel/auth/user_namespace.go @@ -23,8 +23,6 @@ import ( // A UserNamespace represents a user namespace. See user_namespaces(7) for // details. -// -// +stateify savable type UserNamespace struct { // parent is this namespace's parent. If this is the root namespace, parent // is nil. The parent pointer is immutable. diff --git a/pkg/sentry/kernel/epoll/BUILD b/pkg/sentry/kernel/epoll/BUILD index 5e8b36ed6..7d491efbc 100644 --- a/pkg/sentry/kernel/epoll/BUILD +++ b/pkg/sentry/kernel/epoll/BUILD @@ -1,11 +1,22 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "epoll_autogen_state", + srcs = [ + "epoll.go", + "epoll_state.go", + ], + out = "epoll_autogen_state.go", + package = "epoll", +) go_library( name = "epoll", srcs = [ "epoll.go", + "epoll_autogen_state.go", "epoll_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll", @@ -18,7 +29,9 @@ go_library( "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/kdefs", + "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index d87e64a1c..b572fcd7e 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -58,8 +58,6 @@ const ( // potentially be reassigned. We also cannot use just the file pointer because // it is possible to have multiple entries for the same file object as long as // they are created with different FDs (i.e., the FDs point to the same file). -// -// +stateify savable type FileIdentifier struct { File *fs.File Fd kdefs.FD @@ -67,8 +65,6 @@ type FileIdentifier struct { // pollEntry holds all the state associated with an event poll entry, that is, // a file being observed by an event poll object. -// -// +stateify savable type pollEntry struct { ilist.Entry file *refs.WeakRef `state:"manual"` @@ -96,8 +92,6 @@ func (p *pollEntry) WeakRefGone() { // EventPoll holds all the state associated with an event poll object, that is, // collection of files to observe and their current state. -// -// +stateify savable type EventPoll struct { fsutil.PipeSeek `state:"zerovalue"` fsutil.NotDirReaddir `state:"zerovalue"` @@ -108,7 +102,7 @@ type EventPoll struct { // Wait queue is used to notify interested parties when the event poll // object itself becomes readable or writable. - waiter.Queue `state:"zerovalue"` + waiter.Queue // files is the map of all the files currently being observed, it is // protected by mu. diff --git a/pkg/sentry/kernel/eventfd/BUILD b/pkg/sentry/kernel/eventfd/BUILD index cc1120b4f..7ec179bd8 100644 --- a/pkg/sentry/kernel/eventfd/BUILD +++ b/pkg/sentry/kernel/eventfd/BUILD @@ -1,19 +1,33 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "eventfd_state", + srcs = [ + "eventfd.go", + ], + out = "eventfd_state.go", + package = "eventfd", +) go_library( name = "eventfd", - srcs = ["eventfd.go"], + srcs = [ + "eventfd.go", + "eventfd_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/eventfd", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/waiter", "//pkg/waiter/fdnotifier", diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index a4ada0e78..bd50bd9fe 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -35,8 +35,6 @@ import ( // EventOperations represents an event with the semantics of Linux's file-based event // notification (eventfd). Eventfds are usually internal to the Sentry but in certain // situations they may be converted into a host-backed eventfd. -// -// +stateify savable type EventOperations struct { fsutil.NoopRelease `state:"nosave"` fsutil.PipeSeek `state:"nosave"` @@ -51,7 +49,7 @@ type EventOperations struct { // Queue is used to notify interested parties when the event object // becomes readable or writable. - wq waiter.Queue `state:"zerovalue"` + wq waiter.Queue `state:"nosave"` // val is the current value of the event counter. val uint64 diff --git a/pkg/sentry/kernel/fd_map.go b/pkg/sentry/kernel/fd_map.go index d5d4aaacb..299506330 100644 --- a/pkg/sentry/kernel/fd_map.go +++ b/pkg/sentry/kernel/fd_map.go @@ -46,8 +46,6 @@ func (f FDs) Less(i, j int) bool { } // FDFlags define flags for an individual descriptor. -// -// +stateify savable type FDFlags struct { // CloseOnExec indicates the descriptor should be closed on exec. CloseOnExec bool @@ -71,16 +69,12 @@ func (f FDFlags) ToLinuxFDFlags() (mask uint) { // descriptor holds the details about a file descriptor, namely a pointer the // file itself and the descriptor flags. -// -// +stateify savable type descriptor struct { file *fs.File flags FDFlags } // FDMap is used to manage File references and flags. -// -// +stateify savable type FDMap struct { refs.AtomicRefCount k *Kernel diff --git a/pkg/sentry/kernel/fs_context.go b/pkg/sentry/kernel/fs_context.go index f3f05e8f5..dbc097696 100644 --- a/pkg/sentry/kernel/fs_context.go +++ b/pkg/sentry/kernel/fs_context.go @@ -25,8 +25,6 @@ import ( // FSContext contains filesystem context. // // This includes umask and working directory. -// -// +stateify savable type FSContext struct { refs.AtomicRefCount diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index b44a26974..a97a43549 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_template_instance( name = "waiter_list", @@ -14,15 +14,29 @@ go_template_instance( }, ) +go_stateify( + name = "futex_state", + srcs = [ + "futex.go", + "waiter_list.go", + ], + out = "futex_state.go", + package = "futex", +) + go_library( name = "futex", srcs = [ "futex.go", + "futex_state.go", "waiter_list.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex", visibility = ["//pkg/sentry:internal"], - deps = ["//pkg/syserror"], + deps = [ + "//pkg/state", + "//pkg/syserror", + ], ) go_test( diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index 4a1f2a0ef..15e3e5e2c 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -196,8 +196,6 @@ func bucketIndexForAddr(addr uintptr) uintptr { } // Manager holds futex state for a single virtual address space. -// -// +stateify savable type Manager struct { buckets [bucketCount]bucket `state:"zerovalue"` } diff --git a/pkg/sentry/kernel/ipc_namespace.go b/pkg/sentry/kernel/ipc_namespace.go index 5eef49f59..a86bda77b 100644 --- a/pkg/sentry/kernel/ipc_namespace.go +++ b/pkg/sentry/kernel/ipc_namespace.go @@ -21,8 +21,6 @@ import ( ) // IPCNamespace represents an IPC namespace. -// -// +stateify savable type IPCNamespace struct { // User namespace which owns this IPC namespace. Immutable. userNS *auth.UserNamespace diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 419a1d473..64439cd9d 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -62,8 +62,6 @@ import ( // Kernel represents an emulated Linux kernel. It must be initialized by calling // Init() or LoadFrom(). -// -// +stateify savable type Kernel struct { // extMu serializes external changes to the Kernel with calls to // Kernel.SaveTo. (Kernel.SaveTo requires that the state of the Kernel @@ -160,7 +158,7 @@ type Kernel struct { // exitErr is the error causing the sandbox to exit, if any. It is // protected by extMu. - exitErr error `state:"nosave"` + exitErr error // danglingEndpoints is used to save / restore tcpip.DanglingEndpoints. danglingEndpoints struct{} `state:".([]tcpip.Endpoint)"` diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go index 06be5a7e1..5dc0f266c 100644 --- a/pkg/sentry/kernel/pending_signals.go +++ b/pkg/sentry/kernel/pending_signals.go @@ -38,8 +38,6 @@ const ( // pendingSignals holds a collection of pending signals. The zero value of // pendingSignals is a valid empty collection. pendingSignals is thread-unsafe; // users must provide synchronization. -// -// +stateify savable type pendingSignals struct { // signals contains all pending signals. // @@ -54,14 +52,11 @@ type pendingSignals struct { } // pendingSignalQueue holds a pendingSignalList for a single signal number. -// -// +stateify savable type pendingSignalQueue struct { pendingSignalList length int } -// +stateify savable type pendingSignal struct { // pendingSignalEntry links into a pendingSignalList. pendingSignalEntry diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD index 19b23c6d2..4600d19bd 100644 --- a/pkg/sentry/kernel/pipe/BUILD +++ b/pkg/sentry/kernel/pipe/BUILD @@ -1,6 +1,20 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "pipe_state", + srcs = [ + "buffers.go", + "node.go", + "pipe.go", + "reader.go", + "reader_writer.go", + "writer.go", + ], + out = "pipe_state.go", + package = "pipe", +) go_library( name = "pipe", @@ -9,6 +23,7 @@ go_library( "device.go", "node.go", "pipe.go", + "pipe_state.go", "reader.go", "reader_writer.go", "writer.go", @@ -19,12 +34,15 @@ go_library( "//pkg/abi/linux", "//pkg/amutex", "//pkg/ilist", + "//pkg/log", + "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/kernel/pipe/buffers.go b/pkg/sentry/kernel/pipe/buffers.go index a82e45c3f..f300537c5 100644 --- a/pkg/sentry/kernel/pipe/buffers.go +++ b/pkg/sentry/kernel/pipe/buffers.go @@ -20,8 +20,6 @@ import ( // Buffer encapsulates a queueable byte buffer that can // easily be truncated. It is designed only for use with pipes. -// -// +stateify savable type Buffer struct { ilist.Entry data []byte diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index 23d692da1..e418cf174 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -24,8 +24,6 @@ import ( ) // inodeOperations wraps fs.InodeOperations operations with common pipe opening semantics. -// -// +stateify savable type inodeOperations struct { fs.InodeOperations diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index ced2559a7..9a21df5b4 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -41,8 +41,6 @@ const DefaultPipeSize = 65536 // Pipe is an encapsulation of a platform-independent pipe. // It manages a buffered byte queue shared between a reader/writer // pair. -// -// +stateify savable type Pipe struct { waiter.Queue `state:"nosave"` diff --git a/pkg/sentry/kernel/pipe/reader.go b/pkg/sentry/kernel/pipe/reader.go index 1fa5e9a32..40d5e4943 100644 --- a/pkg/sentry/kernel/pipe/reader.go +++ b/pkg/sentry/kernel/pipe/reader.go @@ -20,8 +20,6 @@ import ( // Reader satisfies the fs.FileOperations interface for read-only pipes. // Reader should be used with !fs.FileFlags.Write to reject writes. -// -// +stateify savable type Reader struct { ReaderWriter } diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index 82607367b..dc642a3a6 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -31,8 +31,6 @@ import ( // read and write requests. This should only be used directly for named pipes. // pipe(2) and pipe2(2) only support unidirectional pipes and should use // either pipe.Reader or pipe.Writer. -// -// +stateify savable type ReaderWriter struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/kernel/pipe/writer.go b/pkg/sentry/kernel/pipe/writer.go index d93324b53..fd13008ac 100644 --- a/pkg/sentry/kernel/pipe/writer.go +++ b/pkg/sentry/kernel/pipe/writer.go @@ -20,8 +20,6 @@ import ( // Writer satisfies the fs.FileOperations interface for write-only pipes. // Writer should be used with !fs.FileFlags.Read to reject reads. -// -// +stateify savable type Writer struct { ReaderWriter } diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index e9e69004d..f1c2c4bf0 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -25,8 +25,6 @@ import ( // ptraceOptions are the subset of options controlling a task's ptrace behavior // that are set by ptrace(PTRACE_SETOPTIONS). -// -// +stateify savable type ptraceOptions struct { // ExitKill is true if the tracee should be sent SIGKILL when the tracer // exits. @@ -187,8 +185,6 @@ func (t *Task) hasTracer() bool { } // ptraceStop is a TaskStop placed on tasks in a ptrace-stop. -// -// +stateify savable type ptraceStop struct { // If frozen is true, the stopped task's tracer is currently operating on // it, so Task.Kill should not remove the stop. diff --git a/pkg/sentry/kernel/rseq.go b/pkg/sentry/kernel/rseq.go index 1f3de58e3..635372993 100644 --- a/pkg/sentry/kernel/rseq.go +++ b/pkg/sentry/kernel/rseq.go @@ -23,8 +23,6 @@ import ( // Restartable sequences, as described in https://lwn.net/Articles/650333/. // RSEQCriticalRegion describes a restartable sequence critical region. -// -// +stateify savable type RSEQCriticalRegion struct { // When a task in this thread group has its CPU preempted (as defined by // platform.ErrContextCPUPreempted) or has a signal delivered to an diff --git a/pkg/sentry/kernel/semaphore/BUILD b/pkg/sentry/kernel/semaphore/BUILD index e7fa44e2c..969145fe1 100644 --- a/pkg/sentry/kernel/semaphore/BUILD +++ b/pkg/sentry/kernel/semaphore/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") go_template_instance( name = "waiter_list", @@ -14,10 +14,21 @@ go_template_instance( }, ) +go_stateify( + name = "semaphore_state", + srcs = [ + "semaphore.go", + "waiter_list.go", + ], + out = "semaphore_autogen_state.go", + package = "semaphore", +) + go_library( name = "semaphore", srcs = [ "semaphore.go", + "semaphore_autogen_state.go", "waiter_list.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/semaphore", @@ -29,6 +40,8 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", + "//pkg/state", + "//pkg/state/statefile", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index aa07946cf..a1ee83ce5 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -42,8 +42,6 @@ const ( ) // Registry maintains a set of semaphores that can be found by key or ID. -// -// +stateify savable type Registry struct { // userNS owning the ipc name this registry belongs to. Immutable. userNS *auth.UserNamespace @@ -54,8 +52,6 @@ type Registry struct { } // Set represents a set of semaphores that can be operated atomically. -// -// +stateify savable type Set struct { // registry owning this sem set. Immutable. registry *Registry @@ -83,8 +79,6 @@ type Set struct { } // sem represents a single semanphore from a set. -// -// +stateify savable type sem struct { value int16 waiters waiterList `state:"zerovalue"` @@ -92,8 +86,6 @@ type sem struct { // waiter represents a caller that is waiting for the semaphore value to // become positive or zero. -// -// +stateify savable type waiter struct { waiterEntry diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index cf4e18805..fa4c7b8f6 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -27,8 +27,6 @@ type SessionID ThreadID type ProcessGroupID ThreadID // Session contains a leader threadgroup and a list of ProcessGroups. -// -// +stateify savable type Session struct { refs refs.AtomicRefCount @@ -78,8 +76,6 @@ func (s *Session) decRef() { } // ProcessGroup contains an originator threadgroup and a parent Session. -// -// +stateify savable type ProcessGroup struct { refs refs.AtomicRefCount // not exported. diff --git a/pkg/sentry/kernel/shm/BUILD b/pkg/sentry/kernel/shm/BUILD index 40e641355..0f88eb0ac 100644 --- a/pkg/sentry/kernel/shm/BUILD +++ b/pkg/sentry/kernel/shm/BUILD @@ -1,12 +1,22 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "shm_state", + srcs = [ + "shm.go", + ], + out = "shm_autogen_state.go", + package = "shm", +) go_library( name = "shm", srcs = [ "device.go", "shm.go", + "shm_autogen_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/shm", visibility = ["//pkg/sentry:internal"], @@ -23,6 +33,7 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 1ac444094..7217e8103 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -72,8 +72,6 @@ const ( // Registry tracks all shared memory segments in an IPC namespace. The registry // provides the mechanisms for creating and finding segments, and reporting // global shm parameters. -// -// +stateify savable type Registry struct { // userNS owns the IPC namespace this registry belong to. Immutable. userNS *auth.UserNamespace @@ -290,8 +288,6 @@ func (r *Registry) remove(s *Shm) { // shmctl(SHM_RMID). // // Shm implements memmap.Mappable and memmap.MappingIdentity. -// -// +stateify savable type Shm struct { // AtomicRefCount tracks the number of references to this segment from // maps. A segment always holds a reference to itself, until it's marked for diff --git a/pkg/sentry/kernel/signal_handlers.go b/pkg/sentry/kernel/signal_handlers.go index 3649f5e4d..21ba4ee70 100644 --- a/pkg/sentry/kernel/signal_handlers.go +++ b/pkg/sentry/kernel/signal_handlers.go @@ -22,8 +22,6 @@ import ( ) // SignalHandlers holds information about signal actions. -// -// +stateify savable type SignalHandlers struct { // mu protects actions, as well as the signal state of all tasks and thread // groups using this SignalHandlers object. (See comment on diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go index 4c7811b6c..e20fa3eb6 100644 --- a/pkg/sentry/kernel/syscalls.go +++ b/pkg/sentry/kernel/syscalls.go @@ -176,8 +176,6 @@ type Stracer interface { // SyscallTable is a lookup table of system calls. Critically, a SyscallTable // is *immutable*. In order to make supporting suspend and resume sane, they // must be uniquely registered and may not change during operation. -// -// +stateify savable type SyscallTable struct { // OS is the operating system that this syscall table implements. OS abi.OS `state:"wait"` diff --git a/pkg/sentry/kernel/syslog.go b/pkg/sentry/kernel/syslog.go index 125312b6a..31541749e 100644 --- a/pkg/sentry/kernel/syslog.go +++ b/pkg/sentry/kernel/syslog.go @@ -23,8 +23,6 @@ import ( // syslog represents a sentry-global kernel log. // // Currently, it contains only fun messages for a dmesg easter egg. -// -// +stateify savable type syslog struct { // mu protects the below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index ae9b3d175..7f6735320 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -52,8 +52,6 @@ import ( // All fields that are "exclusive to the task goroutine" can only be accessed // by the task goroutine while it is running. The task goroutine does not // require synchronization to read or write these fields. -// -// +stateify savable type Task struct { taskNode diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 38f7826e2..a61283267 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -349,7 +349,6 @@ func (t *Task) unstopVforkParent() { } } -// +stateify savable type runSyscallAfterPtraceEventClone struct { vforkChild *Task @@ -367,7 +366,6 @@ func (r *runSyscallAfterPtraceEventClone) execute(t *Task) taskRunState { return (*runSyscallExit)(nil) } -// +stateify savable type runSyscallAfterVforkStop struct { // childTID has the same meaning as // runSyscallAfterPtraceEventClone.vforkChildTID. @@ -473,8 +471,6 @@ func (t *Task) Unshare(opts *SharingOptions) error { // current MM. (Normally, CLONE_VFORK is used in conjunction with CLONE_VM, so // that the child and parent share mappings until the child execve()s into a // new process image or exits.) -// -// +stateify savable type vforkStop struct{} // StopIgnoresKill implements TaskStop.Killable. diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index 9a59cbd33..5c563ba08 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -35,8 +35,6 @@ var ErrNoSyscalls = errors.New("no syscall table found") type Auxmap map[string]interface{} // TaskContext is the subset of a task's data that is provided by the loader. -// -// +stateify savable type TaskContext struct { // Name is the thread name set by the prctl(PR_SET_NAME) system call. Name string diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index 385299b24..2285847a2 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -73,8 +73,6 @@ import ( // execStop is a TaskStop that a task sets on itself when it wants to execve // and is waiting for the other tasks in its thread group to exit first. -// -// +stateify savable type execStop struct{} // Killable implements TaskStop.Killable. @@ -121,8 +119,6 @@ func (t *Task) Execve(newTC *TaskContext) (*SyscallControl, error) { // The runSyscallAfterExecStop state continues execve(2) after all siblings of // a thread in the execve syscall have exited. -// -// +stateify savable type runSyscallAfterExecStop struct { tc *TaskContext } diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index b16844e91..d6604f37b 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -38,8 +38,6 @@ import ( // An ExitStatus is a value communicated from an exiting task or thread group // to the party that reaps it. -// -// +stateify savable type ExitStatus struct { // Code is the numeric value passed to the call to exit or exit_group that // caused the exit. If the exit was not caused by such a call, Code is 0. @@ -224,8 +222,6 @@ func (t *Task) advanceExitStateLocked(oldExit, newExit TaskExitState) { } // runExit is the entry point into the task exit path. -// -// +stateify savable type runExit struct{} func (*runExit) execute(t *Task) taskRunState { @@ -233,7 +229,6 @@ func (*runExit) execute(t *Task) taskRunState { return (*runExitMain)(nil) } -// +stateify savable type runExitMain struct{} func (*runExitMain) execute(t *Task) taskRunState { @@ -536,7 +531,6 @@ func (t *Task) reparentLocked(parent *Task) { // tracer (if one exists) and reaps the leader immediately. In Linux, this is // in fs/exec.c:de_thread(); in the sentry, this is in Task.promoteLocked(). -// +stateify savable type runExitNotify struct{} func (*runExitNotify) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_resources.go b/pkg/sentry/kernel/task_resources.go index 0832bf989..4ca25664a 100644 --- a/pkg/sentry/kernel/task_resources.go +++ b/pkg/sentry/kernel/task_resources.go @@ -21,8 +21,6 @@ import ( // TaskResources is the subset of a task's data provided by its creator that is // not provided by the loader. -// -// +stateify savable type TaskResources struct { // SignalMask is the set of signals whose delivery is currently blocked. // diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 8dd0ef6ea..a03fa6ac0 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -131,8 +131,6 @@ func (t *Task) doStop() { // The runApp state checks for interrupts before executing untrusted // application code. -// -// +stateify savable type runApp struct{} func (*runApp) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_sched.go b/pkg/sentry/kernel/task_sched.go index 49141ab74..b50139077 100644 --- a/pkg/sentry/kernel/task_sched.go +++ b/pkg/sentry/kernel/task_sched.go @@ -65,8 +65,6 @@ const ( // TaskGoroutineSchedInfo contains task goroutine scheduling state which must // be read and updated atomically. -// -// +stateify savable type TaskGoroutineSchedInfo struct { // Timestamp was the value of Kernel.cpuClock when this // TaskGoroutineSchedInfo was last updated. diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 62ec530be..91f6c0874 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -748,8 +748,6 @@ func (t *Task) CopyInSignalStack(addr usermem.Addr) (arch.SignalStack, error) { // groupStop is a TaskStop placed on tasks that have received a stop signal // (SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU). (The term "group-stop" originates from // the ptrace man page.) -// -// +stateify savable type groupStop struct{} // Killable implements TaskStop.Killable. @@ -883,8 +881,6 @@ func (t *Task) signalStop(target *Task, code int32, status int32) { } // The runInterrupt state handles conditions indicated by interrupts. -// -// +stateify savable type runInterrupt struct{} func (*runInterrupt) execute(t *Task) taskRunState { @@ -1024,7 +1020,6 @@ func (*runInterrupt) execute(t *Task) taskRunState { return (*runApp)(nil) } -// +stateify savable type runInterruptAfterSignalDeliveryStop struct{} func (*runInterruptAfterSignalDeliveryStop) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index 3b9652504..79f4ff60c 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -241,7 +241,6 @@ func (t *Task) doSyscallEnter(sysno uintptr, args arch.SyscallArguments) taskRun return t.doSyscallInvoke(sysno, args) } -// +stateify savable type runSyscallAfterSyscallEnterStop struct{} func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState { @@ -261,7 +260,6 @@ func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState { return t.doSyscallInvoke(sysno, args) } -// +stateify savable type runSyscallAfterSysemuStop struct{} func (*runSyscallAfterSysemuStop) execute(t *Task) taskRunState { @@ -296,7 +294,6 @@ func (t *Task) doSyscallInvoke(sysno uintptr, args arch.SyscallArguments) taskRu return (*runSyscallExit)(nil).execute(t) } -// +stateify savable type runSyscallReinvoke struct{} func (*runSyscallReinvoke) execute(t *Task) taskRunState { @@ -313,7 +310,6 @@ func (*runSyscallReinvoke) execute(t *Task) taskRunState { return t.doSyscallInvoke(sysno, args) } -// +stateify savable type runSyscallExit struct{} func (*runSyscallExit) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index 441b8a822..8fffd3446 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -28,8 +28,6 @@ import ( // groups" are usually called "processes" in userspace documentation.) // // ThreadGroup is a superset of Linux's struct signal_struct. -// -// +stateify savable type ThreadGroup struct { threadGroupNode diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index 844213c35..440da9dad 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -50,8 +50,6 @@ func (tid ThreadID) String() string { const InitTID ThreadID = 1 // A TaskSet comprises all tasks in a system. -// -// +stateify savable type TaskSet struct { // mu protects all relationships betweens tasks and thread groups in the // TaskSet. (mu is approximately equivalent to Linux's tasklist_lock.) @@ -112,8 +110,6 @@ func (ts *TaskSet) forEachThreadGroupLocked(f func(tg *ThreadGroup)) { // // N.B. A task is said to be visible in a PID namespace if the PID namespace // contains a thread ID that maps to that task. -// -// +stateify savable type PIDNamespace struct { // owner is the TaskSet that this PID namespace belongs to. The owner // pointer is immutable. @@ -267,8 +263,6 @@ func (ns *PIDNamespace) UserNamespace() *auth.UserNamespace { // (threadGroupNode is an anonymous field in ThreadGroup; this is to expose // threadGroupEntry's methods on ThreadGroup to make it implement // threadGroupLinker.) -// -// +stateify savable type threadGroupNode struct { // pidns is the PID namespace containing the thread group and all of its // member tasks. The pidns pointer is immutable. @@ -388,8 +382,6 @@ func (tg *ThreadGroup) ID() ThreadID { // A taskNode defines the relationship between a task and the rest of the // system. The comments on threadGroupNode also apply to taskNode. -// -// +stateify savable type taskNode struct { // tg is the thread group that this task belongs to. The tg pointer is // immutable. diff --git a/pkg/sentry/kernel/time/BUILD b/pkg/sentry/kernel/time/BUILD index 5d8db2273..b3ed42aa4 100644 --- a/pkg/sentry/kernel/time/BUILD +++ b/pkg/sentry/kernel/time/BUILD @@ -1,18 +1,30 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "time_state", + srcs = [ + "time.go", + ], + out = "time_state.go", + package = "time", +) go_library( name = "time", srcs = [ "context.go", "time.go", + "time_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/log", "//pkg/sentry/context", + "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go index 6eadd2878..c223c2f19 100644 --- a/pkg/sentry/kernel/time/time.go +++ b/pkg/sentry/kernel/time/time.go @@ -42,8 +42,6 @@ const ( // // Time may represent time with respect to any clock and may not have any // meaning in the real world. -// -// +stateify savable type Time struct { ns int64 } @@ -288,8 +286,6 @@ type TimerListener interface { } // Setting contains user-controlled mutable Timer properties. -// -// +stateify savable type Setting struct { // Enabled is true if the timer is running. Enabled bool @@ -375,8 +371,6 @@ func (s Setting) advancedTo(now Time) (Setting, uint64) { // // Timers should be created using NewTimer and must be cleaned up by calling // Timer.Destroy when no longer used. -// -// +stateify savable type Timer struct { // clock is the time source. clock is immutable. clock Clock diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go index df5dbe128..4de8ac13b 100644 --- a/pkg/sentry/kernel/timekeeper.go +++ b/pkg/sentry/kernel/timekeeper.go @@ -25,8 +25,6 @@ import ( ) // Timekeeper manages all of the kernel clocks. -// -// +stateify savable type Timekeeper struct { // clocks are the clock sources. // diff --git a/pkg/sentry/kernel/timer.go b/pkg/sentry/kernel/timer.go index 534d03d0f..03a3310be 100644 --- a/pkg/sentry/kernel/timer.go +++ b/pkg/sentry/kernel/timer.go @@ -26,8 +26,6 @@ import ( // timekeeperClock is a ktime.Clock that reads time from a // kernel.Timekeeper-managed clock. -// -// +stateify savable type timekeeperClock struct { tk *Timekeeper c sentrytime.ClockID @@ -51,8 +49,6 @@ func (tc *timekeeperClock) Now() ktime.Time { // tgClock is a ktime.Clock that measures the time a thread group has spent // executing. -// -// +stateify savable type tgClock struct { tg *ThreadGroup @@ -159,8 +155,6 @@ func (tc *taskClock) Now() ktime.Time { } // signalNotifier is a ktime.Listener that sends signals to a ThreadGroup. -// -// +stateify savable type signalNotifier struct { tg *ThreadGroup signal linux.Signal @@ -185,8 +179,6 @@ func (s *signalNotifier) Notify(exp uint64) { func (s *signalNotifier) Destroy() {} // TimerManager is a collection of supported process cpu timers. -// -// +stateify savable type TimerManager struct { // Clocks used to drive thread group execution time timers. virtClock *tgClock diff --git a/pkg/sentry/kernel/uts_namespace.go b/pkg/sentry/kernel/uts_namespace.go index 7e0fe0d21..58e9b4d1b 100644 --- a/pkg/sentry/kernel/uts_namespace.go +++ b/pkg/sentry/kernel/uts_namespace.go @@ -22,8 +22,6 @@ import ( // UTSNamespace represents a UTS namespace, a holder of two system identifiers: // the hostname and domain name. -// -// +stateify savable type UTSNamespace struct { // mu protects all fields below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/vdso.go b/pkg/sentry/kernel/vdso.go index 971e8bc59..0bacbea49 100644 --- a/pkg/sentry/kernel/vdso.go +++ b/pkg/sentry/kernel/vdso.go @@ -52,8 +52,6 @@ type vdsoParams struct { // Everything in the struct is 8 bytes for easy alignment. // // It must be kept in sync with params in vdso/vdso_time.cc. -// -// +stateify savable type VDSOParamPage struct { // The parameter page is fr, allocated from platform.Memory(). platform platform.Platform diff --git a/pkg/sentry/limits/BUILD b/pkg/sentry/limits/BUILD index 90f4395d4..3ce41cacc 100644 --- a/pkg/sentry/limits/BUILD +++ b/pkg/sentry/limits/BUILD @@ -1,12 +1,22 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "limits_state", + srcs = [ + "limits.go", + ], + out = "limits_state.go", + package = "limits", +) go_library( name = "limits", srcs = [ "context.go", "limits.go", + "limits_state.go", "linux.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/limits", @@ -14,6 +24,7 @@ go_library( deps = [ "//pkg/abi/linux", "//pkg/sentry/context", + "//pkg/state", ], ) diff --git a/pkg/sentry/limits/limits.go b/pkg/sentry/limits/limits.go index 02c8b60e3..4230ba958 100644 --- a/pkg/sentry/limits/limits.go +++ b/pkg/sentry/limits/limits.go @@ -47,8 +47,6 @@ const ( const Infinity = ^uint64(0) // Limit specifies a system limit. -// -// +stateify savable type Limit struct { // Cur specifies the current limit. Cur uint64 @@ -57,8 +55,6 @@ type Limit struct { } // LimitSet represents the Limits that correspond to each LimitType. -// -// +stateify savable type LimitSet struct { mu sync.Mutex `state:"nosave"` data map[LimitType]Limit diff --git a/pkg/sentry/loader/BUILD b/pkg/sentry/loader/BUILD index 0beb4561b..e63052c6d 100644 --- a/pkg/sentry/loader/BUILD +++ b/pkg/sentry/loader/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("@io_bazel_rules_go//go:def.bzl", "go_embed_data") -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") go_embed_data( name = "vdso_bin", @@ -10,12 +10,23 @@ go_embed_data( var = "vdsoBin", ) +go_stateify( + name = "loader_state", + srcs = [ + "vdso.go", + "vdso_state.go", + ], + out = "loader_state.go", + package = "loader", +) + go_library( name = "loader", srcs = [ "elf.go", "interpreter.go", "loader.go", + "loader_state.go", "vdso.go", "vdso_state.go", ":vdso_bin", @@ -29,6 +40,7 @@ go_library( "//pkg/cpuid", "//pkg/log", "//pkg/rand", + "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", @@ -43,6 +55,7 @@ go_library( "//pkg/sentry/uniqueid", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index a06e27ac9..2e8693f8e 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -193,8 +193,6 @@ func validateVDSO(ctx context.Context, f *fs.File, size uint64) (elfInfo, error) // // NOTE: to support multiple architectures or operating systems, this // would need to contain a VDSO for each. -// -// +stateify savable type VDSO struct { // ParamPage is the VDSO parameter page. This page should be updated to // inform the VDSO for timekeeping data. diff --git a/pkg/sentry/loader/vdso_state.go b/pkg/sentry/loader/vdso_state.go index dc71e1c2d..92004ad9e 100644 --- a/pkg/sentry/loader/vdso_state.go +++ b/pkg/sentry/loader/vdso_state.go @@ -18,7 +18,6 @@ import ( "debug/elf" ) -// +stateify savable type elfProgHeader struct { Type elf.ProgType Flags elf.ProgFlag diff --git a/pkg/sentry/memmap/BUILD b/pkg/sentry/memmap/BUILD index c9e0b95a0..2e367e189 100644 --- a/pkg/sentry/memmap/BUILD +++ b/pkg/sentry/memmap/BUILD @@ -1,7 +1,18 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "memmap_state", + srcs = [ + "mappable_range.go", + "mapping_set.go", + "mapping_set_impl.go", + ], + out = "memmap_state.go", + package = "memmap", +) go_template_instance( name = "mappable_range", @@ -35,6 +46,7 @@ go_library( "mapping_set.go", "mapping_set_impl.go", "memmap.go", + "memmap_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/memmap", visibility = ["//pkg/sentry:internal"], @@ -44,6 +56,7 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/platform", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/memmap/mapping_set.go b/pkg/sentry/memmap/mapping_set.go index c9483905d..0cd42ffbf 100644 --- a/pkg/sentry/memmap/mapping_set.go +++ b/pkg/sentry/memmap/mapping_set.go @@ -35,8 +35,6 @@ import ( type MappingsOfRange map[MappingOfRange]struct{} // MappingOfRange represents a mapping of a MappableRange. -// -// +stateify savable type MappingOfRange struct { MappingSpace MappingSpace AddrRange usermem.AddrRange diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index bbdfae247..3f396986a 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -1,7 +1,24 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "mm_state", + srcs = [ + "aio_context.go", + "aio_context_state.go", + "file_refcount_set.go", + "io_list.go", + "mm.go", + "pma_set.go", + "save_restore.go", + "special_mappable.go", + "vma_set.go", + ], + out = "mm_state.go", + package = "mm", +) go_template_instance( name = "file_refcount_set", @@ -84,6 +101,7 @@ go_library( "lifecycle.go", "metadata.go", "mm.go", + "mm_state.go", "pma.go", "pma_set.go", "proc_pid_maps.go", @@ -113,6 +131,7 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/sync", "//pkg/syserror", "//pkg/tcpip/buffer", diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index b42156d45..992bde5a5 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -28,8 +28,6 @@ import ( ) // aioManager creates and manages asynchronous I/O contexts. -// -// +stateify savable type aioManager struct { // mu protects below. mu sync.Mutex `state:"nosave"` @@ -91,16 +89,12 @@ func (a *aioManager) lookupAIOContext(id uint64) (*AIOContext, bool) { } // ioResult is a completed I/O operation. -// -// +stateify savable type ioResult struct { data interface{} ioEntry } // AIOContext is a single asynchronous I/O context. -// -// +stateify savable type AIOContext struct { // done is the notification channel used for all requests. done chan struct{} `state:"nosave"` @@ -196,8 +190,6 @@ func (ctx *AIOContext) WaitChannel() (chan struct{}, bool) { // aioMappable implements memmap.MappingIdentity and memmap.Mappable for AIO // ring buffers. -// -// +stateify savable type aioMappable struct { refs.AtomicRefCount diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index 3299ae164..ce8097b7f 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -46,8 +46,6 @@ import ( ) // MemoryManager implements a virtual address space. -// -// +stateify savable type MemoryManager struct { // p is the platform. // @@ -209,8 +207,6 @@ type MemoryManager struct { } // vma represents a virtual memory area. -// -// +stateify savable type vma struct { // mappable is the virtual memory object mapped by this vma. If mappable is // nil, the vma represents a private anonymous mapping. @@ -350,8 +346,6 @@ func (v *vma) loadRealPerms(b int) { } // pma represents a platform mapping area. -// -// +stateify savable type pma struct { // file is the file mapped by this pma. Only pmas for which file == // platform.Platform.Memory() may be saved. pmas hold a reference to the @@ -386,7 +380,6 @@ type pma struct { internalMappings safemem.BlockSeq `state:"nosave"` } -// +stateify savable type privateRefs struct { mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index aa2f87107..9d3614034 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -28,8 +28,6 @@ import ( // semantics similar to Linux's mm/mmap.c:_install_special_mapping(), except // that SpecialMappable takes ownership of the memory that it represents // (_install_special_mapping() does not.) -// -// +stateify savable type SpecialMappable struct { refs.AtomicRefCount diff --git a/pkg/sentry/platform/BUILD b/pkg/sentry/platform/BUILD index af9ba5394..15a7fbbc3 100644 --- a/pkg/sentry/platform/BUILD +++ b/pkg/sentry/platform/BUILD @@ -1,7 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "platform_state", + srcs = [ + "file_range.go", + ], + out = "platform_state.go", + package = "platform", +) go_template_instance( name = "file_range", @@ -21,6 +30,7 @@ go_library( "file_range.go", "mmap_min_addr.go", "platform.go", + "platform_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/platform", visibility = ["//pkg/sentry:internal"], @@ -34,6 +44,7 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/platform/filemem/BUILD b/pkg/sentry/platform/filemem/BUILD index 2a5982763..dadba1d38 100644 --- a/pkg/sentry/platform/filemem/BUILD +++ b/pkg/sentry/platform/filemem/BUILD @@ -1,7 +1,18 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "filemem_autogen_state", + srcs = [ + "filemem.go", + "filemem_state.go", + "usage_set.go", + ], + out = "filemem_autogen_state.go", + package = "filemem", +) go_template_instance( name = "usage_set", @@ -27,6 +38,7 @@ go_library( name = "filemem", srcs = [ "filemem.go", + "filemem_autogen_state.go", "filemem_state.go", "filemem_unsafe.go", "usage_set.go", diff --git a/pkg/sentry/platform/filemem/filemem.go b/pkg/sentry/platform/filemem/filemem.go index feb020ef8..870274ae1 100644 --- a/pkg/sentry/platform/filemem/filemem.go +++ b/pkg/sentry/platform/filemem/filemem.go @@ -155,8 +155,6 @@ type FileMem struct { } // usage tracks usage information. -// -// +stateify savable type usageInfo struct { // kind is the usage kind. kind usage.MemoryKind diff --git a/pkg/sentry/socket/BUILD b/pkg/sentry/socket/BUILD index a320fca0b..929787aa0 100644 --- a/pkg/sentry/socket/BUILD +++ b/pkg/sentry/socket/BUILD @@ -1,10 +1,22 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "socket_state", + srcs = [ + "socket.go", + ], + out = "socket_state_autogen.go", + package = "socket", +) go_library( name = "socket", - srcs = ["socket.go"], + srcs = [ + "socket.go", + "socket_state_autogen.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket", visibility = ["//pkg/sentry:internal"], deps = [ @@ -17,6 +29,7 @@ go_library( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserr", "//pkg/tcpip", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/socket/control/BUILD b/pkg/sentry/socket/control/BUILD index c4874fdfb..faf2b4c27 100644 --- a/pkg/sentry/socket/control/BUILD +++ b/pkg/sentry/socket/control/BUILD @@ -1,14 +1,26 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") -go_library( - name = "control", - srcs = ["control.go"], - importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control", +go_stateify( + name = "control_state", + srcs = [ + "control.go", + ], + out = "control_state.go", imports = [ "gvisor.googlesource.com/gvisor/pkg/sentry/fs", ], + package = "control", +) + +go_library( + name = "control", + srcs = [ + "control.go", + "control_state.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", @@ -19,6 +31,7 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/kdefs", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", ], diff --git a/pkg/sentry/socket/control/control.go b/pkg/sentry/socket/control/control.go index c31182e69..17ecdd11c 100644 --- a/pkg/sentry/socket/control/control.go +++ b/pkg/sentry/socket/control/control.go @@ -51,8 +51,6 @@ type SCMRights interface { // RightsFiles represents a SCM_RIGHTS socket control message. A reference is // maintained for each fs.File and is release either when an FD is created or // when the Release method is called. -// -// +stateify savable type RightsFiles []*fs.File // NewSCMRights creates a new SCM_RIGHTS socket control message representation @@ -130,8 +128,6 @@ func PackRights(t *kernel.Task, rights SCMRights, cloexec bool, buf []byte) []by } // scmCredentials represents an SCM_CREDENTIALS socket control message. -// -// +stateify savable type scmCredentials struct { t *kernel.Task kuid auth.KUID diff --git a/pkg/sentry/socket/epsocket/BUILD b/pkg/sentry/socket/epsocket/BUILD index 49af8db85..7ad5e88c5 100644 --- a/pkg/sentry/socket/epsocket/BUILD +++ b/pkg/sentry/socket/epsocket/BUILD @@ -1,12 +1,24 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "epsocket_state", + srcs = [ + "epsocket.go", + "save_restore.go", + "stack.go", + ], + out = "epsocket_state.go", + package = "epsocket", +) go_library( name = "epsocket", srcs = [ "device.go", "epsocket.go", + "epsocket_state.go", "provider.go", "save_restore.go", "stack.go", @@ -19,6 +31,7 @@ go_library( "//pkg/abi/linux", "//pkg/binary", "//pkg/log", + "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/device", @@ -31,6 +44,7 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index f969a1d7c..a2927e1b9 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -95,8 +95,6 @@ type commonEndpoint interface { // SocketOperations encapsulates all the state needed to represent a network stack // endpoint in the kernel context. -// -// +stateify savable type SocketOperations struct { socket.ReceiveTimeout fsutil.PipeSeek `state:"nosave"` diff --git a/pkg/sentry/socket/epsocket/stack.go b/pkg/sentry/socket/epsocket/stack.go index 12b4b4767..ec1d96ccb 100644 --- a/pkg/sentry/socket/epsocket/stack.go +++ b/pkg/sentry/socket/epsocket/stack.go @@ -26,8 +26,6 @@ import ( ) // Stack implements inet.Stack for netstack/tcpip/stack.Stack. -// -// +stateify savable type Stack struct { Stack *stack.Stack `state:"manual"` } diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD index d623718b3..227ca3926 100644 --- a/pkg/sentry/socket/hostinet/BUILD +++ b/pkg/sentry/socket/hostinet/BUILD @@ -1,12 +1,24 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "hostinet_state", + srcs = [ + "save_restore.go", + "socket.go", + "stack.go", + ], + out = "hostinet_autogen_state.go", + package = "hostinet", +) go_library( name = "hostinet", srcs = [ "device.go", "hostinet.go", + "hostinet_autogen_state.go", "save_restore.go", "socket.go", "socket_unsafe.go", @@ -30,6 +42,7 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/socket/netlink/BUILD b/pkg/sentry/socket/netlink/BUILD index b852165f7..b23a243f7 100644 --- a/pkg/sentry/socket/netlink/BUILD +++ b/pkg/sentry/socket/netlink/BUILD @@ -1,11 +1,21 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "netlink_state", + srcs = [ + "socket.go", + ], + out = "netlink_state.go", + package = "netlink", +) go_library( name = "netlink", srcs = [ "message.go", + "netlink_state.go", "provider.go", "socket.go", ], @@ -26,6 +36,7 @@ go_library( "//pkg/sentry/socket/netlink/port", "//pkg/sentry/socket/unix", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/netlink/port/BUILD b/pkg/sentry/socket/netlink/port/BUILD index 3a7dbc5ed..ba6f686e4 100644 --- a/pkg/sentry/socket/netlink/port/BUILD +++ b/pkg/sentry/socket/netlink/port/BUILD @@ -1,12 +1,23 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "port_state", + srcs = ["port.go"], + out = "port_state.go", + package = "port", +) go_library( name = "port", - srcs = ["port.go"], + srcs = [ + "port.go", + "port_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/port", visibility = ["//pkg/sentry:internal"], + deps = ["//pkg/state"], ) go_test( diff --git a/pkg/sentry/socket/netlink/port/port.go b/pkg/sentry/socket/netlink/port/port.go index 1c5d4c3a5..4ccf0b84c 100644 --- a/pkg/sentry/socket/netlink/port/port.go +++ b/pkg/sentry/socket/netlink/port/port.go @@ -32,8 +32,6 @@ import ( const maxPorts = 10000 // Manager allocates netlink port IDs. -// -// +stateify savable type Manager struct { // mu protects the fields below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/socket/netlink/route/BUILD b/pkg/sentry/socket/netlink/route/BUILD index e1bcfe252..726469fc9 100644 --- a/pkg/sentry/socket/netlink/route/BUILD +++ b/pkg/sentry/socket/netlink/route/BUILD @@ -1,19 +1,32 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "route_state", + srcs = ["protocol.go"], + out = "route_state.go", + package = "route", +) go_library( name = "route", - srcs = ["protocol.go"], + srcs = [ + "protocol.go", + "route_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/route", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", "//pkg/sentry/context", + "//pkg/sentry/fs", "//pkg/sentry/inet", "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/socket/netlink", + "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserr", ], ) diff --git a/pkg/sentry/socket/netlink/route/protocol.go b/pkg/sentry/socket/netlink/route/protocol.go index 55a76e916..e8030c518 100644 --- a/pkg/sentry/socket/netlink/route/protocol.go +++ b/pkg/sentry/socket/netlink/route/protocol.go @@ -43,8 +43,6 @@ func typeKind(typ uint16) commandKind { } // Protocol implements netlink.Protocol. -// -// +stateify savable type Protocol struct{} var _ netlink.Protocol = (*Protocol)(nil) diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index e15d1546c..0b8f528d0 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -51,8 +51,6 @@ var netlinkSocketDevice = device.NewAnonDevice() // to/from the kernel. // // Socket implements socket.Socket. -// -// +stateify savable type Socket struct { socket.ReceiveTimeout fsutil.PipeSeek `state:"nosave"` diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index 54fe64595..bd4858a34 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -195,8 +195,6 @@ func NewDirent(ctx context.Context, d *device.Device) *fs.Dirent { // // Care must be taken when copying ReceiveTimeout as it contains atomic // variables. -// -// +stateify savable type ReceiveTimeout struct { // ns is length of the timeout in nanoseconds. // diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index 9fe681e9a..7d04d6b6b 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -1,6 +1,15 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "unix_state", + srcs = [ + "unix.go", + ], + out = "unix_state.go", + package = "unix", +) go_library( name = "unix", @@ -8,6 +17,7 @@ go_library( "device.go", "io.go", "unix.go", + "unix_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix", visibility = ["//pkg/sentry:internal"], @@ -27,6 +37,7 @@ go_library( "//pkg/sentry/socket/control", "//pkg/sentry/socket/epsocket", "//pkg/sentry/usermem", + "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 5b6411f97..27bacbbc3 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -42,8 +42,6 @@ import ( // SocketOperations is a Unix socket. It is similar to an epsocket, except it is backed // by a unix.Endpoint instead of a tcpip.Endpoint. -// -// +stateify savable type SocketOperations struct { refs.AtomicRefCount socket.ReceiveTimeout diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index e4450a093..574621ad2 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -1,6 +1,18 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "linux_state", + srcs = [ + "sys_aio.go", + "sys_futex.go", + "sys_poll.go", + "sys_time.go", + ], + out = "linux_state.go", + package = "linux", +) go_library( name = "linux", @@ -8,6 +20,7 @@ go_library( "error.go", "flags.go", "linux64.go", + "linux_state.go", "sigset.go", "sys_aio.go", "sys_capability.go", @@ -53,6 +66,7 @@ go_library( "//pkg/abi/linux", "//pkg/binary", "//pkg/bpf", + "//pkg/eventchannel", "//pkg/log", "//pkg/metric", "//pkg/rand", @@ -60,6 +74,7 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", + "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/lock", "//pkg/sentry/fs/timerfd", "//pkg/sentry/kernel", @@ -70,6 +85,7 @@ go_library( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/pipe", "//pkg/sentry/kernel/sched", + "//pkg/sentry/kernel/semaphore", "//pkg/sentry/kernel/shm", "//pkg/sentry/kernel/time", "//pkg/sentry/limits", @@ -81,6 +97,8 @@ go_library( "//pkg/sentry/syscalls", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/state", + "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/syscalls/linux/sys_aio.go b/pkg/sentry/syscalls/linux/sys_aio.go index 54e4afa9e..fc3397081 100644 --- a/pkg/sentry/syscalls/linux/sys_aio.go +++ b/pkg/sentry/syscalls/linux/sys_aio.go @@ -69,8 +69,6 @@ type ioCallback struct { } // ioEvent describes an I/O result. -// -// +stateify savable type ioEvent struct { Data uint64 Obj uint64 diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index 1a0e1f5fb..57762d058 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -132,8 +132,6 @@ func (f futexChecker) Op(addr uintptr, opIn uint32) (bool, error) { // futexWaitRestartBlock encapsulates the state required to restart futex(2) // via restart_syscall(2). -// -// +stateify savable type futexWaitRestartBlock struct { duration time.Duration diff --git a/pkg/sentry/syscalls/linux/sys_poll.go b/pkg/sentry/syscalls/linux/sys_poll.go index b9bdefadb..d4dbfd285 100644 --- a/pkg/sentry/syscalls/linux/sys_poll.go +++ b/pkg/sentry/syscalls/linux/sys_poll.go @@ -274,8 +274,6 @@ func copyOutTimevalRemaining(t *kernel.Task, startNs ktime.Time, timeout time.Du // pollRestartBlock encapsulates the state required to restart poll(2) via // restart_syscall(2). -// -// +stateify savable type pollRestartBlock struct { pfdAddr usermem.Addr nfds uint diff --git a/pkg/sentry/syscalls/linux/sys_time.go b/pkg/sentry/syscalls/linux/sys_time.go index 8e6683444..dcee694b2 100644 --- a/pkg/sentry/syscalls/linux/sys_time.go +++ b/pkg/sentry/syscalls/linux/sys_time.go @@ -168,8 +168,6 @@ func Time(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallC // clockNanosleepRestartBlock encapsulates the state required to restart // clock_nanosleep(2) via restart_syscall(2). -// -// +stateify savable type clockNanosleepRestartBlock struct { c ktime.Clock duration time.Duration diff --git a/pkg/sentry/usage/BUILD b/pkg/sentry/usage/BUILD index 868dfd400..edee44d96 100644 --- a/pkg/sentry/usage/BUILD +++ b/pkg/sentry/usage/BUILD @@ -1,6 +1,17 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "usage_state", + srcs = [ + "cpu.go", + "io.go", + "memory.go", + ], + out = "usage_state.go", + package = "usage", +) go_library( name = "usage", @@ -10,6 +21,7 @@ go_library( "memory.go", "memory_unsafe.go", "usage.go", + "usage_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/usage", visibility = [ @@ -17,6 +29,9 @@ go_library( ], deps = [ "//pkg/bits", + "//pkg/log", "//pkg/sentry/memutil", + "//pkg/state", + "@org_golang_x_sys//unix:go_default_library", ], ) diff --git a/pkg/sentry/usage/cpu.go b/pkg/sentry/usage/cpu.go index ed7b04b9e..1c2cc90e1 100644 --- a/pkg/sentry/usage/cpu.go +++ b/pkg/sentry/usage/cpu.go @@ -20,8 +20,6 @@ import ( // CPUStats contains the subset of struct rusage fields that relate to CPU // scheduling. -// -// +stateify savable type CPUStats struct { // UserTime is the amount of time spent executing application code. UserTime time.Duration diff --git a/pkg/sentry/usage/io.go b/pkg/sentry/usage/io.go index 49faa507d..a05053c32 100644 --- a/pkg/sentry/usage/io.go +++ b/pkg/sentry/usage/io.go @@ -19,8 +19,6 @@ import ( ) // IO contains I/O-related statistics. -// -// +stateify savable type IO struct { // CharsRead is the number of bytes read by read syscalls. CharsRead uint64 diff --git a/pkg/sentry/usermem/BUILD b/pkg/sentry/usermem/BUILD index 69ba919e0..9dd1cd2b5 100644 --- a/pkg/sentry/usermem/BUILD +++ b/pkg/sentry/usermem/BUILD @@ -1,7 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "usermem_state", + srcs = [ + "access_type.go", + "addr.go", + "addr_range.go", + "addr_range_seq_unsafe.go", + ], + out = "usermem_state.go", + package = "usermem", +) go_template_instance( name = "addr_range", @@ -24,6 +36,7 @@ go_library( "bytes_io.go", "bytes_io_unsafe.go", "usermem.go", + "usermem_state.go", "usermem_x86.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/usermem", @@ -34,6 +47,7 @@ go_library( "//pkg/log", "//pkg/sentry/context", "//pkg/sentry/safemem", + "//pkg/state", "//pkg/syserror", "//pkg/tcpip/buffer", ], diff --git a/pkg/sentry/usermem/access_type.go b/pkg/sentry/usermem/access_type.go index 75346d854..7eabecf30 100644 --- a/pkg/sentry/usermem/access_type.go +++ b/pkg/sentry/usermem/access_type.go @@ -20,8 +20,6 @@ import ( // AccessType specifies memory access types. This is used for // setting mapping permissions, as well as communicating faults. -// -// +stateify savable type AccessType struct { // Read is read access. Read bool diff --git a/pkg/sentry/usermem/addr.go b/pkg/sentry/usermem/addr.go index fc94bee80..d175fdc74 100644 --- a/pkg/sentry/usermem/addr.go +++ b/pkg/sentry/usermem/addr.go @@ -19,8 +19,6 @@ import ( ) // Addr represents a generic virtual address. -// -// +stateify savable type Addr uintptr // AddLength adds the given length to start and returns the result. ok is true diff --git a/pkg/tcpip/BUILD b/pkg/tcpip/BUILD index 5153bd3b4..391d801d0 100644 --- a/pkg/tcpip/BUILD +++ b/pkg/tcpip/BUILD @@ -1,13 +1,26 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "tcpip_state", + srcs = [ + "tcpip.go", + ], + out = "tcpip_state.go", + package = "tcpip", +) go_library( name = "tcpip", - srcs = ["tcpip.go"], + srcs = [ + "tcpip.go", + "tcpip_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip", visibility = ["//visibility:public"], deps = [ + "//pkg/state", "//pkg/tcpip/buffer", "//pkg/waiter", ], diff --git a/pkg/tcpip/buffer/BUILD b/pkg/tcpip/buffer/BUILD index 11a725423..efeb6a448 100644 --- a/pkg/tcpip/buffer/BUILD +++ b/pkg/tcpip/buffer/BUILD @@ -1,15 +1,26 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "buffer_state", + srcs = [ + "view.go", + ], + out = "buffer_state.go", + package = "buffer", +) go_library( name = "buffer", srcs = [ + "buffer_state.go", "prependable.go", "view.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer", visibility = ["//visibility:public"], + deps = ["//pkg/state"], ) go_test( diff --git a/pkg/tcpip/buffer/view.go b/pkg/tcpip/buffer/view.go index bbb4e1d24..a5774a327 100644 --- a/pkg/tcpip/buffer/view.go +++ b/pkg/tcpip/buffer/view.go @@ -54,8 +54,6 @@ func (v *View) ToVectorisedView(views [1]View) VectorisedView { // VectorisedView is a vectorised version of View using non contigous memory. // It supports all the convenience methods supported by View. -// -// +stateify savable type VectorisedView struct { views []View size int diff --git a/pkg/tcpip/header/BUILD b/pkg/tcpip/header/BUILD index 8f22ba3a5..3aa2cfb24 100644 --- a/pkg/tcpip/header/BUILD +++ b/pkg/tcpip/header/BUILD @@ -1,6 +1,15 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "tcp_header_state", + srcs = [ + "tcp.go", + ], + out = "tcp_header_state.go", + package = "header", +) go_library( name = "header", @@ -16,11 +25,13 @@ go_library( "ipv6.go", "ipv6_fragment.go", "tcp.go", + "tcp_header_state.go", "udp.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/header", visibility = ["//visibility:public"], deps = [ + "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/seqnum", ], diff --git a/pkg/tcpip/header/tcp.go b/pkg/tcpip/header/tcp.go index 6689a6dc5..a95d282b0 100644 --- a/pkg/tcpip/header/tcp.go +++ b/pkg/tcpip/header/tcp.go @@ -120,8 +120,6 @@ type TCPSynOptions struct { } // SACKBlock represents a single contiguous SACK block. -// -// +stateify savable type SACKBlock struct { // Start indicates the lowest sequence number in the block. Start seqnum.Value @@ -133,8 +131,6 @@ type SACKBlock struct { // TCPOptions are used to parse and cache the TCP segment options for a non // syn/syn-ack segment. -// -// +stateify savable type TCPOptions struct { // TS is true if the TimeStamp option is enabled. TS bool diff --git a/pkg/tcpip/network/fragmentation/BUILD b/pkg/tcpip/network/fragmentation/BUILD index 83b4d253f..ac97ebe43 100644 --- a/pkg/tcpip/network/fragmentation/BUILD +++ b/pkg/tcpip/network/fragmentation/BUILD @@ -1,7 +1,14 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "fragmentation_state", + srcs = ["reassembler_list.go"], + out = "fragmentation_state.go", + package = "fragmentation", +) go_template_instance( name = "reassembler_list", @@ -19,6 +26,7 @@ go_library( srcs = [ "frag_heap.go", "fragmentation.go", + "fragmentation_state.go", "reassembler.go", "reassembler_list.go", ], @@ -26,6 +34,7 @@ go_library( visibility = ["//:sandbox"], deps = [ "//pkg/log", + "//pkg/state", "//pkg/tcpip/buffer", ], ) diff --git a/pkg/tcpip/seqnum/BUILD b/pkg/tcpip/seqnum/BUILD index c5c889239..a75869dac 100644 --- a/pkg/tcpip/seqnum/BUILD +++ b/pkg/tcpip/seqnum/BUILD @@ -1,12 +1,25 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "seqnum_state", + srcs = [ + "seqnum.go", + ], + out = "seqnum_state.go", + package = "seqnum", +) go_library( name = "seqnum", - srcs = ["seqnum.go"], + srcs = [ + "seqnum.go", + "seqnum_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/seqnum", visibility = [ "//visibility:public", ], + deps = ["//pkg/state"], ) diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index af0aec85c..eb1e4645d 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -213,8 +213,6 @@ const ( // FullAddress represents a full transport node address, as required by the // Connect() and Bind() methods. -// -// +stateify savable type FullAddress struct { // NIC is the ID of the NIC this address refers to. // @@ -258,8 +256,6 @@ func (s SlicePayload) Size() int { } // A ControlMessages contains socket control messages for IP sockets. -// -// +stateify savable type ControlMessages struct { // HasTimestamp indicates whether Timestamp is valid/set. HasTimestamp bool diff --git a/pkg/tcpip/transport/ping/BUILD b/pkg/tcpip/transport/ping/BUILD index 117532fea..28e3e1700 100644 --- a/pkg/tcpip/transport/ping/BUILD +++ b/pkg/tcpip/transport/ping/BUILD @@ -1,7 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "ping_state", + srcs = [ + "endpoint.go", + "endpoint_state.go", + "ping_packet_list.go", + ], + out = "ping_state.go", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], + package = "ping", +) go_template_instance( name = "ping_packet_list", @@ -20,13 +32,14 @@ go_library( "endpoint.go", "endpoint_state.go", "ping_packet_list.go", + "ping_state.go", "protocol.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/ping", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/sleep", + "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/ping/endpoint.go b/pkg/tcpip/transport/ping/endpoint.go index a22684de9..f15e44b61 100644 --- a/pkg/tcpip/transport/ping/endpoint.go +++ b/pkg/tcpip/transport/ping/endpoint.go @@ -26,7 +26,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// +stateify savable type pingPacket struct { pingPacketEntry senderAddress tcpip.FullAddress diff --git a/pkg/tcpip/transport/queue/BUILD b/pkg/tcpip/transport/queue/BUILD index 6dcec312e..fb878ad36 100644 --- a/pkg/tcpip/transport/queue/BUILD +++ b/pkg/tcpip/transport/queue/BUILD @@ -1,14 +1,27 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "queue_state", + srcs = [ + "queue.go", + ], + out = "queue_state.go", + package = "queue", +) go_library( name = "queue", - srcs = ["queue.go"], + srcs = [ + "queue.go", + "queue_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/queue", visibility = ["//:sandbox"], deps = [ "//pkg/ilist", + "//pkg/state", "//pkg/tcpip", "//pkg/waiter", ], diff --git a/pkg/tcpip/transport/queue/queue.go b/pkg/tcpip/transport/queue/queue.go index eb9ee8a3f..6a17441ae 100644 --- a/pkg/tcpip/transport/queue/queue.go +++ b/pkg/tcpip/transport/queue/queue.go @@ -33,8 +33,6 @@ type Entry interface { } // Queue is a buffer queue. -// -// +stateify savable type Queue struct { ReaderQueue *waiter.Queue WriterQueue *waiter.Queue diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index 9ebae6cc7..6a7153e4d 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -1,7 +1,27 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "tcp_state", + srcs = [ + "endpoint.go", + "endpoint_state.go", + "rcv.go", + "reno.go", + "segment.go", + "segment_heap.go", + "segment_queue.go", + "segment_state.go", + "snd.go", + "snd_state.go", + "tcp_segment_list.go", + ], + out = "tcp_state.go", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], + package = "tcp", +) go_template_instance( name = "tcp_segment_list", @@ -33,14 +53,15 @@ go_library( "snd.go", "snd_state.go", "tcp_segment_list.go", + "tcp_state.go", "timer.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/tcp", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/rand", "//pkg/sleep", + "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index de1883d84..5b8a1e20f 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -54,8 +54,6 @@ const ( ) // SACKInfo holds TCP SACK related information for a given endpoint. -// -// +stateify savable type SACKInfo struct { // Blocks is the maximum number of SACK blocks we track // per endpoint. @@ -71,8 +69,6 @@ type SACKInfo struct { // have concurrent goroutines make calls into the endpoint, they are properly // synchronized. The protocol implementation, however, runs in a single // goroutine. -// -// +stateify savable type endpoint struct { // workMu is used to arbitrate which goroutine may perform protocol // work. Only the main protocol goroutine is expected to call Lock() on diff --git a/pkg/tcpip/transport/tcp/rcv.go b/pkg/tcpip/transport/tcp/rcv.go index 92ef9c6f7..b22a00ce1 100644 --- a/pkg/tcpip/transport/tcp/rcv.go +++ b/pkg/tcpip/transport/tcp/rcv.go @@ -22,8 +22,6 @@ import ( // receiver holds the state necessary to receive TCP segments and turn them // into a stream of bytes. -// -// +stateify savable type receiver struct { ep *endpoint diff --git a/pkg/tcpip/transport/tcp/reno.go b/pkg/tcpip/transport/tcp/reno.go index 03ae8d747..60f170a27 100644 --- a/pkg/tcpip/transport/tcp/reno.go +++ b/pkg/tcpip/transport/tcp/reno.go @@ -16,8 +16,6 @@ package tcp // renoState stores the variables related to TCP New Reno congestion // control algorithm. -// -// +stateify savable type renoState struct { s *sender } diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index 8dccea2ba..40928ba2c 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -36,8 +36,6 @@ const ( // segment represents a TCP segment. It holds the payload and parsed TCP segment // information, and can be added to intrusive lists. // segment is mostly immutable, the only field allowed to change is viewToDeliver. -// -// +stateify savable type segment struct { segmentEntry refCnt int32 diff --git a/pkg/tcpip/transport/tcp/segment_queue.go b/pkg/tcpip/transport/tcp/segment_queue.go index 6a2d7bc0b..2ddcf5f10 100644 --- a/pkg/tcpip/transport/tcp/segment_queue.go +++ b/pkg/tcpip/transport/tcp/segment_queue.go @@ -21,8 +21,6 @@ import ( ) // segmentQueue is a bounded, thread-safe queue of TCP segments. -// -// +stateify savable type segmentQueue struct { mu sync.Mutex `state:"nosave"` list segmentList `state:"wait"` diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index 376e81846..e38686e1b 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -54,8 +54,6 @@ type congestionControl interface { } // sender holds the state necessary to send TCP segments. -// -// +stateify savable type sender struct { ep *endpoint @@ -135,8 +133,6 @@ type sender struct { } // fastRecovery holds information related to fast recovery from a packet loss. -// -// +stateify savable type fastRecovery struct { // active whether the endpoint is in fast recovery. The following fields // are only meaningful when active is true. diff --git a/pkg/tcpip/transport/tcp/snd_state.go b/pkg/tcpip/transport/tcp/snd_state.go index d536839af..33c8867f4 100644 --- a/pkg/tcpip/transport/tcp/snd_state.go +++ b/pkg/tcpip/transport/tcp/snd_state.go @@ -18,7 +18,6 @@ import ( "time" ) -// +stateify savable type unixTime struct { second int64 nano int64 diff --git a/pkg/tcpip/transport/udp/BUILD b/pkg/tcpip/transport/udp/BUILD index 1a3a62d3d..790dd55a3 100644 --- a/pkg/tcpip/transport/udp/BUILD +++ b/pkg/tcpip/transport/udp/BUILD @@ -1,7 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "udp_state", + srcs = [ + "endpoint.go", + "endpoint_state.go", + "udp_packet_list.go", + ], + out = "udp_state.go", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], + package = "udp", +) go_template_instance( name = "udp_packet_list", @@ -21,12 +33,13 @@ go_library( "endpoint_state.go", "protocol.go", "udp_packet_list.go", + "udp_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/udp", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/sleep", + "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 03fb76f92..2a32c3a87 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -25,7 +25,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// +stateify savable type udpPacket struct { udpPacketEntry senderAddress tcpip.FullAddress @@ -50,8 +49,6 @@ const ( // between users of the endpoint and the protocol implementation; it is legal to // have concurrent goroutines make calls into the endpoint, they are properly // synchronized. -// -// +stateify savable type endpoint struct { // The following fields are initialized at creation time and do not // change throughout the lifetime of the endpoint. diff --git a/pkg/tcpip/transport/unix/BUILD b/pkg/tcpip/transport/unix/BUILD index dae0bd079..676f2cf92 100644 --- a/pkg/tcpip/transport/unix/BUILD +++ b/pkg/tcpip/transport/unix/BUILD @@ -1,6 +1,17 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") + +go_stateify( + name = "unix_state", + srcs = [ + "connectioned.go", + "connectionless.go", + "unix.go", + ], + out = "unix_state.go", + package = "unix", +) go_library( name = "unix", @@ -9,11 +20,14 @@ go_library( "connectioned_state.go", "connectionless.go", "unix.go", + "unix_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix", visibility = ["//:sandbox"], deps = [ "//pkg/ilist", + "//pkg/log", + "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/transport/queue", diff --git a/pkg/tcpip/transport/unix/connectioned.go b/pkg/tcpip/transport/unix/connectioned.go index dd7c03cf1..0e63186b2 100644 --- a/pkg/tcpip/transport/unix/connectioned.go +++ b/pkg/tcpip/transport/unix/connectioned.go @@ -85,8 +85,6 @@ type ConnectingEndpoint interface { // path != "" && acceptedChan != nil => bound and listening. // // Only one of these will be true at any moment. -// -// +stateify savable type connectionedEndpoint struct { baseEndpoint diff --git a/pkg/tcpip/transport/unix/connectionless.go b/pkg/tcpip/transport/unix/connectionless.go index 2a6ec8b4b..3276ddcd0 100644 --- a/pkg/tcpip/transport/unix/connectionless.go +++ b/pkg/tcpip/transport/unix/connectionless.go @@ -25,8 +25,6 @@ import ( // // Specifically, this means datagram unix sockets not created with // socketpair(2). -// -// +stateify savable type connectionlessEndpoint struct { baseEndpoint } diff --git a/pkg/tcpip/transport/unix/unix.go b/pkg/tcpip/transport/unix/unix.go index 8e4af3139..190a1ccdb 100644 --- a/pkg/tcpip/transport/unix/unix.go +++ b/pkg/tcpip/transport/unix/unix.go @@ -60,8 +60,6 @@ type CredentialsControlMessage interface { } // A ControlMessages represents a collection of socket control messages. -// -// +stateify savable type ControlMessages struct { // Rights is a control message containing FDs. Rights RightsControlMessage @@ -237,8 +235,6 @@ type BoundEndpoint interface { } // message represents a message passed over a Unix domain socket. -// -// +stateify savable type message struct { ilist.Entry @@ -310,8 +306,6 @@ type Receiver interface { } // queueReceiver implements Receiver for datagram sockets. -// -// +stateify savable type queueReceiver struct { readQueue *queue.Queue } @@ -375,8 +369,6 @@ func (q *queueReceiver) RecvMaxQueueSize() int64 { func (*queueReceiver) Release() {} // streamQueueReceiver implements Receiver for stream sockets. -// -// +stateify savable type streamQueueReceiver struct { queueReceiver @@ -587,7 +579,6 @@ type ConnectedEndpoint interface { Release() } -// +stateify savable type connectedEndpoint struct { // endpoint represents the subset of the Endpoint functionality needed by // the connectedEndpoint. It is implemented by both connectionedEndpoint @@ -680,8 +671,6 @@ func (*connectedEndpoint) Release() {} // unix domain socket Endpoint implementations. // // Not to be used on its own. -// -// +stateify savable type baseEndpoint struct { *waiter.Queue diff --git a/pkg/waiter/BUILD b/pkg/waiter/BUILD index 5e611c54f..8256acdb4 100644 --- a/pkg/waiter/BUILD +++ b/pkg/waiter/BUILD @@ -1,13 +1,28 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") + +go_stateify( + name = "waiter_state", + srcs = [ + "waiter.go", + ], + out = "waiter_state.go", + package = "waiter", +) go_library( name = "waiter", - srcs = ["waiter.go"], + srcs = [ + "waiter.go", + "waiter_state.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/waiter", visibility = ["//visibility:public"], - deps = ["//pkg/ilist"], + deps = [ + "//pkg/ilist", + "//pkg/state", + ], ) go_test( diff --git a/pkg/waiter/waiter.go b/pkg/waiter/waiter.go index 9825880ca..9b189bb9e 100644 --- a/pkg/waiter/waiter.go +++ b/pkg/waiter/waiter.go @@ -157,8 +157,6 @@ func NewChannelEntry(c chan struct{}) (Entry, chan struct{}) { // notifiers can notify them when events happen. // // The zero value for waiter.Queue is an empty queue ready for use. -// -// +stateify savable type Queue struct { list ilist.List `state:"zerovalue"` mu sync.RWMutex `state:"nosave"` -- cgit v1.2.3 From cf44aff6e08b0e19935d5cd98455b4af98fd8794 Mon Sep 17 00:00:00 2001 From: Brian Geffon Date: Thu, 2 Aug 2018 08:09:03 -0700 Subject: Add seccomp(2) support. Add support for the seccomp syscall and the flag SECCOMP_FILTER_FLAG_TSYNC. PiperOrigin-RevId: 207101507 Change-Id: I5eb8ba9d5ef71b0e683930a6429182726dc23175 --- pkg/sentry/kernel/BUILD | 1 + pkg/sentry/kernel/seccomp.go | 62 ++++++++++++++++++++---- pkg/sentry/kernel/task.go | 19 ++++++-- pkg/sentry/kernel/task_clone.go | 5 +- pkg/sentry/kernel/task_syscall.go | 4 +- pkg/sentry/syscalls/linux/BUILD | 1 + pkg/sentry/syscalls/linux/linux64.go | 1 + pkg/sentry/syscalls/linux/sys_prctl.go | 32 +------------ pkg/sentry/syscalls/linux/sys_seccomp.go | 82 ++++++++++++++++++++++++++++++++ 9 files changed, 160 insertions(+), 47 deletions(-) create mode 100644 pkg/sentry/syscalls/linux/sys_seccomp.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index c4a7dacb2..1c1633068 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -47,6 +47,7 @@ go_stateify( ], out = "kernel_autogen_state.go", imports = [ + "gvisor.googlesource.com/gvisor/pkg/bpf", "gvisor.googlesource.com/gvisor/pkg/sentry/arch", "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", "gvisor.googlesource.com/gvisor/pkg/tcpip", diff --git a/pkg/sentry/kernel/seccomp.go b/pkg/sentry/kernel/seccomp.go index b7c4a507f..d77c05e2f 100644 --- a/pkg/sentry/kernel/seccomp.go +++ b/pkg/sentry/kernel/seccomp.go @@ -144,10 +144,15 @@ func (t *Task) evaluateSyscallFilters(sysno int32, args arch.SyscallArguments, i input := data.asBPFInput() ret := uint32(linux.SECCOMP_RET_ALLOW) + f := t.syscallFilters.Load() + if f == nil { + return ret + } + // "Every filter successfully installed will be evaluated (in reverse // order) for each system call the task makes." - kernel/seccomp.c - for i := len(t.syscallFilters) - 1; i >= 0; i-- { - thisRet, err := bpf.Exec(t.syscallFilters[i], input) + for i := len(f.([]bpf.Program)) - 1; i >= 0; i-- { + thisRet, err := bpf.Exec(f.([]bpf.Program)[i], input) if err != nil { t.Debugf("seccomp-bpf filter %d returned error: %v", i, err) thisRet = linux.SECCOMP_RET_KILL @@ -180,15 +185,53 @@ func (t *Task) AppendSyscallFilter(p bpf.Program) error { // maxSyscallFilterInstructions. (This restriction is inherited from // Linux.) totalLength := p.Length() - for _, f := range t.syscallFilters { - totalLength += f.Length() + 4 + var newFilters []bpf.Program + + // While syscallFilters are an atomic.Value we must take the mutex to + // prevent our read-copy-update from happening while another task + // is syncing syscall filters to us, this keeps the filters in a + // consistent state. + t.mu.Lock() + defer t.mu.Unlock() + if sf := t.syscallFilters.Load(); sf != nil { + oldFilters := sf.([]bpf.Program) + for _, f := range oldFilters { + totalLength += f.Length() + 4 + } + newFilters = append(newFilters, oldFilters...) } + if totalLength > maxSyscallFilterInstructions { return syserror.ENOMEM } - t.mu.Lock() - defer t.mu.Unlock() - t.syscallFilters = append(t.syscallFilters, p) + + newFilters = append(newFilters, p) + t.syscallFilters.Store(newFilters) + return nil +} + +// SyncSyscallFiltersToThreadGroup will copy this task's filters to all other +// threads in our thread group. +func (t *Task) SyncSyscallFiltersToThreadGroup() error { + f := t.syscallFilters.Load() + + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + + // Note: No new privs is always assumed to be set. + for ot := t.tg.tasks.Front(); ot != nil; ot = ot.Next() { + if ot.ThreadID() != t.ThreadID() { + // We must take the other task's mutex to prevent it from + // appending to its own syscall filters while we're syncing. + ot.mu.Lock() + var copiedFilters []bpf.Program + if f != nil { + copiedFilters = append(copiedFilters, f.([]bpf.Program)...) + } + ot.syscallFilters.Store(copiedFilters) + ot.mu.Unlock() + } + } return nil } @@ -196,9 +239,8 @@ func (t *Task) AppendSyscallFilter(p bpf.Program) error { // seccomp syscall filtering mode, appropriate for both prctl(PR_GET_SECCOMP) // and /proc/[pid]/status. func (t *Task) SeccompMode() int { - t.mu.Lock() - defer t.mu.Unlock() - if len(t.syscallFilters) > 0 { + f := t.syscallFilters.Load() + if f != nil && len(f.([]bpf.Program)) > 0 { return linux.SECCOMP_MODE_FILTER } return linux.SECCOMP_MODE_NONE diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 7f6735320..e705260da 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -355,11 +355,11 @@ type Task struct { parentDeathSignal linux.Signal // syscallFilters is all seccomp-bpf syscall filters applicable to the - // task, in the order in which they were installed. + // task, in the order in which they were installed. The type of the atomic + // is []bpf.Program. Writing needs to be protected by mu. // - // syscallFilters is protected by mu. syscallFilters is owned by the task - // goroutine. - syscallFilters []bpf.Program + // syscallFilters is owned by the task goroutine. + syscallFilters atomic.Value `state:".([]bpf.Program)"` // If cleartid is non-zero, treat it as a pointer to a ThreadID in the // task's virtual address space; when the task exits, set the pointed-to @@ -469,6 +469,17 @@ func (t *Task) loadLogPrefix(prefix string) { t.logPrefix.Store(prefix) } +func (t *Task) saveSyscallFilters() []bpf.Program { + if f := t.syscallFilters.Load(); f != nil { + return f.([]bpf.Program) + } + return nil +} + +func (t *Task) loadSyscallFilters(filters []bpf.Program) { + t.syscallFilters.Store(filters) +} + // afterLoad is invoked by stateify. func (t *Task) afterLoad() { t.interruptChan = make(chan struct{}, 1) diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index a61283267..3b77a4965 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -280,7 +280,10 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { // "If fork/clone and execve are allowed by @prog, any child processes will // be constrained to the same filters and system call ABI as the parent." - // Documentation/prctl/seccomp_filter.txt - nt.syscallFilters = append([]bpf.Program(nil), t.syscallFilters...) + if f := t.syscallFilters.Load(); f != nil { + copiedFilters := append([]bpf.Program(nil), f.([]bpf.Program)...) + nt.syscallFilters.Store(copiedFilters) + } if opts.Vfork { nt.vforkParent = t } diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index 79f4ff60c..92ca0acd9 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -194,7 +194,7 @@ func (t *Task) doSyscall() taskRunState { // Check seccomp filters. The nil check is for performance (as seccomp use // is rare), not needed for correctness. - if t.syscallFilters != nil { + if t.syscallFilters.Load() != nil { switch r := t.checkSeccompSyscall(int32(sysno), args, usermem.Addr(t.Arch().IP())); r { case seccompResultDeny: t.Debugf("Syscall %d: denied by seccomp", sysno) @@ -334,7 +334,7 @@ func (t *Task) doVsyscall(addr usermem.Addr, sysno uintptr) taskRunState { // to syscall ABI because they both use RDI, RSI, and RDX for the first three // arguments and none of the vsyscalls uses more than two arguments. args := t.Arch().SyscallArgs() - if t.syscallFilters != nil { + if t.syscallFilters.Load() != nil { switch r := t.checkSeccompSyscall(int32(sysno), args, addr); r { case seccompResultDeny: t.Debugf("vsyscall %d, caller %x: denied by seccomp", sysno, t.Arch().Value(caller)) diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index 574621ad2..32fca3811 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -42,6 +42,7 @@ go_library( "sys_rlimit.go", "sys_rusage.go", "sys_sched.go", + "sys_seccomp.go", "sys_sem.go", "sys_shm.go", "sys_signal.go", diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index edfcdca3f..c102af101 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -359,6 +359,7 @@ var AMD64 = &kernel.SyscallTable{ 312: syscalls.CapError(linux.CAP_SYS_PTRACE), // Kcmp, requires cap_sys_ptrace 313: syscalls.CapError(linux.CAP_SYS_MODULE), // FinitModule, requires cap_sys_module // "Backports." + 317: Seccomp, 318: GetRandom, }, diff --git a/pkg/sentry/syscalls/linux/sys_prctl.go b/pkg/sentry/syscalls/linux/sys_prctl.go index 2ca7471cf..911fef658 100644 --- a/pkg/sentry/syscalls/linux/sys_prctl.go +++ b/pkg/sentry/syscalls/linux/sys_prctl.go @@ -18,29 +18,13 @@ import ( "syscall" "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/bpf" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// userSockFprog is equivalent to Linux's struct sock_fprog on amd64. -type userSockFprog struct { - // Len is the length of the filter in BPF instructions. - Len uint16 - - _ [6]byte // padding for alignment - - // Filter is a user pointer to the struct sock_filter array that makes up - // the filter program. Filter is a uint64 rather than a usermem.Addr - // because usermem.Addr is actually uintptr, which is not a fixed-size - // type, and encoding/binary.Read objects to this. - Filter uint64 -} - // Prctl implements linux syscall prctl(2). // It has a list of subfunctions which operate on the process. The arguments are // all based on each subfunction. @@ -143,20 +127,8 @@ func Prctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall // Unsupported mode. return 0, nil, syscall.EINVAL } - var fprog userSockFprog - if _, err := t.CopyIn(args[2].Pointer(), &fprog); err != nil { - return 0, nil, err - } - filter := make([]linux.BPFInstruction, int(fprog.Len)) - if _, err := t.CopyIn(usermem.Addr(fprog.Filter), &filter); err != nil { - return 0, nil, err - } - compiledFilter, err := bpf.Compile(filter) - if err != nil { - t.Debugf("Invalid seccomp-bpf filter: %v", err) - return 0, nil, syscall.EINVAL - } - return 0, nil, t.AppendSyscallFilter(compiledFilter) + + return 0, nil, seccomp(t, linux.SECCOMP_SET_MODE_FILTER, 0, args[2].Pointer()) case linux.PR_GET_SECCOMP: return uintptr(t.SeccompMode()), nil, nil diff --git a/pkg/sentry/syscalls/linux/sys_seccomp.go b/pkg/sentry/syscalls/linux/sys_seccomp.go new file mode 100644 index 000000000..4323a4df4 --- /dev/null +++ b/pkg/sentry/syscalls/linux/sys_seccomp.go @@ -0,0 +1,82 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package linux + +import ( + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/bpf" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" +) + +// userSockFprog is equivalent to Linux's struct sock_fprog on amd64. +type userSockFprog struct { + // Len is the length of the filter in BPF instructions. + Len uint16 + + _ [6]byte // padding for alignment + + // Filter is a user pointer to the struct sock_filter array that makes up + // the filter program. Filter is a uint64 rather than a usermem.Addr + // because usermem.Addr is actually uintptr, which is not a fixed-size + // type, and encoding/binary.Read objects to this. + Filter uint64 +} + +// seccomp applies a seccomp policy to the current task. +func seccomp(t *kernel.Task, mode, flags uint64, addr usermem.Addr) error { + // We only support SECCOMP_SET_MODE_FILTER at the moment. + if mode != linux.SECCOMP_SET_MODE_FILTER { + // Unsupported mode. + return syscall.EINVAL + } + + tsync := flags&linux.SECCOMP_FILTER_FLAG_TSYNC != 0 + + // The only flag we support now is SECCOMP_FILTER_FLAG_TSYNC. + if flags&^linux.SECCOMP_FILTER_FLAG_TSYNC != 0 { + // Unsupported flag. + return syscall.EINVAL + } + + var fprog userSockFprog + if _, err := t.CopyIn(addr, &fprog); err != nil { + return err + } + filter := make([]linux.BPFInstruction, int(fprog.Len)) + if _, err := t.CopyIn(usermem.Addr(fprog.Filter), &filter); err != nil { + return err + } + compiledFilter, err := bpf.Compile(filter) + if err != nil { + t.Debugf("Invalid seccomp-bpf filter: %v", err) + return syscall.EINVAL + } + + err = t.AppendSyscallFilter(compiledFilter) + if err == nil && tsync { + // Now we must copy this seccomp program to all other threads. + err = t.SyncSyscallFiltersToThreadGroup() + } + return err +} + +// Seccomp implements linux syscall seccomp(2). +func Seccomp(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + return 0, nil, seccomp(t, args[0].Uint64(), args[1].Uint64(), args[2].Pointer()) +} -- cgit v1.2.3 From 57d0fcbdbf7e9d2d573ce8d4ca2f72b82f778d63 Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Thu, 2 Aug 2018 10:41:44 -0700 Subject: Automated rollback of changelist 207037226 PiperOrigin-RevId: 207125440 Change-Id: I6c572afb4d693ee72a0c458a988b0e96d191cd49 --- pkg/abi/BUILD | 13 +----- pkg/abi/linux/BUILD | 16 +------ pkg/abi/linux/bpf.go | 2 + pkg/abi/linux/tty.go | 2 + pkg/bpf/BUILD | 17 +------- pkg/bpf/interpreter.go | 2 + pkg/cpuid/BUILD | 15 +------ pkg/cpuid/cpuid.go | 2 + pkg/ilist/BUILD | 15 +------ pkg/ilist/list.go | 4 ++ pkg/segment/range.go | 2 + pkg/segment/set.go | 5 +++ pkg/sentry/arch/BUILD | 18 +------- pkg/sentry/arch/arch.go | 2 + pkg/sentry/arch/arch_amd64.go | 2 + pkg/sentry/arch/arch_state_x86.go | 1 + pkg/sentry/arch/arch_x86.go | 2 + pkg/sentry/arch/auxv.go | 2 + pkg/sentry/arch/signal_amd64.go | 6 +++ pkg/sentry/context/contexttest/BUILD | 17 +------- pkg/sentry/fs/BUILD | 36 +--------------- pkg/sentry/fs/ashmem/BUILD | 17 +------- pkg/sentry/fs/ashmem/area.go | 8 ++-- pkg/sentry/fs/ashmem/device.go | 22 +++++----- pkg/sentry/fs/ashmem/pin_board.go | 2 + pkg/sentry/fs/attr.go | 12 ++++++ pkg/sentry/fs/binder/BUILD | 13 +----- pkg/sentry/fs/binder/binder.go | 26 ++++++----- pkg/sentry/fs/dentry.go | 4 ++ pkg/sentry/fs/dev/BUILD | 20 +-------- pkg/sentry/fs/dev/dev.go | 2 + pkg/sentry/fs/dev/fs.go | 2 + pkg/sentry/fs/dev/full.go | 2 + pkg/sentry/fs/dev/null.go | 3 ++ pkg/sentry/fs/dev/random.go | 1 + pkg/sentry/fs/dirent.go | 2 + pkg/sentry/fs/dirent_cache.go | 2 + pkg/sentry/fs/fdpipe/BUILD | 31 +------------ pkg/sentry/fs/fdpipe/pipe.go | 2 + pkg/sentry/fs/file.go | 2 + pkg/sentry/fs/file_overlay.go | 4 ++ pkg/sentry/fs/filesystems.go | 2 + pkg/sentry/fs/filetest/BUILD | 18 +------- pkg/sentry/fs/flags.go | 2 + pkg/sentry/fs/fsutil/BUILD | 20 +-------- pkg/sentry/fs/fsutil/dirty_set.go | 2 + pkg/sentry/fs/fsutil/handle.go | 2 + pkg/sentry/fs/fsutil/host_file_mapper.go | 2 + pkg/sentry/fs/fsutil/inode.go | 6 +++ pkg/sentry/fs/fsutil/inode_cached.go | 2 + pkg/sentry/fs/gofer/BUILD | 23 +--------- pkg/sentry/fs/gofer/file.go | 2 + pkg/sentry/fs/gofer/fs.go | 2 + pkg/sentry/fs/gofer/inode.go | 4 ++ pkg/sentry/fs/gofer/session.go | 3 ++ pkg/sentry/fs/host/BUILD | 27 +----------- pkg/sentry/fs/host/descriptor.go | 2 + pkg/sentry/fs/host/file.go | 2 + pkg/sentry/fs/host/fs.go | 6 ++- pkg/sentry/fs/host/inode.go | 4 ++ pkg/sentry/fs/host/socket.go | 2 + pkg/sentry/fs/inode.go | 4 ++ pkg/sentry/fs/inode_inotify.go | 2 + pkg/sentry/fs/inotify.go | 2 + pkg/sentry/fs/inotify_event.go | 2 + pkg/sentry/fs/inotify_watch.go | 2 + pkg/sentry/fs/lock/BUILD | 15 +------ pkg/sentry/fs/lock/lock.go | 6 ++- pkg/sentry/fs/mount.go | 4 ++ pkg/sentry/fs/mount_overlay.go | 4 ++ pkg/sentry/fs/mounts.go | 2 + pkg/sentry/fs/overlay.go | 2 + pkg/sentry/fs/proc/BUILD | 34 +-------------- pkg/sentry/fs/proc/cpuinfo.go | 2 + pkg/sentry/fs/proc/exec_args.go | 2 + pkg/sentry/fs/proc/fds.go | 6 +++ pkg/sentry/fs/proc/file.go | 1 + pkg/sentry/fs/proc/filesystems.go | 2 + pkg/sentry/fs/proc/fs.go | 2 + pkg/sentry/fs/proc/loadavg.go | 2 + pkg/sentry/fs/proc/meminfo.go | 2 + pkg/sentry/fs/proc/mounts.go | 4 ++ pkg/sentry/fs/proc/proc.go | 4 ++ pkg/sentry/fs/proc/seqfile/BUILD | 30 ++----------- pkg/sentry/fs/proc/seqfile/seqfile.go | 4 ++ pkg/sentry/fs/proc/stat.go | 2 + pkg/sentry/fs/proc/sys.go | 5 +++ pkg/sentry/fs/proc/sys_net.go | 2 + pkg/sentry/fs/proc/task.go | 20 +++++++++ pkg/sentry/fs/proc/uid_gid_map.go | 3 ++ pkg/sentry/fs/proc/uptime.go | 2 + pkg/sentry/fs/proc/version.go | 2 + pkg/sentry/fs/ramfs/BUILD | 21 +-------- pkg/sentry/fs/ramfs/dir.go | 2 + pkg/sentry/fs/ramfs/ramfs.go | 2 + pkg/sentry/fs/ramfs/socket.go | 2 + pkg/sentry/fs/ramfs/symlink.go | 2 + pkg/sentry/fs/ramfs/test/BUILD | 18 +------- pkg/sentry/fs/sys/BUILD | 14 +----- pkg/sentry/fs/sys/fs.go | 2 + pkg/sentry/fs/sys/sys.go | 5 ++- pkg/sentry/fs/timerfd/BUILD | 18 +------- pkg/sentry/fs/timerfd/timerfd.go | 4 +- pkg/sentry/fs/tmpfs/BUILD | 17 +------- pkg/sentry/fs/tmpfs/file_regular.go | 2 + pkg/sentry/fs/tmpfs/fs.go | 2 + pkg/sentry/fs/tmpfs/inode_file.go | 2 + pkg/sentry/fs/tmpfs/tmpfs.go | 8 ++++ pkg/sentry/fs/tty/BUILD | 20 +-------- pkg/sentry/fs/tty/dir.go | 18 +++++--- pkg/sentry/fs/tty/fs.go | 4 ++ pkg/sentry/fs/tty/inode.go | 2 + pkg/sentry/fs/tty/line_discipline.go | 6 +++ pkg/sentry/fs/tty/master.go | 4 ++ pkg/sentry/fs/tty/queue.go | 4 +- pkg/sentry/fs/tty/slave.go | 4 ++ pkg/sentry/fs/tty/terminal.go | 2 + pkg/sentry/inet/BUILD | 15 +------ pkg/sentry/inet/inet.go | 2 + pkg/sentry/kernel/BUILD | 60 +++----------------------- pkg/sentry/kernel/abstract_socket_namespace.go | 3 ++ pkg/sentry/kernel/auth/BUILD | 17 +------- pkg/sentry/kernel/auth/credentials.go | 2 + pkg/sentry/kernel/auth/id_map.go | 2 + pkg/sentry/kernel/auth/user_namespace.go | 2 + pkg/sentry/kernel/epoll/BUILD | 15 +------ pkg/sentry/kernel/epoll/epoll.go | 8 +++- pkg/sentry/kernel/eventfd/BUILD | 18 +------- pkg/sentry/kernel/eventfd/eventfd.go | 4 +- pkg/sentry/kernel/fd_map.go | 6 +++ pkg/sentry/kernel/fs_context.go | 2 + pkg/sentry/kernel/futex/BUILD | 18 +------- pkg/sentry/kernel/futex/futex.go | 2 + pkg/sentry/kernel/ipc_namespace.go | 2 + pkg/sentry/kernel/kernel.go | 4 +- pkg/sentry/kernel/pending_signals.go | 5 +++ pkg/sentry/kernel/pipe/BUILD | 20 +-------- pkg/sentry/kernel/pipe/buffers.go | 2 + pkg/sentry/kernel/pipe/node.go | 2 + pkg/sentry/kernel/pipe/pipe.go | 2 + pkg/sentry/kernel/pipe/reader.go | 2 + pkg/sentry/kernel/pipe/reader_writer.go | 2 + pkg/sentry/kernel/pipe/writer.go | 2 + pkg/sentry/kernel/ptrace.go | 4 ++ pkg/sentry/kernel/rseq.go | 2 + pkg/sentry/kernel/semaphore/BUILD | 15 +------ pkg/sentry/kernel/semaphore/semaphore.go | 8 ++++ pkg/sentry/kernel/sessions.go | 4 ++ pkg/sentry/kernel/shm/BUILD | 13 +----- pkg/sentry/kernel/shm/shm.go | 4 ++ pkg/sentry/kernel/signal_handlers.go | 2 + pkg/sentry/kernel/syscalls.go | 2 + pkg/sentry/kernel/syslog.go | 2 + pkg/sentry/kernel/task.go | 2 + pkg/sentry/kernel/task_clone.go | 4 ++ pkg/sentry/kernel/task_context.go | 2 + pkg/sentry/kernel/task_exec.go | 4 ++ pkg/sentry/kernel/task_exit.go | 6 +++ pkg/sentry/kernel/task_resources.go | 2 + pkg/sentry/kernel/task_run.go | 2 + pkg/sentry/kernel/task_sched.go | 2 + pkg/sentry/kernel/task_signals.go | 5 +++ pkg/sentry/kernel/task_syscall.go | 4 ++ pkg/sentry/kernel/thread_group.go | 2 + pkg/sentry/kernel/threads.go | 8 ++++ pkg/sentry/kernel/time/BUILD | 14 +----- pkg/sentry/kernel/time/time.go | 6 +++ pkg/sentry/kernel/timekeeper.go | 2 + pkg/sentry/kernel/timer.go | 8 ++++ pkg/sentry/kernel/uts_namespace.go | 2 + pkg/sentry/kernel/vdso.go | 2 + pkg/sentry/limits/BUILD | 13 +----- pkg/sentry/limits/limits.go | 4 ++ pkg/sentry/loader/BUILD | 15 +------ pkg/sentry/loader/vdso.go | 2 + pkg/sentry/loader/vdso_state.go | 1 + pkg/sentry/memmap/BUILD | 15 +------ pkg/sentry/memmap/mapping_set.go | 2 + pkg/sentry/mm/BUILD | 21 +-------- pkg/sentry/mm/aio_context.go | 8 ++++ pkg/sentry/mm/mm.go | 7 +++ pkg/sentry/mm/special_mappable.go | 2 + pkg/sentry/platform/BUILD | 13 +----- pkg/sentry/platform/filemem/BUILD | 14 +----- pkg/sentry/platform/filemem/filemem.go | 2 + pkg/sentry/socket/BUILD | 17 +------- pkg/sentry/socket/control/BUILD | 23 +++------- pkg/sentry/socket/control/control.go | 4 ++ pkg/sentry/socket/epsocket/BUILD | 16 +------ pkg/sentry/socket/epsocket/epsocket.go | 2 + pkg/sentry/socket/epsocket/stack.go | 2 + pkg/sentry/socket/hostinet/BUILD | 15 +------ pkg/sentry/socket/netlink/BUILD | 13 +----- pkg/sentry/socket/netlink/port/BUILD | 15 +------ pkg/sentry/socket/netlink/port/port.go | 2 + pkg/sentry/socket/netlink/route/BUILD | 17 +------- pkg/sentry/socket/netlink/route/protocol.go | 2 + pkg/sentry/socket/netlink/socket.go | 2 + pkg/sentry/socket/socket.go | 2 + pkg/sentry/socket/unix/BUILD | 13 +----- pkg/sentry/socket/unix/unix.go | 2 + pkg/sentry/syscalls/linux/BUILD | 20 +-------- pkg/sentry/syscalls/linux/sys_aio.go | 2 + pkg/sentry/syscalls/linux/sys_futex.go | 2 + pkg/sentry/syscalls/linux/sys_poll.go | 2 + pkg/sentry/syscalls/linux/sys_time.go | 2 + pkg/sentry/usage/BUILD | 17 +------- pkg/sentry/usage/cpu.go | 2 + pkg/sentry/usage/io.go | 2 + pkg/sentry/usermem/BUILD | 16 +------ pkg/sentry/usermem/access_type.go | 2 + pkg/sentry/usermem/addr.go | 2 + pkg/tcpip/BUILD | 17 +------- pkg/tcpip/buffer/BUILD | 13 +----- pkg/tcpip/buffer/view.go | 2 + pkg/tcpip/header/BUILD | 13 +----- pkg/tcpip/header/tcp.go | 4 ++ pkg/tcpip/network/fragmentation/BUILD | 11 +---- pkg/tcpip/seqnum/BUILD | 17 +------- pkg/tcpip/tcpip.go | 4 ++ pkg/tcpip/transport/ping/BUILD | 17 +------- pkg/tcpip/transport/ping/endpoint.go | 1 + pkg/tcpip/transport/queue/BUILD | 17 +------- pkg/tcpip/transport/queue/queue.go | 2 + pkg/tcpip/transport/tcp/BUILD | 25 +---------- pkg/tcpip/transport/tcp/endpoint.go | 4 ++ pkg/tcpip/transport/tcp/rcv.go | 2 + pkg/tcpip/transport/tcp/reno.go | 2 + pkg/tcpip/transport/tcp/segment.go | 2 + pkg/tcpip/transport/tcp/segment_queue.go | 2 + pkg/tcpip/transport/tcp/snd.go | 4 ++ pkg/tcpip/transport/tcp/snd_state.go | 1 + pkg/tcpip/transport/udp/BUILD | 17 +------- pkg/tcpip/transport/udp/endpoint.go | 3 ++ pkg/tcpip/transport/unix/BUILD | 16 +------ pkg/tcpip/transport/unix/connectioned.go | 2 + pkg/tcpip/transport/unix/connectionless.go | 2 + pkg/tcpip/transport/unix/unix.go | 11 +++++ pkg/waiter/BUILD | 21 ++------- pkg/waiter/waiter.go | 2 + 240 files changed, 664 insertions(+), 1109 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/BUILD b/pkg/abi/BUILD index f1e6bac67..c014d2c4b 100644 --- a/pkg/abi/BUILD +++ b/pkg/abi/BUILD @@ -1,24 +1,13 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "abi_state", - srcs = [ - "abi.go", - ], - out = "abi_state.go", - package = "abi", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "abi", srcs = [ "abi.go", - "abi_state.go", "flag.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/abi", visibility = ["//:sandbox"], - deps = ["//pkg/state"], ) diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index 38b4829c9..ac4ceefbc 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -4,19 +4,7 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "linux_state", - srcs = [ - "binder.go", - "bpf.go", - "time.go", - "tty.go", - ], - out = "linux_state.go", - package = "linux", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "linux", @@ -41,7 +29,6 @@ go_library( "ipc.go", "limits.go", "linux.go", - "linux_state.go", "mm.go", "netdevice.go", "netlink.go", @@ -67,6 +54,5 @@ go_library( "//pkg/abi", "//pkg/binary", "//pkg/bits", - "//pkg/state", ], ) diff --git a/pkg/abi/linux/bpf.go b/pkg/abi/linux/bpf.go index f597ef4f5..80e5b1af1 100644 --- a/pkg/abi/linux/bpf.go +++ b/pkg/abi/linux/bpf.go @@ -15,6 +15,8 @@ package linux // BPFInstruction is a raw BPF virtual machine instruction. +// +// +stateify savable type BPFInstruction struct { // OpCode is the operation to execute. OpCode uint16 diff --git a/pkg/abi/linux/tty.go b/pkg/abi/linux/tty.go index 84b6ccc87..b640f7627 100644 --- a/pkg/abi/linux/tty.go +++ b/pkg/abi/linux/tty.go @@ -38,6 +38,8 @@ type Termios struct { // KernelTermios is struct ktermios/struct termios2, defined in // uapi/asm-generic/termbits.h. +// +// +stateify savable type KernelTermios struct { InputFlags uint32 OutputFlags uint32 diff --git a/pkg/bpf/BUILD b/pkg/bpf/BUILD index 403270049..564df3af5 100644 --- a/pkg/bpf/BUILD +++ b/pkg/bpf/BUILD @@ -1,21 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "bpf_state", - srcs = [ - "interpreter.go", - ], - out = "bpf_state.go", - package = "bpf", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "bpf", srcs = [ "bpf.go", - "bpf_state.go", "decoder.go", "input_bytes.go", "interpreter.go", @@ -23,10 +13,7 @@ go_library( ], importpath = "gvisor.googlesource.com/gvisor/pkg/bpf", visibility = ["//visibility:public"], - deps = [ - "//pkg/abi/linux", - "//pkg/state", - ], + deps = ["//pkg/abi/linux"], ) go_test( diff --git a/pkg/bpf/interpreter.go b/pkg/bpf/interpreter.go index b7dee86a8..111ada9d1 100644 --- a/pkg/bpf/interpreter.go +++ b/pkg/bpf/interpreter.go @@ -88,6 +88,8 @@ func (e Error) Error() string { } // Program is a BPF program that has been validated for consistency. +// +// +stateify savable type Program struct { instructions []linux.BPFInstruction } diff --git a/pkg/cpuid/BUILD b/pkg/cpuid/BUILD index 9a0ca1b33..46fc4703b 100644 --- a/pkg/cpuid/BUILD +++ b/pkg/cpuid/BUILD @@ -1,27 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "cpuid_state", - srcs = ["cpuid.go"], - out = "cpuid_state.go", - package = "cpuid", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "cpuid", srcs = [ "cpu_amd64.s", "cpuid.go", - "cpuid_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/cpuid", visibility = ["//:sandbox"], - deps = [ - "//pkg/log", - "//pkg/state", - ], + deps = ["//pkg/log"], ) go_test( diff --git a/pkg/cpuid/cpuid.go b/pkg/cpuid/cpuid.go index b486ab037..e91e34dc7 100644 --- a/pkg/cpuid/cpuid.go +++ b/pkg/cpuid/cpuid.go @@ -409,6 +409,8 @@ func (f Feature) flagString(cpuinfoOnly bool) string { } // FeatureSet is a set of Features for a cpu. +// +// +stateify savable type FeatureSet struct { // Set is the set of features that are enabled in this FeatureSet. Set map[Feature]bool diff --git a/pkg/ilist/BUILD b/pkg/ilist/BUILD index e32f26ffa..b26a39132 100644 --- a/pkg/ilist/BUILD +++ b/pkg/ilist/BUILD @@ -1,28 +1,15 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "list_state", - srcs = [ - "interface_list.go", - ], - out = "interface_list_state.go", - package = "ilist", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "ilist", srcs = [ "interface_list.go", - "interface_list_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/ilist", visibility = ["//visibility:public"], - deps = [ - "//pkg/state", - ], ) go_template_instance( diff --git a/pkg/ilist/list.go b/pkg/ilist/list.go index 5efb6c072..a88b82196 100644 --- a/pkg/ilist/list.go +++ b/pkg/ilist/list.go @@ -36,6 +36,8 @@ type Linker interface { // for e := l.Front(); e != nil; e = e.Next() { // // do something with e. // } +// +// +stateify savable type List struct { head Linker tail Linker @@ -155,6 +157,8 @@ func (l *List) Remove(e Linker) { // Entry is a default implementation of Linker. Users can add anonymous fields // of this type to their structs to make them automatically implement the // methods needed by List. +// +// +stateify savable type Entry struct { next Linker prev Linker diff --git a/pkg/segment/range.go b/pkg/segment/range.go index 5ff30d489..34c067265 100644 --- a/pkg/segment/range.go +++ b/pkg/segment/range.go @@ -18,6 +18,8 @@ package segment type T uint64 // A Range represents a contiguous range of T. +// +// +stateify savable type Range struct { // Start is the inclusive start of the range. Start T diff --git a/pkg/segment/set.go b/pkg/segment/set.go index 6eed1d930..cffec2a2c 100644 --- a/pkg/segment/set.go +++ b/pkg/segment/set.go @@ -88,6 +88,8 @@ const ( // A Set is a mapping of segments with non-overlapping Range keys. The zero // value for a Set is an empty set. Set values are not safely movable nor // copyable. Set is thread-compatible. +// +// +stateify savable type Set struct { root node `state:".(*SegmentDataSlices)"` } @@ -596,6 +598,7 @@ func (s *Set) ApplyContiguous(r Range, fn func(seg Iterator)) GapIterator { } } +// +stateify savable type node struct { // An internal binary tree node looks like: // @@ -1317,6 +1320,8 @@ func (n *node) writeDebugString(buf *bytes.Buffer, prefix string) { // SegmentDataSlices represents segments from a set as slices of start, end, and // values. SegmentDataSlices is primarily used as an intermediate representation // for save/restore and the layout here is optimized for that. +// +// +stateify savable type SegmentDataSlices struct { Start []Key End []Key diff --git a/pkg/sentry/arch/BUILD b/pkg/sentry/arch/BUILD index 0a2a35400..314b3e962 100644 --- a/pkg/sentry/arch/BUILD +++ b/pkg/sentry/arch/BUILD @@ -1,21 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "arch_state", - srcs = [ - "arch.go", - "arch_amd64.go", - "arch_state_x86.go", - "arch_x86.go", - "auxv.go", - "signal_amd64.go", - ], - out = "arch_state.go", - package = "arch", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "arch", @@ -24,7 +10,6 @@ go_library( "arch.go", "arch_amd64.go", "arch_amd64.s", - "arch_state.go", "arch_state_x86.go", "arch_x86.go", "auxv.go", @@ -46,7 +31,6 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/limits", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go index 0189e958d..21cb84502 100644 --- a/pkg/sentry/arch/arch.go +++ b/pkg/sentry/arch/arch.go @@ -254,6 +254,8 @@ const ( // MemoryManager. // // Note that "highest address" below is always exclusive. +// +// +stateify savable type MmapLayout struct { // MinAddr is the lowest mappable address. MinAddr usermem.Addr diff --git a/pkg/sentry/arch/arch_amd64.go b/pkg/sentry/arch/arch_amd64.go index 23526fe8e..f1e408af9 100644 --- a/pkg/sentry/arch/arch_amd64.go +++ b/pkg/sentry/arch/arch_amd64.go @@ -95,6 +95,8 @@ const ( ) // context64 represents an AMD64 context. +// +// +stateify savable type context64 struct { State sigFPState []x86FPState // fpstate to be restored on sigreturn. diff --git a/pkg/sentry/arch/arch_state_x86.go b/pkg/sentry/arch/arch_state_x86.go index cb38d098a..e9c23a06b 100644 --- a/pkg/sentry/arch/arch_state_x86.go +++ b/pkg/sentry/arch/arch_state_x86.go @@ -56,6 +56,7 @@ func (s *State) afterLoad() { copy(s.x86FPState, old) } +// +stateify savable type syscallPtraceRegs struct { R15 uint64 R14 uint64 diff --git a/pkg/sentry/arch/arch_x86.go b/pkg/sentry/arch/arch_x86.go index 5cc4f8377..b35eec53c 100644 --- a/pkg/sentry/arch/arch_x86.go +++ b/pkg/sentry/arch/arch_x86.go @@ -153,6 +153,8 @@ func NewFloatingPointData() *FloatingPointData { // State contains the common architecture bits for X86 (the build tag of this // file ensures it's only built on x86). +// +// +stateify savable type State struct { // The system registers. Regs syscall.PtraceRegs `state:".(syscallPtraceRegs)"` diff --git a/pkg/sentry/arch/auxv.go b/pkg/sentry/arch/auxv.go index 70e0e35b7..81cfb4a01 100644 --- a/pkg/sentry/arch/auxv.go +++ b/pkg/sentry/arch/auxv.go @@ -19,6 +19,8 @@ import ( ) // An AuxEntry represents an entry in an ELF auxiliary vector. +// +// +stateify savable type AuxEntry struct { Key uint64 Value usermem.Addr diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index c1d743f38..e81717e8b 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -28,6 +28,8 @@ import ( // SignalAct represents the action that should be taken when a signal is // delivered, and is equivalent to struct sigaction on 64-bit x86. +// +// +stateify savable type SignalAct struct { Handler uint64 Flags uint64 @@ -47,6 +49,8 @@ func (s *SignalAct) DeserializeTo(other *SignalAct) { // SignalStack represents information about a user stack, and is equivalent to // stack_t on 64-bit x86. +// +// +stateify savable type SignalStack struct { Addr uint64 Flags uint32 @@ -66,6 +70,8 @@ func (s *SignalStack) DeserializeTo(other *SignalStack) { // SignalInfo represents information about a signal being delivered, and is // equivalent to struct siginfo on 64-bit x86. +// +// +stateify savable type SignalInfo struct { Signo int32 // Signal number Errno int32 // Errno value diff --git a/pkg/sentry/context/contexttest/BUILD b/pkg/sentry/context/contexttest/BUILD index 591b11a4d..01bb40b04 100644 --- a/pkg/sentry/context/contexttest/BUILD +++ b/pkg/sentry/context/contexttest/BUILD @@ -1,23 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "contexttest_state", - srcs = [ - "contexttest.go", - ], - out = "contexttest_state.go", - package = "contexttest", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "contexttest", testonly = 1, - srcs = [ - "contexttest.go", - "contexttest_state.go", - ], + srcs = ["contexttest.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest", visibility = ["//pkg/sentry:internal"], deps = [ @@ -28,6 +16,5 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/platform/ptrace", "//pkg/sentry/uniqueid", - "//pkg/state", ], ) diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index e3c9a9b70..18cd5ae8e 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -1,40 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "fs_state", - srcs = [ - "attr.go", - "dentry.go", - "dirent.go", - "dirent_cache.go", - "dirent_list.go", - "dirent_state.go", - "file.go", - "file_overlay.go", - "file_state.go", - "filesystems.go", - "flags.go", - "inode.go", - "inode_inotify.go", - "inode_operations.go", - "inode_overlay.go", - "inotify.go", - "inotify_event.go", - "inotify_watch.go", - "mock.go", - "mount.go", - "mount_overlay.go", - "mount_state.go", - "mounts.go", - "overlay.go", - "path.go", - ], - out = "fs_state.go", - package = "fs", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "fs", @@ -54,7 +21,6 @@ go_library( "filesystems.go", "flags.go", "fs.go", - "fs_state.go", "inode.go", "inode_inotify.go", "inode_operations.go", diff --git a/pkg/sentry/fs/ashmem/BUILD b/pkg/sentry/fs/ashmem/BUILD index 9f166799a..dc893d22f 100644 --- a/pkg/sentry/fs/ashmem/BUILD +++ b/pkg/sentry/fs/ashmem/BUILD @@ -1,26 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") -go_stateify( - name = "ashmem_state", - srcs = [ - "area.go", - "device.go", - "pin_board.go", - "uint64_range.go", - "uint64_set.go", - ], - out = "ashmem_state.go", - package = "ashmem", -) - go_library( name = "ashmem", srcs = [ "area.go", - "ashmem_state.go", "device.go", "pin_board.go", "uint64_range.go", @@ -41,7 +27,6 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", ], diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index e4f76f0d0..bfd7f2762 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -39,10 +39,12 @@ const ( ) // Area implements fs.FileOperations. +// +// +stateify savable type Area struct { - fsutil.NoFsync - fsutil.DeprecatedFileOperations - fsutil.NotDirReaddir + fsutil.NoFsync `state:"nosave"` + fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.NotDirReaddir `state:"nosave"` ad *Device diff --git a/pkg/sentry/fs/ashmem/device.go b/pkg/sentry/fs/ashmem/device.go index c5b51d4a7..d0986fa11 100644 --- a/pkg/sentry/fs/ashmem/device.go +++ b/pkg/sentry/fs/ashmem/device.go @@ -27,17 +27,19 @@ import ( ) // Device implements fs.InodeOperations. +// +// +stateify savable type Device struct { - fsutil.DeprecatedFileOperations - fsutil.InodeNoExtendedAttributes - fsutil.InodeNotDirectory - fsutil.InodeNotRenameable - fsutil.InodeNotSocket - fsutil.InodeNotSymlink - fsutil.NoFsync - fsutil.NoMappable - fsutil.NoopWriteOut - fsutil.NotDirReaddir + fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotRenameable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.NoFsync `state:"nosave"` + fsutil.NoMappable `state:"nosave"` + fsutil.NoopWriteOut `state:"nosave"` + fsutil.NotDirReaddir `state:"nosave"` mu sync.Mutex `state:"nosave"` unstable fs.UnstableAttr diff --git a/pkg/sentry/fs/ashmem/pin_board.go b/pkg/sentry/fs/ashmem/pin_board.go index c7fb3822c..ecba395a0 100644 --- a/pkg/sentry/fs/ashmem/pin_board.go +++ b/pkg/sentry/fs/ashmem/pin_board.go @@ -56,6 +56,8 @@ func (setFunctions) Split(Range, noValue, uint64) (noValue, noValue) { // segment.Set is used for implementation where segments represent // ranges of pinned bytes, while gaps represent ranges of unpinned // bytes. All ranges are page-aligned. +// +// +stateify savable type PinBoard struct { Set } diff --git a/pkg/sentry/fs/attr.go b/pkg/sentry/fs/attr.go index 56a2ad6f7..4178f18b2 100644 --- a/pkg/sentry/fs/attr.go +++ b/pkg/sentry/fs/attr.go @@ -91,6 +91,8 @@ func (n InodeType) String() string { // StableAttr contains Inode attributes that will be stable throughout the // lifetime of the Inode. +// +// +stateify savable type StableAttr struct { // Type is the InodeType of a InodeOperations. Type InodeType @@ -150,6 +152,8 @@ func IsCharDevice(s StableAttr) bool { // UnstableAttr contains Inode attributes that may change over the lifetime // of the Inode. +// +// +stateify savable type UnstableAttr struct { // Size is the file size in bytes. Size int64 @@ -186,6 +190,8 @@ func WithCurrentTime(ctx context.Context, u UnstableAttr) UnstableAttr { } // AttrMask contains fields to mask StableAttr and UnstableAttr. +// +// +stateify savable type AttrMask struct { Type bool DeviceID bool @@ -227,6 +233,8 @@ func (a AttrMask) Union(b AttrMask) AttrMask { } // PermMask are file access permissions. +// +// +stateify savable type PermMask struct { // Read indicates reading is permitted. Read bool @@ -280,6 +288,8 @@ func (p PermMask) SupersetOf(other PermMask) bool { // FilePermissions represents the permissions of a file, with // Read/Write/Execute bits for user, group, and other. +// +// +stateify savable type FilePermissions struct { User PermMask Group PermMask @@ -370,6 +380,8 @@ func (f FilePermissions) AnyRead() bool { } // FileOwner represents ownership of a file. +// +// +stateify savable type FileOwner struct { UID auth.KUID GID auth.KGID diff --git a/pkg/sentry/fs/binder/BUILD b/pkg/sentry/fs/binder/BUILD index ec3928baf..a077b91d2 100644 --- a/pkg/sentry/fs/binder/BUILD +++ b/pkg/sentry/fs/binder/BUILD @@ -1,25 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "binder_state", - srcs = ["binder.go"], - out = "binder_state.go", - package = "binder", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "binder", srcs = [ "binder.go", - "binder_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/binder", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/log", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", @@ -30,8 +21,6 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", - "//pkg/tcpip/transport/unix", ], ) diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index 3f87b6b08..502a262dd 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -40,15 +40,17 @@ const ( ) // Device implements fs.InodeOperations. +// +// +stateify savable type Device struct { - fsutil.InodeNoExtendedAttributes - fsutil.InodeNotDirectory - fsutil.InodeNotRenameable - fsutil.InodeNotSocket - fsutil.InodeNotSymlink - fsutil.NoMappable - fsutil.NoopWriteOut - fsutil.DeprecatedFileOperations + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotRenameable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.NoMappable `state:"nosave"` + fsutil.NoopWriteOut `state:"nosave"` + fsutil.DeprecatedFileOperations `state:"nosave"` // mu protects unstable. mu sync.Mutex `state:"nosave"` @@ -186,10 +188,12 @@ func (bd *Device) StatFS(context.Context) (fs.Info, error) { } // Proc implements fs.FileOperations and fs.IoctlGetter. +// +// +stateify savable type Proc struct { - fsutil.NoFsync - fsutil.DeprecatedFileOperations - fsutil.NotDirReaddir + fsutil.NoFsync `state:"nosave"` + fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.NotDirReaddir `state:"nosave"` bd *Device task *kernel.Task diff --git a/pkg/sentry/fs/dentry.go b/pkg/sentry/fs/dentry.go index d42e8da81..b347468ff 100644 --- a/pkg/sentry/fs/dentry.go +++ b/pkg/sentry/fs/dentry.go @@ -21,6 +21,8 @@ import ( ) // DentAttr is the metadata of a directory entry. It is a subset of StableAttr. +// +// +stateify savable type DentAttr struct { // Type is the InodeType of an Inode. Type InodeType @@ -154,6 +156,8 @@ func GenericReaddir(ctx *DirCtx, s *SortedDentryMap) (int, error) { } // SortedDentryMap is a sorted map of names and fs.DentAttr entries. +// +// +stateify savable type SortedDentryMap struct { // names is always kept in sorted-order. names []string diff --git a/pkg/sentry/fs/dev/BUILD b/pkg/sentry/fs/dev/BUILD index ea41615fd..fc069bb5f 100644 --- a/pkg/sentry/fs/dev/BUILD +++ b/pkg/sentry/fs/dev/BUILD @@ -1,25 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "dev_state", - srcs = [ - "dev.go", - "fs.go", - "full.go", - "null.go", - "random.go", - ], - out = "dev_state.go", - package = "dev", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "dev", srcs = [ "dev.go", - "dev_state.go", "device.go", "fs.go", "full.go", @@ -30,8 +16,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/amutex", - "//pkg/log", "//pkg/rand", "//pkg/sentry/context", "//pkg/sentry/device", @@ -45,9 +29,7 @@ go_library( "//pkg/sentry/mm", "//pkg/sentry/platform", "//pkg/sentry/safemem", - "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/fs/dev/dev.go b/pkg/sentry/fs/dev/dev.go index 36c61bfc2..3f4f2a40a 100644 --- a/pkg/sentry/fs/dev/dev.go +++ b/pkg/sentry/fs/dev/dev.go @@ -27,6 +27,8 @@ import ( ) // Dev is the root node. +// +// +stateify savable type Dev struct { ramfs.Dir } diff --git a/pkg/sentry/fs/dev/fs.go b/pkg/sentry/fs/dev/fs.go index 3c79f3782..2ae49be4e 100644 --- a/pkg/sentry/fs/dev/fs.go +++ b/pkg/sentry/fs/dev/fs.go @@ -29,6 +29,8 @@ const binderEnabledKey = "binder_enabled" const ashmemEnabledKey = "ashmem_enabled" // filesystem is a devtmpfs. +// +// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index e13eb6c03..492b8eb3a 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -26,6 +26,8 @@ import ( ) // fullDevice is used to implement /dev/full. +// +// +stateify savable type fullDevice struct { ramfs.Entry } diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 66b8ba967..2977c8670 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -29,6 +29,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) +// +stateify savable type nullDevice struct { ramfs.Entry } @@ -54,6 +55,7 @@ func (n *nullDevice) Truncate(context.Context, *fs.Inode, int64) error { return nil } +// +stateify savable type zeroDevice struct { nullDevice } @@ -80,6 +82,7 @@ func (zd *zeroDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.F }), nil } +// +stateify savable type zeroFileOperations struct { fs.FileOperations } diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index 33a045a05..47b76218f 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -24,6 +24,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) +// +stateify savable type randomDevice struct { ramfs.Entry } diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go index f9bf2fba6..4658d044f 100644 --- a/pkg/sentry/fs/dirent.go +++ b/pkg/sentry/fs/dirent.go @@ -81,6 +81,8 @@ var renameMu sync.RWMutex // // Dirents currently do not attempt to free entries that lack application references under // memory pressure. +// +// +stateify savable type Dirent struct { // AtomicRefCount is our reference count. refs.AtomicRefCount diff --git a/pkg/sentry/fs/dirent_cache.go b/pkg/sentry/fs/dirent_cache.go index e786e4f65..c680e4828 100644 --- a/pkg/sentry/fs/dirent_cache.go +++ b/pkg/sentry/fs/dirent_cache.go @@ -25,6 +25,8 @@ import ( // // A nil DirentCache corresponds to a cache with size 0. All methods can be // called, but nothing is actually cached. +// +// +stateify savable type DirentCache struct { // Maximum size of the cache. This must be saved manually, to handle the case // when cache is nil. diff --git a/pkg/sentry/fs/fdpipe/BUILD b/pkg/sentry/fs/fdpipe/BUILD index 4fcb06f1f..ffe4204bc 100644 --- a/pkg/sentry/fs/fdpipe/BUILD +++ b/pkg/sentry/fs/fdpipe/BUILD @@ -1,54 +1,27 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "pipe_state", - srcs = [ - "pipe.go", - "pipe_state.go", - ], - out = "pipe_autogen_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/sentry/fs"], - package = "fdpipe", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "fdpipe", srcs = [ "pipe.go", - "pipe_autogen_state.go", "pipe_opener.go", "pipe_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fdpipe", + imports = ["gvisor.googlesource.com/gvisor/pkg/sentry/fs"], visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/abi/linux", - "//pkg/amutex", "//pkg/fd", "//pkg/log", - "//pkg/metric", - "//pkg/p9", - "//pkg/refs", "//pkg/secio", "//pkg/sentry/context", - "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", - "//pkg/sentry/fs/lock", - "//pkg/sentry/kernel/auth", - "//pkg/sentry/kernel/time", - "//pkg/sentry/memmap", - "//pkg/sentry/platform", "//pkg/sentry/safemem", - "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", - "//pkg/tcpip", - "//pkg/tcpip/transport/unix", - "//pkg/unet", "//pkg/waiter", "//pkg/waiter/fdnotifier", ], diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index 7b318e35f..2e34604e6 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -34,6 +34,8 @@ import ( ) // pipeOperations are the fs.FileOperations of a host pipe. +// +// +stateify savable type pipeOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 6d93ef760..8e535a618 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -47,6 +47,8 @@ const FileMaxOffset = math.MaxInt64 // and write(2). // // FIXME: Split synchronization from cancellation. +// +// +stateify savable type File struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index 36b2cf75e..113962368 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -60,6 +60,8 @@ func overlayFile(ctx context.Context, inode *Inode, flags FileFlags) (*File, err } // overlayFileOperations implements FileOperations for a file in an overlay. +// +// +stateify savable type overlayFileOperations struct { // upperMu protects upper below. In contrast lower is stable. upperMu sync.Mutex `state:"nosave"` @@ -375,6 +377,8 @@ func readdirOne(ctx context.Context, d *Dirent) (map[string]DentAttr, error) { // overlayMappingIdentity wraps a MappingIdentity, and also holds a reference // on a file during its lifetime. +// +// +stateify savable type overlayMappingIdentity struct { refs.AtomicRefCount id memmap.MappingIdentity diff --git a/pkg/sentry/fs/filesystems.go b/pkg/sentry/fs/filesystems.go index 200e792f4..5a1e7a270 100644 --- a/pkg/sentry/fs/filesystems.go +++ b/pkg/sentry/fs/filesystems.go @@ -125,6 +125,8 @@ func GetFilesystems() []Filesystem { } // MountSourceFlags represents all mount option flags as a struct. +// +// +stateify savable type MountSourceFlags struct { // ReadOnly corresponds to mount(2)'s "MS_RDONLY" and indicates that // the filesystem should be mounted read-only. diff --git a/pkg/sentry/fs/filetest/BUILD b/pkg/sentry/fs/filetest/BUILD index f481c57fb..d137fee4c 100644 --- a/pkg/sentry/fs/filetest/BUILD +++ b/pkg/sentry/fs/filetest/BUILD @@ -1,34 +1,20 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "filetest_state", - srcs = [ - "filetest.go", - ], - out = "filetest_state.go", - package = "filetest", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "filetest", testonly = 1, - srcs = [ - "filetest.go", - "filetest_state.go", - ], + srcs = ["filetest.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/filetest", visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/flags.go b/pkg/sentry/fs/flags.go index da0ff58af..1aa271560 100644 --- a/pkg/sentry/fs/flags.go +++ b/pkg/sentry/fs/flags.go @@ -19,6 +19,8 @@ import ( ) // FileFlags encodes file flags. +// +// +stateify savable type FileFlags struct { // Direct indicates that I/O should be done directly. Direct bool diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index 6eea64298..3512bae6f 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -1,24 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "fsutil_state", - srcs = [ - "dirty_set_impl.go", - "file.go", - "file_range_set_impl.go", - "frame_ref_set_impl.go", - "handle.go", - "host_file_mapper.go", - "host_file_mapper_state.go", - "inode.go", - "inode_cached.go", - ], - out = "fsutil_state.go", - package = "fsutil", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "dirty_set_impl", @@ -84,7 +67,6 @@ go_library( "frame_ref_set.go", "frame_ref_set_impl.go", "fsutil.go", - "fsutil_state.go", "handle.go", "host_file_mapper.go", "host_file_mapper_state.go", diff --git a/pkg/sentry/fs/fsutil/dirty_set.go b/pkg/sentry/fs/fsutil/dirty_set.go index 9c6c98542..8e31e48fd 100644 --- a/pkg/sentry/fs/fsutil/dirty_set.go +++ b/pkg/sentry/fs/fsutil/dirty_set.go @@ -32,6 +32,8 @@ import ( // DirtyInfo is the value type of DirtySet, and represents information about a // Mappable offset that is dirty (the cached data for that offset is newer than // its source). +// +// +stateify savable type DirtyInfo struct { // Keep is true if the represented offset is concurrently writable, such // that writing the data for that offset back to the source does not diff --git a/pkg/sentry/fs/fsutil/handle.go b/pkg/sentry/fs/fsutil/handle.go index 149c0f84a..e7efd3c0f 100644 --- a/pkg/sentry/fs/fsutil/handle.go +++ b/pkg/sentry/fs/fsutil/handle.go @@ -27,6 +27,8 @@ import ( // // FIXME: Remove Handle entirely in favor of individual fs.File // implementations using simple generic utilities. +// +// +stateify savable type Handle struct { NoopRelease `state:"nosave"` NoIoctl `state:"nosave"` diff --git a/pkg/sentry/fs/fsutil/host_file_mapper.go b/pkg/sentry/fs/fsutil/host_file_mapper.go index d0a27fc1c..9c1e2f76f 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper.go @@ -29,6 +29,8 @@ import ( // HostFileMapper caches mappings of an arbitrary host file descriptor. It is // used by implementations of memmap.Mappable that represent a host file // descriptor. +// +// +stateify savable type HostFileMapper struct { // HostFile conceptually breaks the file into pieces called chunks, of // size and alignment chunkSize, and caches mappings of the file on a chunk diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index e1ad07df2..177396fdc 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -31,6 +31,8 @@ func NewSimpleInodeOperations(i InodeSimpleAttributes) fs.InodeOperations { } // simpleInodeOperations is a simple implementation of Inode. +// +// +stateify savable type simpleInodeOperations struct { DeprecatedFileOperations `state:"nosave"` InodeNotDirectory `state:"nosave"` @@ -48,6 +50,8 @@ type simpleInodeOperations struct { // InodeSimpleAttributes implements a subset of the Inode interface. It provides // read-only access to attributes. +// +// +stateify savable type InodeSimpleAttributes struct { // FSType is the filesystem type reported by StatFS. FSType uint64 @@ -110,6 +114,8 @@ func (*InodeSimpleAttributes) Truncate(context.Context, *fs.Inode, int64) error // // Users need not initialize Xattrs to non-nil (it will be initialized // when the first extended attribute is set. +// +// +stateify savable type InMemoryAttributes struct { Unstable fs.UnstableAttr Xattrs map[string][]byte diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index cba642a8f..0a320e2d8 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -55,6 +55,8 @@ import ( // // Implementations of InodeOperations.WriteOut must call Sync to write out // in-memory modifications of data and metadata to the CachedFileObject. +// +// +stateify savable type CachingInodeOperations struct { // backingFile is a handle to a cached file object. backingFile CachedFileObject diff --git a/pkg/sentry/fs/gofer/BUILD b/pkg/sentry/fs/gofer/BUILD index 1277379e7..cb17339c9 100644 --- a/pkg/sentry/fs/gofer/BUILD +++ b/pkg/sentry/fs/gofer/BUILD @@ -1,21 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "gofer_state", - srcs = [ - "file.go", - "file_state.go", - "fs.go", - "inode.go", - "inode_state.go", - "session.go", - "session_state.go", - ], - out = "gofer_state.go", - package = "gofer", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "gofer", @@ -27,7 +12,6 @@ go_library( "file.go", "file_state.go", "fs.go", - "gofer_state.go", "handles.go", "inode.go", "inode_state.go", @@ -41,7 +25,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/amutex", "//pkg/fd", "//pkg/log", "//pkg/metric", @@ -54,15 +37,11 @@ go_library( "//pkg/sentry/fs/fdpipe", "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/host", - "//pkg/sentry/fs/lock", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", - "//pkg/sentry/platform", "//pkg/sentry/safemem", - "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index 039618808..46a6bbd5d 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -33,6 +33,8 @@ import ( var openedWX = metric.MustCreateNewUint64Metric("/gofer/opened_write_execute_file", true /* sync */, "Number of times a writable+executable file was opened from a gofer.") // fileOperations implements fs.FileOperations for a remote file system. +// +// +stateify savable type fileOperations struct { fsutil.NoIoctl `state:"nosave"` waiter.AlwaysReady `state:"nosave"` diff --git a/pkg/sentry/fs/gofer/fs.go b/pkg/sentry/fs/gofer/fs.go index dd5d43c47..3ae93f059 100644 --- a/pkg/sentry/fs/gofer/fs.go +++ b/pkg/sentry/fs/gofer/fs.go @@ -83,6 +83,8 @@ var ( ) // filesystem is a 9p client. +// +// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go index df584c382..7fc8f77b0 100644 --- a/pkg/sentry/fs/gofer/inode.go +++ b/pkg/sentry/fs/gofer/inode.go @@ -35,6 +35,8 @@ import ( ) // inodeOperations implements fs.InodeOperations. +// +// +stateify savable type inodeOperations struct { fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` @@ -68,6 +70,8 @@ type inodeOperations struct { // circular load dependency between it and inodeOperations). Even with // lazy loading, this approach defines the dependencies between objects // and the expected load behavior more concretely. +// +// +stateify savable type inodeFileState struct { // s is common file system state for Gofers. s *session `state:"wait"` diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index b6841526a..648a11435 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -27,6 +27,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/unet" ) +// +stateify savable type endpointMap struct { mu sync.RWMutex `state:"nosave"` // TODO: Make map with private unix sockets savable. @@ -63,6 +64,8 @@ func (e *endpointMap) get(key device.MultiDeviceKey) unix.BoundEndpoint { } // session holds state for each 9p session established during sys_mount. +// +// +stateify savable type session struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 23ec66f50..29c79284a 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -1,23 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "host_state", - srcs = [ - "control.go", - "descriptor.go", - "descriptor_state.go", - "file.go", - "fs.go", - "inode.go", - "inode_state.go", - "socket.go", - "socket_state.go", - ], - out = "host_state.go", - package = "host", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "host", @@ -28,7 +11,6 @@ go_library( "device.go", "file.go", "fs.go", - "host_state.go", "inode.go", "inode_state.go", "ioctl_unsafe.go", @@ -42,7 +24,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/amutex", "//pkg/fd", "//pkg/log", "//pkg/refs", @@ -52,20 +33,14 @@ go_library( "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", - "//pkg/sentry/fs/lock", "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", - "//pkg/sentry/platform", "//pkg/sentry/safemem", - "//pkg/sentry/socket", "//pkg/sentry/socket/control", "//pkg/sentry/socket/unix", - "//pkg/sentry/uniqueid", - "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/link/rawfile", diff --git a/pkg/sentry/fs/host/descriptor.go b/pkg/sentry/fs/host/descriptor.go index 613bd06e8..3aee4d11c 100644 --- a/pkg/sentry/fs/host/descriptor.go +++ b/pkg/sentry/fs/host/descriptor.go @@ -25,6 +25,8 @@ import ( ) // descriptor wraps a host fd. +// +// +stateify savable type descriptor struct { // donated is true if the host fd was donated by another process. donated bool diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index bdf844337..f9bef6d93 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -37,6 +37,8 @@ import ( ) // fileOperations implements fs.FileOperations for a host file descriptor. +// +// +stateify savable type fileOperations struct { fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/host/fs.go b/pkg/sentry/fs/host/fs.go index 974700636..e46ae433c 100644 --- a/pkg/sentry/fs/host/fs.go +++ b/pkg/sentry/fs/host/fs.go @@ -51,6 +51,8 @@ const maxTraversals = 10 // to lock down the configurations. This filesystem should only be mounted at root. // // Think twice before exposing this to applications. +// +// +stateify savable type Filesystem struct { // whitelist is a set of host paths to whitelist. paths []string @@ -266,8 +268,10 @@ func newMountSource(ctx context.Context, root string, mounter fs.FileOwner, file } // superOperations implements fs.MountSourceOperations. +// +// +stateify savable type superOperations struct { - fs.SimpleMountSourceOperations `state:"nosave"` + fs.SimpleMountSourceOperations // root is the path of the mount point. All inode mappings // are relative to this root. diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index 226bc5164..761ccde33 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -34,6 +34,8 @@ import ( // inodeOperations implements fs.InodeOperations for an fs.Inodes backed // by a host file descriptor. +// +// +stateify savable type inodeOperations struct { fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` @@ -65,6 +67,8 @@ type inodeOperations struct { // circular load dependency between it and inodeOperations). Even with // lazy loading, this approach defines the dependencies between objects // and the expected load behavior more concretely. +// +// +stateify savable type inodeFileState struct { // Common file system state. mops *superOperations `state:"wait"` diff --git a/pkg/sentry/fs/host/socket.go b/pkg/sentry/fs/host/socket.go index f4689f51f..1d93eb1e3 100644 --- a/pkg/sentry/fs/host/socket.go +++ b/pkg/sentry/fs/host/socket.go @@ -34,6 +34,8 @@ import ( ) // endpoint encapsulates the state needed to represent a host Unix socket. +// +// +stateify savable type endpoint struct { queue waiter.Queue `state:"nosave"` diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index 6c8e6f188..d0dbce5dd 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -28,6 +28,8 @@ import ( // Inode is a file system object that can be simultaneously referenced by different // components of the VFS (Dirent, fs.File, etc). +// +// +stateify savable type Inode struct { // AtomicRefCount is our reference count. refs.AtomicRefCount @@ -58,6 +60,8 @@ type Inode struct { // Note that in Linux fcntl(2) and flock(2) locks are _not_ cooperative, because race and // deadlock conditions make merging them prohibitive. We do the same and keep them oblivious // to each other but provide a "context" as a convenient container. +// +// +stateify savable type LockCtx struct { // Posix is a set of POSIX-style regional advisory locks, see fcntl(2). Posix lock.Locks diff --git a/pkg/sentry/fs/inode_inotify.go b/pkg/sentry/fs/inode_inotify.go index 358bbecdf..683140afe 100644 --- a/pkg/sentry/fs/inode_inotify.go +++ b/pkg/sentry/fs/inode_inotify.go @@ -20,6 +20,8 @@ import ( ) // Watches is the collection of inotify watches on an inode. +// +// +stateify savable type Watches struct { // mu protects the fields below. mu sync.RWMutex `state:"nosave"` diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 6f5e8ce5e..2aabdded8 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -34,6 +34,8 @@ import ( // // Lock ordering: // Inotify.mu -> Inode.Watches.mu -> Watch.mu -> Inotify.evMu +// +// +stateify savable type Inotify struct { // Unique identifier for this inotify instance. We don't just reuse the // inotify fd because fds can be duped. These should not be exposed to the diff --git a/pkg/sentry/fs/inotify_event.go b/pkg/sentry/fs/inotify_event.go index 217915ba4..e9b5e0f56 100644 --- a/pkg/sentry/fs/inotify_event.go +++ b/pkg/sentry/fs/inotify_event.go @@ -28,6 +28,8 @@ import ( const inotifyEventBaseSize = 16 // Event represents a struct inotify_event from linux. +// +// +stateify savable type Event struct { ilist.Entry diff --git a/pkg/sentry/fs/inotify_watch.go b/pkg/sentry/fs/inotify_watch.go index 8904ef544..3e1959e83 100644 --- a/pkg/sentry/fs/inotify_watch.go +++ b/pkg/sentry/fs/inotify_watch.go @@ -27,6 +27,8 @@ import ( // holding an extra ref on each dirent known (by inotify) to point to the // inode. These are known as pins. For a full discussion, see // fs/g3doc/inotify.md. +// +// +stateify savable type Watch struct { // Inotify instance which owns this watch. owner *Inotify diff --git a/pkg/sentry/fs/lock/BUILD b/pkg/sentry/fs/lock/BUILD index 2607d7ed3..3159ff1da 100644 --- a/pkg/sentry/fs/lock/BUILD +++ b/pkg/sentry/fs/lock/BUILD @@ -1,18 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "lock_state", - srcs = [ - "lock.go", - "lock_range.go", - "lock_set.go", - ], - out = "lock_state.go", - package = "lock", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "lock_range", @@ -49,13 +38,11 @@ go_library( "lock_range.go", "lock_set.go", "lock_set_functions.go", - "lock_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/lock", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/log", - "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/lock/lock.go b/pkg/sentry/fs/lock/lock.go index 24d54c989..e9b376eb6 100644 --- a/pkg/sentry/fs/lock/lock.go +++ b/pkg/sentry/fs/lock/lock.go @@ -88,6 +88,8 @@ const LockEOF = math.MaxUint64 // // A Lock may be downgraded from a write lock to a read lock only if // the write lock's uid is the same as the read lock. +// +// +stateify savable type Lock struct { // Readers are the set of read lock holders identified by UniqueID. // If len(Readers) > 0 then HasWriter must be false. @@ -103,6 +105,8 @@ type Lock struct { } // Locks is a thread-safe wrapper around a LockSet. +// +// +stateify savable type Locks struct { // mu protects locks below. mu sync.Mutex `state:"nosave"` @@ -111,7 +115,7 @@ type Locks struct { locks LockSet // blockedQueue is the queue of waiters that are waiting on a lock. - blockedQueue waiter.Queue + blockedQueue waiter.Queue `state:"zerovalue"` } // Blocker is the interface used for blocking locks. Passing a nil Blocker diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index eb1897174..4ede767f9 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -101,6 +101,8 @@ func (i InodeMappings) String() string { // (e.g. cannot be mounted at different locations). // // TODO: Move mount-specific information out of MountSource. +// +// +stateify savable type MountSource struct { refs.AtomicRefCount @@ -260,6 +262,8 @@ func NewNonCachingMountSource(filesystem Filesystem, flags MountSourceFlags) *Mo } // SimpleMountSourceOperations implements MountSourceOperations. +// +// +stateify savable type SimpleMountSourceOperations struct { keep bool } diff --git a/pkg/sentry/fs/mount_overlay.go b/pkg/sentry/fs/mount_overlay.go index 1be81e3a1..d135e8a37 100644 --- a/pkg/sentry/fs/mount_overlay.go +++ b/pkg/sentry/fs/mount_overlay.go @@ -18,6 +18,8 @@ import "gvisor.googlesource.com/gvisor/pkg/sentry/context" // overlayMountSourceOperations implements MountSourceOperations for an overlay // mount point. +// +// +stateify savable type overlayMountSourceOperations struct { upper *MountSource lower *MountSource @@ -72,6 +74,8 @@ func (o *overlayMountSourceOperations) Destroy() { } // type overlayFilesystem is the filesystem for overlay mounts. +// +// +stateify savable type overlayFilesystem struct{} // Name implements Filesystem.Name. diff --git a/pkg/sentry/fs/mounts.go b/pkg/sentry/fs/mounts.go index 87da4ee0e..144d3427d 100644 --- a/pkg/sentry/fs/mounts.go +++ b/pkg/sentry/fs/mounts.go @@ -32,6 +32,8 @@ import ( const DefaultTraversalLimit = 10 // MountNamespace defines a collection of mounts. +// +// +stateify savable type MountNamespace struct { refs.AtomicRefCount diff --git a/pkg/sentry/fs/overlay.go b/pkg/sentry/fs/overlay.go index 7357d6401..af13dc8c7 100644 --- a/pkg/sentry/fs/overlay.go +++ b/pkg/sentry/fs/overlay.go @@ -145,6 +145,8 @@ func newOverlayInode(ctx context.Context, o *overlayEntry, msrc *MountSource) *I } // overlayEntry is the overlay metadata of an Inode. It implements Mappable. +// +// +stateify savable type overlayEntry struct { // lowerExists is true if an Inode exists for this file in the lower // filesystem. If lowerExists is true, then the overlay must create diff --git a/pkg/sentry/fs/proc/BUILD b/pkg/sentry/fs/proc/BUILD index 870df47b2..2d9f07f2f 100644 --- a/pkg/sentry/fs/proc/BUILD +++ b/pkg/sentry/fs/proc/BUILD @@ -1,32 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "proc_state", - srcs = [ - "cpuinfo.go", - "exec_args.go", - "fds.go", - "file.go", - "filesystems.go", - "fs.go", - "loadavg.go", - "meminfo.go", - "mounts.go", - "net.go", - "proc.go", - "stat.go", - "sys.go", - "sys_net.go", - "task.go", - "uid_gid_map.go", - "uptime.go", - "version.go", - ], - out = "proc_state.go", - package = "proc", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "proc", @@ -42,7 +16,6 @@ go_library( "mounts.go", "net.go", "proc.go", - "proc_state.go", "rpcinet_proc.go", "stat.go", "sys.go", @@ -56,9 +29,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/amutex", - "//pkg/log", - "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/proc/device", @@ -73,8 +43,6 @@ go_library( "//pkg/sentry/socket/rpcinet", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", - "//pkg/syserr", "//pkg/syserror", ], ) diff --git a/pkg/sentry/fs/proc/cpuinfo.go b/pkg/sentry/fs/proc/cpuinfo.go index f80aaa5b1..4dfec03a4 100644 --- a/pkg/sentry/fs/proc/cpuinfo.go +++ b/pkg/sentry/fs/proc/cpuinfo.go @@ -27,6 +27,8 @@ import ( // cpuinfo is a file describing the CPU capabilities. // // Presently cpuinfo never changes, so it doesn't need to be a SeqFile. +// +// +stateify savable type cpuinfo struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/exec_args.go b/pkg/sentry/fs/proc/exec_args.go index 0e1523bf1..a69cbaa0e 100644 --- a/pkg/sentry/fs/proc/exec_args.go +++ b/pkg/sentry/fs/proc/exec_args.go @@ -37,6 +37,8 @@ const ( // execArgFile is a file containing the exec args (either cmdline or environ) // for a given task. +// +// +stateify savable type execArgFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index 194a9c12a..cca8f874c 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -138,6 +138,8 @@ func (f *fd) Close() error { } // fdDir implements /proc/TID/fd. +// +// +stateify savable type fdDir struct { ramfs.Dir @@ -197,6 +199,8 @@ func (f *fdDir) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset } // fdInfo is a single file in /proc/TID/fdinfo/. +// +// +stateify savable type fdInfo struct { ramfs.File @@ -229,6 +233,8 @@ func (*fdInfo) Truncate(ctx context.Context, inode *fs.Inode, size int64) error // fdInfoDir implements /proc/TID/fdinfo. It embeds an fdDir, but overrides // Lookup and Readdir. +// +// +stateify savable type fdInfoDir struct { ramfs.Dir diff --git a/pkg/sentry/fs/proc/file.go b/pkg/sentry/fs/proc/file.go index 9a433cdf8..4b2d08e75 100644 --- a/pkg/sentry/fs/proc/file.go +++ b/pkg/sentry/fs/proc/file.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) +// +stateify savable type file struct { fs.InodeOperations diff --git a/pkg/sentry/fs/proc/filesystems.go b/pkg/sentry/fs/proc/filesystems.go index 37db9cf9c..49b92fd8a 100644 --- a/pkg/sentry/fs/proc/filesystems.go +++ b/pkg/sentry/fs/proc/filesystems.go @@ -24,6 +24,8 @@ import ( ) // filesystemsData backs /proc/filesystems. +// +// +stateify savable type filesystemsData struct{} // NeedsUpdate returns true on the first generation. The set of registered file diff --git a/pkg/sentry/fs/proc/fs.go b/pkg/sentry/fs/proc/fs.go index 3aadd6ac4..061824b8c 100644 --- a/pkg/sentry/fs/proc/fs.go +++ b/pkg/sentry/fs/proc/fs.go @@ -22,6 +22,8 @@ import ( ) // filesystem is a procfs. +// +// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/proc/loadavg.go b/pkg/sentry/fs/proc/loadavg.go index 7583b6ccd..6fac251d2 100644 --- a/pkg/sentry/fs/proc/loadavg.go +++ b/pkg/sentry/fs/proc/loadavg.go @@ -23,6 +23,8 @@ import ( ) // loadavgData backs /proc/loadavg. +// +// +stateify savable type loadavgData struct{} // NeedsUpdate implements seqfile.SeqSource.NeedsUpdate. diff --git a/pkg/sentry/fs/proc/meminfo.go b/pkg/sentry/fs/proc/meminfo.go index 49cb0faed..53dfd59ef 100644 --- a/pkg/sentry/fs/proc/meminfo.go +++ b/pkg/sentry/fs/proc/meminfo.go @@ -26,6 +26,8 @@ import ( ) // meminfoData backs /proc/meminfo. +// +// +stateify savable type meminfoData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/proc/mounts.go b/pkg/sentry/fs/proc/mounts.go index 108432f4e..2b8167c28 100644 --- a/pkg/sentry/fs/proc/mounts.go +++ b/pkg/sentry/fs/proc/mounts.go @@ -71,6 +71,8 @@ func forEachMountSource(t *kernel.Task, fn func(string, *fs.MountSource)) { } // mountInfoFile is used to implement /proc/[pid]/mountinfo. +// +// +stateify savable type mountInfoFile struct { t *kernel.Task } @@ -152,6 +154,8 @@ func (mif *mountInfoFile) ReadSeqFileData(ctx context.Context, handle seqfile.Se } // mountsFile is used to implement /proc/[pid]/mountinfo. +// +// +stateify savable type mountsFile struct { t *kernel.Task } diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index b2a8d639c..07029a7bb 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -33,6 +33,8 @@ import ( ) // proc is a root proc node. +// +// +stateify savable type proc struct { ramfs.Dir @@ -47,6 +49,8 @@ type proc struct { // stubProcFSFile is a file type that can be used to return file contents // which are constant. This file is not writable and will always have mode // 0444. +// +// +stateify savable type stubProcFSFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/seqfile/BUILD b/pkg/sentry/fs/proc/seqfile/BUILD index c84f7e20d..53c475652 100644 --- a/pkg/sentry/fs/proc/seqfile/BUILD +++ b/pkg/sentry/fs/proc/seqfile/BUILD @@ -1,22 +1,10 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "seqfile_state", - srcs = [ - "seqfile.go", - ], - out = "seqfile_state.go", - package = "seqfile", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "seqfile", - srcs = [ - "seqfile.go", - "seqfile_state.go", - ], + srcs = ["seqfile.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile", visibility = ["//pkg/sentry:internal"], deps = [ @@ -26,26 +14,16 @@ go_library( "//pkg/sentry/fs/ramfs", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", ], ) -go_stateify( - name = "seqfile_test_state", - srcs = ["seqfile_test.go"], - out = "seqfile_test_state.go", - package = "seqfile", -) - go_test( name = "seqfile_test", size = "small", - srcs = [ - "seqfile_test.go", - "seqfile_test_state.go", - ], + srcs = ["seqfile_test.go"], embed = [":seqfile"], deps = [ + "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs/test", diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index c08565f8a..51cae5e37 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -30,6 +30,8 @@ import ( type SeqHandle interface{} // SeqData holds the data for one unit in the file. +// +// +stateify savable type SeqData struct { // The data to be returned to the user. Buf []byte @@ -82,6 +84,8 @@ func (s *SeqGenerationCounter) IsCurrent(generation int64) bool { } // SeqFile is used to provide dynamic files that can be ordered by record. +// +// +stateify savable type SeqFile struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/stat.go b/pkg/sentry/fs/proc/stat.go index 284f3e52b..bf7650211 100644 --- a/pkg/sentry/fs/proc/stat.go +++ b/pkg/sentry/fs/proc/stat.go @@ -25,6 +25,8 @@ import ( ) // statData backs /proc/stat. +// +// +stateify savable type statData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index aab891c53..a2d36ca23 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -28,6 +28,8 @@ import ( ) // hostname is a file containing the system hostname. +// +// +stateify savable type hostname struct { ramfs.Entry } @@ -52,6 +54,8 @@ func (p *proc) newHostname(ctx context.Context, msrc *fs.MountSource) *fs.Inode } // mmapMinAddrData backs /proc/sys/vm/mmap_min_addr. +// +// +stateify savable type mmapMinAddrData struct { k *kernel.Kernel } @@ -74,6 +78,7 @@ func (d *mmapMinAddrData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHand }, 0 } +// +stateify savable type overcommitMemory struct{} func (*overcommitMemory) NeedsUpdate(generation int64) bool { diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index f3a5043f8..beb25be20 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -33,6 +33,7 @@ const ( tcpWMem ) +// +stateify savable type tcpMem struct { ramfs.Entry s inet.Stack @@ -100,6 +101,7 @@ func (m *tcpMem) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, return n, cperr } +// +stateify savable type tcpSack struct { ramfs.Entry s inet.Stack diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index efc635946..748ca4320 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -52,6 +52,8 @@ func getTaskMM(t *kernel.Task) (*mm.MemoryManager, error) { } // taskDir represents a task-level directory. +// +// +stateify savable type taskDir struct { ramfs.Dir @@ -92,6 +94,8 @@ func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace } // subtasks represents a /proc/TID/task directory. +// +// +stateify savable type subtasks struct { ramfs.Dir @@ -167,6 +171,8 @@ func (s *subtasks) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, off } // exe is an fs.InodeOperations symlink for the /proc/PID/exe file. +// +// +stateify savable type exe struct { ramfs.Symlink @@ -226,6 +232,8 @@ func (e *exe) Readlink(ctx context.Context, inode *fs.Inode) (string, error) { // namespaceFile represents a file in the namespacefs, such as the files in // /proc//ns. +// +// +stateify savable type namespaceFile struct { ramfs.Symlink @@ -274,6 +282,8 @@ func newNamespaceDir(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { } // mapsData implements seqfile.SeqSource for /proc/[pid]/maps. +// +// +stateify savable type mapsData struct { t *kernel.Task } @@ -311,6 +321,7 @@ func (md *mapsData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([ return []seqfile.SeqData{}, 0 } +// +stateify savable type taskStatData struct { t *kernel.Task @@ -391,6 +402,8 @@ func (s *taskStatData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) } // statmData implements seqfile.SeqSource for /proc/[pid]/statm. +// +// +stateify savable type statmData struct { t *kernel.Task } @@ -425,6 +438,8 @@ func (s *statmData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([ } // statusData implements seqfile.SeqSource for /proc/[pid]/status. +// +// +stateify savable type statusData struct { t *kernel.Task pidns *kernel.PIDNamespace @@ -490,6 +505,7 @@ type ioUsage interface { IOUsage() *usage.IO } +// +stateify savable type ioData struct { ioUsage } @@ -530,6 +546,8 @@ func (i *ioData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([]se // On Linux, /proc/[pid]/comm is writable, and writing to the comm file changes // the thread name. We don't implement this yet as there are no known users of // this feature. +// +// +stateify savable type comm struct { ramfs.Entry @@ -559,6 +577,8 @@ func (c *comm) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, off } // auxvec is a file containing the auxiliary vector for a task. +// +// +stateify savable type auxvec struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index 85acb5163..9811d9c9d 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -29,6 +29,8 @@ import ( // An idMapSeqSource is a seqfile.SeqSource that returns UID or GID mappings // from a task's user namespace. +// +// +stateify savable type idMapSeqSource struct { t *kernel.Task gids bool @@ -70,6 +72,7 @@ type idMapSeqHandle struct { value int } +// +stateify savable type idMapSeqFile struct { seqfile.SeqFile } diff --git a/pkg/sentry/fs/proc/uptime.go b/pkg/sentry/fs/proc/uptime.go index 4679d5821..f3a9b81df 100644 --- a/pkg/sentry/fs/proc/uptime.go +++ b/pkg/sentry/fs/proc/uptime.go @@ -27,6 +27,8 @@ import ( ) // uptime is a file containing the system uptime. +// +// +stateify savable type uptime struct { ramfs.Entry diff --git a/pkg/sentry/fs/proc/version.go b/pkg/sentry/fs/proc/version.go index c0f2e87e3..00f6a2afd 100644 --- a/pkg/sentry/fs/proc/version.go +++ b/pkg/sentry/fs/proc/version.go @@ -23,6 +23,8 @@ import ( ) // versionData backs /proc/version. +// +// +stateify savable type versionData struct { // k is the owning Kernel. k *kernel.Kernel diff --git a/pkg/sentry/fs/ramfs/BUILD b/pkg/sentry/fs/ramfs/BUILD index d84f2c624..5230157fe 100644 --- a/pkg/sentry/fs/ramfs/BUILD +++ b/pkg/sentry/fs/ramfs/BUILD @@ -1,19 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "ramfs_state", - srcs = [ - "dir.go", - "file.go", - "ramfs.go", - "socket.go", - "symlink.go", - ], - out = "ramfs_state.go", - package = "ramfs", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "ramfs", @@ -21,7 +8,6 @@ go_library( "dir.go", "file.go", "ramfs.go", - "ramfs_state.go", "socket.go", "symlink.go", "tree.go", @@ -29,12 +15,8 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs", visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/amutex", - "//pkg/log", - "//pkg/refs", "//pkg/secio", "//pkg/sentry/context", - "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", @@ -42,7 +24,6 @@ go_library( "//pkg/sentry/memmap", "//pkg/sentry/safemem", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index 19d5612ed..04432f28c 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -44,6 +44,8 @@ type CreateOps struct { } // Dir represents a single directory in the filesystem. +// +// +stateify savable type Dir struct { Entry diff --git a/pkg/sentry/fs/ramfs/ramfs.go b/pkg/sentry/fs/ramfs/ramfs.go index d6cfaf753..13e72e775 100644 --- a/pkg/sentry/fs/ramfs/ramfs.go +++ b/pkg/sentry/fs/ramfs/ramfs.go @@ -60,6 +60,8 @@ var ( // Entry represents common internal state for file and directory nodes. // This may be used by other packages to easily create ramfs files. +// +// +stateify savable type Entry struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoMappable `state:"nosave"` diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index b0c79325f..93427a1ff 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -21,6 +21,8 @@ import ( ) // Socket represents a socket. +// +// +stateify savable type Socket struct { Entry diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index 9bbf78619..1c54d9991 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -22,6 +22,8 @@ import ( ) // Symlink represents a symlink. +// +// +stateify savable type Symlink struct { Entry diff --git a/pkg/sentry/fs/ramfs/test/BUILD b/pkg/sentry/fs/ramfs/test/BUILD index 57fee45e2..187eac49d 100644 --- a/pkg/sentry/fs/ramfs/test/BUILD +++ b/pkg/sentry/fs/ramfs/test/BUILD @@ -1,30 +1,16 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "test_state", - srcs = [ - "test.go", - ], - out = "test_state.go", - package = "test", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "test", testonly = 1, - srcs = [ - "test.go", - "test_state.go", - ], + srcs = ["test.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/sentry/context", - "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs", - "//pkg/state", ], ) diff --git a/pkg/sentry/fs/sys/BUILD b/pkg/sentry/fs/sys/BUILD index 095ff1f25..bc24e980e 100644 --- a/pkg/sentry/fs/sys/BUILD +++ b/pkg/sentry/fs/sys/BUILD @@ -1,16 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "sys_state", - srcs = [ - "fs.go", - "sys.go", - ], - out = "sys_state.go", - package = "sys", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "sys", @@ -18,7 +8,6 @@ go_library( "device.go", "fs.go", "sys.go", - "sys_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/sys", visibility = ["//pkg/sentry:internal"], @@ -28,6 +17,5 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/fs/ramfs", "//pkg/sentry/usermem", - "//pkg/state", ], ) diff --git a/pkg/sentry/fs/sys/fs.go b/pkg/sentry/fs/sys/fs.go index c6d5f7fd8..625525540 100644 --- a/pkg/sentry/fs/sys/fs.go +++ b/pkg/sentry/fs/sys/fs.go @@ -20,6 +20,8 @@ import ( ) // filesystem is a sysfs. +// +// +stateify savable type filesystem struct{} func init() { diff --git a/pkg/sentry/fs/sys/sys.go b/pkg/sentry/fs/sys/sys.go index ccf56f644..b9b2fb4a1 100644 --- a/pkg/sentry/fs/sys/sys.go +++ b/pkg/sentry/fs/sys/sys.go @@ -22,12 +22,13 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -type Dir struct { +// +stateify savable +type dir struct { ramfs.Dir } func newDir(ctx context.Context, msrc *fs.MountSource, contents map[string]*fs.Inode) *fs.Inode { - d := &Dir{} + d := &dir{} d.InitDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) return fs.NewInode(d, msrc, fs.StableAttr{ DeviceID: sysfsDevice.DeviceID(), diff --git a/pkg/sentry/fs/timerfd/BUILD b/pkg/sentry/fs/timerfd/BUILD index 8b1b7872e..ffdd7e0dc 100644 --- a/pkg/sentry/fs/timerfd/BUILD +++ b/pkg/sentry/fs/timerfd/BUILD @@ -1,33 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "timerfd_state", - srcs = [ - "timerfd.go", - ], - out = "timerfd_state.go", - package = "timerfd", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "timerfd", - srcs = [ - "timerfd.go", - "timerfd_state.go", - ], + srcs = ["timerfd.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/timerfd", visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/fs/timerfd/timerfd.go b/pkg/sentry/fs/timerfd/timerfd.go index ae58f6fd7..767db95a0 100644 --- a/pkg/sentry/fs/timerfd/timerfd.go +++ b/pkg/sentry/fs/timerfd/timerfd.go @@ -30,6 +30,8 @@ import ( ) // TimerOperations implements fs.FileOperations for timerfds. +// +// +stateify savable type TimerOperations struct { fsutil.ZeroSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` @@ -38,7 +40,7 @@ type TimerOperations struct { fsutil.NoMMap `state:"nosave"` fsutil.NoIoctl `state:"nosave"` - events waiter.Queue `state:"nosave"` + events waiter.Queue `state:"zerovalue"` timer *ktime.Timer // val is the number of timer expirations since the last successful call to diff --git a/pkg/sentry/fs/tmpfs/BUILD b/pkg/sentry/fs/tmpfs/BUILD index 473ab4296..cfe11ab02 100644 --- a/pkg/sentry/fs/tmpfs/BUILD +++ b/pkg/sentry/fs/tmpfs/BUILD @@ -1,18 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tmpfs_state", - srcs = [ - "file_regular.go", - "fs.go", - "inode_file.go", - "tmpfs.go", - ], - out = "tmpfs_state.go", - package = "tmpfs", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tmpfs", @@ -22,13 +10,11 @@ go_library( "fs.go", "inode_file.go", "tmpfs.go", - "tmpfs_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tmpfs", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/log", "//pkg/sentry/context", "//pkg/sentry/device", "//pkg/sentry/fs", @@ -41,7 +27,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/tcpip/transport/unix", "//pkg/waiter", ], diff --git a/pkg/sentry/fs/tmpfs/file_regular.go b/pkg/sentry/fs/tmpfs/file_regular.go index 9811d90bc..342688f81 100644 --- a/pkg/sentry/fs/tmpfs/file_regular.go +++ b/pkg/sentry/fs/tmpfs/file_regular.go @@ -25,6 +25,8 @@ import ( // regularFileOperations implements fs.FileOperations for a regular // tmpfs file. +// +// +stateify savable type regularFileOperations struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go index 5bd9ade52..ca620e65e 100644 --- a/pkg/sentry/fs/tmpfs/fs.go +++ b/pkg/sentry/fs/tmpfs/fs.go @@ -47,6 +47,8 @@ const ( var modeRegexp = regexp.MustCompile("0[0-7][0-7][0-7]") // Filesystem is a tmpfs. +// +// +stateify savable type Filesystem struct{} func init() { diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 4e803c9ff..1e4fe47d2 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -43,6 +43,8 @@ import ( // include an InvalidatorRegion associated with that reference. When the // referenced portion of the file is removed (with Truncate), the associated // InvalidatorRegion is invalidated. +// +// +stateify savable type fileInodeOperations struct { fsutil.DeprecatedFileOperations `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 1cc7ae491..10cb5451d 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -49,6 +49,8 @@ func rename(ctx context.Context, oldParent *fs.Inode, oldName string, newParent } // Dir is a directory. +// +// +stateify savable type Dir struct { ramfs.Dir @@ -122,6 +124,8 @@ func (*Dir) StatFS(context.Context) (fs.Info, error) { } // Symlink is a symlink. +// +// +stateify savable type Symlink struct { ramfs.Symlink } @@ -149,6 +153,8 @@ func (s *Symlink) StatFS(context.Context) (fs.Info, error) { } // Socket is a socket. +// +// +stateify savable type Socket struct { ramfs.Socket } @@ -176,6 +182,8 @@ func (s *Socket) StatFS(context.Context) (fs.Info, error) { } // Fifo is a tmpfs named pipe. +// +// +stateify savable type Fifo struct { ramfs.Entry } diff --git a/pkg/sentry/fs/tty/BUILD b/pkg/sentry/fs/tty/BUILD index 363897b2c..3c446eef4 100644 --- a/pkg/sentry/fs/tty/BUILD +++ b/pkg/sentry/fs/tty/BUILD @@ -1,22 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tty_state", - srcs = [ - "dir.go", - "fs.go", - "inode.go", - "line_discipline.go", - "master.go", - "queue.go", - "slave.go", - "terminal.go", - ], - out = "tty_state.go", - package = "tty", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tty", @@ -29,7 +13,6 @@ go_library( "queue.go", "slave.go", "terminal.go", - "tty_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tty", visibility = ["//pkg/sentry:internal"], @@ -44,7 +27,6 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index 2c5b2aed6..c91091db4 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -49,14 +49,16 @@ import ( // corresponding Dirents hold on their parent (this directory). // // dirInodeOperations implements fs.InodeOperations. +// +// +stateify savable type dirInodeOperations struct { - fsutil.DeprecatedFileOperations - fsutil.InodeNotSocket - fsutil.InodeNotRenameable - fsutil.InodeNotSymlink - fsutil.InodeNoExtendedAttributes - fsutil.NoMappable - fsutil.NoopWriteOut + fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotRenameable `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.NoMappable `state:"nosave"` + fsutil.NoopWriteOut `state:"nosave"` // msrc is the super block this directory is on. // @@ -348,6 +350,8 @@ func (d *dirInodeOperations) masterClose(t *Terminal) { // // This is nearly identical to fsutil.DirFileOperations, except that it takes // df.di.mu in IterateDir. +// +// +stateify savable type dirFileOperations struct { waiter.AlwaysReady `state:"nosave"` fsutil.NoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/tty/fs.go b/pkg/sentry/fs/tty/fs.go index dbaffe95e..e28635607 100644 --- a/pkg/sentry/fs/tty/fs.go +++ b/pkg/sentry/fs/tty/fs.go @@ -28,6 +28,8 @@ var ptsDevice = device.NewAnonDevice() // // This devpts is always in the new "multi-instance" mode. i.e., it contains a // ptmx device tied to this mount. +// +// +stateify savable type filesystem struct{} func init() { @@ -69,6 +71,8 @@ func (f *filesystem) Mount(ctx context.Context, device string, flags fs.MountSou } // superOperations implements fs.MountSourceOperations, preventing caching. +// +// +stateify savable type superOperations struct{} // Revalidate implements fs.DirentOperations.Revalidate. diff --git a/pkg/sentry/fs/tty/inode.go b/pkg/sentry/fs/tty/inode.go index 04b9a7727..c0fa2b407 100644 --- a/pkg/sentry/fs/tty/inode.go +++ b/pkg/sentry/fs/tty/inode.go @@ -31,6 +31,8 @@ import ( // // * fs.InodeOperations.Release // * fs.InodeOperations.GetFile +// +// +stateify savable type inodeOperations struct { fsutil.DeprecatedFileOperations `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` diff --git a/pkg/sentry/fs/tty/line_discipline.go b/pkg/sentry/fs/tty/line_discipline.go index f094635f5..d243ee40e 100644 --- a/pkg/sentry/fs/tty/line_discipline.go +++ b/pkg/sentry/fs/tty/line_discipline.go @@ -72,6 +72,8 @@ const ( // termiosMu // inQueue.mu // outQueue.mu +// +// +stateify savable type lineDiscipline struct { // inQueue is the input queue of the terminal. inQueue queue @@ -183,6 +185,8 @@ type transformer interface { // outputQueueTransformer implements transformer. It performs line discipline // transformations on the output queue. +// +// +stateify savable type outputQueueTransformer struct{} // transform does output processing for one end of the pty. See @@ -254,6 +258,8 @@ func (*outputQueueTransformer) transform(l *lineDiscipline, q *queue, buf []byte // inputQueueTransformer implements transformer. It performs line discipline // transformations on the input queue. +// +// +stateify savable type inputQueueTransformer struct{} // transform does input processing for one end of the pty. Characters read are diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index 74cdbe874..c7198e218 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -27,6 +27,8 @@ import ( // masterInodeOperations are the fs.InodeOperations for the master end of the // Terminal (ptmx file). +// +// +stateify savable type masterInodeOperations struct { inodeOperations @@ -96,6 +98,8 @@ func (mi *masterInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flag } // masterFileOperations are the fs.FileOperations for the master end of a terminal. +// +// +stateify savable type masterFileOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/tty/queue.go b/pkg/sentry/fs/tty/queue.go index 026d5e077..42c105abc 100644 --- a/pkg/sentry/fs/tty/queue.go +++ b/pkg/sentry/fs/tty/queue.go @@ -32,11 +32,13 @@ import ( // processed (i.e. undergo termios transformations) as they are added to the // read buffer. The read buffer is readable when its length is nonzero and // readable is true. +// +// +stateify savable type queue struct { // mu protects everything in queue. mu sync.Mutex `state:"nosave"` - waiter.Queue `state:"nosave"` + waiter.Queue `state:"zerovalue"` // readBuf is buffer of data ready to be read when readable is true. // This data has been processed. diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index f5eec726e..1c562b172 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -27,6 +27,8 @@ import ( // slaveInodeOperations are the fs.InodeOperations for the slave end of the // Terminal (pts file). +// +// +stateify savable type slaveInodeOperations struct { inodeOperations @@ -86,6 +88,8 @@ func (si *slaveInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags } // slaveFileOperations are the fs.FileOperations for the slave end of a terminal. +// +// +stateify savable type slaveFileOperations struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/fs/tty/terminal.go b/pkg/sentry/fs/tty/terminal.go index fa5b00409..3cb135124 100644 --- a/pkg/sentry/fs/tty/terminal.go +++ b/pkg/sentry/fs/tty/terminal.go @@ -21,6 +21,8 @@ import ( ) // Terminal is a pseudoterminal. +// +// +stateify savable type Terminal struct { refs.AtomicRefCount diff --git a/pkg/sentry/inet/BUILD b/pkg/sentry/inet/BUILD index eaf8f15b2..159c50efb 100644 --- a/pkg/sentry/inet/BUILD +++ b/pkg/sentry/inet/BUILD @@ -3,26 +3,15 @@ package( licenses = ["notice"], # Apache 2.0 ) -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "inet_state", - srcs = ["inet.go"], - out = "inet_state.go", - package = "inet", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "inet", srcs = [ "context.go", "inet.go", - "inet_state.go", "test_stack.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/inet", - deps = [ - "//pkg/sentry/context", - "//pkg/state", - ], + deps = ["//pkg/sentry/context"], ) diff --git a/pkg/sentry/inet/inet.go b/pkg/sentry/inet/inet.go index e4b326993..e54a61196 100644 --- a/pkg/sentry/inet/inet.go +++ b/pkg/sentry/inet/inet.go @@ -87,6 +87,8 @@ type InterfaceAddr struct { } // TCPBufferSize contains settings controlling TCP buffer sizing. +// +// +stateify savable type TCPBufferSize struct { // Min is the minimum size. Min int diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 1c1633068..69a3fbc45 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -1,59 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "kernel_state", - srcs = [ - "abstract_socket_namespace.go", - "fd_map.go", - "fs_context.go", - "ipc_namespace.go", - "kernel.go", - "kernel_state.go", - "pending_signals.go", - "pending_signals_state.go", - "process_group_list.go", - "ptrace.go", - "rseq.go", - "session_list.go", - "sessions.go", - "signal.go", - "signal_handlers.go", - "syscalls.go", - "syscalls_state.go", - "syslog.go", - "task.go", - "task_clone.go", - "task_context.go", - "task_exec.go", - "task_exit.go", - "task_list.go", - "task_resources.go", - "task_run.go", - "task_sched.go", - "task_signals.go", - "task_start.go", - "task_syscall.go", - "thread_group.go", - "threads.go", - "timekeeper.go", - "timekeeper_state.go", - "timer.go", - "uts_namespace.go", - "vdso.go", - "version.go", - ], - out = "kernel_autogen_state.go", - imports = [ - "gvisor.googlesource.com/gvisor/pkg/bpf", - "gvisor.googlesource.com/gvisor/pkg/sentry/arch", - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs", - "gvisor.googlesource.com/gvisor/pkg/tcpip", - ], - package = "kernel", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "pending_signals_list", @@ -119,7 +67,6 @@ go_library( "fs_context.go", "ipc_namespace.go", "kernel.go", - "kernel_autogen_state.go", "kernel_state.go", "pending_signals.go", "pending_signals_list.go", @@ -165,6 +112,11 @@ go_library( "version.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel", + imports = [ + "gvisor.googlesource.com/gvisor/pkg/bpf", + "gvisor.googlesource.com/gvisor/pkg/sentry/arch", + "gvisor.googlesource.com/gvisor/pkg/tcpip", + ], visibility = ["//:sandbox"], deps = [ "//pkg/abi", diff --git a/pkg/sentry/kernel/abstract_socket_namespace.go b/pkg/sentry/kernel/abstract_socket_namespace.go index 014c4a3bf..d6d1d341d 100644 --- a/pkg/sentry/kernel/abstract_socket_namespace.go +++ b/pkg/sentry/kernel/abstract_socket_namespace.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) +// +stateify savable type abstractEndpoint struct { ep unix.BoundEndpoint wr *refs.WeakRef @@ -39,6 +40,8 @@ func (e *abstractEndpoint) WeakRefGone() { } // AbstractSocketNamespace is used to implement the Linux abstract socket functionality. +// +// +stateify savable type AbstractSocketNamespace struct { mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/auth/BUILD b/pkg/sentry/kernel/auth/BUILD index 5b7b30557..a81085372 100644 --- a/pkg/sentry/kernel/auth/BUILD +++ b/pkg/sentry/kernel/auth/BUILD @@ -1,20 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "auth_state", - srcs = [ - "credentials.go", - "id.go", - "id_map_range.go", - "id_map_set.go", - "user_namespace.go", - ], - out = "auth_state.go", - package = "auth", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_template_instance( name = "id_map_range", @@ -48,7 +35,6 @@ go_library( name = "auth", srcs = [ "auth.go", - "auth_state.go", "capability_set.go", "context.go", "credentials.go", @@ -66,7 +52,6 @@ go_library( "//pkg/bits", "//pkg/log", "//pkg/sentry/context", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go index f6fb05285..f18f7dac9 100644 --- a/pkg/sentry/kernel/auth/credentials.go +++ b/pkg/sentry/kernel/auth/credentials.go @@ -21,6 +21,8 @@ import ( // Credentials contains information required to authorize privileged operations // in a user namespace. +// +// +stateify savable type Credentials struct { // Real/effective/saved user/group IDs in the root user namespace. None of // these should ever be NoID. diff --git a/pkg/sentry/kernel/auth/id_map.go b/pkg/sentry/kernel/auth/id_map.go index 6adb33530..bd0090e0f 100644 --- a/pkg/sentry/kernel/auth/id_map.go +++ b/pkg/sentry/kernel/auth/id_map.go @@ -77,6 +77,8 @@ func (ns *UserNamespace) allIDsMapped(m *idMapSet, start, end uint32) bool { // An IDMapEntry represents a mapping from a range of contiguous IDs in a user // namespace to an equally-sized range of contiguous IDs in the namespace's // parent. +// +// +stateify savable type IDMapEntry struct { // FirstID is the first ID in the range in the namespace. FirstID uint32 diff --git a/pkg/sentry/kernel/auth/user_namespace.go b/pkg/sentry/kernel/auth/user_namespace.go index 0980aeadf..d359f3f31 100644 --- a/pkg/sentry/kernel/auth/user_namespace.go +++ b/pkg/sentry/kernel/auth/user_namespace.go @@ -23,6 +23,8 @@ import ( // A UserNamespace represents a user namespace. See user_namespaces(7) for // details. +// +// +stateify savable type UserNamespace struct { // parent is this namespace's parent. If this is the root namespace, parent // is nil. The parent pointer is immutable. diff --git a/pkg/sentry/kernel/epoll/BUILD b/pkg/sentry/kernel/epoll/BUILD index 7d491efbc..5e8b36ed6 100644 --- a/pkg/sentry/kernel/epoll/BUILD +++ b/pkg/sentry/kernel/epoll/BUILD @@ -1,22 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "epoll_autogen_state", - srcs = [ - "epoll.go", - "epoll_state.go", - ], - out = "epoll_autogen_state.go", - package = "epoll", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "epoll", srcs = [ "epoll.go", - "epoll_autogen_state.go", "epoll_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll", @@ -29,9 +18,7 @@ go_library( "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/kdefs", - "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/waiter", ], ) diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index b572fcd7e..d87e64a1c 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -58,6 +58,8 @@ const ( // potentially be reassigned. We also cannot use just the file pointer because // it is possible to have multiple entries for the same file object as long as // they are created with different FDs (i.e., the FDs point to the same file). +// +// +stateify savable type FileIdentifier struct { File *fs.File Fd kdefs.FD @@ -65,6 +67,8 @@ type FileIdentifier struct { // pollEntry holds all the state associated with an event poll entry, that is, // a file being observed by an event poll object. +// +// +stateify savable type pollEntry struct { ilist.Entry file *refs.WeakRef `state:"manual"` @@ -92,6 +96,8 @@ func (p *pollEntry) WeakRefGone() { // EventPoll holds all the state associated with an event poll object, that is, // collection of files to observe and their current state. +// +// +stateify savable type EventPoll struct { fsutil.PipeSeek `state:"zerovalue"` fsutil.NotDirReaddir `state:"zerovalue"` @@ -102,7 +108,7 @@ type EventPoll struct { // Wait queue is used to notify interested parties when the event poll // object itself becomes readable or writable. - waiter.Queue + waiter.Queue `state:"zerovalue"` // files is the map of all the files currently being observed, it is // protected by mu. diff --git a/pkg/sentry/kernel/eventfd/BUILD b/pkg/sentry/kernel/eventfd/BUILD index 7ec179bd8..cc1120b4f 100644 --- a/pkg/sentry/kernel/eventfd/BUILD +++ b/pkg/sentry/kernel/eventfd/BUILD @@ -1,33 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "eventfd_state", - srcs = [ - "eventfd.go", - ], - out = "eventfd_state.go", - package = "eventfd", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "eventfd", - srcs = [ - "eventfd.go", - "eventfd_state.go", - ], + srcs = ["eventfd.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/eventfd", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", "//pkg/waiter/fdnotifier", diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index bd50bd9fe..a4ada0e78 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -35,6 +35,8 @@ import ( // EventOperations represents an event with the semantics of Linux's file-based event // notification (eventfd). Eventfds are usually internal to the Sentry but in certain // situations they may be converted into a host-backed eventfd. +// +// +stateify savable type EventOperations struct { fsutil.NoopRelease `state:"nosave"` fsutil.PipeSeek `state:"nosave"` @@ -49,7 +51,7 @@ type EventOperations struct { // Queue is used to notify interested parties when the event object // becomes readable or writable. - wq waiter.Queue `state:"nosave"` + wq waiter.Queue `state:"zerovalue"` // val is the current value of the event counter. val uint64 diff --git a/pkg/sentry/kernel/fd_map.go b/pkg/sentry/kernel/fd_map.go index 299506330..d5d4aaacb 100644 --- a/pkg/sentry/kernel/fd_map.go +++ b/pkg/sentry/kernel/fd_map.go @@ -46,6 +46,8 @@ func (f FDs) Less(i, j int) bool { } // FDFlags define flags for an individual descriptor. +// +// +stateify savable type FDFlags struct { // CloseOnExec indicates the descriptor should be closed on exec. CloseOnExec bool @@ -69,12 +71,16 @@ func (f FDFlags) ToLinuxFDFlags() (mask uint) { // descriptor holds the details about a file descriptor, namely a pointer the // file itself and the descriptor flags. +// +// +stateify savable type descriptor struct { file *fs.File flags FDFlags } // FDMap is used to manage File references and flags. +// +// +stateify savable type FDMap struct { refs.AtomicRefCount k *Kernel diff --git a/pkg/sentry/kernel/fs_context.go b/pkg/sentry/kernel/fs_context.go index dbc097696..f3f05e8f5 100644 --- a/pkg/sentry/kernel/fs_context.go +++ b/pkg/sentry/kernel/fs_context.go @@ -25,6 +25,8 @@ import ( // FSContext contains filesystem context. // // This includes umask and working directory. +// +// +stateify savable type FSContext struct { refs.AtomicRefCount diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index a97a43549..b44a26974 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "waiter_list", @@ -14,29 +14,15 @@ go_template_instance( }, ) -go_stateify( - name = "futex_state", - srcs = [ - "futex.go", - "waiter_list.go", - ], - out = "futex_state.go", - package = "futex", -) - go_library( name = "futex", srcs = [ "futex.go", - "futex_state.go", "waiter_list.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex", visibility = ["//pkg/sentry:internal"], - deps = [ - "//pkg/state", - "//pkg/syserror", - ], + deps = ["//pkg/syserror"], ) go_test( diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index 15e3e5e2c..4a1f2a0ef 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -196,6 +196,8 @@ func bucketIndexForAddr(addr uintptr) uintptr { } // Manager holds futex state for a single virtual address space. +// +// +stateify savable type Manager struct { buckets [bucketCount]bucket `state:"zerovalue"` } diff --git a/pkg/sentry/kernel/ipc_namespace.go b/pkg/sentry/kernel/ipc_namespace.go index a86bda77b..5eef49f59 100644 --- a/pkg/sentry/kernel/ipc_namespace.go +++ b/pkg/sentry/kernel/ipc_namespace.go @@ -21,6 +21,8 @@ import ( ) // IPCNamespace represents an IPC namespace. +// +// +stateify savable type IPCNamespace struct { // User namespace which owns this IPC namespace. Immutable. userNS *auth.UserNamespace diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 64439cd9d..419a1d473 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -62,6 +62,8 @@ import ( // Kernel represents an emulated Linux kernel. It must be initialized by calling // Init() or LoadFrom(). +// +// +stateify savable type Kernel struct { // extMu serializes external changes to the Kernel with calls to // Kernel.SaveTo. (Kernel.SaveTo requires that the state of the Kernel @@ -158,7 +160,7 @@ type Kernel struct { // exitErr is the error causing the sandbox to exit, if any. It is // protected by extMu. - exitErr error + exitErr error `state:"nosave"` // danglingEndpoints is used to save / restore tcpip.DanglingEndpoints. danglingEndpoints struct{} `state:".([]tcpip.Endpoint)"` diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go index 5dc0f266c..06be5a7e1 100644 --- a/pkg/sentry/kernel/pending_signals.go +++ b/pkg/sentry/kernel/pending_signals.go @@ -38,6 +38,8 @@ const ( // pendingSignals holds a collection of pending signals. The zero value of // pendingSignals is a valid empty collection. pendingSignals is thread-unsafe; // users must provide synchronization. +// +// +stateify savable type pendingSignals struct { // signals contains all pending signals. // @@ -52,11 +54,14 @@ type pendingSignals struct { } // pendingSignalQueue holds a pendingSignalList for a single signal number. +// +// +stateify savable type pendingSignalQueue struct { pendingSignalList length int } +// +stateify savable type pendingSignal struct { // pendingSignalEntry links into a pendingSignalList. pendingSignalEntry diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD index 4600d19bd..19b23c6d2 100644 --- a/pkg/sentry/kernel/pipe/BUILD +++ b/pkg/sentry/kernel/pipe/BUILD @@ -1,20 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "pipe_state", - srcs = [ - "buffers.go", - "node.go", - "pipe.go", - "reader.go", - "reader_writer.go", - "writer.go", - ], - out = "pipe_state.go", - package = "pipe", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "pipe", @@ -23,7 +9,6 @@ go_library( "device.go", "node.go", "pipe.go", - "pipe_state.go", "reader.go", "reader_writer.go", "writer.go", @@ -34,15 +19,12 @@ go_library( "//pkg/abi/linux", "//pkg/amutex", "//pkg/ilist", - "//pkg/log", - "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/kernel/pipe/buffers.go b/pkg/sentry/kernel/pipe/buffers.go index f300537c5..a82e45c3f 100644 --- a/pkg/sentry/kernel/pipe/buffers.go +++ b/pkg/sentry/kernel/pipe/buffers.go @@ -20,6 +20,8 @@ import ( // Buffer encapsulates a queueable byte buffer that can // easily be truncated. It is designed only for use with pipes. +// +// +stateify savable type Buffer struct { ilist.Entry data []byte diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index e418cf174..23d692da1 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -24,6 +24,8 @@ import ( ) // inodeOperations wraps fs.InodeOperations operations with common pipe opening semantics. +// +// +stateify savable type inodeOperations struct { fs.InodeOperations diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index 9a21df5b4..ced2559a7 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -41,6 +41,8 @@ const DefaultPipeSize = 65536 // Pipe is an encapsulation of a platform-independent pipe. // It manages a buffered byte queue shared between a reader/writer // pair. +// +// +stateify savable type Pipe struct { waiter.Queue `state:"nosave"` diff --git a/pkg/sentry/kernel/pipe/reader.go b/pkg/sentry/kernel/pipe/reader.go index 40d5e4943..1fa5e9a32 100644 --- a/pkg/sentry/kernel/pipe/reader.go +++ b/pkg/sentry/kernel/pipe/reader.go @@ -20,6 +20,8 @@ import ( // Reader satisfies the fs.FileOperations interface for read-only pipes. // Reader should be used with !fs.FileFlags.Write to reject writes. +// +// +stateify savable type Reader struct { ReaderWriter } diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index dc642a3a6..82607367b 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -31,6 +31,8 @@ import ( // read and write requests. This should only be used directly for named pipes. // pipe(2) and pipe2(2) only support unidirectional pipes and should use // either pipe.Reader or pipe.Writer. +// +// +stateify savable type ReaderWriter struct { fsutil.PipeSeek `state:"nosave"` fsutil.NotDirReaddir `state:"nosave"` diff --git a/pkg/sentry/kernel/pipe/writer.go b/pkg/sentry/kernel/pipe/writer.go index fd13008ac..d93324b53 100644 --- a/pkg/sentry/kernel/pipe/writer.go +++ b/pkg/sentry/kernel/pipe/writer.go @@ -20,6 +20,8 @@ import ( // Writer satisfies the fs.FileOperations interface for write-only pipes. // Writer should be used with !fs.FileFlags.Read to reject reads. +// +// +stateify savable type Writer struct { ReaderWriter } diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index f1c2c4bf0..e9e69004d 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -25,6 +25,8 @@ import ( // ptraceOptions are the subset of options controlling a task's ptrace behavior // that are set by ptrace(PTRACE_SETOPTIONS). +// +// +stateify savable type ptraceOptions struct { // ExitKill is true if the tracee should be sent SIGKILL when the tracer // exits. @@ -185,6 +187,8 @@ func (t *Task) hasTracer() bool { } // ptraceStop is a TaskStop placed on tasks in a ptrace-stop. +// +// +stateify savable type ptraceStop struct { // If frozen is true, the stopped task's tracer is currently operating on // it, so Task.Kill should not remove the stop. diff --git a/pkg/sentry/kernel/rseq.go b/pkg/sentry/kernel/rseq.go index 635372993..1f3de58e3 100644 --- a/pkg/sentry/kernel/rseq.go +++ b/pkg/sentry/kernel/rseq.go @@ -23,6 +23,8 @@ import ( // Restartable sequences, as described in https://lwn.net/Articles/650333/. // RSEQCriticalRegion describes a restartable sequence critical region. +// +// +stateify savable type RSEQCriticalRegion struct { // When a task in this thread group has its CPU preempted (as defined by // platform.ErrContextCPUPreempted) or has a signal delivered to an diff --git a/pkg/sentry/kernel/semaphore/BUILD b/pkg/sentry/kernel/semaphore/BUILD index 969145fe1..e7fa44e2c 100644 --- a/pkg/sentry/kernel/semaphore/BUILD +++ b/pkg/sentry/kernel/semaphore/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "waiter_list", @@ -14,21 +14,10 @@ go_template_instance( }, ) -go_stateify( - name = "semaphore_state", - srcs = [ - "semaphore.go", - "waiter_list.go", - ], - out = "semaphore_autogen_state.go", - package = "semaphore", -) - go_library( name = "semaphore", srcs = [ "semaphore.go", - "semaphore_autogen_state.go", "waiter_list.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/semaphore", @@ -40,8 +29,6 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", - "//pkg/state", - "//pkg/state/statefile", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index a1ee83ce5..aa07946cf 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -42,6 +42,8 @@ const ( ) // Registry maintains a set of semaphores that can be found by key or ID. +// +// +stateify savable type Registry struct { // userNS owning the ipc name this registry belongs to. Immutable. userNS *auth.UserNamespace @@ -52,6 +54,8 @@ type Registry struct { } // Set represents a set of semaphores that can be operated atomically. +// +// +stateify savable type Set struct { // registry owning this sem set. Immutable. registry *Registry @@ -79,6 +83,8 @@ type Set struct { } // sem represents a single semanphore from a set. +// +// +stateify savable type sem struct { value int16 waiters waiterList `state:"zerovalue"` @@ -86,6 +92,8 @@ type sem struct { // waiter represents a caller that is waiting for the semaphore value to // become positive or zero. +// +// +stateify savable type waiter struct { waiterEntry diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index fa4c7b8f6..cf4e18805 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -27,6 +27,8 @@ type SessionID ThreadID type ProcessGroupID ThreadID // Session contains a leader threadgroup and a list of ProcessGroups. +// +// +stateify savable type Session struct { refs refs.AtomicRefCount @@ -76,6 +78,8 @@ func (s *Session) decRef() { } // ProcessGroup contains an originator threadgroup and a parent Session. +// +// +stateify savable type ProcessGroup struct { refs refs.AtomicRefCount // not exported. diff --git a/pkg/sentry/kernel/shm/BUILD b/pkg/sentry/kernel/shm/BUILD index 0f88eb0ac..40e641355 100644 --- a/pkg/sentry/kernel/shm/BUILD +++ b/pkg/sentry/kernel/shm/BUILD @@ -1,22 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "shm_state", - srcs = [ - "shm.go", - ], - out = "shm_autogen_state.go", - package = "shm", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "shm", srcs = [ "device.go", "shm.go", - "shm_autogen_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/shm", visibility = ["//pkg/sentry:internal"], @@ -33,7 +23,6 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 7217e8103..1ac444094 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -72,6 +72,8 @@ const ( // Registry tracks all shared memory segments in an IPC namespace. The registry // provides the mechanisms for creating and finding segments, and reporting // global shm parameters. +// +// +stateify savable type Registry struct { // userNS owns the IPC namespace this registry belong to. Immutable. userNS *auth.UserNamespace @@ -288,6 +290,8 @@ func (r *Registry) remove(s *Shm) { // shmctl(SHM_RMID). // // Shm implements memmap.Mappable and memmap.MappingIdentity. +// +// +stateify savable type Shm struct { // AtomicRefCount tracks the number of references to this segment from // maps. A segment always holds a reference to itself, until it's marked for diff --git a/pkg/sentry/kernel/signal_handlers.go b/pkg/sentry/kernel/signal_handlers.go index 21ba4ee70..3649f5e4d 100644 --- a/pkg/sentry/kernel/signal_handlers.go +++ b/pkg/sentry/kernel/signal_handlers.go @@ -22,6 +22,8 @@ import ( ) // SignalHandlers holds information about signal actions. +// +// +stateify savable type SignalHandlers struct { // mu protects actions, as well as the signal state of all tasks and thread // groups using this SignalHandlers object. (See comment on diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go index e20fa3eb6..4c7811b6c 100644 --- a/pkg/sentry/kernel/syscalls.go +++ b/pkg/sentry/kernel/syscalls.go @@ -176,6 +176,8 @@ type Stracer interface { // SyscallTable is a lookup table of system calls. Critically, a SyscallTable // is *immutable*. In order to make supporting suspend and resume sane, they // must be uniquely registered and may not change during operation. +// +// +stateify savable type SyscallTable struct { // OS is the operating system that this syscall table implements. OS abi.OS `state:"wait"` diff --git a/pkg/sentry/kernel/syslog.go b/pkg/sentry/kernel/syslog.go index 31541749e..125312b6a 100644 --- a/pkg/sentry/kernel/syslog.go +++ b/pkg/sentry/kernel/syslog.go @@ -23,6 +23,8 @@ import ( // syslog represents a sentry-global kernel log. // // Currently, it contains only fun messages for a dmesg easter egg. +// +// +stateify savable type syslog struct { // mu protects the below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index e705260da..19029adb1 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -52,6 +52,8 @@ import ( // All fields that are "exclusive to the task goroutine" can only be accessed // by the task goroutine while it is running. The task goroutine does not // require synchronization to read or write these fields. +// +// +stateify savable type Task struct { taskNode diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 3b77a4965..526165af0 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -352,6 +352,7 @@ func (t *Task) unstopVforkParent() { } } +// +stateify savable type runSyscallAfterPtraceEventClone struct { vforkChild *Task @@ -369,6 +370,7 @@ func (r *runSyscallAfterPtraceEventClone) execute(t *Task) taskRunState { return (*runSyscallExit)(nil) } +// +stateify savable type runSyscallAfterVforkStop struct { // childTID has the same meaning as // runSyscallAfterPtraceEventClone.vforkChildTID. @@ -474,6 +476,8 @@ func (t *Task) Unshare(opts *SharingOptions) error { // current MM. (Normally, CLONE_VFORK is used in conjunction with CLONE_VM, so // that the child and parent share mappings until the child execve()s into a // new process image or exits.) +// +// +stateify savable type vforkStop struct{} // StopIgnoresKill implements TaskStop.Killable. diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index 5c563ba08..9a59cbd33 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -35,6 +35,8 @@ var ErrNoSyscalls = errors.New("no syscall table found") type Auxmap map[string]interface{} // TaskContext is the subset of a task's data that is provided by the loader. +// +// +stateify savable type TaskContext struct { // Name is the thread name set by the prctl(PR_SET_NAME) system call. Name string diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index 2285847a2..385299b24 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -73,6 +73,8 @@ import ( // execStop is a TaskStop that a task sets on itself when it wants to execve // and is waiting for the other tasks in its thread group to exit first. +// +// +stateify savable type execStop struct{} // Killable implements TaskStop.Killable. @@ -119,6 +121,8 @@ func (t *Task) Execve(newTC *TaskContext) (*SyscallControl, error) { // The runSyscallAfterExecStop state continues execve(2) after all siblings of // a thread in the execve syscall have exited. +// +// +stateify savable type runSyscallAfterExecStop struct { tc *TaskContext } diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index d6604f37b..b16844e91 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -38,6 +38,8 @@ import ( // An ExitStatus is a value communicated from an exiting task or thread group // to the party that reaps it. +// +// +stateify savable type ExitStatus struct { // Code is the numeric value passed to the call to exit or exit_group that // caused the exit. If the exit was not caused by such a call, Code is 0. @@ -222,6 +224,8 @@ func (t *Task) advanceExitStateLocked(oldExit, newExit TaskExitState) { } // runExit is the entry point into the task exit path. +// +// +stateify savable type runExit struct{} func (*runExit) execute(t *Task) taskRunState { @@ -229,6 +233,7 @@ func (*runExit) execute(t *Task) taskRunState { return (*runExitMain)(nil) } +// +stateify savable type runExitMain struct{} func (*runExitMain) execute(t *Task) taskRunState { @@ -531,6 +536,7 @@ func (t *Task) reparentLocked(parent *Task) { // tracer (if one exists) and reaps the leader immediately. In Linux, this is // in fs/exec.c:de_thread(); in the sentry, this is in Task.promoteLocked(). +// +stateify savable type runExitNotify struct{} func (*runExitNotify) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_resources.go b/pkg/sentry/kernel/task_resources.go index 4ca25664a..0832bf989 100644 --- a/pkg/sentry/kernel/task_resources.go +++ b/pkg/sentry/kernel/task_resources.go @@ -21,6 +21,8 @@ import ( // TaskResources is the subset of a task's data provided by its creator that is // not provided by the loader. +// +// +stateify savable type TaskResources struct { // SignalMask is the set of signals whose delivery is currently blocked. // diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index a03fa6ac0..8dd0ef6ea 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -131,6 +131,8 @@ func (t *Task) doStop() { // The runApp state checks for interrupts before executing untrusted // application code. +// +// +stateify savable type runApp struct{} func (*runApp) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_sched.go b/pkg/sentry/kernel/task_sched.go index b50139077..49141ab74 100644 --- a/pkg/sentry/kernel/task_sched.go +++ b/pkg/sentry/kernel/task_sched.go @@ -65,6 +65,8 @@ const ( // TaskGoroutineSchedInfo contains task goroutine scheduling state which must // be read and updated atomically. +// +// +stateify savable type TaskGoroutineSchedInfo struct { // Timestamp was the value of Kernel.cpuClock when this // TaskGoroutineSchedInfo was last updated. diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 91f6c0874..62ec530be 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -748,6 +748,8 @@ func (t *Task) CopyInSignalStack(addr usermem.Addr) (arch.SignalStack, error) { // groupStop is a TaskStop placed on tasks that have received a stop signal // (SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU). (The term "group-stop" originates from // the ptrace man page.) +// +// +stateify savable type groupStop struct{} // Killable implements TaskStop.Killable. @@ -881,6 +883,8 @@ func (t *Task) signalStop(target *Task, code int32, status int32) { } // The runInterrupt state handles conditions indicated by interrupts. +// +// +stateify savable type runInterrupt struct{} func (*runInterrupt) execute(t *Task) taskRunState { @@ -1020,6 +1024,7 @@ func (*runInterrupt) execute(t *Task) taskRunState { return (*runApp)(nil) } +// +stateify savable type runInterruptAfterSignalDeliveryStop struct{} func (*runInterruptAfterSignalDeliveryStop) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index 92ca0acd9..f0373c375 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -241,6 +241,7 @@ func (t *Task) doSyscallEnter(sysno uintptr, args arch.SyscallArguments) taskRun return t.doSyscallInvoke(sysno, args) } +// +stateify savable type runSyscallAfterSyscallEnterStop struct{} func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState { @@ -260,6 +261,7 @@ func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState { return t.doSyscallInvoke(sysno, args) } +// +stateify savable type runSyscallAfterSysemuStop struct{} func (*runSyscallAfterSysemuStop) execute(t *Task) taskRunState { @@ -294,6 +296,7 @@ func (t *Task) doSyscallInvoke(sysno uintptr, args arch.SyscallArguments) taskRu return (*runSyscallExit)(nil).execute(t) } +// +stateify savable type runSyscallReinvoke struct{} func (*runSyscallReinvoke) execute(t *Task) taskRunState { @@ -310,6 +313,7 @@ func (*runSyscallReinvoke) execute(t *Task) taskRunState { return t.doSyscallInvoke(sysno, args) } +// +stateify savable type runSyscallExit struct{} func (*runSyscallExit) execute(t *Task) taskRunState { diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index 8fffd3446..441b8a822 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -28,6 +28,8 @@ import ( // groups" are usually called "processes" in userspace documentation.) // // ThreadGroup is a superset of Linux's struct signal_struct. +// +// +stateify savable type ThreadGroup struct { threadGroupNode diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index 440da9dad..844213c35 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -50,6 +50,8 @@ func (tid ThreadID) String() string { const InitTID ThreadID = 1 // A TaskSet comprises all tasks in a system. +// +// +stateify savable type TaskSet struct { // mu protects all relationships betweens tasks and thread groups in the // TaskSet. (mu is approximately equivalent to Linux's tasklist_lock.) @@ -110,6 +112,8 @@ func (ts *TaskSet) forEachThreadGroupLocked(f func(tg *ThreadGroup)) { // // N.B. A task is said to be visible in a PID namespace if the PID namespace // contains a thread ID that maps to that task. +// +// +stateify savable type PIDNamespace struct { // owner is the TaskSet that this PID namespace belongs to. The owner // pointer is immutable. @@ -263,6 +267,8 @@ func (ns *PIDNamespace) UserNamespace() *auth.UserNamespace { // (threadGroupNode is an anonymous field in ThreadGroup; this is to expose // threadGroupEntry's methods on ThreadGroup to make it implement // threadGroupLinker.) +// +// +stateify savable type threadGroupNode struct { // pidns is the PID namespace containing the thread group and all of its // member tasks. The pidns pointer is immutable. @@ -382,6 +388,8 @@ func (tg *ThreadGroup) ID() ThreadID { // A taskNode defines the relationship between a task and the rest of the // system. The comments on threadGroupNode also apply to taskNode. +// +// +stateify savable type taskNode struct { // tg is the thread group that this task belongs to. The tg pointer is // immutable. diff --git a/pkg/sentry/kernel/time/BUILD b/pkg/sentry/kernel/time/BUILD index b3ed42aa4..5d8db2273 100644 --- a/pkg/sentry/kernel/time/BUILD +++ b/pkg/sentry/kernel/time/BUILD @@ -1,30 +1,18 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "time_state", - srcs = [ - "time.go", - ], - out = "time_state.go", - package = "time", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "time", srcs = [ "context.go", "time.go", - "time_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", - "//pkg/log", "//pkg/sentry/context", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go index c223c2f19..6eadd2878 100644 --- a/pkg/sentry/kernel/time/time.go +++ b/pkg/sentry/kernel/time/time.go @@ -42,6 +42,8 @@ const ( // // Time may represent time with respect to any clock and may not have any // meaning in the real world. +// +// +stateify savable type Time struct { ns int64 } @@ -286,6 +288,8 @@ type TimerListener interface { } // Setting contains user-controlled mutable Timer properties. +// +// +stateify savable type Setting struct { // Enabled is true if the timer is running. Enabled bool @@ -371,6 +375,8 @@ func (s Setting) advancedTo(now Time) (Setting, uint64) { // // Timers should be created using NewTimer and must be cleaned up by calling // Timer.Destroy when no longer used. +// +// +stateify savable type Timer struct { // clock is the time source. clock is immutable. clock Clock diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go index 4de8ac13b..df5dbe128 100644 --- a/pkg/sentry/kernel/timekeeper.go +++ b/pkg/sentry/kernel/timekeeper.go @@ -25,6 +25,8 @@ import ( ) // Timekeeper manages all of the kernel clocks. +// +// +stateify savable type Timekeeper struct { // clocks are the clock sources. // diff --git a/pkg/sentry/kernel/timer.go b/pkg/sentry/kernel/timer.go index 03a3310be..534d03d0f 100644 --- a/pkg/sentry/kernel/timer.go +++ b/pkg/sentry/kernel/timer.go @@ -26,6 +26,8 @@ import ( // timekeeperClock is a ktime.Clock that reads time from a // kernel.Timekeeper-managed clock. +// +// +stateify savable type timekeeperClock struct { tk *Timekeeper c sentrytime.ClockID @@ -49,6 +51,8 @@ func (tc *timekeeperClock) Now() ktime.Time { // tgClock is a ktime.Clock that measures the time a thread group has spent // executing. +// +// +stateify savable type tgClock struct { tg *ThreadGroup @@ -155,6 +159,8 @@ func (tc *taskClock) Now() ktime.Time { } // signalNotifier is a ktime.Listener that sends signals to a ThreadGroup. +// +// +stateify savable type signalNotifier struct { tg *ThreadGroup signal linux.Signal @@ -179,6 +185,8 @@ func (s *signalNotifier) Notify(exp uint64) { func (s *signalNotifier) Destroy() {} // TimerManager is a collection of supported process cpu timers. +// +// +stateify savable type TimerManager struct { // Clocks used to drive thread group execution time timers. virtClock *tgClock diff --git a/pkg/sentry/kernel/uts_namespace.go b/pkg/sentry/kernel/uts_namespace.go index 58e9b4d1b..7e0fe0d21 100644 --- a/pkg/sentry/kernel/uts_namespace.go +++ b/pkg/sentry/kernel/uts_namespace.go @@ -22,6 +22,8 @@ import ( // UTSNamespace represents a UTS namespace, a holder of two system identifiers: // the hostname and domain name. +// +// +stateify savable type UTSNamespace struct { // mu protects all fields below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/vdso.go b/pkg/sentry/kernel/vdso.go index 0bacbea49..971e8bc59 100644 --- a/pkg/sentry/kernel/vdso.go +++ b/pkg/sentry/kernel/vdso.go @@ -52,6 +52,8 @@ type vdsoParams struct { // Everything in the struct is 8 bytes for easy alignment. // // It must be kept in sync with params in vdso/vdso_time.cc. +// +// +stateify savable type VDSOParamPage struct { // The parameter page is fr, allocated from platform.Memory(). platform platform.Platform diff --git a/pkg/sentry/limits/BUILD b/pkg/sentry/limits/BUILD index 3ce41cacc..90f4395d4 100644 --- a/pkg/sentry/limits/BUILD +++ b/pkg/sentry/limits/BUILD @@ -1,22 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "limits_state", - srcs = [ - "limits.go", - ], - out = "limits_state.go", - package = "limits", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "limits", srcs = [ "context.go", "limits.go", - "limits_state.go", "linux.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/limits", @@ -24,7 +14,6 @@ go_library( deps = [ "//pkg/abi/linux", "//pkg/sentry/context", - "//pkg/state", ], ) diff --git a/pkg/sentry/limits/limits.go b/pkg/sentry/limits/limits.go index 4230ba958..02c8b60e3 100644 --- a/pkg/sentry/limits/limits.go +++ b/pkg/sentry/limits/limits.go @@ -47,6 +47,8 @@ const ( const Infinity = ^uint64(0) // Limit specifies a system limit. +// +// +stateify savable type Limit struct { // Cur specifies the current limit. Cur uint64 @@ -55,6 +57,8 @@ type Limit struct { } // LimitSet represents the Limits that correspond to each LimitType. +// +// +stateify savable type LimitSet struct { mu sync.Mutex `state:"nosave"` data map[LimitType]Limit diff --git a/pkg/sentry/loader/BUILD b/pkg/sentry/loader/BUILD index e63052c6d..0beb4561b 100644 --- a/pkg/sentry/loader/BUILD +++ b/pkg/sentry/loader/BUILD @@ -1,7 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("@io_bazel_rules_go//go:def.bzl", "go_embed_data") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") +load("//tools/go_stateify:defs.bzl", "go_library") go_embed_data( name = "vdso_bin", @@ -10,23 +10,12 @@ go_embed_data( var = "vdsoBin", ) -go_stateify( - name = "loader_state", - srcs = [ - "vdso.go", - "vdso_state.go", - ], - out = "loader_state.go", - package = "loader", -) - go_library( name = "loader", srcs = [ "elf.go", "interpreter.go", "loader.go", - "loader_state.go", "vdso.go", "vdso_state.go", ":vdso_bin", @@ -40,7 +29,6 @@ go_library( "//pkg/cpuid", "//pkg/log", "//pkg/rand", - "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/fs", @@ -55,7 +43,6 @@ go_library( "//pkg/sentry/uniqueid", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index 2e8693f8e..a06e27ac9 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -193,6 +193,8 @@ func validateVDSO(ctx context.Context, f *fs.File, size uint64) (elfInfo, error) // // NOTE: to support multiple architectures or operating systems, this // would need to contain a VDSO for each. +// +// +stateify savable type VDSO struct { // ParamPage is the VDSO parameter page. This page should be updated to // inform the VDSO for timekeeping data. diff --git a/pkg/sentry/loader/vdso_state.go b/pkg/sentry/loader/vdso_state.go index 92004ad9e..dc71e1c2d 100644 --- a/pkg/sentry/loader/vdso_state.go +++ b/pkg/sentry/loader/vdso_state.go @@ -18,6 +18,7 @@ import ( "debug/elf" ) +// +stateify savable type elfProgHeader struct { Type elf.ProgType Flags elf.ProgFlag diff --git a/pkg/sentry/memmap/BUILD b/pkg/sentry/memmap/BUILD index 2e367e189..c9e0b95a0 100644 --- a/pkg/sentry/memmap/BUILD +++ b/pkg/sentry/memmap/BUILD @@ -1,18 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "memmap_state", - srcs = [ - "mappable_range.go", - "mapping_set.go", - "mapping_set_impl.go", - ], - out = "memmap_state.go", - package = "memmap", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "mappable_range", @@ -46,7 +35,6 @@ go_library( "mapping_set.go", "mapping_set_impl.go", "memmap.go", - "memmap_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/memmap", visibility = ["//pkg/sentry:internal"], @@ -56,7 +44,6 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/platform", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/memmap/mapping_set.go b/pkg/sentry/memmap/mapping_set.go index 0cd42ffbf..c9483905d 100644 --- a/pkg/sentry/memmap/mapping_set.go +++ b/pkg/sentry/memmap/mapping_set.go @@ -35,6 +35,8 @@ import ( type MappingsOfRange map[MappingOfRange]struct{} // MappingOfRange represents a mapping of a MappableRange. +// +// +stateify savable type MappingOfRange struct { MappingSpace MappingSpace AddrRange usermem.AddrRange diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index 3f396986a..bbdfae247 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -1,24 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "mm_state", - srcs = [ - "aio_context.go", - "aio_context_state.go", - "file_refcount_set.go", - "io_list.go", - "mm.go", - "pma_set.go", - "save_restore.go", - "special_mappable.go", - "vma_set.go", - ], - out = "mm_state.go", - package = "mm", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "file_refcount_set", @@ -101,7 +84,6 @@ go_library( "lifecycle.go", "metadata.go", "mm.go", - "mm_state.go", "pma.go", "pma_set.go", "proc_pid_maps.go", @@ -131,7 +113,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/sync", "//pkg/syserror", "//pkg/tcpip/buffer", diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index 992bde5a5..b42156d45 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -28,6 +28,8 @@ import ( ) // aioManager creates and manages asynchronous I/O contexts. +// +// +stateify savable type aioManager struct { // mu protects below. mu sync.Mutex `state:"nosave"` @@ -89,12 +91,16 @@ func (a *aioManager) lookupAIOContext(id uint64) (*AIOContext, bool) { } // ioResult is a completed I/O operation. +// +// +stateify savable type ioResult struct { data interface{} ioEntry } // AIOContext is a single asynchronous I/O context. +// +// +stateify savable type AIOContext struct { // done is the notification channel used for all requests. done chan struct{} `state:"nosave"` @@ -190,6 +196,8 @@ func (ctx *AIOContext) WaitChannel() (chan struct{}, bool) { // aioMappable implements memmap.MappingIdentity and memmap.Mappable for AIO // ring buffers. +// +// +stateify savable type aioMappable struct { refs.AtomicRefCount diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index ce8097b7f..3299ae164 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -46,6 +46,8 @@ import ( ) // MemoryManager implements a virtual address space. +// +// +stateify savable type MemoryManager struct { // p is the platform. // @@ -207,6 +209,8 @@ type MemoryManager struct { } // vma represents a virtual memory area. +// +// +stateify savable type vma struct { // mappable is the virtual memory object mapped by this vma. If mappable is // nil, the vma represents a private anonymous mapping. @@ -346,6 +350,8 @@ func (v *vma) loadRealPerms(b int) { } // pma represents a platform mapping area. +// +// +stateify savable type pma struct { // file is the file mapped by this pma. Only pmas for which file == // platform.Platform.Memory() may be saved. pmas hold a reference to the @@ -380,6 +386,7 @@ type pma struct { internalMappings safemem.BlockSeq `state:"nosave"` } +// +stateify savable type privateRefs struct { mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index 9d3614034..aa2f87107 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -28,6 +28,8 @@ import ( // semantics similar to Linux's mm/mmap.c:_install_special_mapping(), except // that SpecialMappable takes ownership of the memory that it represents // (_install_special_mapping() does not.) +// +// +stateify savable type SpecialMappable struct { refs.AtomicRefCount diff --git a/pkg/sentry/platform/BUILD b/pkg/sentry/platform/BUILD index 15a7fbbc3..af9ba5394 100644 --- a/pkg/sentry/platform/BUILD +++ b/pkg/sentry/platform/BUILD @@ -1,16 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "platform_state", - srcs = [ - "file_range.go", - ], - out = "platform_state.go", - package = "platform", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_template_instance( name = "file_range", @@ -30,7 +21,6 @@ go_library( "file_range.go", "mmap_min_addr.go", "platform.go", - "platform_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/platform", visibility = ["//pkg/sentry:internal"], @@ -44,7 +34,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", ], ) diff --git a/pkg/sentry/platform/filemem/BUILD b/pkg/sentry/platform/filemem/BUILD index dadba1d38..2a5982763 100644 --- a/pkg/sentry/platform/filemem/BUILD +++ b/pkg/sentry/platform/filemem/BUILD @@ -1,18 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "filemem_autogen_state", - srcs = [ - "filemem.go", - "filemem_state.go", - "usage_set.go", - ], - out = "filemem_autogen_state.go", - package = "filemem", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "usage_set", @@ -38,7 +27,6 @@ go_library( name = "filemem", srcs = [ "filemem.go", - "filemem_autogen_state.go", "filemem_state.go", "filemem_unsafe.go", "usage_set.go", diff --git a/pkg/sentry/platform/filemem/filemem.go b/pkg/sentry/platform/filemem/filemem.go index 870274ae1..feb020ef8 100644 --- a/pkg/sentry/platform/filemem/filemem.go +++ b/pkg/sentry/platform/filemem/filemem.go @@ -155,6 +155,8 @@ type FileMem struct { } // usage tracks usage information. +// +// +stateify savable type usageInfo struct { // kind is the usage kind. kind usage.MemoryKind diff --git a/pkg/sentry/socket/BUILD b/pkg/sentry/socket/BUILD index 929787aa0..a320fca0b 100644 --- a/pkg/sentry/socket/BUILD +++ b/pkg/sentry/socket/BUILD @@ -1,22 +1,10 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "socket_state", - srcs = [ - "socket.go", - ], - out = "socket_state_autogen.go", - package = "socket", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "socket", - srcs = [ - "socket.go", - "socket_state_autogen.go", - ], + srcs = ["socket.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket", visibility = ["//pkg/sentry:internal"], deps = [ @@ -29,7 +17,6 @@ go_library( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/tcpip", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/socket/control/BUILD b/pkg/sentry/socket/control/BUILD index faf2b4c27..c4874fdfb 100644 --- a/pkg/sentry/socket/control/BUILD +++ b/pkg/sentry/socket/control/BUILD @@ -1,26 +1,14 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "control_state", - srcs = [ - "control.go", - ], - out = "control_state.go", - imports = [ - "gvisor.googlesource.com/gvisor/pkg/sentry/fs", - ], - package = "control", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "control", - srcs = [ - "control.go", - "control_state.go", - ], + srcs = ["control.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control", + imports = [ + "gvisor.googlesource.com/gvisor/pkg/sentry/fs", + ], visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", @@ -31,7 +19,6 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/kdefs", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/transport/unix", ], diff --git a/pkg/sentry/socket/control/control.go b/pkg/sentry/socket/control/control.go index 17ecdd11c..c31182e69 100644 --- a/pkg/sentry/socket/control/control.go +++ b/pkg/sentry/socket/control/control.go @@ -51,6 +51,8 @@ type SCMRights interface { // RightsFiles represents a SCM_RIGHTS socket control message. A reference is // maintained for each fs.File and is release either when an FD is created or // when the Release method is called. +// +// +stateify savable type RightsFiles []*fs.File // NewSCMRights creates a new SCM_RIGHTS socket control message representation @@ -128,6 +130,8 @@ func PackRights(t *kernel.Task, rights SCMRights, cloexec bool, buf []byte) []by } // scmCredentials represents an SCM_CREDENTIALS socket control message. +// +// +stateify savable type scmCredentials struct { t *kernel.Task kuid auth.KUID diff --git a/pkg/sentry/socket/epsocket/BUILD b/pkg/sentry/socket/epsocket/BUILD index 7ad5e88c5..49af8db85 100644 --- a/pkg/sentry/socket/epsocket/BUILD +++ b/pkg/sentry/socket/epsocket/BUILD @@ -1,24 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "epsocket_state", - srcs = [ - "epsocket.go", - "save_restore.go", - "stack.go", - ], - out = "epsocket_state.go", - package = "epsocket", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "epsocket", srcs = [ "device.go", "epsocket.go", - "epsocket_state.go", "provider.go", "save_restore.go", "stack.go", @@ -31,7 +19,6 @@ go_library( "//pkg/abi/linux", "//pkg/binary", "//pkg/log", - "//pkg/refs", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/device", @@ -44,7 +31,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index a2927e1b9..f969a1d7c 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -95,6 +95,8 @@ type commonEndpoint interface { // SocketOperations encapsulates all the state needed to represent a network stack // endpoint in the kernel context. +// +// +stateify savable type SocketOperations struct { socket.ReceiveTimeout fsutil.PipeSeek `state:"nosave"` diff --git a/pkg/sentry/socket/epsocket/stack.go b/pkg/sentry/socket/epsocket/stack.go index ec1d96ccb..12b4b4767 100644 --- a/pkg/sentry/socket/epsocket/stack.go +++ b/pkg/sentry/socket/epsocket/stack.go @@ -26,6 +26,8 @@ import ( ) // Stack implements inet.Stack for netstack/tcpip/stack.Stack. +// +// +stateify savable type Stack struct { Stack *stack.Stack `state:"manual"` } diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD index 227ca3926..d623718b3 100644 --- a/pkg/sentry/socket/hostinet/BUILD +++ b/pkg/sentry/socket/hostinet/BUILD @@ -1,24 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "hostinet_state", - srcs = [ - "save_restore.go", - "socket.go", - "stack.go", - ], - out = "hostinet_autogen_state.go", - package = "hostinet", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "hostinet", srcs = [ "device.go", "hostinet.go", - "hostinet_autogen_state.go", "save_restore.go", "socket.go", "socket_unsafe.go", @@ -42,7 +30,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip/transport/unix", diff --git a/pkg/sentry/socket/netlink/BUILD b/pkg/sentry/socket/netlink/BUILD index b23a243f7..b852165f7 100644 --- a/pkg/sentry/socket/netlink/BUILD +++ b/pkg/sentry/socket/netlink/BUILD @@ -1,21 +1,11 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "netlink_state", - srcs = [ - "socket.go", - ], - out = "netlink_state.go", - package = "netlink", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "netlink", srcs = [ "message.go", - "netlink_state.go", "provider.go", "socket.go", ], @@ -36,7 +26,6 @@ go_library( "//pkg/sentry/socket/netlink/port", "//pkg/sentry/socket/unix", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/netlink/port/BUILD b/pkg/sentry/socket/netlink/port/BUILD index ba6f686e4..3a7dbc5ed 100644 --- a/pkg/sentry/socket/netlink/port/BUILD +++ b/pkg/sentry/socket/netlink/port/BUILD @@ -1,23 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "port_state", - srcs = ["port.go"], - out = "port_state.go", - package = "port", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "port", - srcs = [ - "port.go", - "port_state.go", - ], + srcs = ["port.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/port", visibility = ["//pkg/sentry:internal"], - deps = ["//pkg/state"], ) go_test( diff --git a/pkg/sentry/socket/netlink/port/port.go b/pkg/sentry/socket/netlink/port/port.go index 4ccf0b84c..1c5d4c3a5 100644 --- a/pkg/sentry/socket/netlink/port/port.go +++ b/pkg/sentry/socket/netlink/port/port.go @@ -32,6 +32,8 @@ import ( const maxPorts = 10000 // Manager allocates netlink port IDs. +// +// +stateify savable type Manager struct { // mu protects the fields below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/socket/netlink/route/BUILD b/pkg/sentry/socket/netlink/route/BUILD index 726469fc9..e1bcfe252 100644 --- a/pkg/sentry/socket/netlink/route/BUILD +++ b/pkg/sentry/socket/netlink/route/BUILD @@ -1,32 +1,19 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "route_state", - srcs = ["protocol.go"], - out = "route_state.go", - package = "route", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "route", - srcs = [ - "protocol.go", - "route_state.go", - ], + srcs = ["protocol.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/route", visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", "//pkg/sentry/context", - "//pkg/sentry/fs", "//pkg/sentry/inet", "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/socket/netlink", - "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", ], ) diff --git a/pkg/sentry/socket/netlink/route/protocol.go b/pkg/sentry/socket/netlink/route/protocol.go index e8030c518..55a76e916 100644 --- a/pkg/sentry/socket/netlink/route/protocol.go +++ b/pkg/sentry/socket/netlink/route/protocol.go @@ -43,6 +43,8 @@ func typeKind(typ uint16) commandKind { } // Protocol implements netlink.Protocol. +// +// +stateify savable type Protocol struct{} var _ netlink.Protocol = (*Protocol)(nil) diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index 0b8f528d0..e15d1546c 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -51,6 +51,8 @@ var netlinkSocketDevice = device.NewAnonDevice() // to/from the kernel. // // Socket implements socket.Socket. +// +// +stateify savable type Socket struct { socket.ReceiveTimeout fsutil.PipeSeek `state:"nosave"` diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index bd4858a34..54fe64595 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -195,6 +195,8 @@ func NewDirent(ctx context.Context, d *device.Device) *fs.Dirent { // // Care must be taken when copying ReceiveTimeout as it contains atomic // variables. +// +// +stateify savable type ReceiveTimeout struct { // ns is length of the timeout in nanoseconds. // diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index 7d04d6b6b..9fe681e9a 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -1,15 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "unix_state", - srcs = [ - "unix.go", - ], - out = "unix_state.go", - package = "unix", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "unix", @@ -17,7 +8,6 @@ go_library( "device.go", "io.go", "unix.go", - "unix_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix", visibility = ["//pkg/sentry:internal"], @@ -37,7 +27,6 @@ go_library( "//pkg/sentry/socket/control", "//pkg/sentry/socket/epsocket", "//pkg/sentry/usermem", - "//pkg/state", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 27bacbbc3..5b6411f97 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -42,6 +42,8 @@ import ( // SocketOperations is a Unix socket. It is similar to an epsocket, except it is backed // by a unix.Endpoint instead of a tcpip.Endpoint. +// +// +stateify savable type SocketOperations struct { refs.AtomicRefCount socket.ReceiveTimeout diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index 32fca3811..bbdfad9da 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -1,18 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "linux_state", - srcs = [ - "sys_aio.go", - "sys_futex.go", - "sys_poll.go", - "sys_time.go", - ], - out = "linux_state.go", - package = "linux", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "linux", @@ -20,7 +8,6 @@ go_library( "error.go", "flags.go", "linux64.go", - "linux_state.go", "sigset.go", "sys_aio.go", "sys_capability.go", @@ -67,7 +54,6 @@ go_library( "//pkg/abi/linux", "//pkg/binary", "//pkg/bpf", - "//pkg/eventchannel", "//pkg/log", "//pkg/metric", "//pkg/rand", @@ -75,7 +61,6 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", - "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/lock", "//pkg/sentry/fs/timerfd", "//pkg/sentry/kernel", @@ -86,7 +71,6 @@ go_library( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/pipe", "//pkg/sentry/kernel/sched", - "//pkg/sentry/kernel/semaphore", "//pkg/sentry/kernel/shm", "//pkg/sentry/kernel/time", "//pkg/sentry/limits", @@ -98,8 +82,6 @@ go_library( "//pkg/sentry/syscalls", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/state", - "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip/transport/unix", "//pkg/waiter", diff --git a/pkg/sentry/syscalls/linux/sys_aio.go b/pkg/sentry/syscalls/linux/sys_aio.go index fc3397081..54e4afa9e 100644 --- a/pkg/sentry/syscalls/linux/sys_aio.go +++ b/pkg/sentry/syscalls/linux/sys_aio.go @@ -69,6 +69,8 @@ type ioCallback struct { } // ioEvent describes an I/O result. +// +// +stateify savable type ioEvent struct { Data uint64 Obj uint64 diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index 57762d058..1a0e1f5fb 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -132,6 +132,8 @@ func (f futexChecker) Op(addr uintptr, opIn uint32) (bool, error) { // futexWaitRestartBlock encapsulates the state required to restart futex(2) // via restart_syscall(2). +// +// +stateify savable type futexWaitRestartBlock struct { duration time.Duration diff --git a/pkg/sentry/syscalls/linux/sys_poll.go b/pkg/sentry/syscalls/linux/sys_poll.go index d4dbfd285..b9bdefadb 100644 --- a/pkg/sentry/syscalls/linux/sys_poll.go +++ b/pkg/sentry/syscalls/linux/sys_poll.go @@ -274,6 +274,8 @@ func copyOutTimevalRemaining(t *kernel.Task, startNs ktime.Time, timeout time.Du // pollRestartBlock encapsulates the state required to restart poll(2) via // restart_syscall(2). +// +// +stateify savable type pollRestartBlock struct { pfdAddr usermem.Addr nfds uint diff --git a/pkg/sentry/syscalls/linux/sys_time.go b/pkg/sentry/syscalls/linux/sys_time.go index dcee694b2..8e6683444 100644 --- a/pkg/sentry/syscalls/linux/sys_time.go +++ b/pkg/sentry/syscalls/linux/sys_time.go @@ -168,6 +168,8 @@ func Time(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallC // clockNanosleepRestartBlock encapsulates the state required to restart // clock_nanosleep(2) via restart_syscall(2). +// +// +stateify savable type clockNanosleepRestartBlock struct { c ktime.Clock duration time.Duration diff --git a/pkg/sentry/usage/BUILD b/pkg/sentry/usage/BUILD index edee44d96..868dfd400 100644 --- a/pkg/sentry/usage/BUILD +++ b/pkg/sentry/usage/BUILD @@ -1,17 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "usage_state", - srcs = [ - "cpu.go", - "io.go", - "memory.go", - ], - out = "usage_state.go", - package = "usage", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "usage", @@ -21,7 +10,6 @@ go_library( "memory.go", "memory_unsafe.go", "usage.go", - "usage_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/usage", visibility = [ @@ -29,9 +17,6 @@ go_library( ], deps = [ "//pkg/bits", - "//pkg/log", "//pkg/sentry/memutil", - "//pkg/state", - "@org_golang_x_sys//unix:go_default_library", ], ) diff --git a/pkg/sentry/usage/cpu.go b/pkg/sentry/usage/cpu.go index 1c2cc90e1..ed7b04b9e 100644 --- a/pkg/sentry/usage/cpu.go +++ b/pkg/sentry/usage/cpu.go @@ -20,6 +20,8 @@ import ( // CPUStats contains the subset of struct rusage fields that relate to CPU // scheduling. +// +// +stateify savable type CPUStats struct { // UserTime is the amount of time spent executing application code. UserTime time.Duration diff --git a/pkg/sentry/usage/io.go b/pkg/sentry/usage/io.go index a05053c32..49faa507d 100644 --- a/pkg/sentry/usage/io.go +++ b/pkg/sentry/usage/io.go @@ -19,6 +19,8 @@ import ( ) // IO contains I/O-related statistics. +// +// +stateify savable type IO struct { // CharsRead is the number of bytes read by read syscalls. CharsRead uint64 diff --git a/pkg/sentry/usermem/BUILD b/pkg/sentry/usermem/BUILD index 9dd1cd2b5..69ba919e0 100644 --- a/pkg/sentry/usermem/BUILD +++ b/pkg/sentry/usermem/BUILD @@ -1,19 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "usermem_state", - srcs = [ - "access_type.go", - "addr.go", - "addr_range.go", - "addr_range_seq_unsafe.go", - ], - out = "usermem_state.go", - package = "usermem", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "addr_range", @@ -36,7 +24,6 @@ go_library( "bytes_io.go", "bytes_io_unsafe.go", "usermem.go", - "usermem_state.go", "usermem_x86.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/usermem", @@ -47,7 +34,6 @@ go_library( "//pkg/log", "//pkg/sentry/context", "//pkg/sentry/safemem", - "//pkg/state", "//pkg/syserror", "//pkg/tcpip/buffer", ], diff --git a/pkg/sentry/usermem/access_type.go b/pkg/sentry/usermem/access_type.go index 7eabecf30..75346d854 100644 --- a/pkg/sentry/usermem/access_type.go +++ b/pkg/sentry/usermem/access_type.go @@ -20,6 +20,8 @@ import ( // AccessType specifies memory access types. This is used for // setting mapping permissions, as well as communicating faults. +// +// +stateify savable type AccessType struct { // Read is read access. Read bool diff --git a/pkg/sentry/usermem/addr.go b/pkg/sentry/usermem/addr.go index d175fdc74..fc94bee80 100644 --- a/pkg/sentry/usermem/addr.go +++ b/pkg/sentry/usermem/addr.go @@ -19,6 +19,8 @@ import ( ) // Addr represents a generic virtual address. +// +// +stateify savable type Addr uintptr // AddLength adds the given length to start and returns the result. ok is true diff --git a/pkg/tcpip/BUILD b/pkg/tcpip/BUILD index 391d801d0..5153bd3b4 100644 --- a/pkg/tcpip/BUILD +++ b/pkg/tcpip/BUILD @@ -1,26 +1,13 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tcpip_state", - srcs = [ - "tcpip.go", - ], - out = "tcpip_state.go", - package = "tcpip", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tcpip", - srcs = [ - "tcpip.go", - "tcpip_state.go", - ], + srcs = ["tcpip.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip", visibility = ["//visibility:public"], deps = [ - "//pkg/state", "//pkg/tcpip/buffer", "//pkg/waiter", ], diff --git a/pkg/tcpip/buffer/BUILD b/pkg/tcpip/buffer/BUILD index efeb6a448..11a725423 100644 --- a/pkg/tcpip/buffer/BUILD +++ b/pkg/tcpip/buffer/BUILD @@ -1,26 +1,15 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "buffer_state", - srcs = [ - "view.go", - ], - out = "buffer_state.go", - package = "buffer", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "buffer", srcs = [ - "buffer_state.go", "prependable.go", "view.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer", visibility = ["//visibility:public"], - deps = ["//pkg/state"], ) go_test( diff --git a/pkg/tcpip/buffer/view.go b/pkg/tcpip/buffer/view.go index a5774a327..bbb4e1d24 100644 --- a/pkg/tcpip/buffer/view.go +++ b/pkg/tcpip/buffer/view.go @@ -54,6 +54,8 @@ func (v *View) ToVectorisedView(views [1]View) VectorisedView { // VectorisedView is a vectorised version of View using non contigous memory. // It supports all the convenience methods supported by View. +// +// +stateify savable type VectorisedView struct { views []View size int diff --git a/pkg/tcpip/header/BUILD b/pkg/tcpip/header/BUILD index 3aa2cfb24..8f22ba3a5 100644 --- a/pkg/tcpip/header/BUILD +++ b/pkg/tcpip/header/BUILD @@ -1,15 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tcp_header_state", - srcs = [ - "tcp.go", - ], - out = "tcp_header_state.go", - package = "header", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "header", @@ -25,13 +16,11 @@ go_library( "ipv6.go", "ipv6_fragment.go", "tcp.go", - "tcp_header_state.go", "udp.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/header", visibility = ["//visibility:public"], deps = [ - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/seqnum", ], diff --git a/pkg/tcpip/header/tcp.go b/pkg/tcpip/header/tcp.go index a95d282b0..6689a6dc5 100644 --- a/pkg/tcpip/header/tcp.go +++ b/pkg/tcpip/header/tcp.go @@ -120,6 +120,8 @@ type TCPSynOptions struct { } // SACKBlock represents a single contiguous SACK block. +// +// +stateify savable type SACKBlock struct { // Start indicates the lowest sequence number in the block. Start seqnum.Value @@ -131,6 +133,8 @@ type SACKBlock struct { // TCPOptions are used to parse and cache the TCP segment options for a non // syn/syn-ack segment. +// +// +stateify savable type TCPOptions struct { // TS is true if the TimeStamp option is enabled. TS bool diff --git a/pkg/tcpip/network/fragmentation/BUILD b/pkg/tcpip/network/fragmentation/BUILD index ac97ebe43..83b4d253f 100644 --- a/pkg/tcpip/network/fragmentation/BUILD +++ b/pkg/tcpip/network/fragmentation/BUILD @@ -1,14 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "fragmentation_state", - srcs = ["reassembler_list.go"], - out = "fragmentation_state.go", - package = "fragmentation", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "reassembler_list", @@ -26,7 +19,6 @@ go_library( srcs = [ "frag_heap.go", "fragmentation.go", - "fragmentation_state.go", "reassembler.go", "reassembler_list.go", ], @@ -34,7 +26,6 @@ go_library( visibility = ["//:sandbox"], deps = [ "//pkg/log", - "//pkg/state", "//pkg/tcpip/buffer", ], ) diff --git a/pkg/tcpip/seqnum/BUILD b/pkg/tcpip/seqnum/BUILD index a75869dac..c5c889239 100644 --- a/pkg/tcpip/seqnum/BUILD +++ b/pkg/tcpip/seqnum/BUILD @@ -1,25 +1,12 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "seqnum_state", - srcs = [ - "seqnum.go", - ], - out = "seqnum_state.go", - package = "seqnum", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "seqnum", - srcs = [ - "seqnum.go", - "seqnum_state.go", - ], + srcs = ["seqnum.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/seqnum", visibility = [ "//visibility:public", ], - deps = ["//pkg/state"], ) diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index eb1e4645d..af0aec85c 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -213,6 +213,8 @@ const ( // FullAddress represents a full transport node address, as required by the // Connect() and Bind() methods. +// +// +stateify savable type FullAddress struct { // NIC is the ID of the NIC this address refers to. // @@ -256,6 +258,8 @@ func (s SlicePayload) Size() int { } // A ControlMessages contains socket control messages for IP sockets. +// +// +stateify savable type ControlMessages struct { // HasTimestamp indicates whether Timestamp is valid/set. HasTimestamp bool diff --git a/pkg/tcpip/transport/ping/BUILD b/pkg/tcpip/transport/ping/BUILD index 28e3e1700..117532fea 100644 --- a/pkg/tcpip/transport/ping/BUILD +++ b/pkg/tcpip/transport/ping/BUILD @@ -1,19 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "ping_state", - srcs = [ - "endpoint.go", - "endpoint_state.go", - "ping_packet_list.go", - ], - out = "ping_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], - package = "ping", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_template_instance( name = "ping_packet_list", @@ -32,14 +20,13 @@ go_library( "endpoint.go", "endpoint_state.go", "ping_packet_list.go", - "ping_state.go", "protocol.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/ping", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/sleep", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/ping/endpoint.go b/pkg/tcpip/transport/ping/endpoint.go index f15e44b61..a22684de9 100644 --- a/pkg/tcpip/transport/ping/endpoint.go +++ b/pkg/tcpip/transport/ping/endpoint.go @@ -26,6 +26,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) +// +stateify savable type pingPacket struct { pingPacketEntry senderAddress tcpip.FullAddress diff --git a/pkg/tcpip/transport/queue/BUILD b/pkg/tcpip/transport/queue/BUILD index fb878ad36..6dcec312e 100644 --- a/pkg/tcpip/transport/queue/BUILD +++ b/pkg/tcpip/transport/queue/BUILD @@ -1,27 +1,14 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "queue_state", - srcs = [ - "queue.go", - ], - out = "queue_state.go", - package = "queue", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "queue", - srcs = [ - "queue.go", - "queue_state.go", - ], + srcs = ["queue.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/queue", visibility = ["//:sandbox"], deps = [ "//pkg/ilist", - "//pkg/state", "//pkg/tcpip", "//pkg/waiter", ], diff --git a/pkg/tcpip/transport/queue/queue.go b/pkg/tcpip/transport/queue/queue.go index 6a17441ae..eb9ee8a3f 100644 --- a/pkg/tcpip/transport/queue/queue.go +++ b/pkg/tcpip/transport/queue/queue.go @@ -33,6 +33,8 @@ type Entry interface { } // Queue is a buffer queue. +// +// +stateify savable type Queue struct { ReaderQueue *waiter.Queue WriterQueue *waiter.Queue diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index 6a7153e4d..9ebae6cc7 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -1,27 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "tcp_state", - srcs = [ - "endpoint.go", - "endpoint_state.go", - "rcv.go", - "reno.go", - "segment.go", - "segment_heap.go", - "segment_queue.go", - "segment_state.go", - "snd.go", - "snd_state.go", - "tcp_segment_list.go", - ], - out = "tcp_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], - package = "tcp", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "tcp_segment_list", @@ -53,15 +33,14 @@ go_library( "snd.go", "snd_state.go", "tcp_segment_list.go", - "tcp_state.go", "timer.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/tcp", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/rand", "//pkg/sleep", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index 5b8a1e20f..de1883d84 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -54,6 +54,8 @@ const ( ) // SACKInfo holds TCP SACK related information for a given endpoint. +// +// +stateify savable type SACKInfo struct { // Blocks is the maximum number of SACK blocks we track // per endpoint. @@ -69,6 +71,8 @@ type SACKInfo struct { // have concurrent goroutines make calls into the endpoint, they are properly // synchronized. The protocol implementation, however, runs in a single // goroutine. +// +// +stateify savable type endpoint struct { // workMu is used to arbitrate which goroutine may perform protocol // work. Only the main protocol goroutine is expected to call Lock() on diff --git a/pkg/tcpip/transport/tcp/rcv.go b/pkg/tcpip/transport/tcp/rcv.go index b22a00ce1..92ef9c6f7 100644 --- a/pkg/tcpip/transport/tcp/rcv.go +++ b/pkg/tcpip/transport/tcp/rcv.go @@ -22,6 +22,8 @@ import ( // receiver holds the state necessary to receive TCP segments and turn them // into a stream of bytes. +// +// +stateify savable type receiver struct { ep *endpoint diff --git a/pkg/tcpip/transport/tcp/reno.go b/pkg/tcpip/transport/tcp/reno.go index 60f170a27..03ae8d747 100644 --- a/pkg/tcpip/transport/tcp/reno.go +++ b/pkg/tcpip/transport/tcp/reno.go @@ -16,6 +16,8 @@ package tcp // renoState stores the variables related to TCP New Reno congestion // control algorithm. +// +// +stateify savable type renoState struct { s *sender } diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index 40928ba2c..8dccea2ba 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -36,6 +36,8 @@ const ( // segment represents a TCP segment. It holds the payload and parsed TCP segment // information, and can be added to intrusive lists. // segment is mostly immutable, the only field allowed to change is viewToDeliver. +// +// +stateify savable type segment struct { segmentEntry refCnt int32 diff --git a/pkg/tcpip/transport/tcp/segment_queue.go b/pkg/tcpip/transport/tcp/segment_queue.go index 2ddcf5f10..6a2d7bc0b 100644 --- a/pkg/tcpip/transport/tcp/segment_queue.go +++ b/pkg/tcpip/transport/tcp/segment_queue.go @@ -21,6 +21,8 @@ import ( ) // segmentQueue is a bounded, thread-safe queue of TCP segments. +// +// +stateify savable type segmentQueue struct { mu sync.Mutex `state:"nosave"` list segmentList `state:"wait"` diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index e38686e1b..376e81846 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -54,6 +54,8 @@ type congestionControl interface { } // sender holds the state necessary to send TCP segments. +// +// +stateify savable type sender struct { ep *endpoint @@ -133,6 +135,8 @@ type sender struct { } // fastRecovery holds information related to fast recovery from a packet loss. +// +// +stateify savable type fastRecovery struct { // active whether the endpoint is in fast recovery. The following fields // are only meaningful when active is true. diff --git a/pkg/tcpip/transport/tcp/snd_state.go b/pkg/tcpip/transport/tcp/snd_state.go index 33c8867f4..d536839af 100644 --- a/pkg/tcpip/transport/tcp/snd_state.go +++ b/pkg/tcpip/transport/tcp/snd_state.go @@ -18,6 +18,7 @@ import ( "time" ) +// +stateify savable type unixTime struct { second int64 nano int64 diff --git a/pkg/tcpip/transport/udp/BUILD b/pkg/tcpip/transport/udp/BUILD index 790dd55a3..1a3a62d3d 100644 --- a/pkg/tcpip/transport/udp/BUILD +++ b/pkg/tcpip/transport/udp/BUILD @@ -1,19 +1,7 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "udp_state", - srcs = [ - "endpoint.go", - "endpoint_state.go", - "udp_packet_list.go", - ], - out = "udp_state.go", - imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], - package = "udp", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_template_instance( name = "udp_packet_list", @@ -33,13 +21,12 @@ go_library( "endpoint_state.go", "protocol.go", "udp_packet_list.go", - "udp_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/udp", + imports = ["gvisor.googlesource.com/gvisor/pkg/tcpip/buffer"], visibility = ["//visibility:public"], deps = [ "//pkg/sleep", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/header", diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 2a32c3a87..03fb76f92 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -25,6 +25,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) +// +stateify savable type udpPacket struct { udpPacketEntry senderAddress tcpip.FullAddress @@ -49,6 +50,8 @@ const ( // between users of the endpoint and the protocol implementation; it is legal to // have concurrent goroutines make calls into the endpoint, they are properly // synchronized. +// +// +stateify savable type endpoint struct { // The following fields are initialized at creation time and do not // change throughout the lifetime of the endpoint. diff --git a/pkg/tcpip/transport/unix/BUILD b/pkg/tcpip/transport/unix/BUILD index 676f2cf92..dae0bd079 100644 --- a/pkg/tcpip/transport/unix/BUILD +++ b/pkg/tcpip/transport/unix/BUILD @@ -1,17 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify") - -go_stateify( - name = "unix_state", - srcs = [ - "connectioned.go", - "connectionless.go", - "unix.go", - ], - out = "unix_state.go", - package = "unix", -) +load("//tools/go_stateify:defs.bzl", "go_library") go_library( name = "unix", @@ -20,14 +9,11 @@ go_library( "connectioned_state.go", "connectionless.go", "unix.go", - "unix_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix", visibility = ["//:sandbox"], deps = [ "//pkg/ilist", - "//pkg/log", - "//pkg/state", "//pkg/tcpip", "//pkg/tcpip/buffer", "//pkg/tcpip/transport/queue", diff --git a/pkg/tcpip/transport/unix/connectioned.go b/pkg/tcpip/transport/unix/connectioned.go index 0e63186b2..dd7c03cf1 100644 --- a/pkg/tcpip/transport/unix/connectioned.go +++ b/pkg/tcpip/transport/unix/connectioned.go @@ -85,6 +85,8 @@ type ConnectingEndpoint interface { // path != "" && acceptedChan != nil => bound and listening. // // Only one of these will be true at any moment. +// +// +stateify savable type connectionedEndpoint struct { baseEndpoint diff --git a/pkg/tcpip/transport/unix/connectionless.go b/pkg/tcpip/transport/unix/connectionless.go index 3276ddcd0..2a6ec8b4b 100644 --- a/pkg/tcpip/transport/unix/connectionless.go +++ b/pkg/tcpip/transport/unix/connectionless.go @@ -25,6 +25,8 @@ import ( // // Specifically, this means datagram unix sockets not created with // socketpair(2). +// +// +stateify savable type connectionlessEndpoint struct { baseEndpoint } diff --git a/pkg/tcpip/transport/unix/unix.go b/pkg/tcpip/transport/unix/unix.go index 190a1ccdb..8e4af3139 100644 --- a/pkg/tcpip/transport/unix/unix.go +++ b/pkg/tcpip/transport/unix/unix.go @@ -60,6 +60,8 @@ type CredentialsControlMessage interface { } // A ControlMessages represents a collection of socket control messages. +// +// +stateify savable type ControlMessages struct { // Rights is a control message containing FDs. Rights RightsControlMessage @@ -235,6 +237,8 @@ type BoundEndpoint interface { } // message represents a message passed over a Unix domain socket. +// +// +stateify savable type message struct { ilist.Entry @@ -306,6 +310,8 @@ type Receiver interface { } // queueReceiver implements Receiver for datagram sockets. +// +// +stateify savable type queueReceiver struct { readQueue *queue.Queue } @@ -369,6 +375,8 @@ func (q *queueReceiver) RecvMaxQueueSize() int64 { func (*queueReceiver) Release() {} // streamQueueReceiver implements Receiver for stream sockets. +// +// +stateify savable type streamQueueReceiver struct { queueReceiver @@ -579,6 +587,7 @@ type ConnectedEndpoint interface { Release() } +// +stateify savable type connectedEndpoint struct { // endpoint represents the subset of the Endpoint functionality needed by // the connectedEndpoint. It is implemented by both connectionedEndpoint @@ -671,6 +680,8 @@ func (*connectedEndpoint) Release() {} // unix domain socket Endpoint implementations. // // Not to be used on its own. +// +// +stateify savable type baseEndpoint struct { *waiter.Queue diff --git a/pkg/waiter/BUILD b/pkg/waiter/BUILD index 8256acdb4..5e611c54f 100644 --- a/pkg/waiter/BUILD +++ b/pkg/waiter/BUILD @@ -1,28 +1,13 @@ package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_stateify", "go_test") - -go_stateify( - name = "waiter_state", - srcs = [ - "waiter.go", - ], - out = "waiter_state.go", - package = "waiter", -) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "waiter", - srcs = [ - "waiter.go", - "waiter_state.go", - ], + srcs = ["waiter.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/waiter", visibility = ["//visibility:public"], - deps = [ - "//pkg/ilist", - "//pkg/state", - ], + deps = ["//pkg/ilist"], ) go_test( diff --git a/pkg/waiter/waiter.go b/pkg/waiter/waiter.go index 9b189bb9e..9825880ca 100644 --- a/pkg/waiter/waiter.go +++ b/pkg/waiter/waiter.go @@ -157,6 +157,8 @@ func NewChannelEntry(c chan struct{}) (Entry, chan struct{}) { // notifiers can notify them when events happen. // // The zero value for waiter.Queue is an empty queue ready for use. +// +// +stateify savable type Queue struct { list ilist.List `state:"zerovalue"` mu sync.RWMutex `state:"nosave"` -- cgit v1.2.3 From c348d0786388ded1a4bad3c98000b4653724c764 Mon Sep 17 00:00:00 2001 From: Zhaozhong Ni Date: Tue, 7 Aug 2018 10:26:17 -0700 Subject: sentry: make epoll.pollEntry wait for the file operation in restore. PiperOrigin-RevId: 207737935 Change-Id: I3a301ece1f1d30909715f36562474e3248b6a0d5 --- pkg/sentry/fs/file.go | 2 +- pkg/sentry/fs/host/inode.go | 2 +- pkg/sentry/kernel/epoll/epoll.go | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 8e535a618..904827a3e 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -91,7 +91,7 @@ type File struct { mu amutex.AbortableMutex `state:"nosave"` // FileOperations implements file system specific behavior for this File. - FileOperations FileOperations + FileOperations FileOperations `state:"wait"` // offset is the File's offset. Updating offset is protected by mu but // can be read atomically via File.Offset() outside of mu. diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index 761ccde33..66c17debb 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -77,7 +77,7 @@ type inodeFileState struct { descriptor *descriptor `state:"wait"` // Event queue for blocking operations. - queue waiter.Queue `state:"nosave"` + queue waiter.Queue `state:"zerovalue"` // sattr is used to restore the inodeOperations. sattr fs.StableAttr `state:"wait"` diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index d87e64a1c..a8eb114c0 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -61,7 +61,7 @@ const ( // // +stateify savable type FileIdentifier struct { - File *fs.File + File *fs.File `state:"wait"` Fd kdefs.FD } -- cgit v1.2.3 From c036da5dffdf6cad912abe2723e69c04b59430b7 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Tue, 7 Aug 2018 13:08:37 -0700 Subject: Hold TaskSet.mu in Task.Parent. PiperOrigin-RevId: 207766238 Change-Id: Id3b66d8fe1f44c3570f67fa5ae7ba16021e35be1 --- pkg/sentry/kernel/threads.go | 2 ++ 1 file changed, 2 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index 844213c35..3d5713106 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -441,6 +441,8 @@ func (t *Task) Timekeeper() *Timekeeper { // Parent returns t's parent. func (t *Task) Parent() *Task { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() return t.parent } -- cgit v1.2.3 From 2e06b23aa61216fcdbefcd6b11a24bca7a456b16 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Thu, 9 Aug 2018 16:49:23 -0700 Subject: Fix missing O_LARGEFILE from O_CREAT files Cleanup some more syscall.O_* references while we're here. PiperOrigin-RevId: 208133460 Change-Id: I48db71a38f817e4f4673977eafcc0e3874eb9a25 --- pkg/abi/linux/file.go | 4 ++++ pkg/sentry/kernel/fd_map.go | 6 ++++-- pkg/sentry/syscalls/linux/flags.go | 14 ++------------ pkg/sentry/syscalls/linux/sys_file.go | 32 ++++++++++++++++++-------------- pkg/sentry/syscalls/linux/sys_pipe.go | 7 ++++--- 5 files changed, 32 insertions(+), 31 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/file.go b/pkg/abi/linux/file.go index f2b7e26ca..509f6b5b3 100644 --- a/pkg/abi/linux/file.go +++ b/pkg/abi/linux/file.go @@ -27,6 +27,10 @@ const ( O_RDONLY = 00000000 O_WRONLY = 00000001 O_RDWR = 00000002 + O_CREAT = 00000100 + O_EXCL = 00000200 + O_NOCTTY = 00000400 + O_TRUNC = 00001000 O_APPEND = 00002000 O_NONBLOCK = 00004000 O_ASYNC = 00020000 diff --git a/pkg/sentry/kernel/fd_map.go b/pkg/sentry/kernel/fd_map.go index d5d4aaacb..cad0b0a20 100644 --- a/pkg/sentry/kernel/fd_map.go +++ b/pkg/sentry/kernel/fd_map.go @@ -53,7 +53,8 @@ type FDFlags struct { CloseOnExec bool } -// ToLinuxFileFlags converts a kernel.FDFlags object to a Linux file flags representation. +// ToLinuxFileFlags converts a kernel.FDFlags object to a Linux file flags +// representation. func (f FDFlags) ToLinuxFileFlags() (mask uint) { if f.CloseOnExec { mask |= linux.O_CLOEXEC @@ -61,7 +62,8 @@ func (f FDFlags) ToLinuxFileFlags() (mask uint) { return } -// ToLinuxFDFlags converts a kernel.FDFlags object to a Linux descriptor flags representation. +// ToLinuxFDFlags converts a kernel.FDFlags object to a Linux descriptor flags +// representation. func (f FDFlags) ToLinuxFDFlags() (mask uint) { if f.CloseOnExec { mask |= linux.FD_CLOEXEC diff --git a/pkg/sentry/syscalls/linux/flags.go b/pkg/sentry/syscalls/linux/flags.go index e8db3e0de..f01483cd3 100644 --- a/pkg/sentry/syscalls/linux/flags.go +++ b/pkg/sentry/syscalls/linux/flags.go @@ -34,8 +34,8 @@ func flagsToPermissions(mask uint) (p fs.PermMask) { return } -// linuxToFlags converts linux file flags to a FileFlags object. -func linuxToFlags(mask uint) (flags fs.FileFlags) { +// linuxToFlags converts Linux file flags to a FileFlags object. +func linuxToFlags(mask uint) fs.FileFlags { return fs.FileFlags{ Direct: mask&linux.O_DIRECT != 0, Sync: mask&linux.O_SYNC != 0, @@ -48,13 +48,3 @@ func linuxToFlags(mask uint) (flags fs.FileFlags) { LargeFile: mask&linux.O_LARGEFILE != 0, } } - -// linuxToSettableFlags converts linux file flags to a SettableFileFlags object. -func linuxToSettableFlags(mask uint) fs.SettableFileFlags { - return fs.SettableFileFlags{ - Direct: mask&linux.O_DIRECT != 0, - NonBlocking: mask&linux.O_NONBLOCK != 0, - Append: mask&linux.O_APPEND != 0, - Async: mask&linux.O_ASYNC != 0, - } -} diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 94b7ac7a5..2cf429f5c 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -164,7 +164,7 @@ func openAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint) (fd u if dirPath { return syserror.ENOTDIR } - if fileFlags.Write && flags&syscall.O_TRUNC != 0 { + if fileFlags.Write && flags&linux.O_TRUNC != 0 { if err := d.Inode.Truncate(t, d, 0); err != nil { return err } @@ -178,7 +178,7 @@ func openAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint) (fd u defer file.DecRef() // Success. - fdFlags := kernel.FDFlags{CloseOnExec: flags&syscall.O_CLOEXEC != 0} + fdFlags := kernel.FDFlags{CloseOnExec: flags&linux.O_CLOEXEC != 0} newFD, err := t.FDMap().NewFDFrom(0, file, fdFlags, t.ThreadGroup().Limits()) if err != nil { return err @@ -302,6 +302,10 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod return syserror.ENOTDIR } + fileFlags := linuxToFlags(flags) + // Linux always adds the O_LARGEFILE flag when running in 64-bit mode. + fileFlags.LargeFile = true + // Does this file exist already? targetDirent, err := t.MountNamespace().FindInode(t, root, d, name, linux.MaxSymlinkTraversals) var newFile *fs.File @@ -311,7 +315,7 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod defer targetDirent.DecRef() // Check if we wanted to create. - if flags&syscall.O_EXCL != 0 { + if flags&linux.O_EXCL != 0 { return syserror.EEXIST } @@ -323,14 +327,14 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod } // Should we truncate the file? - if flags&syscall.O_TRUNC != 0 { + if flags&linux.O_TRUNC != 0 { if err := targetDirent.Inode.Truncate(t, targetDirent, 0); err != nil { return err } } // Create a new fs.File. - newFile, err = targetDirent.Inode.GetFile(t, targetDirent, linuxToFlags(flags)) + newFile, err = targetDirent.Inode.GetFile(t, targetDirent, fileFlags) if err != nil { return syserror.ConvertIntr(err, kernel.ERESTARTSYS) } @@ -346,7 +350,7 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod // Attempt a creation. perms := fs.FilePermsFromMode(mode &^ linux.FileMode(t.FSContext().Umask())) - newFile, err = d.Create(t, root, name, linuxToFlags(flags), perms) + newFile, err = d.Create(t, root, name, fileFlags, perms) if err != nil { // No luck, bail. return err @@ -356,7 +360,7 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod } // Success. - fdFlags := kernel.FDFlags{CloseOnExec: flags&syscall.O_CLOEXEC != 0} + fdFlags := kernel.FDFlags{CloseOnExec: flags&linux.O_CLOEXEC != 0} newFD, err := t.FDMap().NewFDFrom(0, newFile, fdFlags, t.ThreadGroup().Limits()) if err != nil { return err @@ -380,7 +384,7 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod func Open(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { addr := args[0].Pointer() flags := uint(args[1].Uint()) - if flags&syscall.O_CREAT != 0 { + if flags&linux.O_CREAT != 0 { mode := linux.FileMode(args[2].ModeT()) n, err := createAt(t, linux.AT_FDCWD, addr, flags, mode) return n, nil, err @@ -394,7 +398,7 @@ func Openat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal dirFD := kdefs.FD(args[0].Int()) addr := args[1].Pointer() flags := uint(args[2].Uint()) - if flags&syscall.O_CREAT != 0 { + if flags&linux.O_CREAT != 0 { mode := linux.FileMode(args[3].ModeT()) n, err := createAt(t, dirFD, addr, flags, mode) return n, nil, err @@ -407,7 +411,7 @@ func Openat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal func Creat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { addr := args[0].Pointer() mode := linux.FileMode(args[1].ModeT()) - n, err := createAt(t, linux.AT_FDCWD, addr, syscall.O_WRONLY|syscall.O_TRUNC, mode) + n, err := createAt(t, linux.AT_FDCWD, addr, linux.O_WRONLY|linux.O_TRUNC, mode) return n, nil, err } @@ -747,7 +751,7 @@ func Dup3(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallC } defer oldFile.DecRef() - err := t.FDMap().NewFDAt(newfd, oldFile, kernel.FDFlags{CloseOnExec: flags&syscall.O_CLOEXEC != 0}, t.ThreadGroup().Limits()) + err := t.FDMap().NewFDAt(newfd, oldFile, kernel.FDFlags{CloseOnExec: flags&linux.O_CLOEXEC != 0}, t.ThreadGroup().Limits()) if err != nil { return 0, nil, err } @@ -802,7 +806,7 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall switch cmd { case linux.F_DUPFD, linux.F_DUPFD_CLOEXEC: from := kdefs.FD(args[2].Int()) - fdFlags := kernel.FDFlags{CloseOnExec: cmd == syscall.F_DUPFD_CLOEXEC} + fdFlags := kernel.FDFlags{CloseOnExec: cmd == linux.F_DUPFD_CLOEXEC} fd, err := t.FDMap().NewFDFrom(from, file, fdFlags, t.ThreadGroup().Limits()) if err != nil { return 0, nil, err @@ -813,13 +817,13 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall case linux.F_SETFD: flags := args[2].Uint() t.FDMap().SetFlags(fd, kernel.FDFlags{ - CloseOnExec: flags&syscall.FD_CLOEXEC != 0, + CloseOnExec: flags&linux.FD_CLOEXEC != 0, }) case linux.F_GETFL: return uintptr(file.Flags().ToLinux()), nil, nil case linux.F_SETFL: flags := uint(args[2].Uint()) - file.SetFlags(linuxToSettableFlags(flags)) + file.SetFlags(linuxToFlags(flags).Settable()) case linux.F_SETLK, linux.F_SETLKW: // In Linux the file system can choose to provide lock operations for an inode. // Normally pipe and socket types lack lock operations. We diverge and use a heavy diff --git a/pkg/sentry/syscalls/linux/sys_pipe.go b/pkg/sentry/syscalls/linux/sys_pipe.go index 3efc06a27..2b544f145 100644 --- a/pkg/sentry/syscalls/linux/sys_pipe.go +++ b/pkg/sentry/syscalls/linux/sys_pipe.go @@ -17,6 +17,7 @@ package linux import ( "syscall" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" @@ -26,7 +27,7 @@ import ( // pipe2 implements the actual system call with flags. func pipe2(t *kernel.Task, addr usermem.Addr, flags uint) (uintptr, error) { - if flags&^(syscall.O_NONBLOCK|syscall.O_CLOEXEC) != 0 { + if flags&^(linux.O_NONBLOCK|linux.O_CLOEXEC) != 0 { return 0, syscall.EINVAL } r, w := pipe.NewConnectedPipe(t, pipe.DefaultPipeSize, usermem.PageSize) @@ -38,14 +39,14 @@ func pipe2(t *kernel.Task, addr usermem.Addr, flags uint) (uintptr, error) { defer w.DecRef() rfd, err := t.FDMap().NewFDFrom(0, r, kernel.FDFlags{ - CloseOnExec: flags&syscall.O_CLOEXEC != 0}, + CloseOnExec: flags&linux.O_CLOEXEC != 0}, t.ThreadGroup().Limits()) if err != nil { return 0, err } wfd, err := t.FDMap().NewFDFrom(0, w, kernel.FDFlags{ - CloseOnExec: flags&syscall.O_CLOEXEC != 0}, + CloseOnExec: flags&linux.O_CLOEXEC != 0}, t.ThreadGroup().Limits()) if err != nil { t.FDMap().Remove(rfd) -- cgit v1.2.3 From 635b0c45933cd841298b0c21a513a9169e849594 Mon Sep 17 00:00:00 2001 From: Kevin Krakauer Date: Wed, 15 Aug 2018 16:24:07 -0700 Subject: runsc fsgofer: Support dynamic serving of filesystems. When multiple containers run inside a sentry, each container has its own root filesystem and set of mounts. Containers are also added after sentry boot rather than all configured and known at boot time. The fsgofer needs to be able to serve the root filesystem of each container. Thus, it must be possible to add filesystems after the fsgofer has already started. This change: * Creates a URPC endpoint within the gofer process that listens for requests to serve new content. * Enables the sentry, when starting a new container, to add the new container's filesystem. * Mounts those new filesystems at separate roots within the sentry. PiperOrigin-RevId: 208903248 Change-Id: Ifa91ec9c8caf5f2f0a9eead83c4a57090ce92068 --- pkg/sentry/kernel/kernel.go | 14 ++- pkg/urpc/urpc.go | 10 +++ runsc/boot/controller.go | 19 +++- runsc/boot/fs.go | 159 ++++++++++++++++++++++++--------- runsc/boot/loader.go | 36 ++++---- runsc/cmd/BUILD | 1 - runsc/cmd/gofer.go | 36 +++----- runsc/container/container.go | 7 ++ runsc/fsgofer/BUILD | 4 + runsc/fsgofer/control.go | 203 +++++++++++++++++++++++++++++++++++++++++++ runsc/fsgofer/fsgofer.go | 5 ++ runsc/sandbox/BUILD | 1 + runsc/sandbox/sandbox.go | 139 ++++++++++++++++++++++------- 13 files changed, 515 insertions(+), 119 deletions(-) create mode 100644 runsc/fsgofer/control.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 419a1d473..cb43fdcdc 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -504,6 +504,14 @@ type CreateProcessArgs struct { // IPCNamespace is the initial IPC namespace. IPCNamespace *IPCNamespace + + // Root optionally contains the dirent that serves as the root for the + // process. If nil, the mount namespace's root is used as the process' + // root. + // + // Anyone setting Root must donate a reference (i.e. increment it) to + // keep it alive until it is decremented by CreateProcess. + Root *fs.Dirent } // NewContext returns a context.Context that represents the task that will be @@ -581,8 +589,12 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { ctx := args.NewContext(k) // Grab the root directory. - root := fs.RootFromContext(ctx) + root := args.Root + if root == nil { + root = fs.RootFromContext(ctx) + } defer root.DecRef() + args.Root = nil // Grab the working directory. wd := root // Default. diff --git a/pkg/urpc/urpc.go b/pkg/urpc/urpc.go index af620b704..1ec06dd4c 100644 --- a/pkg/urpc/urpc.go +++ b/pkg/urpc/urpc.go @@ -63,6 +63,10 @@ func (r RemoteError) Error() string { // file as a result of an RPC. These are not actually serialized, rather they // are sent via an accompanying SCM_RIGHTS message (plumbed through the unet // package). +// +// When embedding a FilePayload in an argument struct, the argument type _must_ +// be a pointer to the struct rather than the struct type itself. This is +// because the urpc package defines pointer methods on FilePayload. type FilePayload struct { Files []*os.File `json:"-"` } @@ -552,6 +556,12 @@ func (c *Client) Call(method string, arg interface{}, result interface{}) error c.mu.Lock() defer c.mu.Unlock() + // If arg is a FilePayload, not a *FilePayload, files won't actually be + // sent, so error out. + if _, ok := arg.(FilePayload); ok { + return fmt.Errorf("argument is a FilePayload, but should be a *FilePayload") + } + // Are there files to send? var fs []*os.File if fp, ok := arg.(filePayloader); ok { diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index fc6ea326a..69e88d8e0 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -17,6 +17,7 @@ package boot import ( "errors" "fmt" + "path" specs "github.com/opencontainers/runtime-spec/specs-go" "gvisor.googlesource.com/gvisor/pkg/control/server" @@ -181,11 +182,15 @@ type StartArgs struct { // CID is the ID of the container to start. CID string + + // FilePayload contains the file descriptor over which the sandbox will + // request files from its root filesystem. + urpc.FilePayload } // Start runs a created container within a sandbox. func (cm *containerManager) Start(args *StartArgs, _ *struct{}) error { - log.Debugf("containerManager.Start") + log.Debugf("containerManager.Start: %+v", args) // Validate arguments. if args == nil { @@ -200,8 +205,18 @@ func (cm *containerManager) Start(args *StartArgs, _ *struct{}) error { if args.CID == "" { return errors.New("start argument missing container ID") } + // Prevent CIDs containing ".." from confusing the sentry when creating + // /containers/ directory. + // TODO: Once we have multiple independant roots, this + // check won't be necessary. + if path.Clean(args.CID) != args.CID { + return fmt.Errorf("container ID shouldn't contain directory traversals such as \"..\": %q", args.CID) + } + if len(args.FilePayload.Files) != 1 { + return fmt.Errorf("start arguments must contain one file for the container root") + } - tgid, err := cm.l.startContainer(args, cm.l.k) + tgid, err := cm.l.startContainer(cm.l.k, args.Spec, args.Conf, args.CID, args.FilePayload.Files[0]) if err != nil { return err } diff --git a/runsc/boot/fs.go b/runsc/boot/fs.go index eea2ec1f5..8996b1398 100644 --- a/runsc/boot/fs.go +++ b/runsc/boot/fs.go @@ -48,6 +48,19 @@ const ( // Device name for root mount. rootDevice = "9pfs-/" + + // childContainersDir is the directory where child container root + // filesystems are mounted. + childContainersDir = "/__runsc_containers__" + + // Filesystems that runsc supports. + bind = "bind" + devpts = "devpts" + devtmpfs = "devtmpfs" + proc = "proc" + sysfs = "sysfs" + tmpfs = "tmpfs" + nonefs = "none" ) type fdDispenser struct { @@ -70,8 +83,15 @@ func (f *fdDispenser) empty() bool { // createMountNamespace creates a mount namespace containing the root filesystem // and all mounts. 'rootCtx' is used to walk directories to find mount points. func createMountNamespace(userCtx context.Context, rootCtx context.Context, spec *specs.Spec, conf *Config, ioFDs []int) (*fs.MountNamespace, error) { + mounts := compileMounts(spec) + // Create a tmpfs mount where we create and mount a root filesystem for + // each child container. + mounts = append(mounts, specs.Mount{ + Type: tmpfs, + Destination: childContainersDir, + }) fds := &fdDispenser{fds: ioFDs} - rootInode, err := createRootMount(rootCtx, spec, conf, fds) + rootInode, err := createRootMount(rootCtx, spec, conf, fds, mounts) if err != nil { return nil, fmt.Errorf("failed to create root mount: %v", err) } @@ -79,7 +99,7 @@ func createMountNamespace(userCtx context.Context, rootCtx context.Context, spec if err != nil { return nil, fmt.Errorf("failed to create root mount namespace: %v", err) } - mounts := compileMounts(spec) + if err := setMounts(rootCtx, conf, mns, fds, mounts); err != nil { return nil, fmt.Errorf("failed to configure mounts: %v", err) } @@ -98,12 +118,12 @@ func compileMounts(spec *specs.Spec) []specs.Mount { // Always mount /dev. mounts = append(mounts, specs.Mount{ - Type: "devtmpfs", + Type: devtmpfs, Destination: "/dev", }) mounts = append(mounts, specs.Mount{ - Type: "devpts", + Type: devpts, Destination: "/dev/pts", }) @@ -129,13 +149,13 @@ func compileMounts(spec *specs.Spec) []specs.Mount { var mandatoryMounts []specs.Mount if !procMounted { mandatoryMounts = append(mandatoryMounts, specs.Mount{ - Type: "proc", + Type: proc, Destination: "/proc", }) } if !sysMounted { mandatoryMounts = append(mandatoryMounts, specs.Mount{ - Type: "sysfs", + Type: sysfs, Destination: "/sys", }) } @@ -149,7 +169,7 @@ func compileMounts(spec *specs.Spec) []specs.Mount { // that. Until then, the /tmp mount will always appear empty at // container creation. mandatoryMounts = append(mandatoryMounts, specs.Mount{ - Type: "tmpfs", + Type: tmpfs, Destination: "/tmp", }) } @@ -165,7 +185,7 @@ func compileMounts(spec *specs.Spec) []specs.Mount { // mount namespace. func setMounts(ctx context.Context, conf *Config, mns *fs.MountNamespace, fds *fdDispenser, mounts []specs.Mount) error { for _, m := range mounts { - if err := mountSubmount(ctx, conf, mns, fds, m, mounts); err != nil { + if err := mountSubmount(ctx, conf, mns, fds, m, mounts, m.Destination); err != nil { return err } } @@ -173,7 +193,7 @@ func setMounts(ctx context.Context, conf *Config, mns *fs.MountNamespace, fds *f } // createRootMount creates the root filesystem. -func createRootMount(ctx context.Context, spec *specs.Spec, conf *Config, fds *fdDispenser) (*fs.Inode, error) { +func createRootMount(ctx context.Context, spec *specs.Spec, conf *Config, fds *fdDispenser, mounts []specs.Mount) (*fs.Inode, error) { // First construct the filesystem from the spec.Root. mf := fs.MountSourceFlags{ReadOnly: spec.Root.Readonly} @@ -207,7 +227,7 @@ func createRootMount(ctx context.Context, spec *specs.Spec, conf *Config, fds *f // We need to overlay the root on top of a ramfs with stub directories // for submount paths. "/dev" "/sys" "/proc" and "/tmp" are always // mounted even if they are not in the spec. - submounts := append(subtargets("/", spec.Mounts), "/dev", "/sys", "/proc", "/tmp") + submounts := append(subtargets("/", mounts), "/dev", "/sys", "/proc", "/tmp") rootInode, err = addSubmountOverlay(ctx, rootInode, submounts) if err != nil { return nil, fmt.Errorf("error adding submount overlay: %v", err) @@ -256,17 +276,17 @@ func getMountNameAndOptions(conf *Config, m specs.Mount, fds *fdDispenser) (stri ) switch m.Type { - case "devpts", "devtmpfs", "proc", "sysfs": + case devpts, devtmpfs, proc, sysfs: fsName = m.Type - case "none": - fsName = "sysfs" - case "tmpfs": + case nonefs: + fsName = sysfs + case tmpfs: fsName = m.Type // tmpfs has some extra supported options that we must pass through. opts, err = parseAndFilterOptions(m.Options, "mode", "uid", "gid") - case "bind": + case bind: switch conf.FileAccess { case FileAccessProxy, FileAccessProxyExclusive: fd := fds.remove() @@ -291,7 +311,7 @@ func getMountNameAndOptions(conf *Config, m specs.Mount, fds *fdDispenser) (stri return fsName, opts, useOverlay, err } -func mountSubmount(ctx context.Context, conf *Config, mns *fs.MountNamespace, fds *fdDispenser, m specs.Mount, mounts []specs.Mount) error { +func mountSubmount(ctx context.Context, conf *Config, mns *fs.MountNamespace, fds *fdDispenser, m specs.Mount, mounts []specs.Mount, dest string) error { // Map mount type to filesystem name, and parse out the options that we are // capable of dealing with. fsName, opts, useOverlay, err := getMountNameAndOptions(conf, m, fds) @@ -342,51 +362,52 @@ func mountSubmount(ctx context.Context, conf *Config, mns *fs.MountNamespace, fd // in the right location, e.g. // mount: /var/run/secrets, may be created in '/run/secrets' if // '/var/run' => '/var'. - if err := mkdirAll(ctx, mns, m.Destination); err != nil { + if err := mkdirAll(ctx, mns, dest); err != nil { return err } root := mns.Root() defer root.DecRef() - dirent, err := mns.FindInode(ctx, root, nil, m.Destination, linux.MaxSymlinkTraversals) + dirent, err := mns.FindInode(ctx, root, nil, dest, linux.MaxSymlinkTraversals) if err != nil { - return fmt.Errorf("failed to find mount destination %q: %v", m.Destination, err) + return fmt.Errorf("failed to find mount destination %q: %v", dest, err) } defer dirent.DecRef() if err := mns.Mount(ctx, dirent, inode); err != nil { - return fmt.Errorf("failed to mount at destination %q: %v", m.Destination, err) + return fmt.Errorf("failed to mount at destination %q: %v", dest, err) } - log.Infof("Mounted %q to %q type %s", m.Source, m.Destination, m.Type) + log.Infof("Mounted %q to %q type %s", m.Source, dest, m.Type) return nil } func mkdirAll(ctx context.Context, mns *fs.MountNamespace, path string) error { + log.Infof("mkdirAll called with path %s", path) root := mns.Root() defer root.DecRef() // Starting at the root, walk the path. parent := root ps := strings.Split(filepath.Clean(path), string(filepath.Separator)) - for i := 0; i < len(ps); i++ { - if ps[i] == "" { + for _, pathElem := range ps { + if pathElem == "" { // This will be case for the first and last element, if the path // begins or ends with '/'. Note that we always treat the path as // absolute, regardless of what the first character contains. continue } - d, err := mns.FindInode(ctx, root, parent, ps[i], fs.DefaultTraversalLimit) + d, err := mns.FindInode(ctx, root, parent, pathElem, fs.DefaultTraversalLimit) if err == syserror.ENOENT { // If we encounter a path that does not exist, then // create it. - if err := parent.CreateDirectory(ctx, root, ps[i], fs.FilePermsFromMode(0755)); err != nil { - return fmt.Errorf("failed to create directory %q: %v", ps[i], err) + if err := parent.CreateDirectory(ctx, root, pathElem, fs.FilePermsFromMode(0755)); err != nil { + return fmt.Errorf("failed to create directory %q: %v", pathElem, err) } - if d, err = parent.Walk(ctx, root, ps[i]); err != nil { - return fmt.Errorf("walk to %q failed: %v", ps[i], err) + if d, err = parent.Walk(ctx, root, pathElem); err != nil { + return fmt.Errorf("walk to %q failed: %v", pathElem, err) } } else if err != nil { - return fmt.Errorf("failed to find inode %q: %v", ps[i], err) + return fmt.Errorf("failed to find inode %q: %v", pathElem, err) } parent = d } @@ -444,7 +465,7 @@ func destinations(mounts []specs.Mount, extra ...string) []string { // mountDevice returns a device string based on the fs type and target // of the mount. func mountDevice(m specs.Mount) string { - if m.Type == "bind" { + if m.Type == bind { // Make a device string that includes the target, which is consistent across // S/R and uniquely identifies the connection. return "9pfs-" + m.Destination @@ -589,7 +610,7 @@ func subtargets(root string, mnts []specs.Mount) []string { // setFileSystemForProcess is used to set up the file system and amend the procArgs accordingly. // procArgs are passed by reference and the FDMap field is modified. -func setFileSystemForProcess(procArgs *kernel.CreateProcessArgs, spec *specs.Spec, conf *Config, ioFDs []int, console bool, creds *auth.Credentials, ls *limits.LimitSet, k *kernel.Kernel) error { +func setFileSystemForProcess(procArgs *kernel.CreateProcessArgs, spec *specs.Spec, conf *Config, ioFDs []int, console bool, creds *auth.Credentials, ls *limits.LimitSet, k *kernel.Kernel, cid string) error { ctx := procArgs.NewContext(k) // Create the FD map, which will set stdin, stdout, and stderr. If @@ -604,27 +625,79 @@ func setFileSystemForProcess(procArgs *kernel.CreateProcessArgs, spec *specs.Spe // won't need ours either way. procArgs.FDMap = fdm + // Use root user to configure mounts. The current user might not have + // permission to do so. + rootProcArgs := kernel.CreateProcessArgs{ + WorkingDirectory: "/", + Credentials: auth.NewRootCredentials(creds.UserNamespace), + Umask: 0022, + MaxSymlinkTraversals: linux.MaxSymlinkTraversals, + } + rootCtx := rootProcArgs.NewContext(k) + // If this is the root container, we also need to setup the root mount // namespace. - if k.RootMountNamespace() == nil { - // Use root user to configure mounts. The current user might not have - // permission to do so. - rootProcArgs := kernel.CreateProcessArgs{ - WorkingDirectory: "/", - Credentials: auth.NewRootCredentials(creds.UserNamespace), - Umask: 0022, - MaxSymlinkTraversals: linux.MaxSymlinkTraversals, - } - rootCtx := rootProcArgs.NewContext(k) - + mns := k.RootMountNamespace() + if mns == nil { // Create the virtual filesystem. mns, err := createMountNamespace(ctx, rootCtx, spec, conf, ioFDs) if err != nil { return fmt.Errorf("error creating mounts: %v", err) } - k.SetRootMountNamespace(mns) + return nil + } + + // Setup a child container. + + // Create the container's root filesystem mount. + log.Infof("Creating new process in child container.") + fds := &fdDispenser{fds: append([]int{}, ioFDs...)} + rootInode, err := createRootMount(rootCtx, spec, conf, fds, nil) + if err != nil { + return fmt.Errorf("error creating filesystem for container: %v", err) + } + + // Make directories for submounts within the container. + rootDir := mns.Root() + defer rootDir.DecRef() + containerRoot := filepath.Join(childContainersDir, cid) + mkdirAll(ctx, mns, containerRoot) + + // Mount the container's root filesystem to the newly created + // mount point. + containerRootDirent, err := mns.FindInode(ctx, rootDir, nil, containerRoot, linux.MaxSymlinkTraversals) + if err != nil { + return fmt.Errorf("failed to find mount destination: %q: %v", containerRoot, err) + } + if err := mns.Mount(ctx, containerRootDirent, rootInode); err != nil { + return fmt.Errorf("failed to mount at destination %q: %v", containerRoot, err) + } + containerRootDirent.DecRef() + + // We have to re-walk to the dirent to find the mounted + // directory. The old dirent is invalid at this point. + containerRootDirent, err = mns.FindInode(ctx, rootDir, nil, containerRoot, linux.MaxSymlinkTraversals) + if err != nil { + return fmt.Errorf("failed to find mount destination2: %q: %v", containerRoot, err) + } + log.Infof("Mounted child's root fs to %q", containerRoot) + + // Mount all submounts. + mounts := compileMounts(spec) + for _, m := range mounts { + // TODO: Enable bind mounts in child containers. + if m.Type == bind { + log.Infof("Bind mounts in child containers are not yet supported: %+v", m) + continue + } + dest := filepath.Join(containerRoot, m.Destination) + if err := mountSubmount(rootCtx, conf, k.RootMountNamespace(), fds, m, mounts, dest); err != nil { + return fmt.Errorf("error mounting filesystem for container: %v", err) + } } + // Set the procArgs root directory. + procArgs.Root = containerRootDirent return nil } diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index f6c7bf223..7debf0ac2 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -19,6 +19,7 @@ import ( "errors" "fmt" "math/rand" + "os" "runtime" "sync" "sync/atomic" @@ -229,7 +230,7 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console // Ensure that signals received are forwarded to the emulated kernel. stopSignalForwarding := sighandling.PrepareForwarding(k, false)() - procArgs, err := newProcess(spec, conf, ioFDs, console, creds, utsns, ipcns, k) + procArgs, err := newProcess(spec, creds, utsns, ipcns, k) if err != nil { return nil, fmt.Errorf("failed to create root process: %v", err) } @@ -250,7 +251,7 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console } // newProcess creates a process that can be run with kernel.CreateProcess. -func newProcess(spec *specs.Spec, conf *Config, ioFDs []int, console bool, creds *auth.Credentials, utsns *kernel.UTSNamespace, ipcns *kernel.IPCNamespace, k *kernel.Kernel) (kernel.CreateProcessArgs, error) { +func newProcess(spec *specs.Spec, creds *auth.Credentials, utsns *kernel.UTSNamespace, ipcns *kernel.IPCNamespace, k *kernel.Kernel) (kernel.CreateProcessArgs, error) { // Create initial limits. ls, err := createLimitSet(spec) if err != nil { @@ -277,7 +278,6 @@ func newProcess(spec *specs.Spec, conf *Config, ioFDs []int, console bool, creds UTSNamespace: utsns, IPCNamespace: ipcns, } - return procArgs, nil } @@ -356,7 +356,8 @@ func (l *Loader) run() error { l.console, l.rootProcArgs.Credentials, l.rootProcArgs.Limits, - l.k) + l.k, + "" /* CID, which isn't needed for the root container */) if err != nil { return err } @@ -376,8 +377,7 @@ func (l *Loader) run() error { // startContainer starts a child container. It returns the thread group ID of // the newly created process. -func (l *Loader) startContainer(args *StartArgs, k *kernel.Kernel) (kernel.ThreadID, error) { - spec := args.Spec +func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config, cid string, file *os.File) (kernel.ThreadID, error) { // Create capabilities. caps, err := specutils.Capabilities(spec.Process.Capabilities) if err != nil { @@ -406,26 +406,24 @@ func (l *Loader) startContainer(args *StartArgs, k *kernel.Kernel) (kernel.Threa // when indicated by the spec. procArgs, err := newProcess( - args.Spec, - args.Conf, - nil, // ioFDs - false, // console + spec, creds, - k.RootUTSNamespace(), - k.RootIPCNamespace(), - k) + l.k.RootUTSNamespace(), + l.k.RootIPCNamespace(), + l.k) if err != nil { return 0, fmt.Errorf("failed to create new process: %v", err) } err = setFileSystemForProcess( &procArgs, - args.Spec, - args.Conf, - nil, + spec, + conf, + []int{int(file.Fd())}, // ioFDs false, creds, procArgs.Limits, - k) + k, + cid) if err != nil { return 0, fmt.Errorf("failed to create new process: %v", err) } @@ -435,7 +433,7 @@ func (l *Loader) startContainer(args *StartArgs, k *kernel.Kernel) (kernel.Threa return 0, fmt.Errorf("failed to create process in sentry: %v", err) } - ts := k.TaskSet() + ts := l.k.TaskSet() tgid := ts.Root.IDOfThreadGroup(tg) if tgid == 0 { return 0, errors.New("failed to get thread group ID of new process") @@ -446,7 +444,7 @@ func (l *Loader) startContainer(args *StartArgs, k *kernel.Kernel) (kernel.Threa l.mu.Lock() defer l.mu.Unlock() - l.containerRootTGIDs[args.CID] = tgid + l.containerRootTGIDs[cid] = tgid return tgid, nil } diff --git a/runsc/cmd/BUILD b/runsc/cmd/BUILD index 18e95284b..c45784749 100644 --- a/runsc/cmd/BUILD +++ b/runsc/cmd/BUILD @@ -36,7 +36,6 @@ go_library( "//pkg/p9", "//pkg/sentry/control", "//pkg/sentry/kernel/auth", - "//pkg/unet", "//pkg/urpc", "//runsc/boot", "//runsc/container", diff --git a/runsc/cmd/gofer.go b/runsc/cmd/gofer.go index 55315c0e8..ed4b1d29c 100644 --- a/runsc/cmd/gofer.go +++ b/runsc/cmd/gofer.go @@ -16,7 +16,6 @@ package cmd import ( "os" - "sync" "syscall" "context" @@ -25,7 +24,6 @@ import ( specs "github.com/opencontainers/runtime-spec/specs-go" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/p9" - "gvisor.googlesource.com/gvisor/pkg/unet" "gvisor.googlesource.com/gvisor/runsc/fsgofer" "gvisor.googlesource.com/gvisor/runsc/specutils" ) @@ -36,6 +34,10 @@ type Gofer struct { bundleDir string ioFDs intFlags applyCaps bool + + // controllerFD is the file descriptor of a stream socket for the + // control server that is donated to this process. + controllerFD int } // Name implements subcommands.Command. @@ -58,11 +60,12 @@ func (g *Gofer) SetFlags(f *flag.FlagSet) { f.StringVar(&g.bundleDir, "bundle", "", "path to the root of the bundle directory, defaults to the current directory") f.Var(&g.ioFDs, "io-fds", "list of FDs to connect 9P servers. They must follow this order: root first, then mounts as defined in the spec") f.BoolVar(&g.applyCaps, "apply-caps", true, "if true, apply capabilities to restrict what the Gofer process can do") + f.IntVar(&g.controllerFD, "controller-fd", -1, "required FD of a stream socket for the control server that must be donated to this process") } // Execute implements subcommands.Command. func (g *Gofer) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) subcommands.ExitStatus { - if g.bundleDir == "" || len(g.ioFDs) < 1 { + if g.bundleDir == "" || len(g.ioFDs) < 1 || g.controllerFD == -1 { f.Usage() return subcommands.ExitUsageError } @@ -134,29 +137,14 @@ func (g *Gofer) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) Fatalf("Too many FDs passed for mounts. mounts: %d, FDs: %d", mountIdx, len(g.ioFDs)) } - runServers(ats, g.ioFDs) - return subcommands.ExitSuccess -} + ctrl, err := fsgofer.NewController(g.controllerFD, g.bundleDir) -func runServers(ats []p9.Attacher, ioFDs []int) { - // Run the loops and wait for all to exit. - var wg sync.WaitGroup - for i, ioFD := range ioFDs { - wg.Add(1) - go func(ioFD int, at p9.Attacher) { - socket, err := unet.NewSocket(ioFD) - if err != nil { - Fatalf("err creating server on FD %d: %v", ioFD, err) - } - s := p9.NewServer(at) - if err := s.Handle(socket); err != nil { - Fatalf("P9 server returned error. Gofer is shutting down. FD: %d, err: %v", ioFD, err) - } - wg.Done() - }(ioFD, ats[i]) + if err := ctrl.Serve(ats, g.ioFDs); err != nil { + Fatalf("Failed to serve via P9: %v", err) } - wg.Wait() - log.Infof("All 9P servers exited.") + ctrl.Wait() + + return subcommands.ExitSuccess } func isReadonlyMount(opts []string) bool { diff --git a/runsc/container/container.go b/runsc/container/container.go index 574075b00..da2ce0d25 100644 --- a/runsc/container/container.go +++ b/runsc/container/container.go @@ -249,6 +249,13 @@ func Create(id string, spec *specs.Spec, conf *boot.Config, bundleDir, consoleSo return nil, err } c.Sandbox = sb.Sandbox + + // Prepare the gofer to serve the container's filesystem. + err = sb.Sandbox.CreateChild(c.ID, bundleDir) + if err != nil { + c.Destroy() + return nil, err + } } c.Status = Created diff --git a/runsc/fsgofer/BUILD b/runsc/fsgofer/BUILD index 24e172f48..0bc682b5f 100644 --- a/runsc/fsgofer/BUILD +++ b/runsc/fsgofer/BUILD @@ -5,6 +5,7 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") go_library( name = "fsgofer", srcs = [ + "control.go", "fsgofer.go", "fsgofer_unsafe.go", ], @@ -14,9 +15,12 @@ go_library( ], deps = [ "//pkg/abi/linux", + "//pkg/control/server", "//pkg/fd", "//pkg/log", "//pkg/p9", + "//pkg/unet", + "//pkg/urpc", "@org_golang_x_sys//unix:go_default_library", ], ) diff --git a/runsc/fsgofer/control.go b/runsc/fsgofer/control.go new file mode 100644 index 000000000..8ce8ee8a0 --- /dev/null +++ b/runsc/fsgofer/control.go @@ -0,0 +1,203 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package fsgofer + +import ( + "fmt" + "path/filepath" + "sync" + + "gvisor.googlesource.com/gvisor/pkg/control/server" + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/p9" + "gvisor.googlesource.com/gvisor/pkg/unet" + "gvisor.googlesource.com/gvisor/pkg/urpc" +) + +// Controller manages the fsgofer's control server. +type Controller struct { + // api holds the control server's URPC endpoints. + api api + + // srv is the control server. + srv *server.Server +} + +// NewController creates a new Controller and starts it listenting +func NewController(fd int, rootBundleDir string) (*Controller, error) { + if !filepath.IsAbs(rootBundleDir) { + return nil, fmt.Errorf("NewController should receive an absolute bundle dir path, but got %q", rootBundleDir) + } + + srv, err := server.CreateFromFD(fd) + if err != nil { + return nil, err + } + + cr := &Controller{srv: srv} + cr.api.rootBundleDir = rootBundleDir + cr.api.bundleDirs = make(map[string]string) + srv.Register(&cr.api) + + if err := srv.StartServing(); err != nil { + return nil, err + } + + return cr, nil +} + +// Wait waits for all the p9 servers to finish, then shuts down the control +// server. +func (cr *Controller) Wait() { + cr.api.p9wg.Wait() + cr.srv.Stop() + log.Infof("All 9P servers exited.") +} + +// Serve starts serving each Attacher in ats via its corresponding file +// descriptor in ioFDs. +func (cr *Controller) Serve(ats []p9.Attacher, ioFDs []int) error { + if len(ats) != len(ioFDs) { + return fmt.Errorf("number of attach points does not match the number of IO FDs (%d and %d)", len(ats), len(ioFDs)) + } + for i, _ := range ats { + cr.api.serve(ats[i], ioFDs[i]) + } + return nil +} + +// api URPC methods. +const ( + // AddBundleDirs readies the gofer to serve from a new bundle + // directory. It should be called during runsc create. + AddBundleDirs = "api.AddBundleDirs" + + // ServeDirectory serves a new directory via the fsgofer. It should be + // called during runsc start. + ServeDirectory = "api.ServeDirectory" +) + +// API defines and implements the URPC endpoints for the gofer. +type api struct { + // p9wg waits for all the goroutines serving the sentry via p9. When its + // counter is 0, the gofer is out of work and exits. + p9wg sync.WaitGroup + + // bundleDirs maps from container ID to bundle directory for each + // container. + bundleDirs map[string]string + + // rootBundleDir is the bundle directory of the root container. + rootBundleDir string +} + +// AddBundleDirsRequest is the URPC argument to AddBundleDirs. +type AddBundleDirsRequest struct { + // BundleDirs is a map of container IDs to bundle directories to add to + // the gofer. + BundleDirs map[string]string +} + +// AddBundleDirsRequest adds bundle directories that for the gofer to serve. +func (api *api) AddBundleDirs(req *AddBundleDirsRequest, _ *struct{}) error { + log.Debugf("fsgofer.AddBundleDirs") + for cid, bd := range req.BundleDirs { + if _, ok := api.bundleDirs[cid]; ok { + return fmt.Errorf("fsgofer already has a bundleDir for container %q", cid) + } + api.bundleDirs[cid] = bd + } + return nil +} + +// ServeDirectoryRequest is the URPC argument to ServeDirectory. +type ServeDirectoryRequest struct { + // Dir is the absolute path to a directory to be served to the sentry. + Dir string + + // IsReadOnly specifies whether the directory should be served in + // read-only mode. + IsReadOnly bool + + // CID is the container ID of the container that needs to serve a + // directory. + CID string + + // FilePayload contains the socket over which the sentry will request + // files from Dir. + urpc.FilePayload +} + +// ServeDirectory begins serving a directory via a file descriptor for the +// sentry. Directories must be added via AddBundleDirsRequest before +// ServeDirectory is called. +func (api *api) ServeDirectory(req *ServeDirectoryRequest, _ *struct{}) error { + log.Debugf("fsgofer.ServeDirectory: %+v", req) + + if req.Dir == "" { + return fmt.Errorf("ServeDirectory should receive a directory argument, but was empty") + } + if req.CID == "" { + return fmt.Errorf("ServeDirectory should receive a CID argument, but was empty") + } + // Prevent CIDs containing ".." from confusing the sentry when creating + // /containers/ directory. + // TODO: Once we have multiple independant roots, this + // check won't be necessary. + if filepath.Clean(req.CID) != req.CID { + return fmt.Errorf("container ID shouldn't contain directory traversals such as \"..\": %q", req.CID) + } + if nFiles := len(req.FilePayload.Files); nFiles != 1 { + return fmt.Errorf("ServeDirectory should receive 1 file descriptor, but got %d", nFiles) + } + + bd, ok := api.bundleDirs[req.CID] + if !ok { + // If there's no entry in bundleDirs for the container ID, this + // is the root container. + bd = api.rootBundleDir + } + + // Relative paths are served relative to the bundle directory. + absDir := req.Dir + if !filepath.IsAbs(absDir) { + absDir = filepath.Join(bd, req.Dir) + } + + // Create the attach point and start serving. + at := NewAttachPoint(absDir, Config{ + ROMount: req.IsReadOnly, + LazyOpenForWrite: true, + }) + api.serve(at, int(req.FilePayload.Files[0].Fd())) + + return nil +} + +// serve begins serving a directory via a file descriptor. +func (api *api) serve(at p9.Attacher, ioFD int) { + api.p9wg.Add(1) + go func(ioFD int, at p9.Attacher) { + socket, err := unet.NewSocket(ioFD) + if err != nil { + panic(fmt.Sprintf("err creating server on FD %d: %v", ioFD, err)) + } + s := p9.NewServer(at) + if err := s.Handle(socket); err != nil { + panic(fmt.Sprintf("P9 server returned error. Gofer is shutting down. FD: %d, err: %v", ioFD, err)) + } + api.p9wg.Done() + }(ioFD, at) +} diff --git a/runsc/fsgofer/fsgofer.go b/runsc/fsgofer/fsgofer.go index 52cdc91a2..38263896a 100644 --- a/runsc/fsgofer/fsgofer.go +++ b/runsc/fsgofer/fsgofer.go @@ -66,6 +66,11 @@ func (f fileType) String() string { return "unknown" } +// ControlSocketAddr generates an abstract unix socket name for the given id. +func ControlSocketAddr(id string) string { + return fmt.Sprintf("\x00runsc-gofer.%s", id) +} + // Config sets configuration options for each attach point. type Config struct { // ROMount is set to true if this is a readonly mount. diff --git a/runsc/sandbox/BUILD b/runsc/sandbox/BUILD index a961c3cc7..cdacc5e22 100644 --- a/runsc/sandbox/BUILD +++ b/runsc/sandbox/BUILD @@ -21,6 +21,7 @@ go_library( "//pkg/sentry/control", "//pkg/urpc", "//runsc/boot", + "//runsc/fsgofer", "//runsc/specutils", "@com_github_kr_pty//:go_default_library", "@com_github_opencontainers_runtime-spec//specs-go:go_default_library", diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 2b043d412..83cc94dc4 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -31,6 +31,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/control" "gvisor.googlesource.com/gvisor/pkg/urpc" "gvisor.googlesource.com/gvisor/runsc/boot" + "gvisor.googlesource.com/gvisor/runsc/fsgofer" "gvisor.googlesource.com/gvisor/runsc/specutils" ) @@ -84,7 +85,7 @@ func Create(id string, spec *specs.Spec, conf *boot.Config, bundleDir, consoleSo // StartRoot starts running the root container process inside the sandbox. func (s *Sandbox) StartRoot(spec *specs.Spec, conf *boot.Config) error { log.Debugf("Start root sandbox %q, pid: %d", s.ID, s.Pid) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return err } @@ -104,21 +105,67 @@ func (s *Sandbox) StartRoot(spec *specs.Spec, conf *boot.Config) error { return nil } +// CreateChild creates a non-root container inside the sandbox. +func (s *Sandbox) CreateChild(cid, bundleDir string) error { + log.Debugf("Create non-root container sandbox %q, pid: %d for container %q with bundle directory %q", s.ID, s.Pid, cid, bundleDir) + + // Connect to the gofer and prepare it to serve from bundleDir for this + // container. + goferConn, err := s.goferConnect() + if err != nil { + return fmt.Errorf("couldn't connect to gofer: %v", err) + } + defer goferConn.Close() + goferReq := fsgofer.AddBundleDirsRequest{BundleDirs: map[string]string{cid: bundleDir}} + if err := goferConn.Call(fsgofer.AddBundleDirs, &goferReq, nil); err != nil { + return fmt.Errorf("error serving new filesystem for non-root container %v: %v", goferReq, err) + } + + return nil +} + // Start starts running a non-root container inside the sandbox. func (s *Sandbox) Start(spec *specs.Spec, conf *boot.Config, cid string) error { log.Debugf("Start non-root container sandbox %q, pid: %d", s.ID, s.Pid) - conn, err := s.connect() + + sandboxConn, err := s.sandboxConnect() + if err != nil { + return fmt.Errorf("couldn't connect to sandbox: %v", err) + } + defer sandboxConn.Close() + goferConn, err := s.goferConnect() + if err != nil { + return fmt.Errorf("couldn't connect to gofer: %v", err) + } + defer goferConn.Close() + + // Create socket that connects the sandbox and gofer. + sandEnd, goferEnd, err := createSocketPair() if err != nil { return err } - defer conn.Close() + defer sandEnd.Close() + defer goferEnd.Close() + + // Tell the Gofer about the new filesystem it needs to serve. + goferReq := fsgofer.ServeDirectoryRequest{ + Dir: spec.Root.Path, + IsReadOnly: spec.Root.Readonly, + CID: cid, + FilePayload: urpc.FilePayload{Files: []*os.File{goferEnd}}, + } + if err := goferConn.Call(fsgofer.ServeDirectory, &goferReq, nil); err != nil { + return fmt.Errorf("error serving new filesystem for non-root container %v: %v", goferReq, err) + } + // Start running the container. args := boot.StartArgs{ - Spec: spec, - Conf: conf, - CID: cid, + Spec: spec, + Conf: conf, + CID: cid, + FilePayload: urpc.FilePayload{Files: []*os.File{sandEnd}}, } - if err := conn.Call(boot.ContainerStart, args, nil); err != nil { + if err := sandboxConn.Call(boot.ContainerStart, &args, nil); err != nil { return fmt.Errorf("error starting non-root container %v: %v", spec.Process.Args, err) } @@ -142,7 +189,7 @@ func (s *Sandbox) Restore(cid string, spec *specs.Spec, conf *boot.Config, f str SandboxID: s.ID, } - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return err } @@ -165,7 +212,7 @@ func (s *Sandbox) Restore(cid string, spec *specs.Spec, conf *boot.Config, f str // given container in this sandbox. func (s *Sandbox) Processes(cid string) ([]*control.Process, error) { log.Debugf("Getting processes for container %q in sandbox %q", cid, s.ID) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return nil, err } @@ -183,7 +230,7 @@ func (s *Sandbox) Processes(cid string) ([]*control.Process, error) { // Execute runs the specified command in the container. func (s *Sandbox) Execute(cid string, e *control.ExecArgs) (syscall.WaitStatus, error) { log.Debugf("Executing new process in container %q in sandbox %q", cid, s.ID) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return 0, s.connError(err) } @@ -203,7 +250,7 @@ func (s *Sandbox) Execute(cid string, e *control.ExecArgs) (syscall.WaitStatus, // Event retrieves stats about the sandbox such as memory and CPU utilization. func (s *Sandbox) Event(cid string) (*boot.Event, error) { log.Debugf("Getting events for container %q in sandbox %q", cid, s.ID) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return nil, err } @@ -219,7 +266,7 @@ func (s *Sandbox) Event(cid string) (*boot.Event, error) { return &e, nil } -func (s *Sandbox) connect() (*urpc.Client, error) { +func (s *Sandbox) sandboxConnect() (*urpc.Client, error) { log.Debugf("Connecting to sandbox %q", s.ID) conn, err := client.ConnectTo(boot.ControlSocketAddr(s.ID)) if err != nil { @@ -228,6 +275,15 @@ func (s *Sandbox) connect() (*urpc.Client, error) { return conn, nil } +func (s *Sandbox) goferConnect() (*urpc.Client, error) { + log.Debugf("Connecting to gofer for sandbox %q", s.ID) + conn, err := client.ConnectTo(fsgofer.ControlSocketAddr(s.ID)) + if err != nil { + return nil, s.connError(err) + } + return conn, nil +} + func (s *Sandbox) connError(err error) error { return fmt.Errorf("error connecting to control server at pid %d: %v", s.Pid, err) } @@ -244,31 +300,45 @@ func (s *Sandbox) createGoferProcess(spec *specs.Spec, conf *boot.Config, bundle // Add root mount and then add any other additional mounts. mountCount := 1 + + // Add additional mounts. for _, m := range spec.Mounts { if specutils.Is9PMount(m) { mountCount++ } } - sandEnds := make([]*os.File, 0, mountCount) goferEnds := make([]*os.File, 0, mountCount) - for i := 0; i < mountCount; i++ { - // Create socket that connects the sandbox and gofer. - fds, err := syscall.Socketpair(syscall.AF_UNIX, syscall.SOCK_STREAM|syscall.SOCK_CLOEXEC, 0) + // nextFD is the next available file descriptor for the gofer process. + // It starts at 3 because 0-2 are used by stdin/stdout/stderr. + var nextFD int + for nextFD = 3; nextFD-3 < mountCount; nextFD++ { + sandEnd, goferEnd, err := createSocketPair() if err != nil { return nil, err } - sandEnds = append(sandEnds, os.NewFile(uintptr(fds[0]), "sandbox io fd")) - - goferEnd := os.NewFile(uintptr(fds[1]), "gofer io fd") defer goferEnd.Close() + sandEnds = append(sandEnds, sandEnd) goferEnds = append(goferEnds, goferEnd) + args = append(args, fmt.Sprintf("--io-fds=%d", nextFD)) + } - args = append(args, fmt.Sprintf("--io-fds=%d", 3+i)) + // Create and donate a file descriptor for the control server. + addr := fsgofer.ControlSocketAddr(s.ID) + serverFD, err := server.CreateSocket(addr) + if err != nil { + return nil, fmt.Errorf("error creating control server socket for sandbox %q: %v", s.ID, err) } + // Add the control server fd. + args = append(args, "--controller-fd="+strconv.Itoa(nextFD)) + nextFD++ + controllerFile := os.NewFile(uintptr(serverFD), "gofer_control_socket_server") + defer controllerFile.Close() + cmd := exec.Command(binPath, args...) cmd.ExtraFiles = goferEnds + cmd.ExtraFiles = append(cmd.ExtraFiles, controllerFile) // Setup any uid/gid mappings, and create or join the configured user // namespace so the gofer's view of the filesystem aligns with the @@ -286,6 +356,15 @@ func (s *Sandbox) createGoferProcess(spec *specs.Spec, conf *boot.Config, bundle return sandEnds, nil } +// createSocketPair creates a pair of files wrapping a socket pair. +func createSocketPair() (*os.File, *os.File, error) { + fds, err := syscall.Socketpair(syscall.AF_UNIX, syscall.SOCK_STREAM|syscall.SOCK_CLOEXEC, 0) + if err != nil { + return nil, nil, err + } + return os.NewFile(uintptr(fds[0]), "sandbox io fd"), os.NewFile(uintptr(fds[1]), "gofer io fd"), nil +} + // createSandboxProcess starts the sandbox as a subprocess by running the "boot" // command, passing in the bundle dir. func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bundleDir, consoleSocket, binPath string, ioFiles []*os.File) error { @@ -296,7 +375,9 @@ func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bund // Create control server socket here and donate FD to child process because // it may be in a different network namespace and won't be reachable from // outside. - fd, err := server.CreateSocket(boot.ControlSocketAddr(s.ID)) + addr := boot.ControlSocketAddr(s.ID) + fd, err := server.CreateSocket(addr) + log.Infof("creating sandbox process with addr: %s", addr) if err != nil { return fmt.Errorf("error creating control server socket for sandbox %q: %v", s.ID, err) } @@ -438,7 +519,7 @@ func (s *Sandbox) waitForCreated(timeout time.Duration) error { if err := specutils.WaitForReady(s.Pid, timeout, ready); err != nil { return fmt.Errorf("unexpected error waiting for sandbox %q, err: %v", s.ID, err) } - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return err } @@ -454,7 +535,7 @@ func (s *Sandbox) waitForCreated(timeout time.Duration) error { func (s *Sandbox) Wait(cid string) (syscall.WaitStatus, error) { log.Debugf("Waiting for container %q in sandbox %q", cid, s.ID) var ws syscall.WaitStatus - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return ws, err } @@ -471,7 +552,7 @@ func (s *Sandbox) Wait(cid string) (syscall.WaitStatus, error) { func (s *Sandbox) WaitPID(pid int32, cid string) (syscall.WaitStatus, error) { log.Debugf("Waiting for PID %d in sandbox %q", pid, s.ID) var ws syscall.WaitStatus - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return ws, err } @@ -536,7 +617,7 @@ func (s *Sandbox) Destroy() error { // Signal sends the signal to a container in the sandbox. func (s *Sandbox) Signal(cid string, sig syscall.Signal) error { log.Debugf("Signal sandbox %q", s.ID) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return err } @@ -556,7 +637,7 @@ func (s *Sandbox) Signal(cid string, sig syscall.Signal) error { // The statefile will be written to f. func (s *Sandbox) Checkpoint(cid string, f *os.File) error { log.Debugf("Checkpoint sandbox %q", s.ID) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return err } @@ -577,7 +658,7 @@ func (s *Sandbox) Checkpoint(cid string, f *os.File) error { // Pause sends the pause call for a container in the sandbox. func (s *Sandbox) Pause(cid string) error { log.Debugf("Pause sandbox %q", s.ID) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return err } @@ -592,7 +673,7 @@ func (s *Sandbox) Pause(cid string) error { // Resume sends the resume call for a container in the sandbox. func (s *Sandbox) Resume(cid string) error { log.Debugf("Resume sandbox %q", s.ID) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return err } @@ -630,7 +711,7 @@ func (s *Sandbox) IsRunning() bool { // Stacks collects and returns all stacks for the sandbox. func (s *Sandbox) Stacks() (string, error) { log.Debugf("Stacks sandbox %q", s.ID) - conn, err := s.connect() + conn, err := s.sandboxConnect() if err != nil { return "", err } -- cgit v1.2.3 From 9c407382b031f16160f83383ef8b0d419457829a Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Tue, 21 Aug 2018 11:15:15 -0700 Subject: Fix races in kernel.(*Task).Value() PiperOrigin-RevId: 209627180 Change-Id: Idc84afd38003427e411df6e75abfabd9174174e1 --- pkg/sentry/kernel/task.go | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 19029adb1..0f83c0a39 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -568,12 +568,18 @@ func (t *Task) Value(key interface{}) interface{} { case CtxPIDNamespace: return t.tg.pidns case CtxUTSNamespace: + t.mu.Lock() + defer t.mu.Unlock() return t.utsns case CtxIPCNamespace: + t.mu.Lock() + defer t.mu.Unlock() return t.ipcns case CtxTask: return t case auth.CtxCredentials: + t.mu.Lock() + defer t.mu.Unlock() return t.creds case context.CtxThreadGroupID: return int32(t.ThreadGroup().ID()) -- cgit v1.2.3 From 64403265a04aa0c8be3ebb652a09f6e2d7a84ca7 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Thu, 23 Aug 2018 16:31:25 -0700 Subject: Implement POSIX per-process interval timers. PiperOrigin-RevId: 210021612 Change-Id: If7c161e6fd08cf17942bfb6bc5a8d2c4e271c61e --- pkg/abi/linux/signal.go | 20 ++ pkg/abi/linux/time.go | 4 + pkg/sentry/arch/signal_amd64.go | 30 +++ pkg/sentry/kernel/BUILD | 2 +- pkg/sentry/kernel/kernel.go | 19 +- pkg/sentry/kernel/pending_signals.go | 17 +- pkg/sentry/kernel/pending_signals_state.go | 21 +- pkg/sentry/kernel/posixtimer.go | 306 +++++++++++++++++++++++++++++ pkg/sentry/kernel/ptrace.go | 2 +- pkg/sentry/kernel/task_exec.go | 16 ++ pkg/sentry/kernel/task_exit.go | 2 +- pkg/sentry/kernel/task_signals.go | 12 +- pkg/sentry/kernel/thread_group.go | 27 ++- pkg/sentry/kernel/time/time.go | 23 ++- pkg/sentry/syscalls/linux/linux64.go | 10 +- pkg/sentry/syscalls/linux/sys_timer.go | 85 ++++++++ pkg/sentry/syscalls/linux/sys_timerfd.go | 33 +--- 17 files changed, 579 insertions(+), 50 deletions(-) create mode 100644 pkg/sentry/kernel/posixtimer.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/signal.go b/pkg/abi/linux/signal.go index fed2a159f..b2c7230c4 100644 --- a/pkg/abi/linux/signal.go +++ b/pkg/abi/linux/signal.go @@ -209,3 +209,23 @@ const ( // POLL_HUP indicates that a device disconnected. POLL_HUP = SI_POLL | 6 ) + +// Sigevent represents struct sigevent. +type Sigevent struct { + Value uint64 // union sigval {int, void*} + Signo int32 + Notify int32 + + // struct sigevent here contains 48-byte union _sigev_un. However, only + // member _tid is significant to the kernel. + Tid int32 + UnRemainder [44]byte +} + +// Possible values for Sigevent.Notify, aka struct sigevent::sigev_notify. +const ( + SIGEV_SIGNAL = 0 + SIGEV_NONE = 1 + SIGEV_THREAD = 2 + SIGEV_THREAD_ID = 4 +) diff --git a/pkg/abi/linux/time.go b/pkg/abi/linux/time.go index 9109a2848..4569f4208 100644 --- a/pkg/abi/linux/time.go +++ b/pkg/abi/linux/time.go @@ -222,3 +222,7 @@ type Tms struct { CUTime ClockT CSTime ClockT } + +// TimerID represents type timer_t, which identifies a POSIX per-process +// interval timer. +type TimerID int32 diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index e81717e8b..9ca4c8ed1 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -176,6 +176,36 @@ func (s *SignalInfo) SetUid(val int32) { usermem.ByteOrder.PutUint32(s.Fields[4:8], uint32(val)) } +// Sigval returns the sigval field, which is aliased to both si_int and si_ptr. +func (s *SignalInfo) Sigval() uint64 { + return usermem.ByteOrder.Uint64(s.Fields[8:16]) +} + +// SetSigval mutates the sigval field. +func (s *SignalInfo) SetSigval(val uint64) { + usermem.ByteOrder.PutUint64(s.Fields[8:16], val) +} + +// TimerID returns the si_timerid field. +func (s *SignalInfo) TimerID() linux.TimerID { + return linux.TimerID(usermem.ByteOrder.Uint32(s.Fields[0:4])) +} + +// SetTimerID sets the si_timerid field. +func (s *SignalInfo) SetTimerID(val linux.TimerID) { + usermem.ByteOrder.PutUint32(s.Fields[0:4], uint32(val)) +} + +// Overrun returns the si_overrun field. +func (s *SignalInfo) Overrun() int32 { + return int32(usermem.ByteOrder.Uint32(s.Fields[4:8])) +} + +// SetOverrun sets the si_overrun field. +func (s *SignalInfo) SetOverrun(val int32) { + usermem.ByteOrder.PutUint32(s.Fields[4:8], uint32(val)) +} + // Addr returns the si_addr field. func (s *SignalInfo) Addr() uint64 { return usermem.ByteOrder.Uint64(s.Fields[0:8]) diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 69a3fbc45..a7b847e94 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -71,6 +71,7 @@ go_library( "pending_signals.go", "pending_signals_list.go", "pending_signals_state.go", + "posixtimer.go", "process_group_list.go", "ptrace.go", "rseq.go", @@ -114,7 +115,6 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel", imports = [ "gvisor.googlesource.com/gvisor/pkg/bpf", - "gvisor.googlesource.com/gvisor/pkg/sentry/arch", "gvisor.googlesource.com/gvisor/pkg/tcpip", ], visibility = ["//:sandbox"], diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index cb43fdcdc..33cd727c6 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -19,9 +19,11 @@ // Lock order (outermost locks must be taken first): // // Kernel.extMu -// TaskSet.mu -// SignalHandlers.mu -// Task.mu +// ThreadGroup.timerMu +// ktime.Timer.mu (for IntervalTimer) +// TaskSet.mu +// SignalHandlers.mu +// Task.mu // // Locking SignalHandlers.mu in multiple SignalHandlers requires locking // TaskSet.mu exclusively first. Locking Task.mu in multiple Tasks at the same @@ -706,8 +708,12 @@ func (k *Kernel) pauseTimeLocked() { if t == t.tg.leader { t.tg.tm.pause() } - // This means we'll iterate FDMaps shared by multiple tasks repeatedly, - // but ktime.Timer.Pause is idempotent so this is harmless. + // This means we'll iterate ThreadGroups and FDMaps shared by multiple + // tasks repeatedly, but ktime.Timer.Pause is idempotent so this is + // harmless. + for _, it := range t.tg.timers { + it.PauseTimer() + } if fdm := t.tr.FDMap; fdm != nil { for _, desc := range fdm.files { if tfd, ok := desc.file.FileOperations.(*timerfd.TimerOperations); ok { @@ -735,6 +741,9 @@ func (k *Kernel) resumeTimeLocked() { if t == t.tg.leader { t.tg.tm.resume() } + for _, it := range t.tg.timers { + it.ResumeTimer() + } if fdm := t.tr.FDMap; fdm != nil { for _, desc := range fdm.files { if tfd, ok := desc.file.FileOperations.(*timerfd.TimerOperations); ok { diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go index 06be5a7e1..bb5db0309 100644 --- a/pkg/sentry/kernel/pending_signals.go +++ b/pkg/sentry/kernel/pending_signals.go @@ -46,7 +46,7 @@ type pendingSignals struct { // Note that signals is zero-indexed, but signal 1 is the first valid // signal, so signals[0] contains signals with signo 1 etc. This offset is // usually handled by using Signal.index(). - signals [linux.SignalMaximum]pendingSignalQueue `state:".([]*arch.SignalInfo)"` + signals [linux.SignalMaximum]pendingSignalQueue `state:".([]savedPendingSignal)"` // Bit i of pendingSet is set iff there is at least one signal with signo // i+1 pending. @@ -66,13 +66,16 @@ type pendingSignal struct { // pendingSignalEntry links into a pendingSignalList. pendingSignalEntry *arch.SignalInfo + + // If timer is not nil, it is the IntervalTimer which sent this signal. + timer *IntervalTimer } // enqueue enqueues the given signal. enqueue returns true on success and false // on failure (if the given signal's queue is full). // // Preconditions: info represents a valid signal. -func (p *pendingSignals) enqueue(info *arch.SignalInfo) bool { +func (p *pendingSignals) enqueue(info *arch.SignalInfo, timer *IntervalTimer) bool { sig := linux.Signal(info.Signo) q := &p.signals[sig.Index()] if sig.IsStandard() { @@ -82,7 +85,7 @@ func (p *pendingSignals) enqueue(info *arch.SignalInfo) bool { } else if q.length >= rtSignalCap { return false } - q.pendingSignalList.PushBack(&pendingSignal{SignalInfo: info}) + q.pendingSignalList.PushBack(&pendingSignal{SignalInfo: info, timer: timer}) q.length++ p.pendingSet |= linux.SignalSetOf(sig) return true @@ -119,12 +122,20 @@ func (p *pendingSignals) dequeueSpecific(sig linux.Signal) *arch.SignalInfo { if q.length == 0 { p.pendingSet &^= linux.SignalSetOf(sig) } + if ps.timer != nil { + ps.timer.updateDequeuedSignalLocked(ps.SignalInfo) + } return ps.SignalInfo } // discardSpecific causes all pending signals with number sig to be discarded. func (p *pendingSignals) discardSpecific(sig linux.Signal) { q := &p.signals[sig.Index()] + for ps := q.pendingSignalList.Front(); ps != nil; ps = ps.Next() { + if ps.timer != nil { + ps.timer.signalRejectedLocked() + } + } q.pendingSignalList.Reset() q.length = 0 p.pendingSet &^= linux.SignalSetOf(sig) diff --git a/pkg/sentry/kernel/pending_signals_state.go b/pkg/sentry/kernel/pending_signals_state.go index af61f6e8e..6d90ed033 100644 --- a/pkg/sentry/kernel/pending_signals_state.go +++ b/pkg/sentry/kernel/pending_signals_state.go @@ -18,20 +18,29 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/arch" ) +// +stateify savable +type savedPendingSignal struct { + si *arch.SignalInfo + timer *IntervalTimer +} + // saveSignals is invoked by stateify. -func (p *pendingSignals) saveSignals() []*arch.SignalInfo { - var pending []*arch.SignalInfo +func (p *pendingSignals) saveSignals() []savedPendingSignal { + var pending []savedPendingSignal for _, q := range p.signals { for ps := q.pendingSignalList.Front(); ps != nil; ps = ps.Next() { - pending = append(pending, ps.SignalInfo) + pending = append(pending, savedPendingSignal{ + si: ps.SignalInfo, + timer: ps.timer, + }) } } return pending } // loadSignals is invoked by stateify. -func (p *pendingSignals) loadSignals(pending []*arch.SignalInfo) { - for _, si := range pending { - p.enqueue(si) +func (p *pendingSignals) loadSignals(pending []savedPendingSignal) { + for _, sps := range pending { + p.enqueue(sps.si, sps.timer) } } diff --git a/pkg/sentry/kernel/posixtimer.go b/pkg/sentry/kernel/posixtimer.go new file mode 100644 index 000000000..0ab958529 --- /dev/null +++ b/pkg/sentry/kernel/posixtimer.go @@ -0,0 +1,306 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "math" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// IntervalTimer represents a POSIX interval timer as described by +// timer_create(2). +// +// +stateify savable +type IntervalTimer struct { + timer *ktime.Timer + + // If target is not nil, it receives signo from timer expirations. If group + // is true, these signals are thread-group-directed. These fields are + // immutable. + target *Task + signo linux.Signal + id linux.TimerID + sigval uint64 + group bool + + // If sigpending is true, a signal to target is already queued, and timer + // expirations should increment overrunCur instead of sending another + // signal. sigpending is protected by target's signal mutex. (If target is + // nil, the timer will never send signals, so sigpending will be unused.) + sigpending bool + + // If sigorphan is true, timer's setting has been changed since sigpending + // last became true, such that overruns should no longer be counted in the + // pending signals si_overrun. sigorphan is protected by target's signal + // mutex. + sigorphan bool + + // overrunCur is the number of overruns that have occurred since the last + // time a signal was sent. overrunCur is protected by target's signal + // mutex. + overrunCur uint64 + + // Consider the last signal sent by this timer that has been dequeued. + // overrunLast is the number of overruns that occurred between when this + // signal was sent and when it was dequeued. Equivalently, overrunLast was + // the value of overrunCur when this signal was dequeued. overrunLast is + // protected by target's signal mutex. + overrunLast uint64 +} + +// DestroyTimer releases it's resources. +func (it *IntervalTimer) DestroyTimer() { + it.timer.Destroy() + it.timerSettingChanged() + // A destroyed IntervalTimer is still potentially reachable via a + // pendingSignal; nil out timer so that it won't be saved. + it.timer = nil +} + +func (it *IntervalTimer) timerSettingChanged() { + if it.target == nil { + return + } + it.target.tg.pidns.owner.mu.RLock() + defer it.target.tg.pidns.owner.mu.RUnlock() + it.target.tg.signalHandlers.mu.Lock() + defer it.target.tg.signalHandlers.mu.Unlock() + it.sigorphan = true + it.overrunCur = 0 + it.overrunLast = 0 +} + +// PauseTimer pauses the associated Timer. +func (it *IntervalTimer) PauseTimer() { + it.timer.Pause() +} + +// ResumeTimer resumes the associated Timer. +func (it *IntervalTimer) ResumeTimer() { + it.timer.Resume() +} + +// Preconditions: it.target's signal mutex must be locked. +func (it *IntervalTimer) updateDequeuedSignalLocked(si *arch.SignalInfo) { + it.sigpending = false + if it.sigorphan { + return + } + it.overrunLast = it.overrunCur + it.overrunCur = 0 + si.SetOverrun(saturateI32FromU64(it.overrunLast)) +} + +// Preconditions: it.target's signal mutex must be locked. +func (it *IntervalTimer) signalRejectedLocked() { + it.sigpending = false + if it.sigorphan { + return + } + it.overrunCur++ +} + +// Notify implements ktime.TimerListener.Notify. +func (it *IntervalTimer) Notify(exp uint64) { + if it.target == nil { + return + } + + it.target.tg.pidns.owner.mu.RLock() + defer it.target.tg.pidns.owner.mu.RUnlock() + it.target.tg.signalHandlers.mu.Lock() + defer it.target.tg.signalHandlers.mu.Unlock() + + if it.sigpending { + it.overrunCur += exp + return + } + + // sigpending must be set before sendSignalTimerLocked() so that it can be + // unset if the signal is discarded (in which case sendSignalTimerLocked() + // will return nil). + it.sigpending = true + it.sigorphan = false + it.overrunCur += exp - 1 + si := &arch.SignalInfo{ + Signo: int32(it.signo), + Code: arch.SignalInfoTimer, + } + si.SetTimerID(it.id) + si.SetSigval(it.sigval) + // si_overrun is set when the signal is dequeued. + if err := it.target.sendSignalTimerLocked(si, it.group, it); err != nil { + it.signalRejectedLocked() + } +} + +// Destroy implements ktime.TimerListener.Destroy. Users of Timer should call +// DestroyTimer instead. +func (it *IntervalTimer) Destroy() { +} + +// IntervalTimerCreate implements timer_create(2). +func (t *Task) IntervalTimerCreate(c ktime.Clock, sigev *linux.Sigevent) (linux.TimerID, error) { + t.tg.timerMu.Lock() + defer t.tg.timerMu.Unlock() + + // Allocate a timer ID. + var id linux.TimerID + end := t.tg.nextTimerID + for { + id = t.tg.nextTimerID + _, ok := t.tg.timers[id] + t.tg.nextTimerID++ + if t.tg.nextTimerID < 0 { + t.tg.nextTimerID = 0 + } + if !ok { + break + } + if t.tg.nextTimerID == end { + return 0, syserror.EAGAIN + } + } + + // "The implementation of the default case where evp [sic] is NULL is + // handled inside glibc, which invokes the underlying system call with a + // suitably populated sigevent structure." - timer_create(2). This is + // misleading; the timer_create syscall also handles a NULL sevp as + // described by the man page + // (kernel/time/posix-timers.c:sys_timer_create(), do_timer_create()). This + // must be handled here instead of the syscall wrapper since sigval is the + // timer ID, which isn't available until we allocate it in this function. + if sigev == nil { + sigev = &linux.Sigevent{ + Signo: int32(linux.SIGALRM), + Notify: linux.SIGEV_SIGNAL, + Value: uint64(id), + } + } + + // Construct the timer. + it := &IntervalTimer{ + id: id, + sigval: sigev.Value, + } + switch sigev.Notify { + case linux.SIGEV_NONE: + // leave it.target = nil + case linux.SIGEV_SIGNAL, linux.SIGEV_THREAD: + // POSIX SIGEV_THREAD semantics are implemented in userspace by libc; + // to the kernel, SIGEV_THREAD and SIGEV_SIGNAL are equivalent. (See + // Linux's kernel/time/posix-timers.c:good_sigevent().) + it.target = t.tg.leader + it.group = true + case linux.SIGEV_THREAD_ID: + t.tg.pidns.owner.mu.RLock() + target, ok := t.tg.pidns.tasks[ThreadID(sigev.Tid)] + t.tg.pidns.owner.mu.RUnlock() + if !ok || target.tg != t.tg { + return 0, syserror.EINVAL + } + it.target = target + default: + return 0, syserror.EINVAL + } + if sigev.Notify != linux.SIGEV_NONE { + it.signo = linux.Signal(sigev.Signo) + if !it.signo.IsValid() { + return 0, syserror.EINVAL + } + } + it.timer = ktime.NewTimer(c, it) + + t.tg.timers[id] = it + return id, nil +} + +// IntervalTimerDelete implements timer_delete(2). +func (t *Task) IntervalTimerDelete(id linux.TimerID) error { + t.tg.timerMu.Lock() + defer t.tg.timerMu.Unlock() + it := t.tg.timers[id] + if it == nil { + return syserror.EINVAL + } + delete(t.tg.timers, id) + it.DestroyTimer() + return nil +} + +// IntervalTimerSettime implements timer_settime(2). +func (t *Task) IntervalTimerSettime(id linux.TimerID, its linux.Itimerspec, abs bool) (linux.Itimerspec, error) { + t.tg.timerMu.Lock() + defer t.tg.timerMu.Unlock() + it := t.tg.timers[id] + if it == nil { + return linux.Itimerspec{}, syserror.EINVAL + } + + newS, err := ktime.SettingFromItimerspec(its, abs, it.timer.Clock()) + if err != nil { + return linux.Itimerspec{}, err + } + tm, oldS := it.timer.SwapAnd(newS, it.timerSettingChanged) + its = ktime.ItimerspecFromSetting(tm, oldS) + return its, nil +} + +// IntervalTimerGettime implements timer_gettime(2). +func (t *Task) IntervalTimerGettime(id linux.TimerID) (linux.Itimerspec, error) { + t.tg.timerMu.Lock() + defer t.tg.timerMu.Unlock() + it := t.tg.timers[id] + if it == nil { + return linux.Itimerspec{}, syserror.EINVAL + } + + tm, s := it.timer.Get() + its := ktime.ItimerspecFromSetting(tm, s) + return its, nil +} + +// IntervalTimerGetoverrun implements timer_getoverrun(2). +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) IntervalTimerGetoverrun(id linux.TimerID) (int32, error) { + t.tg.timerMu.Lock() + defer t.tg.timerMu.Unlock() + it := t.tg.timers[id] + if it == nil { + return 0, syserror.EINVAL + } + // By timer_create(2) invariant, either it.target == nil (in which case + // it.overrunLast is immutably 0) or t.tg == it.target.tg; and the fact + // that t is executing timer_getoverrun(2) means that t.tg can't be + // completing execve, so t.tg.signalHandlers can't be changing, allowing us + // to lock t.tg.signalHandlers.mu without holding the TaskSet mutex. + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + // This is consistent with Linux after 78c9c4dfbf8c ("posix-timers: + // Sanitize overrun handling"). + return saturateI32FromU64(it.overrunLast), nil +} + +func saturateI32FromU64(x uint64) int32 { + if x > math.MaxInt32 { + return math.MaxInt32 + } + return int32(x) +} diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index e9e69004d..1a0d1876d 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -627,7 +627,7 @@ func (t *Task) ptraceClone(kind ptraceCloneKind, child *Task, opts *CloneOptions // running, so we don't have to. child.pendingSignals.enqueue(&arch.SignalInfo{ Signo: int32(linux.SIGSTOP), - }) + }, nil) child.tg.signalHandlers.mu.Unlock() } } diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index 385299b24..bb3d0bd02 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -143,6 +143,22 @@ func (r *runSyscallAfterExecStop) execute(t *Task) taskRunState { oldTID = tracer.tg.pidns.tids[t] } t.promoteLocked() + // "POSIX timers are not preserved (timer_create(2))." - execve(2). Handle + // this first since POSIX timers are protected by the signal mutex, which + // we're about to change. Note that we have to stop and destroy timers + // without holding any mutexes to avoid circular lock ordering. + var its []*IntervalTimer + t.tg.signalHandlers.mu.Lock() + for _, it := range t.tg.timers { + its = append(its, it) + } + t.tg.timers = make(map[linux.TimerID]*IntervalTimer) + t.tg.signalHandlers.mu.Unlock() + t.tg.pidns.owner.mu.Unlock() + for _, it := range its { + it.DestroyTimer() + } + t.tg.pidns.owner.mu.Lock() // "During an execve(2), the dispositions of handled signals are reset to // the default; the dispositions of ignored signals are left unchanged. ... // [The] signal mask is preserved across execve(2). ... [The] pending diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index b16844e91..b37fcf4c1 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -129,7 +129,7 @@ func (t *Task) killLocked() { // enqueueing an actual siginfo, such that // kernel/signal.c:collect_signal() initializes si_code to SI_USER. Code: arch.SignalInfoUser, - }) + }, nil) t.interrupt() } diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 62ec530be..4a66bce6b 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -396,6 +396,10 @@ func (tg *ThreadGroup) SendTimerSignal(info *arch.SignalInfo, includeSys bool) e } func (t *Task) sendSignalLocked(info *arch.SignalInfo, group bool) error { + return t.sendSignalTimerLocked(info, group, nil) +} + +func (t *Task) sendSignalTimerLocked(info *arch.SignalInfo, group bool, timer *IntervalTimer) error { if t.exitState == TaskExitDead { return syserror.ESRCH } @@ -429,6 +433,9 @@ func (t *Task) sendSignalLocked(info *arch.SignalInfo, group bool) error { ignored := computeAction(sig, t.tg.signalHandlers.actions[sig]) == SignalActionIgnore if linux.SignalSetOf(sig)&t.tr.SignalMask == 0 && ignored && !t.hasTracer() { t.Debugf("Discarding ignored signal %d", sig) + if timer != nil { + timer.signalRejectedLocked() + } return nil } @@ -436,11 +443,14 @@ func (t *Task) sendSignalLocked(info *arch.SignalInfo, group bool) error { if group { q = &t.tg.pendingSignals } - if !q.enqueue(info) { + if !q.enqueue(info, timer) { if sig.IsRealtime() { return syserror.EAGAIN } t.Debugf("Discarding duplicate signal %d", sig) + if timer != nil { + timer.signalRejectedLocked() + } return nil } diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index 441b8a822..13dce08ce 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -155,6 +155,14 @@ type ThreadGroup struct { // tm contains process timers. TimerManager fields are immutable. tm TimerManager + // timers is the thread group's POSIX interval timers. nextTimerID is the + // TimerID at which allocation should begin searching for an unused ID. + // + // timers and nextTimerID are protected by timerMu. + timerMu sync.Mutex `state:"nosave"` + timers map[linux.TimerID]*IntervalTimer + nextTimerID linux.TimerID + // exitedCPUStats is the CPU usage for all exited tasks in the thread // group. exitedCPUStats is protected by the TaskSet mutex. exitedCPUStats usage.CPUStats @@ -218,6 +226,7 @@ func NewThreadGroup(ns *PIDNamespace, sh *SignalHandlers, terminationSignal linu limits: limits, } tg.tm = newTimerManager(tg, monotonicClock) + tg.timers = make(map[linux.TimerID]*IntervalTimer) tg.rscr.Store(&RSEQCriticalRegion{}) return tg } @@ -252,9 +261,23 @@ func (tg *ThreadGroup) Limits() *limits.LimitSet { // release releases the thread group's resources. func (tg *ThreadGroup) release() { - // This must be done without holding the TaskSet mutex since thread group - // timers call SendSignal with Timer.mu locked. + // These must be done without holding the TaskSet or signal mutexes since + // timers send signals with Timer.mu locked. + tg.tm.destroy() + + var its []*IntervalTimer + tg.pidns.owner.mu.Lock() + tg.signalHandlers.mu.Lock() + for _, it := range tg.timers { + its = append(its, it) + } + tg.timers = make(map[linux.TimerID]*IntervalTimer) // nil maps can't be saved + tg.signalHandlers.mu.Unlock() + tg.pidns.owner.mu.Unlock() + for _, it := range its { + it.DestroyTimer() + } } // forEachChildThreadGroupLocked indicates over all child ThreadGroups. diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go index 6eadd2878..1f6fed007 100644 --- a/pkg/sentry/kernel/time/time.go +++ b/pkg/sentry/kernel/time/time.go @@ -320,8 +320,8 @@ func SettingFromSpec(value time.Duration, interval time.Duration, c Clock) (Sett }, nil } -// SettingFromAbsSpec converts a (value, interval) pair to a Setting based on a -// reading from c. value is interpreted as an absolute time. +// SettingFromAbsSpec converts a (value, interval) pair to a Setting. value is +// interpreted as an absolute time. func SettingFromAbsSpec(value Time, interval time.Duration) (Setting, error) { if value.Before(ZeroTime) { return Setting{}, syserror.EINVAL @@ -336,6 +336,16 @@ func SettingFromAbsSpec(value Time, interval time.Duration) (Setting, error) { }, nil } +// SettingFromItimerspec converts a linux.Itimerspec to a Setting. If abs is +// true, its.Value is interpreted as an absolute time. Otherwise, it is +// interpreted as a time relative to c.Now(). +func SettingFromItimerspec(its linux.Itimerspec, abs bool, c Clock) (Setting, error) { + if abs { + return SettingFromAbsSpec(FromTimespec(its.Value), its.Interval.ToDuration()) + } + return SettingFromSpec(its.Value.ToDuration(), its.Interval.ToDuration(), c) +} + // SpecFromSetting converts a timestamp and a Setting to a (relative value, // interval) pair, as used by most Linux syscalls that return a struct // itimerval or struct itimerspec. @@ -346,6 +356,15 @@ func SpecFromSetting(now Time, s Setting) (value, period time.Duration) { return s.Next.Sub(now), s.Period } +// ItimerspecFromSetting converts a Setting to a linux.Itimerspec. +func ItimerspecFromSetting(now Time, s Setting) linux.Itimerspec { + val, iv := SpecFromSetting(now, s) + return linux.Itimerspec{ + Interval: linux.DurationToTimespec(iv), + Value: linux.DurationToTimespec(val), + } +} + // advancedTo returns an updated Setting and a number of expirations after // the associated Clock indicates a time of now. // diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index c102af101..4465549ad 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -266,11 +266,11 @@ var AMD64 = &kernel.SyscallTable{ 219: RestartSyscall, // 220: Semtimedop, TODO 221: Fadvise64, - // 222: TimerCreate, TODO - // 223: TimerSettime, TODO - // 224: TimerGettime, TODO - // 225: TimerGetoverrun, TODO - // 226: TimerDelete, TODO + 222: TimerCreate, + 223: TimerSettime, + 224: TimerGettime, + 225: TimerGetoverrun, + 226: TimerDelete, 227: ClockSettime, 228: ClockGettime, 229: ClockGetres, diff --git a/pkg/sentry/syscalls/linux/sys_timer.go b/pkg/sentry/syscalls/linux/sys_timer.go index 4ed077626..aaed75c81 100644 --- a/pkg/sentry/syscalls/linux/sys_timer.go +++ b/pkg/sentry/syscalls/linux/sys_timer.go @@ -166,3 +166,88 @@ func Alarm(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall return uintptr(sec), nil, nil } + +// TimerCreate implements linux syscall timer_create(2). +func TimerCreate(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + clockID := args[0].Int() + sevp := args[1].Pointer() + timerIDp := args[2].Pointer() + + c, err := getClock(t, clockID) + if err != nil { + return 0, nil, err + } + + var sev *linux.Sigevent + if sevp != 0 { + sev = &linux.Sigevent{} + if _, err = t.CopyIn(sevp, sev); err != nil { + return 0, nil, err + } + } + + id, err := t.IntervalTimerCreate(c, sev) + if err != nil { + return 0, nil, err + } + + if _, err := t.CopyOut(timerIDp, &id); err != nil { + t.IntervalTimerDelete(id) + return 0, nil, err + } + + return uintptr(id), nil, nil +} + +// TimerSettime implements linux syscall timer_settime(2). +func TimerSettime(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + timerID := linux.TimerID(args[0].Value) + flags := args[1].Int() + newValAddr := args[2].Pointer() + oldValAddr := args[3].Pointer() + + var newVal linux.Itimerspec + if _, err := t.CopyIn(newValAddr, &newVal); err != nil { + return 0, nil, err + } + oldVal, err := t.IntervalTimerSettime(timerID, newVal, flags&linux.TIMER_ABSTIME != 0) + if err != nil { + return 0, nil, err + } + if oldValAddr != 0 { + if _, err := t.CopyOut(oldValAddr, &oldVal); err != nil { + return 0, nil, err + } + } + return 0, nil, nil +} + +// TimerGettime implements linux syscall timer_gettime(2). +func TimerGettime(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + timerID := linux.TimerID(args[0].Value) + curValAddr := args[1].Pointer() + + curVal, err := t.IntervalTimerGettime(timerID) + if err != nil { + return 0, nil, err + } + _, err = t.CopyOut(curValAddr, &curVal) + return 0, nil, err +} + +// TimerGetoverrun implements linux syscall timer_getoverrun(2). +func TimerGetoverrun(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + timerID := linux.TimerID(args[0].Value) + + o, err := t.IntervalTimerGetoverrun(timerID) + if err != nil { + return 0, nil, err + } + return uintptr(o), nil, nil +} + +// TimerDelete implements linux syscall timer_delete(2). +func TimerDelete(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + timerID := linux.TimerID(args[0].Value) + return 0, nil, t.IntervalTimerDelete(timerID) +} diff --git a/pkg/sentry/syscalls/linux/sys_timerfd.go b/pkg/sentry/syscalls/linux/sys_timerfd.go index cb81d42b9..92c6a3d60 100644 --- a/pkg/sentry/syscalls/linux/sys_timerfd.go +++ b/pkg/sentry/syscalls/linux/sys_timerfd.go @@ -85,28 +85,18 @@ func TimerfdSettime(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kerne if _, err := t.CopyIn(newValAddr, &newVal); err != nil { return 0, nil, err } - var s ktime.Setting - var err error - if flags&linux.TFD_TIMER_ABSTIME != 0 { - s, err = ktime.SettingFromAbsSpec(ktime.FromTimespec(newVal.Value), - newVal.Interval.ToDuration()) - } else { - s, err = ktime.SettingFromSpec(newVal.Value.ToDuration(), - newVal.Interval.ToDuration(), tf.Clock()) - } + newS, err := ktime.SettingFromItimerspec(newVal, flags&linux.TFD_TIMER_ABSTIME != 0, tf.Clock()) if err != nil { return 0, nil, err } - valueNS, intervalNS := ktime.SpecFromSetting(tf.SetTime(s)) - if oldValAddr == 0 { - return 0, nil, nil - } - oldVal := linux.Itimerspec{ - Interval: linux.DurationToTimespec(intervalNS), - Value: linux.DurationToTimespec(valueNS), + tm, oldS := tf.SetTime(newS) + if oldValAddr != 0 { + oldVal := ktime.ItimerspecFromSetting(tm, oldS) + if _, err := t.CopyOut(oldValAddr, &oldVal); err != nil { + return 0, nil, err + } } - _, err = t.CopyOut(oldValAddr, &oldVal) - return 0, nil, err + return 0, nil, nil } // TimerfdGettime implements Linux syscall timerfd_gettime(2). @@ -125,11 +115,8 @@ func TimerfdGettime(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kerne return 0, nil, syserror.EINVAL } - valueNS, intervalNS := ktime.SpecFromSetting(tf.GetTime()) - curVal := linux.Itimerspec{ - Interval: linux.DurationToTimespec(intervalNS), - Value: linux.DurationToTimespec(valueNS), - } + tm, s := tf.GetTime() + curVal := ktime.ItimerspecFromSetting(tm, s) _, err := t.CopyOut(curValAddr, &curVal) return 0, nil, err } -- cgit v1.2.3 From f0492d45aa31e32f8a04b13b7bf53e0161e1afb6 Mon Sep 17 00:00:00 2001 From: Brian Geffon Date: Mon, 27 Aug 2018 17:20:36 -0700 Subject: Add /proc/sys/kernel/shm[all,max,mni]. PiperOrigin-RevId: 210459956 Change-Id: I51859b90fa967631e0a54a390abc3b5541fbee66 --- pkg/abi/linux/shm.go | 11 +++++++++++ pkg/sentry/fs/proc/sys.go | 6 ++++++ pkg/sentry/kernel/shm/shm.go | 34 ++++++++-------------------------- 3 files changed, 25 insertions(+), 26 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/shm.go b/pkg/abi/linux/shm.go index 9149ed094..f50b3c2e2 100644 --- a/pkg/abi/linux/shm.go +++ b/pkg/abi/linux/shm.go @@ -14,6 +14,8 @@ package linux +import "math" + // shmat(2) flags. Source: include/uapi/linux/shm.h const ( SHM_RDONLY = 010000 // Read-only access. @@ -38,6 +40,15 @@ const ( SHM_INFO = 14 ) +// SHM defaults as specified by linux. Source: include/uapi/linux/shm.h +const ( + SHMMIN = 1 + SHMMNI = 4096 + SHMMAX = math.MaxUint64 - 1<<24 + SHMALL = math.MaxUint64 - 1<<24 + SHMSEG = 4096 +) + // ShmidDS is equivalent to struct shmid64_ds. Source: // include/uapi/asm-generic/shmbuf.h type ShmidDS struct { diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index a2d36ca23..384b4ffe1 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -17,7 +17,9 @@ package proc import ( "fmt" "io" + "strconv" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile" @@ -102,6 +104,10 @@ func (p *proc) newKernelDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode d := &ramfs.Dir{} d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) d.AddChild(ctx, "hostname", p.newHostname(ctx, msrc)) + + d.AddChild(ctx, "shmmax", p.newStubProcFSFile(ctx, msrc, []byte(strconv.FormatUint(linux.SHMMAX, 10)))) + d.AddChild(ctx, "shmall", p.newStubProcFSFile(ctx, msrc, []byte(strconv.FormatUint(linux.SHMALL, 10)))) + d.AddChild(ctx, "shmmni", p.newStubProcFSFile(ctx, msrc, []byte(strconv.FormatUint(linux.SHMMNI, 10)))) return newFile(d, msrc, fs.SpecialDirectory, nil) } diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 1ac444094..77973951e 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -35,7 +35,6 @@ package shm import ( "fmt" - "math" "sync" "gvisor.googlesource.com/gvisor/pkg/abi/linux" @@ -52,23 +51,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/syserror" ) -// Various limits for shared memory segments. -const ( - // shmsTotalMaxPages is the system-wide limit on all shared memory segments, measured - // in number of pages. - shmsTotalMaxPages = math.MaxInt64 // SHMALL - - // shmMaxSize is the maximum size of a single segment, in bytes. - shmMaxSize = math.MaxInt64 // SHMMAX - - // shmMinSize is the minimum specifiable size of a segment, effectively - // yielding a size rounded up to the next page size. Measured in bytes. - shmMinSize = 1 // SHMMIN - - // shmsTotalMax is the maximum number of segments on the system. - shmsTotalMax = 4096 // SHMMNI -) - // Registry tracks all shared memory segments in an IPC namespace. The registry // provides the mechanisms for creating and finding segments, and reporting // global shm parameters. @@ -119,7 +101,7 @@ func (r *Registry) findByKey(key int32) *Shm { // FindOrCreate looks up or creates a segment in the registry. It's functionally // analogous to open(2). func (r *Registry) FindOrCreate(ctx context.Context, pid, key int32, size uint64, mode linux.FileMode, private, create, exclusive bool) (*Shm, error) { - if create && (size < shmMinSize || size > shmMaxSize) { + if create && (size < linux.SHMMIN || size > linux.SHMMAX) { // "A new segment was to be created and size is less than SHMMIN or // greater than SHMMAX." - man shmget(2) return nil, syserror.EINVAL @@ -128,7 +110,7 @@ func (r *Registry) FindOrCreate(ctx context.Context, pid, key int32, size uint64 r.mu.Lock() defer r.mu.Unlock() - if len(r.shms) >= shmsTotalMax { + if len(r.shms) >= linux.SHMMNI { // "All possible shared memory IDs have been taken (SHMMNI) ..." // - man shmget(2) return nil, syserror.ENOSPC @@ -179,7 +161,7 @@ func (r *Registry) FindOrCreate(ctx context.Context, pid, key int32, size uint64 return nil, syserror.EINVAL } - if numPages := sizeAligned / usermem.PageSize; r.totalPages+numPages > shmsTotalMaxPages { + if numPages := sizeAligned / usermem.PageSize; r.totalPages+numPages > linux.SHMALL { // "... allocating a segment of the requested size would cause the // system to exceed the system-wide limit on shared memory (SHMALL)." // - man shmget(2) @@ -245,11 +227,11 @@ func (r *Registry) newShm(ctx context.Context, pid, key int32, creator fs.FileOw // system. See shmctl(IPC_INFO). func (r *Registry) IPCInfo() *linux.ShmParams { return &linux.ShmParams{ - ShmMax: shmMaxSize, - ShmMin: shmMinSize, - ShmMni: shmsTotalMax, - ShmSeg: shmsTotalMax, // Linux also sets this to SHMMNI. - ShmAll: shmsTotalMaxPages, + ShmMax: linux.SHMMAX, + ShmMin: linux.SHMMIN, + ShmMni: linux.SHMMNI, + ShmSeg: linux.SHMSEG, + ShmAll: linux.SHMALL, } } -- cgit v1.2.3 From 52e6714146b46487a1260fab838a6ba193029845 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Tue, 28 Aug 2018 17:25:13 -0700 Subject: fasync: don't keep mutex after return PiperOrigin-RevId: 210637533 Change-Id: I3536c3f9efb54732a0d8ada8bc299142b2c1682f --- pkg/sentry/kernel/fasync/fasync.go | 1 + 1 file changed, 1 insertion(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/fasync/fasync.go b/pkg/sentry/kernel/fasync/fasync.go index 15218fb5a..69c7970fa 100644 --- a/pkg/sentry/kernel/fasync/fasync.go +++ b/pkg/sentry/kernel/fasync/fasync.go @@ -47,6 +47,7 @@ type FileAsync struct { func (a *FileAsync) Callback(e *waiter.Entry) { a.mu.Lock() if a.e.Callback == nil { + a.mu.Unlock() return } t := a.recipientT -- cgit v1.2.3 From b1c1afa3ccc499df3fd15814d2b6cf9005bc2ab1 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Fri, 31 Aug 2018 13:06:16 -0700 Subject: Delete the long-obsolete kernel.TaskMaybe interface. PiperOrigin-RevId: 211131855 Change-Id: Ia7799561ccd65d16269e0ae6f408ab53749bca37 --- pkg/sentry/kernel/task.go | 12 ------------ 1 file changed, 12 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 0f83c0a39..21be3120e 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -499,13 +499,6 @@ func (t *Task) afterLoad() { // struct. const copyScratchBufferLen = 52 -// TaskMaybe is the interface for extracting Tasks out of things which may be -// or contain Task objects. -type TaskMaybe interface { - // ExtractTask returns the Task. - ExtractTask() *Task -} - // CopyScratchBuffer returns a scratch buffer to be used in CopyIn/CopyOut // functions. It must only be used within those functions and can only be used // by the task goroutine; it exists to improve performance and thus @@ -525,11 +518,6 @@ func (t *Task) FutexWaiter() *futex.Waiter { return t.futexWaiter } -// ExtractTask implements TaskMaybe.ExtractTask. -func (t *Task) ExtractTask() *Task { - return t -} - // TaskContext returns t's TaskContext. // // Precondition: The caller must be running on the task goroutine, or t.mu must -- cgit v1.2.3 From 098046ba193b839d69c059f7a0e68c89409b4237 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Fri, 31 Aug 2018 13:57:02 -0700 Subject: Disintegrate kernel.TaskResources. This allows us to call kernel.FDMap.DecRef without holding mutexes cleanly. PiperOrigin-RevId: 211139657 Change-Id: Ie59d5210fb9282e1950e2e40323df7264a01bcec --- pkg/sentry/kernel/BUILD | 1 - pkg/sentry/kernel/kernel.go | 35 ++++++---- pkg/sentry/kernel/ptrace.go | 4 +- pkg/sentry/kernel/task.go | 131 +++++++++++++++++++++++++---------- pkg/sentry/kernel/task_clone.go | 64 ++++++++++++----- pkg/sentry/kernel/task_exec.go | 2 +- pkg/sentry/kernel/task_exit.go | 13 ++-- pkg/sentry/kernel/task_log.go | 2 +- pkg/sentry/kernel/task_resources.go | 132 ------------------------------------ pkg/sentry/kernel/task_signals.go | 24 +++---- pkg/sentry/kernel/task_start.go | 75 ++++++++++++-------- 11 files changed, 230 insertions(+), 253 deletions(-) delete mode 100644 pkg/sentry/kernel/task_resources.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index a7b847e94..0bc735550 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -95,7 +95,6 @@ go_library( "task_list.go", "task_log.go", "task_net.go", - "task_resources.go", "task_run.go", "task_sched.go", "task_signals.go", diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 33cd727c6..c2b5c7269 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -332,7 +332,8 @@ func (ts *TaskSet) flushWritesToFiles(ctx context.Context) error { ts.mu.RLock() defer ts.mu.RUnlock() for t := range ts.Root.tids { - if fdmap := t.FDMap(); fdmap != nil { + // We can skip locking Task.mu here since the kernel is paused. + if fdmap := t.fds; fdmap != nil { for _, desc := range fdmap.files { if flags := desc.file.Flags(); !flags.Write { continue @@ -381,7 +382,8 @@ func (ts *TaskSet) unregisterEpollWaiters() { ts.mu.RLock() defer ts.mu.RUnlock() for t := range ts.Root.tids { - if fdmap := t.FDMap(); fdmap != nil { + // We can skip locking Task.mu here since the kernel is paused. + if fdmap := t.fds; fdmap != nil { for _, desc := range fdmap.files { if desc.file != nil { if e, ok := desc.file.FileOperations.(*epoll.EventPoll); ok { @@ -625,20 +627,23 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { if err != nil { return nil, err } - tr := newTaskResources(args.FDMap, newFSContext(root, wd, args.Umask)) - // NewTask unconditionally takes ownership of tr, so we never have to call - // tr.release. + + // Take a reference on the FDMap, which will be transferred to + // TaskSet.NewTask(). + args.FDMap.IncRef() // Create the task. config := &TaskConfig{ - Kernel: k, - ThreadGroup: tg, - TaskContext: tc, - TaskResources: tr, - Credentials: args.Credentials, - UTSNamespace: args.UTSNamespace, - IPCNamespace: args.IPCNamespace, - AllowedCPUMask: sched.NewFullCPUSet(k.applicationCores), + Kernel: k, + ThreadGroup: tg, + TaskContext: tc, + FSContext: newFSContext(root, wd, args.Umask), + FDMap: args.FDMap, + Credentials: args.Credentials, + AllowedCPUMask: sched.NewFullCPUSet(k.applicationCores), + UTSNamespace: args.UTSNamespace, + IPCNamespace: args.IPCNamespace, + AbstractSocketNamespace: NewAbstractSocketNamespace(), // FIXME } t, err := k.tasks.NewTask(config) if err != nil { @@ -714,7 +719,7 @@ func (k *Kernel) pauseTimeLocked() { for _, it := range t.tg.timers { it.PauseTimer() } - if fdm := t.tr.FDMap; fdm != nil { + if fdm := t.fds; fdm != nil { for _, desc := range fdm.files { if tfd, ok := desc.file.FileOperations.(*timerfd.TimerOperations); ok { tfd.PauseTimer() @@ -744,7 +749,7 @@ func (k *Kernel) resumeTimeLocked() { for _, it := range t.tg.timers { it.ResumeTimer() } - if fdm := t.tr.FDMap; fdm != nil { + if fdm := t.fds; fdm != nil { for _, desc := range fdm.files { if tfd, ok := desc.file.FileOperations.(*timerfd.TimerOperations); ok { tfd.ResumeTimer() diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index 1a0d1876d..e21a25ae6 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -992,9 +992,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { if addr != linux.SignalSetSize { return syserror.EINVAL } - target.mu.Lock() - defer target.mu.Unlock() - _, err := t.CopyOut(data, target.tr.SignalMask) + _, err := t.CopyOut(data, target.SignalMask()) return err case linux.PTRACE_SETSIGMASK: diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 21be3120e..32db0bf48 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -99,6 +99,19 @@ type Task struct { // ThreadGroup.signalHandlers. pendingSignals pendingSignals + // signalMask is the set of signals whose delivery is currently blocked. + // + // signalMask is accessed using atomic memory operations, and is protected + // by the signal mutex (such that reading signalMask is safe if either the + // signal mutex is locked or if atomic memory operations are used, while + // writing signalMask requires both). signalMask is owned by the task + // goroutine. + signalMask linux.SignalSet + + // FIXME: An equivalent to task_struct::real_blocked is needed + // to prevent signals that are ignored, but transiently unblocked by + // sigtimedwait(2), from being dropped in Task.sendSignalTimerLocked. + // If haveSavedSignalMask is true, savedSignalMask is the signal mask that // should be applied after the task has either delivered one signal to a // user handler or is about to resume execution in the untrusted @@ -182,25 +195,30 @@ type Task struct { // syscallRestartBlock is exclusive to the task goroutine. syscallRestartBlock SyscallRestartBlock + // p provides the mechanism by which the task runs code in userspace. The p + // interface object is immutable. + p platform.Context `state:"nosave"` + + // k is the Kernel that this task belongs to. The k pointer is immutable. + k *Kernel + // mu protects some of the following fields. mu sync.Mutex `state:"nosave"` - // tc and tr form the majority of the task's data. + // tc holds task data provided by the ELF loader. // - // tc and tr are protected by mu. tc and tr are owned by the task - // goroutine. tr.signalMask is protected by the signal mutex and must be - // written using atomic memory operations (such that reading tr.signalMask - // is safe if the signal mutex is locked or if atomic memory operations are - // used), but is also owned by the task goroutine. + // tc is protected by mu, and is owned by the task goroutine. tc TaskContext - tr TaskResources - // p provides the mechanism by which the task runs code in userspace. The p - // interface object is immutable. - p platform.Context `state:"nosave"` + // fsc is the task's filesystem context. + // + // fsc is protected by mu, and is owned by the task goroutine. + fsc *FSContext - // k is the Kernel that this task belongs to. The k pointer is immutable. - k *Kernel + // fds is the task's file descriptor table. + // + // fds is protected by mu, and is owned by the task goroutine. + fds *FDMap // If vforkParent is not nil, it is the task that created this task with // vfork() or clone(CLONE_VFORK), and should have its vforkStop ended when @@ -351,6 +369,11 @@ type Task struct { // ipcns is protected by mu. ipcns *IPCNamespace + // abstractSockets tracks abstract sockets that are in use. + // + // abstractSockets is protected by mu. + abstractSockets *AbstractSocketNamespace + // parentDeathSignal is sent to this task's thread group when its parent exits. // // parentDeathSignal is protected by mu. @@ -518,29 +541,6 @@ func (t *Task) FutexWaiter() *futex.Waiter { return t.futexWaiter } -// TaskContext returns t's TaskContext. -// -// Precondition: The caller must be running on the task goroutine, or t.mu must -// be locked. -func (t *Task) TaskContext() *TaskContext { - return &t.tc -} - -// TaskResources returns t's TaskResources. -// -// Precondition: The caller must be running on the task goroutine, or t.mu must -// be locked. -func (t *Task) TaskResources() *TaskResources { - return &t.tr -} - -// WithMuLocked executes f with t.mu locked. -func (t *Task) WithMuLocked(f func(*Task)) { - t.mu.Lock() - defer t.mu.Unlock() - f(t) -} - // Kernel returns the Kernel containing t. func (t *Task) Kernel() *Kernel { return t.k @@ -572,7 +572,7 @@ func (t *Task) Value(key interface{}) interface{} { case context.CtxThreadGroupID: return int32(t.ThreadGroup().ID()) case fs.CtxRoot: - return t.FSContext().RootDirectory() + return t.fsc.RootDirectory() case inet.CtxStack: return t.NetworkContext() case ktime.CtxRealtimeClock: @@ -619,3 +619,62 @@ func (t *Task) SyscallRestartBlock() SyscallRestartBlock { t.syscallRestartBlock = nil return r } + +// IsChrooted returns true if the root directory of t's FSContext is not the +// root directory of t's MountNamespace. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) IsChrooted() bool { + realRoot := t.k.mounts.Root() + defer realRoot.DecRef() + root := t.fsc.RootDirectory() + if root != nil { + defer root.DecRef() + } + return root != realRoot +} + +// TaskContext returns t's TaskContext. +// +// Precondition: The caller must be running on the task goroutine, or t.mu must +// be locked. +func (t *Task) TaskContext() *TaskContext { + return &t.tc +} + +// FSContext returns t's FSContext. FSContext does not take an additional +// reference on the returned FSContext. +// +// Precondition: The caller must be running on the task goroutine, or t.mu must +// be locked. +func (t *Task) FSContext() *FSContext { + return t.fsc +} + +// FDMap returns t's FDMap. FDMap does not take an additional reference on the +// returned FDMap. +// +// Precondition: The caller must be running on the task goroutine, or t.mu must +// be locked. +func (t *Task) FDMap() *FDMap { + return t.fds +} + +// WithMuLocked executes f with t.mu locked. +func (t *Task) WithMuLocked(f func(*Task)) { + t.mu.Lock() + defer t.mu.Unlock() + f(t) +} + +// MountNamespace returns t's MountNamespace. MountNamespace does not take an +// additional reference on the returned MountNamespace. +func (t *Task) MountNamespace() *fs.MountNamespace { + return t.k.mounts +} + +// AbstractSockets returns t's AbstractSocketNamespace. +func (t *Task) AbstractSockets() *AbstractSocketNamespace { + return t.abstractSockets +} diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 526165af0..46c688b20 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -213,6 +213,22 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { tc.Arch.StateData().Regs.Fs_base = uint64(opts.TLS) } + var fsc *FSContext + if opts.NewFSContext { + fsc = t.fsc.Fork() + } else { + fsc = t.fsc + fsc.IncRef() + } + + var fds *FDMap + if opts.NewFiles { + fds = t.fds.Fork() + } else { + fds = t.fds + fds.IncRef() + } + pidns := t.tg.pidns if t.childPIDNamespace != nil { pidns = t.childPIDNamespace @@ -227,17 +243,21 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { } tg = NewThreadGroup(pidns, sh, opts.TerminationSignal, tg.limits.GetCopy(), t.k.monotonicClock) } + cfg := &TaskConfig{ - Kernel: t.k, - ThreadGroup: tg, - TaskContext: tc, - TaskResources: t.tr.Fork(!opts.NewFiles, !opts.NewFSContext), - Niceness: t.Niceness(), - Credentials: creds.Fork(), - NetworkNamespaced: t.netns, - AllowedCPUMask: t.CPUMask(), - UTSNamespace: utsns, - IPCNamespace: ipcns, + Kernel: t.k, + ThreadGroup: tg, + SignalMask: t.SignalMask(), + TaskContext: tc, + FSContext: fsc, + FDMap: fds, + Credentials: creds.Fork(), + Niceness: t.Niceness(), + NetworkNamespaced: t.netns, + AllowedCPUMask: t.CPUMask(), + UTSNamespace: utsns, + IPCNamespace: ipcns, + AbstractSocketNamespace: t.abstractSockets, } if opts.NewThreadGroup { cfg.Parent = t @@ -435,15 +455,17 @@ func (t *Task) Unshare(opts *SharingOptions) error { t.childPIDNamespace = t.tg.pidns.NewChild(t.UserNamespace()) } t.mu.Lock() - defer t.mu.Unlock() + // Can't defer unlock: DecRefs must occur without holding t.mu. if opts.NewNetworkNamespace { if !haveCapSysAdmin { + t.mu.Unlock() return syserror.EPERM } t.netns = true } if opts.NewUTSNamespace { if !haveCapSysAdmin { + t.mu.Unlock() return syserror.EPERM } // Note that this must happen after NewUserNamespace, so the @@ -452,21 +474,29 @@ func (t *Task) Unshare(opts *SharingOptions) error { } if opts.NewIPCNamespace { if !haveCapSysAdmin { + t.mu.Unlock() return syserror.EPERM } // Note that "If CLONE_NEWIPC is set, then create the process in a new IPC // namespace" t.ipcns = NewIPCNamespace(t.creds.UserNamespace) } + var oldfds *FDMap if opts.NewFiles { - oldFDMap := t.tr.FDMap - t.tr.FDMap = oldFDMap.Fork() - oldFDMap.DecRef() + oldfds = t.fds + t.fds = oldfds.Fork() } + var oldfsc *FSContext if opts.NewFSContext { - oldFS := t.tr.FSContext - t.tr.FSContext = oldFS.Fork() - oldFS.DecRef() + oldfsc = t.fsc + t.fsc = oldfsc.Fork() + } + t.mu.Unlock() + if oldfds != nil { + oldfds.DecRef() + } + if oldfsc != nil { + oldfsc.DecRef() } return nil } diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index bb3d0bd02..1b760aba4 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -194,7 +194,7 @@ func (r *runSyscallAfterExecStop) execute(t *Task) taskRunState { t.tg.pidns.owner.mu.Unlock() // Remove FDs with the CloseOnExec flag set. - t.FDMap().RemoveIf(func(file *fs.File, flags FDFlags) bool { + t.fds.RemoveIf(func(file *fs.File, flags FDFlags) bool { return flags.CloseOnExec }) diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index b37fcf4c1..a1b24e1c6 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -253,21 +253,22 @@ func (*runExitMain) execute(t *Task) taskRunState { } } - // Deactivate the address space before releasing the MM. + // Deactivate the address space and update max RSS before releasing the + // task's MM. t.Deactivate() - - // Update the max resident set size before releasing t.tc.mm. t.tg.pidns.owner.mu.Lock() t.updateRSSLocked() t.tg.pidns.owner.mu.Unlock() - - // Release all of the task's resources. t.mu.Lock() t.tc.release() - t.tr.release() t.mu.Unlock() + + // Releasing the MM unblocks a blocked CLONE_VFORK parent. t.unstopVforkParent() + t.fsc.DecRef() + t.fds.DecRef() + // If this is the last task to exit from the thread group, release the // thread group's resources. if lastExiter { diff --git a/pkg/sentry/kernel/task_log.go b/pkg/sentry/kernel/task_log.go index 18efacb19..1769da210 100644 --- a/pkg/sentry/kernel/task_log.go +++ b/pkg/sentry/kernel/task_log.go @@ -63,7 +63,7 @@ func (t *Task) DebugDumpState() { if mm := t.MemoryManager(); mm != nil { t.Debugf("Mappings:\n%s", mm) } - t.Debugf("FDMap:\n%s", t.FDMap()) + t.Debugf("FDMap:\n%s", t.fds) } // debugDumpRegisters logs register state at log level debug. diff --git a/pkg/sentry/kernel/task_resources.go b/pkg/sentry/kernel/task_resources.go deleted file mode 100644 index 0832bf989..000000000 --- a/pkg/sentry/kernel/task_resources.go +++ /dev/null @@ -1,132 +0,0 @@ -// Copyright 2018 Google Inc. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package kernel - -import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" -) - -// TaskResources is the subset of a task's data provided by its creator that is -// not provided by the loader. -// -// +stateify savable -type TaskResources struct { - // SignalMask is the set of signals whose delivery is currently blocked. - // - // FIXME: Determine if we also need RealSignalMask - SignalMask linux.SignalSet - - // FSContext is the filesystem context. - *FSContext - - // FDMap provides access to files to the task. - *FDMap - - // Tracks abstract sockets that are in use. - AbstractSockets *AbstractSocketNamespace -} - -// newTaskResources returns a new TaskResources, taking an additional reference -// on fdm. -func newTaskResources(fdm *FDMap, fc *FSContext) *TaskResources { - fdm.IncRef() - return &TaskResources{ - FDMap: fdm, - FSContext: fc, - AbstractSockets: NewAbstractSocketNamespace(), - } -} - -// release releases all resources held by the TaskResources. release is called -// by the task when it exits. -func (tr *TaskResources) release() { - tr.FDMap.DecRef() - tr.FDMap = nil - tr.FSContext.DecRef() - tr.FSContext = nil - tr.AbstractSockets = nil -} - -// Fork returns a duplicate of tr. -// -// FIXME: Preconditions: When tr is owned by a Task, that task's -// signal mutex must be locked, or Fork must be called by the task's goroutine. -func (tr *TaskResources) Fork(shareFiles bool, shareFSContext bool) *TaskResources { - var fdmap *FDMap - if shareFiles { - fdmap = tr.FDMap - fdmap.IncRef() - } else { - fdmap = tr.FDMap.Fork() - } - - var fsc *FSContext - if shareFSContext { - fsc = tr.FSContext - fsc.IncRef() - } else { - fsc = tr.FSContext.Fork() - } - - return &TaskResources{ - SignalMask: tr.SignalMask, - FDMap: fdmap, - FSContext: fsc, - AbstractSockets: tr.AbstractSockets, - } -} - -// FDMap returns t's FDMap. -// -// Preconditions: The caller must be running on the task goroutine, or t.mu -// must be locked. -func (t *Task) FDMap() *FDMap { - return t.tr.FDMap -} - -// FSContext returns t's FSContext. -// -// Preconditions: The caller must be running on the task goroutine, or t.mu -// must be locked. -func (t *Task) FSContext() *FSContext { - return t.tr.FSContext -} - -// MountNamespace returns t's MountNamespace. MountNamespace does not take an additional -// reference on the returned MountNamespace. -func (t *Task) MountNamespace() *fs.MountNamespace { - return t.k.mounts -} - -// AbstractSockets returns t's AbstractSocketNamespace. -func (t *Task) AbstractSockets() *AbstractSocketNamespace { - return t.tr.AbstractSockets -} - -// IsChrooted returns true if the root directory of t's FSContext is not the -// root directory of t's MountNamespace. -// -// Preconditions: The caller must be running on the task goroutine, or t.mu -// must be locked. -func (t *Task) IsChrooted() bool { - realRoot := t.k.mounts.Root() - defer realRoot.DecRef() - root := t.tr.FSContext.RootDirectory() - if root != nil { - defer root.DecRef() - } - return root != realRoot -} diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 4a66bce6b..58a1bc0bd 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -124,10 +124,10 @@ var StopSignals = linux.MakeSignalSet(linux.SIGSTOP, linux.SIGTSTP, linux.SIGTTI // // Preconditions: t.tg.signalHandlers.mu must be locked. func (t *Task) dequeueSignalLocked() *arch.SignalInfo { - if info := t.pendingSignals.dequeue(t.tr.SignalMask); info != nil { + if info := t.pendingSignals.dequeue(t.signalMask); info != nil { return info } - return t.tg.pendingSignals.dequeue(t.tr.SignalMask) + return t.tg.pendingSignals.dequeue(t.signalMask) } // TakeSignal returns a pending signal not blocked by mask. Signal handlers are @@ -252,7 +252,7 @@ func (t *Task) deliverSignalToHandler(info *arch.SignalInfo, act arch.SignalAct) // handler should run with the current mask, but sigreturn should restore // the saved one. st := &arch.Stack{t.Arch(), t.MemoryManager(), sp} - mask := t.tr.SignalMask + mask := t.signalMask if t.haveSavedSignalMask { mask = t.savedSignalMask } @@ -262,7 +262,7 @@ func (t *Task) deliverSignalToHandler(info *arch.SignalInfo, act arch.SignalAct) t.haveSavedSignalMask = false // Add our signal mask. - newMask := t.tr.SignalMask | act.Mask + newMask := t.signalMask | act.Mask if !act.IsNoDefer() { newMask |= linux.SignalSetOf(linux.Signal(info.Signo)) } @@ -431,7 +431,7 @@ func (t *Task) sendSignalTimerLocked(info *arch.SignalInfo, group bool, timer *I // Linux's kernel/signal.c:__send_signal() => prepare_signal() => // sig_ignored(). ignored := computeAction(sig, t.tg.signalHandlers.actions[sig]) == SignalActionIgnore - if linux.SignalSetOf(sig)&t.tr.SignalMask == 0 && ignored && !t.hasTracer() { + if linux.SignalSetOf(sig)&t.signalMask == 0 && ignored && !t.hasTracer() { t.Debugf("Discarding ignored signal %d", sig) if timer != nil { timer.signalRejectedLocked() @@ -515,7 +515,7 @@ func (tg *ThreadGroup) applySignalSideEffectsLocked(sig linux.Signal) { // Preconditions: The signal mutex must be locked. func (t *Task) canReceiveSignalLocked(sig linux.Signal) bool { // - Do not choose tasks that are blocking the signal. - if linux.SignalSetOf(sig)&t.tr.SignalMask != 0 { + if linux.SignalSetOf(sig)&t.signalMask != 0 { return false } // - No need to check Task.exitState, as the exit path sets every bit in the @@ -564,21 +564,21 @@ func (t *Task) forceSignal(sig linux.Signal, unconditional bool) { } func (t *Task) forceSignalLocked(sig linux.Signal, unconditional bool) { - blocked := linux.SignalSetOf(sig)&t.tr.SignalMask != 0 + blocked := linux.SignalSetOf(sig)&t.signalMask != 0 act := t.tg.signalHandlers.actions[sig] ignored := act.Handler == arch.SignalActIgnore if blocked || ignored || unconditional { act.Handler = arch.SignalActDefault t.tg.signalHandlers.actions[sig] = act if blocked { - t.setSignalMaskLocked(t.tr.SignalMask &^ linux.SignalSetOf(sig)) + t.setSignalMaskLocked(t.signalMask &^ linux.SignalSetOf(sig)) } } } // SignalMask returns a copy of t's signal mask. func (t *Task) SignalMask() linux.SignalSet { - return linux.SignalSet(atomic.LoadUint64((*uint64)(&t.tr.SignalMask))) + return linux.SignalSet(atomic.LoadUint64((*uint64)(&t.signalMask))) } // SetSignalMask sets t's signal mask. @@ -595,8 +595,8 @@ func (t *Task) SetSignalMask(mask linux.SignalSet) { // Preconditions: The signal mutex must be locked. func (t *Task) setSignalMaskLocked(mask linux.SignalSet) { - oldMask := t.tr.SignalMask - atomic.StoreUint64((*uint64)(&t.tr.SignalMask), uint64(mask)) + oldMask := t.signalMask + atomic.StoreUint64((*uint64)(&t.signalMask), uint64(mask)) // If the new mask blocks any signals that were not blocked by the old // mask, and at least one such signal is pending in tg.pendingSignals, and @@ -1076,7 +1076,7 @@ func (*runInterruptAfterSignalDeliveryStop) execute(t *Task) taskRunState { t.tg.signalHandlers.mu.Lock() t.tg.pidns.owner.mu.Unlock() // If the signal is masked, re-queue it. - if linux.SignalSetOf(sig)&t.tr.SignalMask != 0 { + if linux.SignalSetOf(sig)&t.signalMask != 0 { t.sendSignalLocked(info, false /* group */) t.tg.signalHandlers.mu.Unlock() return (*runInterrupt)(nil) diff --git a/pkg/sentry/kernel/task_start.go b/pkg/sentry/kernel/task_start.go index c97dee8fc..6ce99d268 100644 --- a/pkg/sentry/kernel/task_start.go +++ b/pkg/sentry/kernel/task_start.go @@ -15,6 +15,7 @@ package kernel import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" @@ -26,7 +27,7 @@ import ( // TaskConfig defines the configuration of a new Task (see below). type TaskConfig struct { // Kernel is the owning Kernel. - *Kernel + Kernel *Kernel // Parent is the new task's parent. Parent may be nil. Parent *Task @@ -36,13 +37,24 @@ type TaskConfig struct { InheritParent *Task // ThreadGroup is the ThreadGroup the new task belongs to. - *ThreadGroup + ThreadGroup *ThreadGroup - // TaskContext is the TaskContext of the new task. - *TaskContext + // SignalMask is the new task's initial signal mask. + SignalMask linux.SignalSet - // TaskResources is the TaskResources of the new task. - *TaskResources + // TaskContext is the TaskContext of the new task. Ownership of the + // TaskContext is transferred to TaskSet.NewTask, whether or not it + // succeeds. + TaskContext *TaskContext + + // FSContext is the FSContext of the new task. A reference must be held on + // FSContext, which is transferred to TaskSet.NewTask whether or not it + // succeeds. + FSContext *FSContext + + // FDMap is the FDMap of the new task. A reference must be held on FDMap, + // which is transferred to TaskSet.NewTask whether or not it succeeds. + FDMap *FDMap // Credentials is the Credentials of the new task. Credentials *auth.Credentials @@ -62,25 +74,27 @@ type TaskConfig struct { // IPCNamespace is the IPCNamespace of the new task. IPCNamespace *IPCNamespace + + // AbstractSocketNamespace is the AbstractSocketNamespace of the new task. + AbstractSocketNamespace *AbstractSocketNamespace } -// NewTask creates a new task defined by TaskConfig. -// Whether or not NewTask is successful, it takes ownership of both TaskContext -// and TaskResources of the TaskConfig. +// NewTask creates a new task defined by cfg. // // NewTask does not start the returned task; the caller must call Task.Start. func (ts *TaskSet) NewTask(cfg *TaskConfig) (*Task, error) { t, err := ts.newTask(cfg) if err != nil { cfg.TaskContext.release() - cfg.TaskResources.release() + cfg.FSContext.DecRef() + cfg.FDMap.DecRef() return nil, err } return t, nil } -// newTask is a helper for TaskSet.NewTask that only takes ownership of TaskContext -// and TaskResources of the TaskConfig if it succeeds. +// newTask is a helper for TaskSet.NewTask that only takes ownership of parts +// of cfg if it succeeds. func (ts *TaskSet) newTask(cfg *TaskConfig) (*Task, error) { tg := cfg.ThreadGroup tc := cfg.TaskContext @@ -90,23 +104,26 @@ func (ts *TaskSet) newTask(cfg *TaskConfig) (*Task, error) { parent: cfg.Parent, children: make(map[*Task]struct{}), }, - runState: (*runApp)(nil), - interruptChan: make(chan struct{}, 1), - signalStack: arch.SignalStack{Flags: arch.SignalStackFlagDisable}, - tc: *tc, - tr: *cfg.TaskResources, - p: cfg.Kernel.Platform.NewContext(), - k: cfg.Kernel, - ptraceTracees: make(map[*Task]struct{}), - allowedCPUMask: cfg.AllowedCPUMask.Copy(), - ioUsage: &usage.IO{}, - creds: cfg.Credentials, - niceness: cfg.Niceness, - netns: cfg.NetworkNamespaced, - utsns: cfg.UTSNamespace, - ipcns: cfg.IPCNamespace, - rseqCPU: -1, - futexWaiter: futex.NewWaiter(), + runState: (*runApp)(nil), + interruptChan: make(chan struct{}, 1), + signalMask: cfg.SignalMask, + signalStack: arch.SignalStack{Flags: arch.SignalStackFlagDisable}, + tc: *tc, + fsc: cfg.FSContext, + fds: cfg.FDMap, + p: cfg.Kernel.Platform.NewContext(), + k: cfg.Kernel, + ptraceTracees: make(map[*Task]struct{}), + allowedCPUMask: cfg.AllowedCPUMask.Copy(), + ioUsage: &usage.IO{}, + creds: cfg.Credentials, + niceness: cfg.Niceness, + netns: cfg.NetworkNamespaced, + utsns: cfg.UTSNamespace, + ipcns: cfg.IPCNamespace, + abstractSockets: cfg.AbstractSocketNamespace, + rseqCPU: -1, + futexWaiter: futex.NewWaiter(), } t.endStopCond.L = &t.tg.signalHandlers.mu t.ptraceTracer.Store((*Task)(nil)) -- cgit v1.2.3 From f8ccfbbed4875e65c78c849cd46afa882ba68ee3 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Fri, 31 Aug 2018 15:43:32 -0700 Subject: Document more task-goroutine-owned fields in kernel.Task. Task.creds can only be changed by the task's own set*id and execve syscalls, and Task namespaces can only be changed by the task's own unshare/setns syscalls. PiperOrigin-RevId: 211156279 Change-Id: I94d57105d34e8739d964400995a8a5d76306b2a0 --- pkg/sentry/kernel/task.go | 19 ++++++++----------- 1 file changed, 8 insertions(+), 11 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 32db0bf48..ae4fd7817 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -354,19 +354,19 @@ type Task struct { // creds is the task's credentials. // - // creds is protected by mu, however the value itself is immutable and - // can only be changed by a copy. After reading the pointer, access - // will proceed outside the scope of mu. + // creds is protected by mu, however the value itself is immutable and can + // only be changed by a copy. After reading the pointer, access will + // proceed outside the scope of mu. creds is owned by the task goroutine. creds *auth.Credentials // utsns is the task's UTS namespace. // - // utsns is protected by mu. + // utsns is protected by mu. utsns is owned by the task goroutine. utsns *UTSNamespace // ipcns is the task's IPC namespace. // - // ipcns is protected by mu. + // ipcns is protected by mu. ipcns is owned by the task goroutine. ipcns *IPCNamespace // abstractSockets tracks abstract sockets that are in use. @@ -547,6 +547,9 @@ func (t *Task) Kernel() *Kernel { } // Value implements context.Context.Value. +// +// Preconditions: The caller must be running on the task goroutine (as implied +// by the requirements of context.Context). func (t *Task) Value(key interface{}) interface{} { switch key { case CtxCanTrace: @@ -556,18 +559,12 @@ func (t *Task) Value(key interface{}) interface{} { case CtxPIDNamespace: return t.tg.pidns case CtxUTSNamespace: - t.mu.Lock() - defer t.mu.Unlock() return t.utsns case CtxIPCNamespace: - t.mu.Lock() - defer t.mu.Unlock() return t.ipcns case CtxTask: return t case auth.CtxCredentials: - t.mu.Lock() - defer t.mu.Unlock() return t.creds case context.CtxThreadGroupID: return int32(t.ThreadGroup().ID()) -- cgit v1.2.3 From c09f9acd7c7a2e85472b1ee47bf26f7c89ded43e Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Tue, 4 Sep 2018 09:18:00 -0700 Subject: Distinguish Element and Linker for ilist. Furthermore, allow for the specification of an ElementMapper. This allows a single "Element" type to exist on multiple inline lists, and work without having to embed the entry type. This is a requisite change for supporting a per-Inode list of Dirents. PiperOrigin-RevId: 211467497 Change-Id: If2768999b43e03fdaecf8ed15f435fe37518d163 --- pkg/ilist/BUILD | 7 ++- pkg/ilist/list.go | 108 ++++++++++++++++++++-------------- pkg/refs/BUILD | 15 ++++- pkg/refs/refcounter.go | 8 +-- pkg/sentry/fs/BUILD | 1 + pkg/sentry/kernel/BUILD | 4 ++ pkg/sentry/kernel/futex/BUILD | 1 + pkg/sentry/kernel/semaphore/BUILD | 1 + pkg/sentry/mm/BUILD | 1 + pkg/tcpip/network/fragmentation/BUILD | 1 + pkg/tcpip/transport/ping/BUILD | 1 + pkg/tcpip/transport/tcp/BUILD | 1 + pkg/tcpip/transport/udp/BUILD | 1 + 13 files changed, 100 insertions(+), 50 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/ilist/BUILD b/pkg/ilist/BUILD index b26a39132..1bd71b800 100644 --- a/pkg/ilist/BUILD +++ b/pkg/ilist/BUILD @@ -28,6 +28,7 @@ go_template_instance( prefix = "direct", template = ":generic_list", types = { + "Element": "*direct", "Linker": "*direct", }, ) @@ -47,6 +48,10 @@ go_template( srcs = [ "list.go", ], - opt_types = ["Linker"], + opt_types = [ + "Element", + "ElementMapper", + "Linker", + ], visibility = ["//visibility:public"], ) diff --git a/pkg/ilist/list.go b/pkg/ilist/list.go index a88b82196..4ae02eee9 100644 --- a/pkg/ilist/list.go +++ b/pkg/ilist/list.go @@ -21,12 +21,34 @@ package ilist // N.B. When substituted in a template instantiation, Linker doesn't need to // be an interface, and in most cases won't be. type Linker interface { - Next() Linker - Prev() Linker - SetNext(Linker) - SetPrev(Linker) + Next() Element + Prev() Element + SetNext(Element) + SetPrev(Element) } +// Element the item that is used at the API level. +// +// N.B. Like Linker, this is unlikely to be an interface in most cases. +type Element interface { + Linker +} + +// ElementMapper provides an identity mapping by default. +// +// This can be replaced to provide a struct that maps elements to linker +// objects, if they are not the same. An ElementMapper is not typically +// required if: Linker is left as is, Element is left as is, or Linker and +// Element are the same type. +type ElementMapper struct{} + +// linkerFor maps an Element to a Linker. +// +// This default implementation should be inlined. +// +//go:nosplit +func (ElementMapper) linkerFor(elem Element) Linker { return elem } + // List is an intrusive list. Entries can be added to or removed from the list // in O(1) time and with no additional memory allocations. // @@ -39,8 +61,8 @@ type Linker interface { // // +stateify savable type List struct { - head Linker - tail Linker + head Element + tail Element } // Reset resets list l to the empty state. @@ -55,22 +77,22 @@ func (l *List) Empty() bool { } // Front returns the first element of list l or nil. -func (l *List) Front() Linker { +func (l *List) Front() Element { return l.head } // Back returns the last element of list l or nil. -func (l *List) Back() Linker { +func (l *List) Back() Element { return l.tail } // PushFront inserts the element e at the front of list l. -func (l *List) PushFront(e Linker) { - e.SetNext(l.head) - e.SetPrev(nil) +func (l *List) PushFront(e Element) { + ElementMapper{}.linkerFor(e).SetNext(l.head) + ElementMapper{}.linkerFor(e).SetPrev(nil) if l.head != nil { - l.head.SetPrev(e) + ElementMapper{}.linkerFor(l.head).SetPrev(e) } else { l.tail = e } @@ -79,12 +101,12 @@ func (l *List) PushFront(e Linker) { } // PushBack inserts the element e at the back of list l. -func (l *List) PushBack(e Linker) { - e.SetNext(nil) - e.SetPrev(l.tail) +func (l *List) PushBack(e Element) { + ElementMapper{}.linkerFor(e).SetNext(nil) + ElementMapper{}.linkerFor(e).SetPrev(l.tail) if l.tail != nil { - l.tail.SetNext(e) + ElementMapper{}.linkerFor(l.tail).SetNext(e) } else { l.head = e } @@ -98,8 +120,8 @@ func (l *List) PushBackList(m *List) { l.head = m.head l.tail = m.tail } else if m.head != nil { - l.tail.SetNext(m.head) - m.head.SetPrev(l.tail) + ElementMapper{}.linkerFor(l.tail).SetNext(m.head) + ElementMapper{}.linkerFor(m.head).SetPrev(l.tail) l.tail = m.tail } @@ -109,46 +131,46 @@ func (l *List) PushBackList(m *List) { } // InsertAfter inserts e after b. -func (l *List) InsertAfter(b, e Linker) { - a := b.Next() - e.SetNext(a) - e.SetPrev(b) - b.SetNext(e) +func (l *List) InsertAfter(b, e Element) { + a := ElementMapper{}.linkerFor(b).Next() + ElementMapper{}.linkerFor(e).SetNext(a) + ElementMapper{}.linkerFor(e).SetPrev(b) + ElementMapper{}.linkerFor(b).SetNext(e) if a != nil { - a.SetPrev(e) + ElementMapper{}.linkerFor(a).SetPrev(e) } else { l.tail = e } } // InsertBefore inserts e before a. -func (l *List) InsertBefore(a, e Linker) { - b := a.Prev() - e.SetNext(a) - e.SetPrev(b) - a.SetPrev(e) +func (l *List) InsertBefore(a, e Element) { + b := ElementMapper{}.linkerFor(a).Prev() + ElementMapper{}.linkerFor(e).SetNext(a) + ElementMapper{}.linkerFor(e).SetPrev(b) + ElementMapper{}.linkerFor(a).SetPrev(e) if b != nil { - b.SetNext(e) + ElementMapper{}.linkerFor(b).SetNext(e) } else { l.head = e } } // Remove removes e from l. -func (l *List) Remove(e Linker) { - prev := e.Prev() - next := e.Next() +func (l *List) Remove(e Element) { + prev := ElementMapper{}.linkerFor(e).Prev() + next := ElementMapper{}.linkerFor(e).Next() if prev != nil { - prev.SetNext(next) + ElementMapper{}.linkerFor(prev).SetNext(next) } else { l.head = next } if next != nil { - next.SetPrev(prev) + ElementMapper{}.linkerFor(next).SetPrev(prev) } else { l.tail = prev } @@ -160,26 +182,26 @@ func (l *List) Remove(e Linker) { // // +stateify savable type Entry struct { - next Linker - prev Linker + next Element + prev Element } // Next returns the entry that follows e in the list. -func (e *Entry) Next() Linker { +func (e *Entry) Next() Element { return e.next } // Prev returns the entry that precedes e in the list. -func (e *Entry) Prev() Linker { +func (e *Entry) Prev() Element { return e.prev } // SetNext assigns 'entry' as the entry that follows e in the list. -func (e *Entry) SetNext(entry Linker) { - e.next = entry +func (e *Entry) SetNext(elem Element) { + e.next = elem } // SetPrev assigns 'entry' as the entry that precedes e in the list. -func (e *Entry) SetPrev(entry Linker) { - e.prev = entry +func (e *Entry) SetPrev(elem Element) { + e.prev = elem } diff --git a/pkg/refs/BUILD b/pkg/refs/BUILD index 3ea877ccf..98150ba8f 100644 --- a/pkg/refs/BUILD +++ b/pkg/refs/BUILD @@ -1,16 +1,29 @@ package(licenses = ["notice"]) # Apache 2.0 +load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +go_template_instance( + name = "weak_ref_list", + out = "weak_ref_list.go", + package = "refs", + prefix = "weakRef", + template = "//pkg/ilist:generic_list", + types = { + "Element": "*WeakRef", + "Linker": "*WeakRef", + }, +) + go_library( name = "refs", srcs = [ "refcounter.go", "refcounter_state.go", + "weak_ref_list.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/refs", visibility = ["//:sandbox"], - deps = ["//pkg/ilist"], ) go_test( diff --git a/pkg/refs/refcounter.go b/pkg/refs/refcounter.go index 0d44c2499..638a93bab 100644 --- a/pkg/refs/refcounter.go +++ b/pkg/refs/refcounter.go @@ -20,8 +20,6 @@ import ( "reflect" "sync" "sync/atomic" - - "gvisor.googlesource.com/gvisor/pkg/ilist" ) // RefCounter is the interface to be implemented by objects that are reference @@ -61,7 +59,7 @@ type WeakRefUser interface { // // +stateify savable type WeakRef struct { - ilist.Entry `state:"nosave"` + weakRefEntry `state:"nosave"` // obj is an atomic value that points to the refCounter. obj atomic.Value `state:".(savedReference)"` @@ -195,7 +193,7 @@ type AtomicRefCount struct { mu sync.Mutex `state:"nosave"` // weakRefs is our collection of weak references. - weakRefs ilist.List `state:"nosave"` + weakRefs weakRefList `state:"nosave"` } // ReadRefs returns the current number of references. The returned count is @@ -276,7 +274,7 @@ func (r *AtomicRefCount) DecRefWithDestructor(destroy func()) { // return false due to the reference count check. r.mu.Lock() for !r.weakRefs.Empty() { - w := r.weakRefs.Front().(*WeakRef) + w := r.weakRefs.Front() // Capture the callback because w cannot be touched // after it's zapped -- the owner is free it reuse it // after that. diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 18cd5ae8e..a949fffbf 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -78,6 +78,7 @@ go_template_instance( template = "//pkg/ilist:generic_list", types = { "Linker": "*Dirent", + "Element": "*Dirent", }, ) diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 0bc735550..7eb2bffeb 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -10,6 +10,7 @@ go_template_instance( prefix = "pendingSignal", template = "//pkg/ilist:generic_list", types = { + "Element": "*pendingSignal", "Linker": "*pendingSignal", }, ) @@ -21,6 +22,7 @@ go_template_instance( prefix = "processGroup", template = "//pkg/ilist:generic_list", types = { + "Element": "*ProcessGroup", "Linker": "*ProcessGroup", }, ) @@ -43,6 +45,7 @@ go_template_instance( prefix = "session", template = "//pkg/ilist:generic_list", types = { + "Element": "*Session", "Linker": "*Session", }, ) @@ -54,6 +57,7 @@ go_template_instance( prefix = "task", template = "//pkg/ilist:generic_list", types = { + "Element": "*Task", "Linker": "*Task", }, ) diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index b44a26974..0ff5b0a95 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -10,6 +10,7 @@ go_template_instance( prefix = "waiter", template = "//pkg/ilist:generic_list", types = { + "Element": "*Waiter", "Linker": "*Waiter", }, ) diff --git a/pkg/sentry/kernel/semaphore/BUILD b/pkg/sentry/kernel/semaphore/BUILD index e7fa44e2c..bdcf4ce5c 100644 --- a/pkg/sentry/kernel/semaphore/BUILD +++ b/pkg/sentry/kernel/semaphore/BUILD @@ -10,6 +10,7 @@ go_template_instance( prefix = "waiter", template = "//pkg/ilist:generic_list", types = { + "Element": "*waiter", "Linker": "*waiter", }, ) diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index bbdfae247..ad9231774 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -67,6 +67,7 @@ go_template_instance( prefix = "io", template = "//pkg/ilist:generic_list", types = { + "Element": "*ioResult", "Linker": "*ioResult", }, ) diff --git a/pkg/tcpip/network/fragmentation/BUILD b/pkg/tcpip/network/fragmentation/BUILD index 83b4d253f..aaabfcb9a 100644 --- a/pkg/tcpip/network/fragmentation/BUILD +++ b/pkg/tcpip/network/fragmentation/BUILD @@ -10,6 +10,7 @@ go_template_instance( prefix = "reassembler", template = "//pkg/ilist:generic_list", types = { + "Element": "*reassembler", "Linker": "*reassembler", }, ) diff --git a/pkg/tcpip/transport/ping/BUILD b/pkg/tcpip/transport/ping/BUILD index 117532fea..982b6795c 100644 --- a/pkg/tcpip/transport/ping/BUILD +++ b/pkg/tcpip/transport/ping/BUILD @@ -10,6 +10,7 @@ go_template_instance( prefix = "pingPacket", template = "//pkg/ilist:generic_list", types = { + "Element": "*pingPacket", "Linker": "*pingPacket", }, ) diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index 8b911c295..c7943f08e 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -10,6 +10,7 @@ go_template_instance( prefix = "segment", template = "//pkg/ilist:generic_list", types = { + "Element": "*segment", "Linker": "*segment", }, ) diff --git a/pkg/tcpip/transport/udp/BUILD b/pkg/tcpip/transport/udp/BUILD index 1a3a62d3d..4225e28dc 100644 --- a/pkg/tcpip/transport/udp/BUILD +++ b/pkg/tcpip/transport/udp/BUILD @@ -10,6 +10,7 @@ go_template_instance( prefix = "udpPacket", template = "//pkg/ilist:generic_list", types = { + "Element": "*udpPacket", "Linker": "*udpPacket", }, ) -- cgit v1.2.3 From 6516b5648b471951e8c4da7869531c9509ba1495 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Thu, 6 Sep 2018 13:46:45 -0700 Subject: createProcessArgs.RootFromContext should return process Root if it exists. It was always returning the MountNamespace root, which may be different from the process Root if the process is in a chroot environment. PiperOrigin-RevId: 211862181 Change-Id: I63bfeb610e2b0affa9fdbdd8147eba3c39014480 --- pkg/sentry/kernel/kernel.go | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index c2b5c7269..31a2f068d 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -553,10 +553,18 @@ func (ctx *createProcessContext) Value(key interface{}) interface{} { case auth.CtxCredentials: return ctx.args.Credentials case fs.CtxRoot: - if ctx.k.mounts == nil { - return nil + if ctx.args.Root != nil { + // Take a refernce on the root dirent that will be + // given to the caller. + ctx.args.Root.IncRef() + return ctx.args.Root } - return ctx.k.mounts.Root() + if ctx.k.mounts != nil { + // MountNamespace.Root() will take a reference on the + // root dirent for us. + return ctx.k.mounts.Root() + } + return nil case ktime.CtxRealtimeClock: return ctx.k.RealtimeClock() case limits.CtxLimits: -- cgit v1.2.3 From f895cb4d8b4b37a563b7a5b9dc92eae552084b44 Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Fri, 7 Sep 2018 10:44:50 -0700 Subject: Use root abstract socket namespace for exec PiperOrigin-RevId: 211999211 Change-Id: I5968dd1a8313d3e49bb6e6614e130107495de41d --- pkg/sentry/control/proc.go | 23 ++++++++------- pkg/sentry/kernel/kernel.go | 41 +++++++++++++++++--------- runsc/boot/loader.go | 72 +++++++++++++++++++-------------------------- 3 files changed, 70 insertions(+), 66 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go index 4848a5d2b..6949a3ae5 100644 --- a/pkg/sentry/control/proc.go +++ b/pkg/sentry/control/proc.go @@ -95,17 +95,18 @@ func (proc *Proc) Exec(args *ExecArgs, waitStatus *uint32) error { proc.Kernel.RootUserNamespace()) initArgs := kernel.CreateProcessArgs{ - Filename: args.Filename, - Argv: args.Argv, - Envv: args.Envv, - WorkingDirectory: args.WorkingDirectory, - Credentials: creds, - FDMap: fdm, - Umask: 0022, - Limits: l, - MaxSymlinkTraversals: linux.MaxSymlinkTraversals, - UTSNamespace: proc.Kernel.RootUTSNamespace(), - IPCNamespace: proc.Kernel.RootIPCNamespace(), + Filename: args.Filename, + Argv: args.Argv, + Envv: args.Envv, + WorkingDirectory: args.WorkingDirectory, + Credentials: creds, + FDMap: fdm, + Umask: 0022, + Limits: l, + MaxSymlinkTraversals: linux.MaxSymlinkTraversals, + UTSNamespace: proc.Kernel.RootUTSNamespace(), + IPCNamespace: proc.Kernel.RootIPCNamespace(), + AbstractSocketNamespace: proc.Kernel.RootAbstractSocketNamespace(), } ctx := initArgs.NewContext(proc.Kernel) mounter := fs.FileOwnerFromContext(ctx) diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 31a2f068d..bc41c3963 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -90,17 +90,18 @@ type Kernel struct { platform.Platform `state:"nosave"` // See InitKernelArgs for the meaning of these fields. - featureSet *cpuid.FeatureSet - timekeeper *Timekeeper - tasks *TaskSet - rootUserNamespace *auth.UserNamespace - networkStack inet.Stack `state:"nosave"` - applicationCores uint - useHostCores bool - extraAuxv []arch.AuxEntry - vdso *loader.VDSO - rootUTSNamespace *UTSNamespace - rootIPCNamespace *IPCNamespace + featureSet *cpuid.FeatureSet + timekeeper *Timekeeper + tasks *TaskSet + rootUserNamespace *auth.UserNamespace + networkStack inet.Stack `state:"nosave"` + applicationCores uint + useHostCores bool + extraAuxv []arch.AuxEntry + vdso *loader.VDSO + rootUTSNamespace *UTSNamespace + rootIPCNamespace *IPCNamespace + rootAbstractSocketNamespace *AbstractSocketNamespace // mounts holds the state of the virtual filesystem. mounts is initially // nil, and must be set by calling Kernel.SetRootMountNamespace before @@ -201,11 +202,14 @@ type InitKernelArgs struct { // Vdso holds the VDSO and its parameter page. Vdso *loader.VDSO - // RootUTSNamespace is the root UTS namepsace. + // RootUTSNamespace is the root UTS namespace. RootUTSNamespace *UTSNamespace - // RootIPCNamespace is the root IPC namepsace. + // RootIPCNamespace is the root IPC namespace. RootIPCNamespace *IPCNamespace + + // RootAbstractSocketNamespace is the root Abstract Socket namespace. + RootAbstractSocketNamespace *AbstractSocketNamespace } // Init initialize the Kernel with no tasks. @@ -231,6 +235,7 @@ func (k *Kernel) Init(args InitKernelArgs) error { k.rootUserNamespace = args.RootUserNamespace k.rootUTSNamespace = args.RootUTSNamespace k.rootIPCNamespace = args.RootIPCNamespace + k.rootAbstractSocketNamespace = args.RootAbstractSocketNamespace k.networkStack = args.NetworkStack k.applicationCores = args.ApplicationCores if args.UseHostCores { @@ -509,6 +514,9 @@ type CreateProcessArgs struct { // IPCNamespace is the initial IPC namespace. IPCNamespace *IPCNamespace + // AbstractSocketNamespace is the initial Abstract Socket namespace. + AbstractSocketNamespace *AbstractSocketNamespace + // Root optionally contains the dirent that serves as the root for the // process. If nil, the mount namespace's root is used as the process' // root. @@ -651,7 +659,7 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { AllowedCPUMask: sched.NewFullCPUSet(k.applicationCores), UTSNamespace: args.UTSNamespace, IPCNamespace: args.IPCNamespace, - AbstractSocketNamespace: NewAbstractSocketNamespace(), // FIXME + AbstractSocketNamespace: args.AbstractSocketNamespace, } t, err := k.tasks.NewTask(config) if err != nil { @@ -839,6 +847,11 @@ func (k *Kernel) RootIPCNamespace() *IPCNamespace { return k.rootIPCNamespace } +// RootAbstractSocketNamespace returns the root AbstractSocketNamespace. +func (k *Kernel) RootAbstractSocketNamespace() *AbstractSocketNamespace { + return k.rootAbstractSocketNamespace +} + // RootMountNamespace returns the MountNamespace. func (k *Kernel) RootMountNamespace() *fs.MountNamespace { k.extMu.Lock() diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index ae2226e12..540cd6188 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -143,6 +143,19 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console } tk.SetClocks(time.NewCalibratedClocks()) + if err := enableStrace(conf); err != nil { + return nil, fmt.Errorf("failed to enable strace: %v", err) + } + + // Create an empty network stack because the network namespace may be empty at + // this point. Netns is configured before Run() is called. Netstack is + // configured using a control uRPC message. Host network is configured inside + // Run(). + networkStack, err := newEmptyNetworkStack(conf, k) + if err != nil { + return nil, fmt.Errorf("failed to create network: %v", err) + } + // Create capabilities. caps, err := specutils.Capabilities(spec.Process.Capabilities) if err != nil { @@ -163,26 +176,6 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console caps, auth.NewRootUserNamespace()) - // Create user namespace. - // TODO: Not clear what domain name should be here. It is - // not configurable from runtime spec. - utsns := kernel.NewUTSNamespace(spec.Hostname, "", creds.UserNamespace) - - ipcns := kernel.NewIPCNamespace(creds.UserNamespace) - - if err := enableStrace(conf); err != nil { - return nil, fmt.Errorf("failed to enable strace: %v", err) - } - - // Create an empty network stack because the network namespace may be empty at - // this point. Netns is configured before Run() is called. Netstack is - // configured using a control uRPC message. Host network is configured inside - // Run(). - networkStack, err := newEmptyNetworkStack(conf, k) - if err != nil { - return nil, fmt.Errorf("failed to create network: %v", err) - } - // Initiate the Kernel object, which is required by the Context passed // to createVFS in order to mount (among other things) procfs. if err = k.Init(kernel.InitKernelArgs{ @@ -191,10 +184,11 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console RootUserNamespace: creds.UserNamespace, NetworkStack: networkStack, // TODO: use number of logical processors from cgroups. - ApplicationCores: uint(runtime.NumCPU()), - Vdso: vdso, - RootUTSNamespace: utsns, - RootIPCNamespace: ipcns, + ApplicationCores: uint(runtime.NumCPU()), + Vdso: vdso, + RootUTSNamespace: kernel.NewUTSNamespace(spec.Hostname, "", creds.UserNamespace), + RootIPCNamespace: kernel.NewIPCNamespace(creds.UserNamespace), + RootAbstractSocketNamespace: kernel.NewAbstractSocketNamespace(), }); err != nil { return nil, fmt.Errorf("error initializing kernel: %v", err) } @@ -244,7 +238,7 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console log.Infof("Panic signal set to %v(%d)", ps, conf.PanicSignal) } - procArgs, err := newProcess(spec, creds, utsns, ipcns, k) + procArgs, err := newProcess(spec, creds, k) if err != nil { return nil, fmt.Errorf("failed to create root process: %v", err) } @@ -265,7 +259,7 @@ func New(spec *specs.Spec, conf *Config, controllerFD int, ioFDs []int, console } // newProcess creates a process that can be run with kernel.CreateProcess. -func newProcess(spec *specs.Spec, creds *auth.Credentials, utsns *kernel.UTSNamespace, ipcns *kernel.IPCNamespace, k *kernel.Kernel) (kernel.CreateProcessArgs, error) { +func newProcess(spec *specs.Spec, creds *auth.Credentials, k *kernel.Kernel) (kernel.CreateProcessArgs, error) { // Create initial limits. ls, err := createLimitSet(spec) if err != nil { @@ -274,15 +268,16 @@ func newProcess(spec *specs.Spec, creds *auth.Credentials, utsns *kernel.UTSName // Create the process arguments. procArgs := kernel.CreateProcessArgs{ - Argv: spec.Process.Args, - Envv: spec.Process.Env, - WorkingDirectory: spec.Process.Cwd, // Defaults to '/' if empty. - Credentials: creds, - Umask: 0022, - Limits: ls, - MaxSymlinkTraversals: linux.MaxSymlinkTraversals, - UTSNamespace: utsns, - IPCNamespace: ipcns, + Argv: spec.Process.Args, + Envv: spec.Process.Env, + WorkingDirectory: spec.Process.Cwd, // Defaults to '/' if empty. + Credentials: creds, + Umask: 0022, + Limits: ls, + MaxSymlinkTraversals: linux.MaxSymlinkTraversals, + UTSNamespace: k.RootUTSNamespace(), + IPCNamespace: k.RootIPCNamespace(), + AbstractSocketNamespace: k.RootAbstractSocketNamespace(), } return procArgs, nil } @@ -421,12 +416,7 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config // TODO New containers should be started in new PID namespaces // when indicated by the spec. - procArgs, err := newProcess( - spec, - creds, - l.k.RootUTSNamespace(), - l.k.RootIPCNamespace(), - l.k) + procArgs, err := newProcess(spec, creds, l.k) if err != nil { return 0, fmt.Errorf("failed to create new process: %v", err) } -- cgit v1.2.3 From 172860a059ce2cff68aa85a3f66319ee52bdec13 Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Fri, 7 Sep 2018 16:58:19 -0700 Subject: Add 'Starting gVisor...' message to syslog This allows applications to verify they are running with gVisor. It also helps debugging when running with a mix of container runtimes. Closes #54 PiperOrigin-RevId: 212059457 Change-Id: I51d9595ee742b58c1f83f3902ab2e2ecbd5cedec --- pkg/sentry/kernel/syslog.go | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/syslog.go b/pkg/sentry/kernel/syslog.go index 125312b6a..6531bd5d2 100644 --- a/pkg/sentry/kernel/syslog.go +++ b/pkg/sentry/kernel/syslog.go @@ -86,14 +86,18 @@ func (s *syslog) Log() []byte { return m } - time := 0.0 + const format = "<6>[%11.6f] %s\n" + + s.msg = append(s.msg, []byte(fmt.Sprintf(format, 0.0, "Starting gVisor..."))...) + + time := 0.1 for i := 0; i < 10; i++ { time += rand.Float64() / 2 - s.msg = append(s.msg, []byte(fmt.Sprintf("<6>[%11.6f] %s\n", time, selectMessage()))...) + s.msg = append(s.msg, []byte(fmt.Sprintf(format, time, selectMessage()))...) } time += rand.Float64() / 2 - s.msg = append(s.msg, []byte(fmt.Sprintf("<6>[%11.6f] Ready!\n", time))...) + s.msg = append(s.msg, []byte(fmt.Sprintf(format, time, "Ready!"))...) // Return a copy. o := make([]byte, len(s.msg)) -- cgit v1.2.3 From adf8f339703922211886d3e5588160f65bc131b3 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Thu, 13 Sep 2018 15:15:33 -0700 Subject: Extend memory usage events to report mapped memory usage. PiperOrigin-RevId: 212887555 Change-Id: I3545383ce903cbe9f00d9b5288d9ef9a049b9f4f --- pkg/sentry/kernel/memevent/memory_events.go | 5 ++++- pkg/sentry/kernel/memevent/memory_events.proto | 4 ++++ 2 files changed, 8 insertions(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/memevent/memory_events.go b/pkg/sentry/kernel/memevent/memory_events.go index ecc9151de..f7a183a1d 100644 --- a/pkg/sentry/kernel/memevent/memory_events.go +++ b/pkg/sentry/kernel/memevent/memory_events.go @@ -94,5 +94,8 @@ func (m *MemoryEvents) emit() { snapshot, _ := usage.MemoryAccounting.Copy() total := totalPlatform + snapshot.Mapped - eventchannel.Emit(&pb.MemoryUsageEvent{Total: total}) + eventchannel.Emit(&pb.MemoryUsageEvent{ + Mapped: snapshot.Mapped, + Total: total, + }) } diff --git a/pkg/sentry/kernel/memevent/memory_events.proto b/pkg/sentry/kernel/memevent/memory_events.proto index e6e0bd628..abc565054 100644 --- a/pkg/sentry/kernel/memevent/memory_events.proto +++ b/pkg/sentry/kernel/memevent/memory_events.proto @@ -22,4 +22,8 @@ message MemoryUsageEvent { // The total memory usage of the sandboxed application in bytes, calculated // using the 'fast' method. uint64 total = 1; + + // Memory used to back memory-mapped regions for files in the application, in + // bytes. This corresponds to the usage.MemoryKind.Mapped memory type. + uint64 mapped = 2; } -- cgit v1.2.3 From 29a7271f5da9fdb7b4a9a6c9ea61421ce6844a73 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Thu, 13 Sep 2018 19:11:12 -0700 Subject: Plumb monotonic time to netstack Netstack needs to be portable, so this seems to be preferable to using raw system calls. PiperOrigin-RevId: 212917409 Change-Id: I7b2073e7db4b4bf75300717ca23aea4c15be944c --- pkg/sentry/kernel/kernel.go | 11 +++++++++++ pkg/tcpip/BUILD | 5 ++++- pkg/tcpip/tcpip.go | 11 +++-------- pkg/tcpip/time_unsafe.go | 43 +++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 61 insertions(+), 9 deletions(-) create mode 100644 pkg/tcpip/time_unsafe.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index bc41c3963..316612b37 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -943,6 +943,8 @@ func (k *Kernel) SetExitError(err error) { } } +var _ tcpip.Clock = (*Kernel)(nil) + // NowNanoseconds implements tcpip.Clock.NowNanoseconds. func (k *Kernel) NowNanoseconds() int64 { now, err := k.timekeeper.GetTime(sentrytime.Realtime) @@ -952,6 +954,15 @@ func (k *Kernel) NowNanoseconds() int64 { return now } +// NowMonotonic implements tcpip.Clock.NowMonotonic. +func (k *Kernel) NowMonotonic() int64 { + now, err := k.timekeeper.GetTime(sentrytime.Monotonic) + if err != nil { + panic("Kernel.NowMonotonic: " + err.Error()) + } + return now +} + // SupervisorContext returns a Context with maximum privileges in k. It should // only be used by goroutines outside the control of the emulated kernel // defined by e. diff --git a/pkg/tcpip/BUILD b/pkg/tcpip/BUILD index 5153bd3b4..daff9a0a0 100644 --- a/pkg/tcpip/BUILD +++ b/pkg/tcpip/BUILD @@ -4,7 +4,10 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") go_library( name = "tcpip", - srcs = ["tcpip.go"], + srcs = [ + "tcpip.go", + "time_unsafe.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip", visibility = ["//visibility:public"], deps = [ diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index 51360b11f..b1bd5117f 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -123,16 +123,11 @@ func (e ErrSaveRejection) Error() string { // time, but never for netstack internal timekeeping. type Clock interface { // NowNanoseconds returns the current real time as a number of - // nanoseconds since some epoch. + // nanoseconds since the Unix epoch. NowNanoseconds() int64 -} - -// StdClock implements Clock with the time package. -type StdClock struct{} -// NowNanoseconds implements Clock.NowNanoseconds. -func (*StdClock) NowNanoseconds() int64 { - return time.Now().UnixNano() + // NowMonotonic returns a monotonic time value. + NowMonotonic() int64 } // Address is a byte slice cast as a string that represents the address of a diff --git a/pkg/tcpip/time_unsafe.go b/pkg/tcpip/time_unsafe.go new file mode 100644 index 000000000..2102e9633 --- /dev/null +++ b/pkg/tcpip/time_unsafe.go @@ -0,0 +1,43 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// +build go1.9 +// +build !go1.12 + +package tcpip + +import ( + _ "time" // Used with go:linkname. + _ "unsafe" // Required for go:linkname. +) + +// StdClock implements Clock with the time package. +type StdClock struct{} + +var _ Clock = (*StdClock)(nil) + +//go:linkname now time.now +func now() (sec int64, nsec int32, mono int64) + +// NowNanoseconds implements Clock.NowNanoseconds. +func (*StdClock) NowNanoseconds() int64 { + sec, nsec, _ := now() + return sec*1e9 + int64(nsec) +} + +// NowMonotonic implements Clock.NowMonotonic. +func (*StdClock) NowMonotonic() int64 { + _, _, mono := now() + return mono +} -- cgit v1.2.3 From 0380bcb3a4125723dc5248f70174ff64fb1942a2 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Fri, 14 Sep 2018 11:09:41 -0700 Subject: Fix interaction between rt_sigtimedwait and ignored signals. PiperOrigin-RevId: 213011782 Change-Id: I716c6ea3c586b0c6c5a892b6390d2d11478bc5af --- pkg/sentry/kernel/task.go | 9 +++-- pkg/sentry/kernel/task_signals.go | 68 ++++++++++++++++++++++++--------- pkg/sentry/syscalls/linux/sys_signal.go | 55 +++----------------------- 3 files changed, 61 insertions(+), 71 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index ae4fd7817..2f6f825ac 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -108,9 +108,12 @@ type Task struct { // goroutine. signalMask linux.SignalSet - // FIXME: An equivalent to task_struct::real_blocked is needed - // to prevent signals that are ignored, but transiently unblocked by - // sigtimedwait(2), from being dropped in Task.sendSignalTimerLocked. + // If the task goroutine is currently executing Task.sigtimedwait, + // realSignalMask is the previous value of signalMask, which has temporarily + // been replaced by Task.sigtimedwait. Otherwise, realSignalMask is 0. + // + // realSignalMask is exclusive to the task goroutine. + realSignalMask linux.SignalSet // If haveSavedSignalMask is true, savedSignalMask is the signal mask that // should be applied after the task has either delivered one signal to a diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 58a1bc0bd..afb010f60 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -19,6 +19,7 @@ package kernel import ( "fmt" "sync/atomic" + "time" "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" @@ -119,25 +120,11 @@ var UnblockableSignals = linux.MakeSignalSet(linux.SIGKILL, linux.SIGSTOP) // StopSignals is the set of signals whose default action is SignalActionStop. var StopSignals = linux.MakeSignalSet(linux.SIGSTOP, linux.SIGTSTP, linux.SIGTTIN, linux.SIGTTOU) -// dequeueSignalLocked returns a pending unmasked signal. If there are no -// pending unmasked signals, dequeueSignalLocked returns nil. +// dequeueSignalLocked returns a pending signal that is *not* included in mask. +// If there are no pending unmasked signals, dequeueSignalLocked returns nil. // // Preconditions: t.tg.signalHandlers.mu must be locked. -func (t *Task) dequeueSignalLocked() *arch.SignalInfo { - if info := t.pendingSignals.dequeue(t.signalMask); info != nil { - return info - } - return t.tg.pendingSignals.dequeue(t.signalMask) -} - -// TakeSignal returns a pending signal not blocked by mask. Signal handlers are -// not affected. If there are no pending signals not blocked by mask, -// TakeSignal returns a nil SignalInfo. -func (t *Task) TakeSignal(mask linux.SignalSet) *arch.SignalInfo { - t.tg.pidns.owner.mu.RLock() - defer t.tg.pidns.owner.mu.RUnlock() - t.tg.signalHandlers.mu.Lock() - defer t.tg.signalHandlers.mu.Unlock() +func (t *Task) dequeueSignalLocked(mask linux.SignalSet) *arch.SignalInfo { if info := t.pendingSignals.dequeue(mask); info != nil { return info } @@ -294,6 +281,49 @@ func (t *Task) SignalReturn(rt bool) (*SyscallControl, error) { return ctrlResume, nil } +// Sigtimedwait implements the semantics of sigtimedwait(2). +// +// Preconditions: The caller must be running on the task goroutine. t.exitState +// < TaskExitZombie. +func (t *Task) Sigtimedwait(set linux.SignalSet, timeout time.Duration) (*arch.SignalInfo, error) { + // set is the set of signals we're interested in; invert it to get the set + // of signals to block. + mask := ^set &^ UnblockableSignals + + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + if info := t.dequeueSignalLocked(mask); info != nil { + return info, nil + } + + if timeout == 0 { + return nil, syserror.EAGAIN + } + + // Unblock signals we're waiting for. Remember the original signal mask so + // that Task.sendSignalTimerLocked doesn't discard ignored signals that + // we're temporarily unblocking. + t.realSignalMask = t.signalMask + t.setSignalMaskLocked(t.signalMask & mask) + + // Wait for a timeout or new signal. + t.tg.signalHandlers.mu.Unlock() + _, err := t.BlockWithTimeout(nil, true, timeout) + t.tg.signalHandlers.mu.Lock() + + // Restore the original signal mask. + t.setSignalMaskLocked(t.realSignalMask) + t.realSignalMask = 0 + + if info := t.dequeueSignalLocked(mask); info != nil { + return info, nil + } + if err == syserror.ETIMEDOUT { + return nil, syserror.EAGAIN + } + return nil, err +} + // SendSignal sends the given signal to t. // // The following errors may be returned: @@ -431,7 +461,7 @@ func (t *Task) sendSignalTimerLocked(info *arch.SignalInfo, group bool, timer *I // Linux's kernel/signal.c:__send_signal() => prepare_signal() => // sig_ignored(). ignored := computeAction(sig, t.tg.signalHandlers.actions[sig]) == SignalActionIgnore - if linux.SignalSetOf(sig)&t.signalMask == 0 && ignored && !t.hasTracer() { + if sigset := linux.SignalSetOf(sig); sigset&t.signalMask == 0 && sigset&t.realSignalMask == 0 && ignored && !t.hasTracer() { t.Debugf("Discarding ignored signal %d", sig) if timer != nil { timer.signalRejectedLocked() @@ -1010,7 +1040,7 @@ func (*runInterrupt) execute(t *Task) taskRunState { } // Are there signals pending? - if info := t.dequeueSignalLocked(); info != nil { + if info := t.dequeueSignalLocked(t.signalMask); info != nil { if linux.SignalSetOf(linux.Signal(info.Signo))&StopSignals != 0 && t.tg.groupStopPhase == groupStopNone { // Indicate that we've dequeued a stop signal before // unlocking the signal mutex; initiateGroupStop will check diff --git a/pkg/sentry/syscalls/linux/sys_signal.go b/pkg/sentry/syscalls/linux/sys_signal.go index 66ecb1299..ecdec5d3a 100644 --- a/pkg/sentry/syscalls/linux/sys_signal.go +++ b/pkg/sentry/syscalls/linux/sys_signal.go @@ -343,44 +343,6 @@ func Pause(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall return 0, nil, syserror.ConvertIntr(t.Block(nil), kernel.ERESTARTNOHAND) } -func sigtimedwait(t *kernel.Task, mask linux.SignalSet, timeout time.Duration) (*arch.SignalInfo, error) { - // Is it already pending? - if info := t.TakeSignal(^mask); info != nil { - return info, nil - } - - // No signals available immediately and asked not to wait. - if timeout == 0 { - return nil, syserror.EAGAIN - } - - // No signals available yet. Temporarily unblock the ones we are interested - // in then wait for either a timeout or a new signal. - oldmask := t.SignalMask() - t.SetSignalMask(oldmask &^ mask) - _, err := t.BlockWithTimeout(nil, true, timeout) - t.SetSignalMask(oldmask) - - // How did the wait go? - switch err { - case syserror.ErrInterrupted: - if info := t.TakeSignal(^mask); info != nil { - // Got one of the signals we were waiting for. - return info, nil - } - // Got a signal we weren't waiting for. - return nil, syserror.EINTR - case syserror.ETIMEDOUT: - // Timed out and still no signals. - return nil, syserror.EAGAIN - default: - // Some other error? Shouldn't be possible. The event channel - // passed to BlockWithTimeout was nil, so the only two ways the - // block could've ended are a timeout or an interrupt. - panic("unreachable") - } -} - // RtSigpending implements linux syscall rt_sigpending(2). func RtSigpending(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { addr := args[0].Pointer() @@ -415,23 +377,18 @@ func RtSigtimedwait(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kerne timeout = time.Duration(math.MaxInt64) } - si, err := sigtimedwait(t, mask, timeout) + si, err := t.Sigtimedwait(mask, timeout) if err != nil { return 0, nil, err } - if si != nil { - if siginfo != 0 { - si.FixSignalCodeForUser() - if _, err := t.CopyOut(siginfo, si); err != nil { - return 0, nil, err - } + if siginfo != 0 { + si.FixSignalCodeForUser() + if _, err := t.CopyOut(siginfo, si); err != nil { + return 0, nil, err } - return uintptr(si.Signo), nil, nil } - - // sigtimedwait's not supposed to return nil si and err... - return 0, nil, nil + return uintptr(si.Signo), nil, nil } // RtSigqueueinfo implements linux syscall rt_sigqueueinfo(2). -- cgit v1.2.3 From ab6fa44588233fa48d1ae0bf7d9b0d9e984a6af0 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Mon, 17 Sep 2018 13:35:00 -0700 Subject: Allow kernel.(*Task).Block to accept an extract only channel PiperOrigin-RevId: 213328293 Change-Id: I4164133e6f709ecdb89ffbb5f7df3324c273860a --- pkg/sentry/fs/lock/lock.go | 2 +- pkg/sentry/kernel/task_block.go | 6 +++--- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/lock/lock.go b/pkg/sentry/fs/lock/lock.go index e9b376eb6..439e645db 100644 --- a/pkg/sentry/fs/lock/lock.go +++ b/pkg/sentry/fs/lock/lock.go @@ -121,7 +121,7 @@ type Locks struct { // Blocker is the interface used for blocking locks. Passing a nil Blocker // will be treated as non-blocking. type Blocker interface { - Block(C chan struct{}) error + Block(C <-chan struct{}) error } const ( diff --git a/pkg/sentry/kernel/task_block.go b/pkg/sentry/kernel/task_block.go index 9fd24f134..6dc7b938e 100644 --- a/pkg/sentry/kernel/task_block.go +++ b/pkg/sentry/kernel/task_block.go @@ -95,7 +95,7 @@ func (t *Task) BlockWithDeadline(C chan struct{}, haveDeadline bool, deadline kt // Most clients should use BlockWithDeadline or BlockWithTimeout instead. // // Preconditions: The caller must be running on the task goroutine. -func (t *Task) BlockWithTimer(C chan struct{}, tchan <-chan struct{}) error { +func (t *Task) BlockWithTimer(C <-chan struct{}, tchan <-chan struct{}) error { return t.block(C, tchan) } @@ -104,13 +104,13 @@ func (t *Task) BlockWithTimer(C chan struct{}, tchan <-chan struct{}) error { // is interrupted. // // Preconditions: The caller must be running on the task goroutine. -func (t *Task) Block(C chan struct{}) error { +func (t *Task) Block(C <-chan struct{}) error { return t.block(C, nil) } // block blocks a task on one of many events. // N.B. defer is too expensive to be used here. -func (t *Task) block(C chan struct{}, timerChan <-chan struct{}) error { +func (t *Task) block(C <-chan struct{}, timerChan <-chan struct{}) error { // Fast path if the request is already done. select { case <-C: -- cgit v1.2.3 From bb88c187c5457df14fa78e5e6b6f48cbc90fb489 Mon Sep 17 00:00:00 2001 From: Kevin Krakauer Date: Mon, 17 Sep 2018 16:24:05 -0700 Subject: runsc: Enable waiting on exited processes. This makes `runsc wait` behave more like waitpid()/wait4() in that: - Once a process has run to completion, you can wait on it and get its exit code. - Processes not waited on will consume memory (like a zombie process) PiperOrigin-RevId: 213358916 Change-Id: I5b5eca41ce71eea68e447380df8c38361a4d1558 --- pkg/sentry/control/proc.go | 14 ++-- pkg/sentry/kernel/kernel.go | 17 ++--- runsc/boot/controller.go | 33 +++------ runsc/boot/loader.go | 114 +++++++++++++++++++++++++++----- runsc/boot/loader_test.go | 25 +++---- runsc/cmd/exec.go | 14 +++- runsc/cmd/wait.go | 4 +- runsc/container/container.go | 8 +-- runsc/container/container_test.go | 4 +- runsc/container/multi_container_test.go | 94 ++++++++++++++++++++++++-- runsc/sandbox/sandbox.go | 7 +- 11 files changed, 248 insertions(+), 86 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go index 19bc76f5c..68d3b179b 100644 --- a/pkg/sentry/control/proc.go +++ b/pkg/sentry/control/proc.go @@ -87,7 +87,7 @@ type ExecArgs struct { // Exec runs a new task. func (proc *Proc) Exec(args *ExecArgs, waitStatus *uint32) error { - newTG, err := proc.execAsync(args) + newTG, _, err := proc.execAsync(args) if err != nil { return err } @@ -100,11 +100,13 @@ func (proc *Proc) Exec(args *ExecArgs, waitStatus *uint32) error { // ExecAsync runs a new task, but doesn't wait for it to finish. It is defined // as a function rather than a method to avoid exposing execAsync as an RPC. -func ExecAsync(proc *Proc, args *ExecArgs) (*kernel.ThreadGroup, error) { +func ExecAsync(proc *Proc, args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadID, error) { return proc.execAsync(args) } -func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, error) { +// execAsync runs a new task, but doesn't wait for it to finish. It returns the +// newly created thread group and its PID. +func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadID, error) { // Import file descriptors. l := limits.NewLimitSet() fdm := proc.Kernel.NewFDMap() @@ -144,7 +146,7 @@ func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, error) { paths := fs.GetPath(initArgs.Envv) f, err := proc.Kernel.RootMountNamespace().ResolveExecutablePath(ctx, initArgs.WorkingDirectory, initArgs.Argv[0], paths) if err != nil { - return nil, fmt.Errorf("error finding executable %q in PATH %v: %v", initArgs.Argv[0], paths, err) + return nil, 0, fmt.Errorf("error finding executable %q in PATH %v: %v", initArgs.Argv[0], paths, err) } initArgs.Filename = f } @@ -156,7 +158,7 @@ func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, error) { // Import the given file FD. This dups the FD as well. file, err := host.ImportFile(ctx, int(f.Fd()), mounter, enableIoctl) if err != nil { - return nil, err + return nil, 0, err } defer file.DecRef() @@ -164,7 +166,7 @@ func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, error) { f.Close() if err := fdm.NewFDAt(kdefs.FD(appFD), file, kernel.FDFlags{}, l); err != nil { - return nil, err + return nil, 0, err } } diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 316612b37..f71e32ac9 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -596,13 +596,13 @@ func (ctx *createProcessContext) Value(key interface{}) interface{} { // // CreateProcess has no analogue in Linux; it is used to create the initial // application task, as well as processes started by the control server. -func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { +func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, ThreadID, error) { k.extMu.Lock() defer k.extMu.Unlock() log.Infof("EXEC: %v", args.Argv) if k.mounts == nil { - return nil, fmt.Errorf("no kernel MountNamespace") + return nil, 0, fmt.Errorf("no kernel MountNamespace") } tg := NewThreadGroup(k.tasks.Root, NewSignalHandlers(), linux.SIGCHLD, args.Limits, k.monotonicClock) @@ -622,7 +622,7 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { var err error wd, err = k.mounts.FindInode(ctx, root, nil, args.WorkingDirectory, args.MaxSymlinkTraversals) if err != nil { - return nil, fmt.Errorf("failed to find initial working directory %q: %v", args.WorkingDirectory, err) + return nil, 0, fmt.Errorf("failed to find initial working directory %q: %v", args.WorkingDirectory, err) } defer wd.DecRef() } @@ -630,10 +630,10 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { if args.Filename == "" { // Was anything provided? if len(args.Argv) == 0 { - return nil, fmt.Errorf("no filename or command provided") + return nil, 0, fmt.Errorf("no filename or command provided") } if !filepath.IsAbs(args.Argv[0]) { - return nil, fmt.Errorf("'%s' is not an absolute path", args.Argv[0]) + return nil, 0, fmt.Errorf("'%s' is not an absolute path", args.Argv[0]) } args.Filename = args.Argv[0] } @@ -641,7 +641,7 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { // Create a fresh task context. tc, err := k.LoadTaskImage(ctx, k.mounts, root, wd, args.MaxSymlinkTraversals, args.Filename, args.Argv, args.Envv, k.featureSet) if err != nil { - return nil, err + return nil, 0, err } // Take a reference on the FDMap, which will be transferred to @@ -663,17 +663,18 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, error) { } t, err := k.tasks.NewTask(config) if err != nil { - return nil, err + return nil, 0, err } // Success. + tgid := k.tasks.Root.IDOfThreadGroup(tg) if k.started { tid := k.tasks.Root.IDOfTask(t) t.Start(tid) } else if k.globalInit == nil { k.globalInit = tg } - return tg, nil + return tg, tgid, nil } // Start starts execution of all tasks in k. diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index 4d41dcd6c..dc9359092 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -242,32 +242,11 @@ type ExecArgs struct { // returns the pid of the new process. func (cm *containerManager) ExecuteAsync(args *ExecArgs, pid *int32) error { log.Debugf("containerManager.ExecuteAsync: %+v", args) - - // Get the container Root Dirent from the Task, since we must run this - // process with the same Root. - cm.l.mu.Lock() - tg, ok := cm.l.containerRootTGs[args.CID] - cm.l.mu.Unlock() - if !ok { - return fmt.Errorf("cannot exec in container %q: no such container", args.CID) - } - tg.Leader().WithMuLocked(func(t *kernel.Task) { - args.Root = t.FSContext().RootDirectory() - }) - if args.Root != nil { - defer args.Root.DecRef() - } - - // Start the process. - proc := control.Proc{Kernel: cm.l.k} - newTG, err := control.ExecAsync(&proc, &args.ExecArgs) + tgid, err := cm.l.executeAsync(&args.ExecArgs, args.CID) if err != nil { - return fmt.Errorf("error executing: %+v: %v", args, err) + return err } - - // Return the pid of the newly-created process. - ts := cm.l.k.TaskSet() - *pid = int32(ts.Root.IDOfThreadGroup(newTG)) + *pid = int32(tgid) return nil } @@ -409,12 +388,16 @@ type WaitPIDArgs struct { // CID is the container ID. CID string + + // ClearStatus determines whether the exit status of the process should + // be cleared when WaitPID returns. + ClearStatus bool } // WaitPID waits for the process with PID 'pid' in the sandbox. func (cm *containerManager) WaitPID(args *WaitPIDArgs, waitStatus *uint32) error { log.Debugf("containerManager.Wait") - return cm.l.waitPID(kernel.ThreadID(args.PID), args.CID, waitStatus) + return cm.l.waitPID(kernel.ThreadID(args.PID), args.CID, args.ClearStatus, waitStatus) } // SignalArgs are arguments to the Signal method. diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 5e9ccb96f..665240ab6 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -31,6 +31,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/cpuid" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/control" "gvisor.googlesource.com/gvisor/pkg/sentry/inet" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" @@ -103,7 +104,7 @@ type Loader struct { // sandboxID is the ID for the whole sandbox. sandboxID string - // mu guards containerRootTGs. + // mu guards containerRootTGs and execProcesses. mu sync.Mutex // containerRootTGs maps container IDs to their root processes. It @@ -111,7 +112,24 @@ type Loader struct { // call methods on particular containers. // // containerRootTGs is guarded by mu. + // + // TODO: When containers are removed via `runsc delete`, + // containerRootTGs should be cleaned up. containerRootTGs map[string]*kernel.ThreadGroup + + // execProcesses maps each invocation of exec to the process it spawns. + // + // execProcesses is guardded by mu. + // + // TODO: When containers are removed via `runsc delete`, + // execProcesses should be cleaned up. + execProcesses map[execID]*kernel.ThreadGroup +} + +// execID uniquely identifies a sentry process. +type execID struct { + cid string + pid kernel.ThreadID } func init() { @@ -385,7 +403,8 @@ func (l *Loader) run() error { } // Create the root container init task. - if _, err := l.k.CreateProcess(l.rootProcArgs); err != nil { + _, _, err := l.k.CreateProcess(l.rootProcArgs) + if err != nil { return fmt.Errorf("failed to create init process: %v", err) } @@ -393,6 +412,11 @@ func (l *Loader) run() error { l.rootProcArgs.FDMap.DecRef() } + if l.execProcesses != nil { + return fmt.Errorf("there shouldn't already be a cache of exec'd processes, but found: %v", l.execProcesses) + } + l.execProcesses = make(map[execID]*kernel.ThreadGroup) + // Start signal forwarding only after an init process is created. l.stopSignalForwarding = l.startSignalForwarding() @@ -467,7 +491,7 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config return fmt.Errorf("error setting executable path for %+v: %v", procArgs, err) } - tg, err := l.k.CreateProcess(procArgs) + tg, _, err := l.k.CreateProcess(procArgs) if err != nil { return fmt.Errorf("failed to create process in sentry: %v", err) } @@ -482,6 +506,40 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config return nil } +func (l *Loader) executeAsync(args *control.ExecArgs, cid string) (kernel.ThreadID, error) { + // Get the container Root Dirent from the Task, since we must run this + // process with the same Root. + l.mu.Lock() + tg, ok := l.containerRootTGs[cid] + l.mu.Unlock() + if !ok { + return 0, fmt.Errorf("cannot exec in container %q: no such container", cid) + } + tg.Leader().WithMuLocked(func(t *kernel.Task) { + args.Root = t.FSContext().RootDirectory() + }) + if args.Root != nil { + defer args.Root.DecRef() + } + + // Start the process. + proc := control.Proc{Kernel: l.k} + tg, tgid, err := control.ExecAsync(&proc, args) + if err != nil { + return 0, fmt.Errorf("error executing: %+v: %v", args, err) + } + + // Insert the process into execProcesses so that we can wait on it + // later. + l.mu.Lock() + defer l.mu.Unlock() + eid := execID{cid: cid, pid: tgid} + l.execProcesses[eid] = tg + log.Debugf("updated execProcesses: %v", l.execProcesses) + + return tgid, nil +} + // TODO: Per-container namespaces must be supported for -pid. // waitContainer waits for the root process of a container to exit. @@ -500,39 +558,59 @@ func (l *Loader) waitContainer(cid string, waitStatus *uint32) error { // consider the container exited. // TODO: Multiple calls to waitContainer() should return // the same exit status. - defer func() { - l.mu.Lock() - defer l.mu.Unlock() - // TODO: Containers don't map 1:1 with their root - // processes. Container exits should be managed explicitly - // rather than via PID. - delete(l.containerRootTGs, cid) - }() - l.wait(tg, waitStatus) + ws := l.wait(tg) + *waitStatus = ws + + l.mu.Lock() + defer l.mu.Unlock() + delete(l.containerRootTGs, cid) + return nil } -func (l *Loader) waitPID(tgid kernel.ThreadID, cid string, waitStatus *uint32) error { +func (l *Loader) waitPID(tgid kernel.ThreadID, cid string, clearStatus bool, waitStatus *uint32) error { // TODO: Containers all currently share a PID namespace. // When per-container PID namespaces are supported, wait should use cid // to find the appropriate PID namespace. /*if cid != l.sandboxID { return errors.New("non-sandbox PID namespaces are not yet implemented") }*/ - // TODO: This won't work if the exec process already exited. - tg := l.k.TaskSet().Root.ThreadGroupWithID(kernel.ThreadID(tgid)) + + // If the process was started via runsc exec, it will have an + // entry in l.execProcesses. + l.mu.Lock() + eid := execID{cid: cid, pid: tgid} + tg, ok := l.execProcesses[eid] + l.mu.Unlock() + if ok { + ws := l.wait(tg) + *waitStatus = ws + if clearStatus { + // Remove tg from the cache. + l.mu.Lock() + delete(l.execProcesses, eid) + log.Debugf("updated execProcesses (removal): %v", l.execProcesses) + l.mu.Unlock() + } + return nil + } + + // This process wasn't created by runsc exec or start, so just find it + // by pid and hope it hasn't exited yet. + tg = l.k.TaskSet().Root.ThreadGroupWithID(kernel.ThreadID(tgid)) if tg == nil { return fmt.Errorf("no thread group with ID %d", tgid) } - l.wait(tg, waitStatus) + ws := l.wait(tg) + *waitStatus = ws return nil } // wait waits for the process with TGID 'tgid' in a container's PID namespace // to exit. -func (l *Loader) wait(tg *kernel.ThreadGroup, waitStatus *uint32) { +func (l *Loader) wait(tg *kernel.ThreadGroup) uint32 { tg.WaitExited() - *waitStatus = tg.ExitStatus().Status() + return tg.ExitStatus().Status() } func (l *Loader) setRootContainerID(cid string) { diff --git a/runsc/boot/loader_test.go b/runsc/boot/loader_test.go index 9398292ff..a8a796445 100644 --- a/runsc/boot/loader_test.go +++ b/runsc/boot/loader_test.go @@ -111,11 +111,11 @@ func createLoader() (*Loader, func(), error) { // TestRun runs a simple application in a sandbox and checks that it succeeds. func TestRun(t *testing.T) { - s, cleanup, err := createLoader() + l, cleanup, err := createLoader() if err != nil { t.Fatalf("error creating loader: %v", err) } - defer s.Destroy() + defer l.Destroy() defer cleanup() // Start a goroutine to read the start chan result, otherwise Run will @@ -124,12 +124,13 @@ func TestRun(t *testing.T) { var wg sync.WaitGroup wg.Add(1) go func() { - resultChanErr = <-s.ctrl.manager.startResultChan + resultChanErr = <-l.ctrl.manager.startResultChan wg.Done() }() - // Run the container.. - if err := s.Run(); err != nil { + // Run the container. + l.setRootContainerID("foo") + if err := l.Run(); err != nil { t.Errorf("error running container: %v", err) } @@ -140,7 +141,7 @@ func TestRun(t *testing.T) { } // Wait for the application to exit. It should succeed. - if status := s.WaitExit(); status.Code != 0 || status.Signo != 0 { + if status := l.WaitExit(); status.Code != 0 || status.Signo != 0 { t.Errorf("application exited with status %+v, want 0", status) } } @@ -148,24 +149,24 @@ func TestRun(t *testing.T) { // TestStartSignal tests that the controller Start message will cause // WaitForStartSignal to return. func TestStartSignal(t *testing.T) { - s, cleanup, err := createLoader() + l, cleanup, err := createLoader() if err != nil { t.Fatalf("error creating loader: %v", err) } - defer s.Destroy() + defer l.Destroy() defer cleanup() // We aren't going to wait on this application, so the control server // needs to be shut down manually. - defer s.ctrl.srv.Stop() + defer l.ctrl.srv.Stop() // Start a goroutine that calls WaitForStartSignal and writes to a // channel when it returns. waitFinished := make(chan struct{}) go func() { - s.WaitForStartSignal() + l.WaitForStartSignal() // Pretend that Run() executed and returned no error. - s.ctrl.manager.startResultChan <- nil + l.ctrl.manager.startResultChan <- nil waitFinished <- struct{}{} }() @@ -181,7 +182,7 @@ func TestStartSignal(t *testing.T) { // Trigger the control server StartRoot method. cid := "foo" - if err := s.ctrl.manager.StartRoot(&cid, nil); err != nil { + if err := l.ctrl.manager.StartRoot(&cid, nil); err != nil { t.Errorf("error calling StartRoot: %v", err) } diff --git a/runsc/cmd/exec.go b/runsc/cmd/exec.go index 0d1fa6e20..957c4f0ff 100644 --- a/runsc/cmd/exec.go +++ b/runsc/cmd/exec.go @@ -49,6 +49,7 @@ type Exec struct { extraKGIDs stringSlice caps stringSlice detach bool + clearStatus bool processPath string pidFile string internalPidFile string @@ -100,6 +101,9 @@ func (ex *Exec) SetFlags(f *flag.FlagSet) { f.StringVar(&ex.pidFile, "pid-file", "", "filename that the container pid will be written to") f.StringVar(&ex.internalPidFile, "internal-pid-file", "", "filename that the container-internal pid will be written to") f.StringVar(&ex.consoleSocket, "console-socket", "", "path to an AF_UNIX socket which will receive a file descriptor referencing the master end of the console's pseudoterminal") + + // clear-status is expected to only be set when we fork due to --detach being set. + f.BoolVar(&ex.clearStatus, "clear-status", true, "clear the status of the exec'd process upon completion") } // Execute implements subcommands.Command.Execute. It starts a process in an @@ -163,7 +167,7 @@ func (ex *Exec) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) } // Wait for the process to exit. - ws, err := c.WaitPID(pid) + ws, err := c.WaitPID(pid, ex.clearStatus) if err != nil { Fatalf("error waiting on pid %d: %v", pid, err) } @@ -194,10 +198,16 @@ func (ex *Exec) execAndWait(waitStatus *syscall.WaitStatus) subcommands.ExitStat // Add the rest of the args, excluding the "detach" flag. for _, a := range os.Args[1:] { - if !strings.Contains(a, "detach") { + if strings.Contains(a, "detach") { + // Replace with the "clear-status" flag, which tells + // the new process it's a detached child and shouldn't + // clear the exit status of the sentry process. + args = append(args, "--clear-status=false") + } else { args = append(args, a) } } + cmd := exec.Command(binPath, args...) // Exec stdio defaults to current process stdio. diff --git a/runsc/cmd/wait.go b/runsc/cmd/wait.go index b41edc725..956349140 100644 --- a/runsc/cmd/wait.go +++ b/runsc/cmd/wait.go @@ -88,14 +88,14 @@ func (wt *Wait) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) waitStatus = ws // Wait on a PID in the root PID namespace. case wt.rootPID != unsetPID: - ws, err := c.WaitRootPID(int32(wt.rootPID)) + ws, err := c.WaitRootPID(int32(wt.rootPID), true /* clearStatus */) if err != nil { Fatalf("error waiting on PID in root PID namespace %d in container %q: %v", wt.rootPID, c.ID, err) } waitStatus = ws // Wait on a PID in the container's PID namespace. case wt.pid != unsetPID: - ws, err := c.WaitPID(int32(wt.pid)) + ws, err := c.WaitPID(int32(wt.pid), true /* clearStatus */) if err != nil { Fatalf("error waiting on PID %d in container %q: %v", wt.pid, c.ID, err) } diff --git a/runsc/container/container.go b/runsc/container/container.go index 792b7967b..a24c6cc31 100644 --- a/runsc/container/container.go +++ b/runsc/container/container.go @@ -398,22 +398,22 @@ func (c *Container) Wait() (syscall.WaitStatus, error) { // WaitRootPID waits for process 'pid' in the sandbox's PID namespace and // returns its WaitStatus. -func (c *Container) WaitRootPID(pid int32) (syscall.WaitStatus, error) { +func (c *Container) WaitRootPID(pid int32, clearStatus bool) (syscall.WaitStatus, error) { log.Debugf("Wait on pid %d in sandbox %q", pid, c.Sandbox.ID) if c.Sandbox == nil || !c.Sandbox.IsRunning() { return 0, fmt.Errorf("container sandbox is not running") } - return c.Sandbox.WaitPID(pid, c.Sandbox.ID) + return c.Sandbox.WaitPID(c.Sandbox.ID, pid, clearStatus) } // WaitPID waits for process 'pid' in the container's PID namespace and returns // its WaitStatus. -func (c *Container) WaitPID(pid int32) (syscall.WaitStatus, error) { +func (c *Container) WaitPID(pid int32, clearStatus bool) (syscall.WaitStatus, error) { log.Debugf("Wait on pid %d in container %q", pid, c.ID) if c.Sandbox == nil || !c.Sandbox.IsRunning() { return 0, fmt.Errorf("container sandbox is not running") } - return c.Sandbox.WaitPID(pid, c.ID) + return c.Sandbox.WaitPID(c.ID, pid, clearStatus) } // Signal sends the signal to the container. diff --git a/runsc/container/container_test.go b/runsc/container/container_test.go index ab1823f1c..5fe80f20f 100644 --- a/runsc/container/container_test.go +++ b/runsc/container/container_test.go @@ -551,7 +551,7 @@ func TestExec(t *testing.T) { args := &control.ExecArgs{ Filename: "/bin/sleep", - Argv: []string{"sleep", "5"}, + Argv: []string{"/bin/sleep", "5"}, WorkingDirectory: "/", KUID: uid, } @@ -1598,7 +1598,7 @@ func (cont *Container) executeSync(args *control.ExecArgs) (syscall.WaitStatus, if err != nil { return 0, fmt.Errorf("error executing: %v", err) } - ws, err := cont.WaitPID(pid) + ws, err := cont.WaitPID(pid, true /* clearStatus */) if err != nil { return 0, fmt.Errorf("error waiting: %v", err) } diff --git a/runsc/container/multi_container_test.go b/runsc/container/multi_container_test.go index 84e0ec080..09888cb86 100644 --- a/runsc/container/multi_container_test.go +++ b/runsc/container/multi_container_test.go @@ -163,16 +163,15 @@ func TestMultiContainerWait(t *testing.T) { go func(c *Container) { defer wg.Done() const pid = 2 - if ws, err := c.WaitPID(pid); err != nil { + if ws, err := c.WaitPID(pid, true /* clearStatus */); err != nil { t.Errorf("failed to wait for PID %d: %v", pid, err) } else if es := ws.ExitStatus(); es != 0 { t.Errorf("PID %d exited with non-zero status %d", pid, es) } - if _, err := c.WaitPID(pid); err == nil { + if _, err := c.WaitPID(pid, true /* clearStatus */); err == nil { t.Errorf("wait for stopped PID %d should fail", pid) } - // TODO: use 'container[1]' when PID namespace is supported. - }(containers[0]) + }(containers[1]) } wg.Wait() @@ -184,6 +183,93 @@ func TestMultiContainerWait(t *testing.T) { } } +// TestExecWait ensures what we can wait containers and individual processes in the +// sandbox that have already exited. +func TestExecWait(t *testing.T) { + rootDir, err := testutil.SetupRootDir() + if err != nil { + t.Fatalf("error creating root dir: %v", err) + } + defer os.RemoveAll(rootDir) + + // The first container should run the entire duration of the test. + cmd1 := []string{"sleep", "100"} + // We'll wait on the second container, which is much shorter lived. + cmd2 := []string{"sleep", "1"} + specs, ids := createSpecs(cmd1, cmd2) + + // Setup the containers. + var containers []*Container + for i, spec := range specs { + conf := testutil.TestConfig() + bundleDir, err := testutil.SetupContainerInRoot(rootDir, spec, conf) + if err != nil { + t.Fatalf("error setting up container: %v", err) + } + defer os.RemoveAll(bundleDir) + cont, err := Create(ids[i], spec, conf, bundleDir, "", "") + if err != nil { + t.Fatalf("error creating container: %v", err) + } + defer cont.Destroy() + if err := cont.Start(conf); err != nil { + t.Fatalf("error starting container: %v", err) + } + containers = append(containers, cont) + } + + // Check via ps that multiple processes are running. + expectedPL := []*control.Process{ + {PID: 1, Cmd: "sleep"}, + {PID: 2, Cmd: "sleep"}, + } + if err := waitForProcessList(containers[0], expectedPL); err != nil { + t.Fatalf("failed to wait for sleep to start: %v", err) + } + + // Wait for the second container to finish. + if err := waitForProcessList(containers[0], expectedPL[:1]); err != nil { + t.Fatalf("failed to wait for second container to stop: %v", err) + } + + // Get the second container exit status. + if ws, err := containers[1].Wait(); err != nil { + t.Fatalf("failed to wait for process %s: %v", containers[1].Spec.Process.Args, err) + } else if es := ws.ExitStatus(); es != 0 { + t.Fatalf("process %s exited with non-zero status %d", containers[1].Spec.Process.Args, es) + } + if _, err := containers[1].Wait(); err == nil { + t.Fatalf("wait for stopped process %s should fail", containers[1].Spec.Process.Args) + } + + // Execute another process in the first container. + args := &control.ExecArgs{ + Filename: "/bin/sleep", + Argv: []string{"/bin/sleep", "1"}, + WorkingDirectory: "/", + KUID: 0, + } + pid, err := containers[0].Execute(args) + if err != nil { + t.Fatalf("error executing: %v", err) + } + + // Wait for the exec'd process to exit. + if err := waitForProcessList(containers[0], expectedPL[:1]); err != nil { + t.Fatalf("failed to wait for second container to stop: %v", err) + } + + // Get the exit status from the exec'd process. + if ws, err := containers[0].WaitPID(pid, true /* clearStatus */); err != nil { + t.Fatalf("failed to wait for process %+v with pid %d: %v", args, pid, err) + } else if es := ws.ExitStatus(); es != 0 { + t.Fatalf("process %+v exited with non-zero status %d", args, es) + } + if _, err := containers[0].WaitPID(pid, true /* clearStatus */); err == nil { + t.Fatalf("wait for stopped process %+v should fail", args) + } +} + // TestMultiContainerMount tests that bind mounts can be used with multiple // containers. func TestMultiContainerMount(t *testing.T) { diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 8c4d0d495..3b10fd20e 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -522,7 +522,7 @@ func (s *Sandbox) Wait(cid string) (syscall.WaitStatus, error) { // WaitPID waits for process 'pid' in the container's sandbox and returns its // WaitStatus. -func (s *Sandbox) WaitPID(pid int32, cid string) (syscall.WaitStatus, error) { +func (s *Sandbox) WaitPID(cid string, pid int32, clearStatus bool) (syscall.WaitStatus, error) { log.Debugf("Waiting for PID %d in sandbox %q", pid, s.ID) var ws syscall.WaitStatus conn, err := s.sandboxConnect() @@ -532,8 +532,9 @@ func (s *Sandbox) WaitPID(pid int32, cid string) (syscall.WaitStatus, error) { defer conn.Close() args := &boot.WaitPIDArgs{ - PID: pid, - CID: cid, + PID: pid, + CID: cid, + ClearStatus: clearStatus, } if err := conn.Call(boot.ContainerWaitPID, args, &ws); err != nil { return ws, fmt.Errorf("error waiting on PID %d in sandbox %q: %v", pid, s.ID, err) -- cgit v1.2.3 From 491faac03b2815ca1bc9b5425c1b3f6291468e20 Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Thu, 27 Sep 2018 15:00:03 -0700 Subject: Implement 'runsc kill --all' In order to implement kill --all correctly, the Sentry needs to track all tasks that belong to a given container. This change introduces ContainerID to the task, that gets inherited by all children. 'kill --all' then iterates over all tasks comparing the ContainerID field to find all processes that need to be signalled. PiperOrigin-RevId: 214841768 Change-Id: I693b2374be8692d88cc441ef13a0ae34abf73ac6 --- pkg/sentry/control/proc.go | 14 ++- pkg/sentry/kernel/kernel.go | 25 +++++ pkg/sentry/kernel/task.go | 12 ++ pkg/sentry/kernel/task_clone.go | 1 + pkg/sentry/kernel/task_start.go | 4 + runsc/boot/controller.go | 29 ++--- runsc/boot/loader.go | 26 +++-- runsc/cmd/kill.go | 15 ++- runsc/container/BUILD | 6 +- runsc/container/container.go | 11 +- runsc/container/container_test.go | 38 +++++-- runsc/container/multi_container_test.go | 190 ++++++++++++++++++++++++++++---- runsc/container/test_app.go | 63 +++++++++++ runsc/sandbox/sandbox.go | 16 ++- 14 files changed, 371 insertions(+), 79 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go index b120471cb..106055e86 100644 --- a/pkg/sentry/control/proc.go +++ b/pkg/sentry/control/proc.go @@ -83,6 +83,9 @@ type ExecArgs struct { // FilePayload determines the files to give to the new process. urpc.FilePayload + + // ContainerID is the container for the process being executed. + ContainerID string } // Exec runs a new task. @@ -133,6 +136,7 @@ func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadI UTSNamespace: proc.Kernel.RootUTSNamespace(), IPCNamespace: proc.Kernel.RootIPCNamespace(), AbstractSocketNamespace: proc.Kernel.RootAbstractSocketNamespace(), + ContainerID: args.ContainerID, } if initArgs.Root != nil { // initArgs must hold a reference on Root. This ref is dropped @@ -182,7 +186,7 @@ type PsArgs struct { // Ps provides a process listing for the running kernel. func (proc *Proc) Ps(args *PsArgs, out *string) error { var p []*Process - if e := Processes(proc.Kernel, &p); e != nil { + if e := Processes(proc.Kernel, "", &p); e != nil { return e } if !args.JSON { @@ -258,8 +262,9 @@ func PrintPIDsJSON(pl []*Process) (string, error) { return string(b), nil } -// Processes retrieves information about processes running in the sandbox. -func Processes(k *kernel.Kernel, out *[]*Process) error { +// Processes retrieves information about processes running in the sandbox with +// the given container id. All processes are returned if 'containerID' is empty. +func Processes(k *kernel.Kernel, containerID string, out *[]*Process) error { ts := k.TaskSet() now := k.RealtimeClock().Now() for _, tg := range ts.Root.ThreadGroups() { @@ -268,6 +273,9 @@ func Processes(k *kernel.Kernel, out *[]*Process) error { if pid == 0 { continue } + if containerID != "" && containerID != tg.Leader().ContainerID() { + continue + } ppid := kernel.ThreadID(0) if tg.Leader().Parent() != nil { diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index f71e32ac9..1ace0b501 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -524,6 +524,9 @@ type CreateProcessArgs struct { // Anyone setting Root must donate a reference (i.e. increment it) to // keep it alive until it is decremented by CreateProcess. Root *fs.Dirent + + // ContainerID is the container that the process belongs to. + ContainerID string } // NewContext returns a context.Context that represents the task that will be @@ -660,6 +663,7 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, ThreadID, UTSNamespace: args.UTSNamespace, IPCNamespace: args.IPCNamespace, AbstractSocketNamespace: args.AbstractSocketNamespace, + ContainerID: args.ContainerID, } t, err := k.tasks.NewTask(config) if err != nil { @@ -818,6 +822,27 @@ func (k *Kernel) SendExternalSignal(info *arch.SignalInfo, context string) { k.sendExternalSignal(info, context) } +// SendContainerSignal sends the given signal to all processes inside the +// namespace that match the given container ID. +func (k *Kernel) SendContainerSignal(cid string, info *arch.SignalInfo) error { + k.extMu.Lock() + defer k.extMu.Unlock() + k.tasks.mu.RLock() + defer k.tasks.mu.RUnlock() + + for t := range k.tasks.Root.tids { + if t == t.tg.leader && t.ContainerID() == cid { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + infoCopy := *info + if err := t.sendSignalLocked(&infoCopy, true /*group*/); err != nil { + return err + } + } + } + return nil +} + // FeatureSet returns the FeatureSet. func (k *Kernel) FeatureSet() *cpuid.FeatureSet { return k.featureSet diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 2f6f825ac..07ad1614c 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -205,6 +205,13 @@ type Task struct { // k is the Kernel that this task belongs to. The k pointer is immutable. k *Kernel + // containerID has no equivalent in Linux; it's used by runsc to track all + // tasks that belong to a given containers since cgroups aren't implemented. + // It's inherited by the children, is immutable, and may be empty. + // + // NOTE: cgroups can be used to track this when implemented. + containerID string + // mu protects some of the following fields. mu sync.Mutex `state:"nosave"` @@ -678,3 +685,8 @@ func (t *Task) MountNamespace() *fs.MountNamespace { func (t *Task) AbstractSockets() *AbstractSocketNamespace { return t.abstractSockets } + +// ContainerID returns t's container ID. +func (t *Task) ContainerID() string { + return t.containerID +} diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 46c688b20..130bd652b 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -258,6 +258,7 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { UTSNamespace: utsns, IPCNamespace: ipcns, AbstractSocketNamespace: t.abstractSockets, + ContainerID: t.ContainerID(), } if opts.NewThreadGroup { cfg.Parent = t diff --git a/pkg/sentry/kernel/task_start.go b/pkg/sentry/kernel/task_start.go index 6ce99d268..6c8d7d316 100644 --- a/pkg/sentry/kernel/task_start.go +++ b/pkg/sentry/kernel/task_start.go @@ -77,6 +77,9 @@ type TaskConfig struct { // AbstractSocketNamespace is the AbstractSocketNamespace of the new task. AbstractSocketNamespace *AbstractSocketNamespace + + // ContainerID is the container the new task belongs to. + ContainerID string } // NewTask creates a new task defined by cfg. @@ -124,6 +127,7 @@ func (ts *TaskSet) newTask(cfg *TaskConfig) (*Task, error) { abstractSockets: cfg.AbstractSocketNamespace, rseqCPU: -1, futexWaiter: futex.NewWaiter(), + containerID: cfg.ContainerID, } t.endStopCond.L = &t.tg.signalHandlers.mu t.ptraceTracer.Store((*Task)(nil)) diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index bc33e028a..116a8369c 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -174,10 +174,17 @@ func (cm *containerManager) StartRoot(cid *string, _ *struct{}) error { return nil } +// ProcessesArgs container arguments to Processes method. +type ProcessesArgs struct { + // CID restricts the result to processes belonging to + // the given container. Empty means all. + CID string +} + // Processes retrieves information about processes running in the sandbox. -func (cm *containerManager) Processes(_, out *[]*control.Process) error { +func (cm *containerManager) Processes(args *ProcessesArgs, out *[]*control.Process) error { log.Debugf("containerManager.Processes") - return control.Processes(cm.l.k, out) + return control.Processes(cm.l.k, args.CID, out) } // StartArgs contains arguments to the Start method. @@ -326,19 +333,11 @@ func (cm *containerManager) Destroy(cid *string, _ *struct{}) error { return nil } -// ExecArgs contains arguments to Execute. -type ExecArgs struct { - control.ExecArgs - - // CID is the ID of the container to exec in. - CID string -} - // ExecuteAsync starts running a command on a created or running sandbox. It // returns the pid of the new process. -func (cm *containerManager) ExecuteAsync(args *ExecArgs, pid *int32) error { +func (cm *containerManager) ExecuteAsync(args *control.ExecArgs, pid *int32) error { log.Debugf("containerManager.ExecuteAsync: %+v", args) - tgid, err := cm.l.executeAsync(&args.ExecArgs, args.CID) + tgid, err := cm.l.executeAsync(args) if err != nil { return err } @@ -503,11 +502,15 @@ type SignalArgs struct { // Signo is the signal to send to the process. Signo int32 + + // All is set when signal should be sent to all processes in the container. + // When false, the signal is sent to the root container process only. + All bool } // Signal sends a signal to the init process of the container. // TODO: Send signal to exec process. func (cm *containerManager) Signal(args *SignalArgs, _ *struct{}) error { log.Debugf("containerManager.Signal") - return cm.l.signal(args.CID, args.Signo) + return cm.l.signal(args.CID, args.Signo, args.All) } diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 9a5d649ab..bd6e146fc 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -270,7 +270,7 @@ func New(id string, spec *specs.Spec, conf *Config, controllerFD, deviceFD int, log.Infof("Panic signal set to %v(%d)", ps, conf.PanicSignal) } - procArgs, err := newProcess(spec, creds, k) + procArgs, err := newProcess(id, spec, creds, k) if err != nil { return nil, fmt.Errorf("failed to create root process: %v", err) } @@ -295,7 +295,7 @@ func New(id string, spec *specs.Spec, conf *Config, controllerFD, deviceFD int, } // newProcess creates a process that can be run with kernel.CreateProcess. -func newProcess(spec *specs.Spec, creds *auth.Credentials, k *kernel.Kernel) (kernel.CreateProcessArgs, error) { +func newProcess(id string, spec *specs.Spec, creds *auth.Credentials, k *kernel.Kernel) (kernel.CreateProcessArgs, error) { // Create initial limits. ls, err := createLimitSet(spec) if err != nil { @@ -314,6 +314,7 @@ func newProcess(spec *specs.Spec, creds *auth.Credentials, k *kernel.Kernel) (ke UTSNamespace: k.RootUTSNamespace(), IPCNamespace: k.RootIPCNamespace(), AbstractSocketNamespace: k.RootAbstractSocketNamespace(), + ContainerID: id, } return procArgs, nil } @@ -465,7 +466,7 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config // TODO New containers should be started in new PID namespaces // when indicated by the spec. - procArgs, err := newProcess(spec, creds, l.k) + procArgs, err := newProcess(cid, spec, creds, l.k) if err != nil { return fmt.Errorf("failed to create new process: %v", err) } @@ -525,14 +526,14 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config return nil } -func (l *Loader) executeAsync(args *control.ExecArgs, cid string) (kernel.ThreadID, error) { +func (l *Loader) executeAsync(args *control.ExecArgs) (kernel.ThreadID, error) { // Get the container Root Dirent from the Task, since we must run this // process with the same Root. l.mu.Lock() - tg, ok := l.containerRootTGs[cid] + tg, ok := l.containerRootTGs[args.ContainerID] l.mu.Unlock() if !ok { - return 0, fmt.Errorf("cannot exec in container %q: no such container", cid) + return 0, fmt.Errorf("cannot exec in container %q: no such container", args.ContainerID) } tg.Leader().WithMuLocked(func(t *kernel.Task) { args.Root = t.FSContext().RootDirectory() @@ -552,7 +553,7 @@ func (l *Loader) executeAsync(args *control.ExecArgs, cid string) (kernel.Thread // later. l.mu.Lock() defer l.mu.Unlock() - eid := execID{cid: cid, pid: tgid} + eid := execID{cid: args.ContainerID, pid: tgid} l.execProcesses[eid] = tg log.Debugf("updated execProcesses: %v", l.execProcesses) @@ -671,8 +672,7 @@ func newEmptyNetworkStack(conf *Config, clock tcpip.Clock) (inet.Stack, error) { } } -// TODO: Support sending signal to all. -func (l *Loader) signal(cid string, signo int32) error { +func (l *Loader) signal(cid string, signo int32, all bool) error { l.mu.Lock() tg, ok := l.containerRootTGs[cid] l.mu.Unlock() @@ -681,5 +681,13 @@ func (l *Loader) signal(cid string, signo int32) error { } si := arch.SignalInfo{Signo: signo} + if all { + // Pause the kernel to prevent new processes from being created while + // the signal is delivered. This prevents process leaks when SIGKILL is + // sent to the entire container. + l.k.Pause() + defer l.k.Unpause() + return l.k.SendContainerSignal(cid, &si) + } return tg.Leader().SendSignal(&si) } diff --git a/runsc/cmd/kill.go b/runsc/cmd/kill.go index 6fa5674f1..af709bc71 100644 --- a/runsc/cmd/kill.go +++ b/runsc/cmd/kill.go @@ -29,7 +29,9 @@ import ( ) // Kill implements subcommands.Command for the "kill" command. -type Kill struct{} +type Kill struct { + all bool +} // Name implements subcommands.Command.Name. func (*Kill) Name() string { @@ -47,15 +49,12 @@ func (*Kill) Usage() string { } // SetFlags implements subcommands.Command.SetFlags. -func (*Kill) SetFlags(f *flag.FlagSet) { - // TODO: Implement this flag. It is defined here just to - // prevent runsc from crashing if it is passed. - var all bool - f.BoolVar(&all, "all", false, "send the specified signal to all processes inside the container") +func (k *Kill) SetFlags(f *flag.FlagSet) { + f.BoolVar(&k.all, "all", false, "send the specified signal to all processes inside the container") } // Execute implements subcommands.Command.Execute. -func (*Kill) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) subcommands.ExitStatus { +func (k *Kill) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) subcommands.ExitStatus { if f.NArg() == 0 || f.NArg() > 2 { f.Usage() return subcommands.ExitUsageError @@ -83,7 +82,7 @@ func (*Kill) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) su } // TODO: Distinguish between already-exited containers and // genuine errors. - if err := c.Signal(sig); err != nil { + if err := c.Signal(sig, k.all); err != nil { Fatalf("%v", err) } return subcommands.ExitSuccess diff --git a/runsc/container/BUILD b/runsc/container/BUILD index d72d05c13..e68fb1e8e 100644 --- a/runsc/container/BUILD +++ b/runsc/container/BUILD @@ -53,6 +53,7 @@ go_test( "//runsc/boot", "//runsc/specutils", "//runsc/test/testutil", + "@com_github_cenkalti_backoff//:go_default_library", "@com_github_opencontainers_runtime-spec//specs-go:go_default_library", "@org_golang_x_sys//unix:go_default_library", ], @@ -61,5 +62,8 @@ go_test( go_binary( name = "test_app", srcs = ["test_app.go"], - deps = ["@com_github_google_subcommands//:go_default_library"], + deps = [ + "//runsc/test/testutil", + "@com_github_google_subcommands//:go_default_library", + ], ) diff --git a/runsc/container/container.go b/runsc/container/container.go index a1b31d861..44b7dad8a 100644 --- a/runsc/container/container.go +++ b/runsc/container/container.go @@ -159,7 +159,7 @@ func Load(rootDir, id string) (*Container, error) { } else if c.Status == Running { // Container state should reflect the actual state of the application, so // we don't consider gofer process here. - if err := c.Signal(syscall.Signal(0)); err != nil { + if err := c.Signal(syscall.Signal(0), false); err != nil { c.changeStatus(Stopped) } } @@ -398,7 +398,8 @@ func (c *Container) Execute(args *control.ExecArgs) (int32, error) { if err := c.requireStatus("execute in", Created, Running); err != nil { return 0, err } - return c.Sandbox.Execute(c.ID, args) + args.ContainerID = c.ID + return c.Sandbox.Execute(args) } // Event returns events for the container. @@ -453,13 +454,13 @@ func (c *Container) WaitPID(pid int32, clearStatus bool) (syscall.WaitStatus, er // Signal sends the signal to the container. // Signal returns an error if the container is already stopped. // TODO: Distinguish different error types. -func (c *Container) Signal(sig syscall.Signal) error { +func (c *Container) Signal(sig syscall.Signal, all bool) error { log.Debugf("Signal container %q: %v", c.ID, sig) if err := c.requireStatus("signal", Running); err != nil { return err } // TODO: Query the container for its state, then save it. - return c.Sandbox.Signal(c.ID, sig) + return c.Sandbox.Signal(c.ID, sig, all) } // Checkpoint sends the checkpoint call to the container. @@ -612,7 +613,7 @@ func (c *Container) waitForStopped() error { b := backoff.WithContext(backoff.NewConstantBackOff(100*time.Millisecond), ctx) op := func() error { if c.isSandboxRunning() { - if err := c.Signal(syscall.Signal(0)); err == nil { + if err := c.Signal(syscall.Signal(0), false); err == nil { return fmt.Errorf("container is still running") } } diff --git a/runsc/container/container_test.go b/runsc/container/container_test.go index efa598202..de1e50a3f 100644 --- a/runsc/container/container_test.go +++ b/runsc/container/container_test.go @@ -30,6 +30,7 @@ import ( "testing" "time" + "github.com/cenkalti/backoff" specs "github.com/opencontainers/runtime-spec/specs-go" "golang.org/x/sys/unix" "gvisor.googlesource.com/gvisor/pkg/abi/linux" @@ -49,21 +50,34 @@ func init() { } // waitForProcessList waits for the given process list to show up in the container. -func waitForProcessList(cont *Container, expected []*control.Process) error { - var got []*control.Process - for start := time.Now(); time.Now().Sub(start) < 10*time.Second; { - var err error - got, err = cont.Processes() +func waitForProcessList(cont *Container, want []*control.Process) error { + cb := func() error { + got, err := cont.Processes() if err != nil { - return fmt.Errorf("error getting process data from container: %v", err) + err = fmt.Errorf("error getting process data from container: %v", err) + return &backoff.PermanentError{Err: err} } - if procListsEqual(got, expected) { - return nil + if !procListsEqual(got, want) { + return fmt.Errorf("container got process list: %s, want: %s", procListToString(got), procListToString(want)) } - // Process might not have started, try again... - time.Sleep(10 * time.Millisecond) + return nil + } + return testutil.Poll(cb, 5*time.Second) +} + +func waitForProcessCount(cont *Container, want int) error { + cb := func() error { + pss, err := cont.Processes() + if err != nil { + err = fmt.Errorf("error getting process data from container: %v", err) + return &backoff.PermanentError{Err: err} + } + if got := len(pss); got != want { + return fmt.Errorf("wrong process count, got: %d, want: %d", got, want) + } + return nil } - return fmt.Errorf("container got process list: %s, want: %s", procListToString(got), procListToString(expected)) + return testutil.Poll(cb, 5*time.Second) } // procListsEqual is used to check whether 2 Process lists are equal for all @@ -345,7 +359,7 @@ func TestLifecycle(t *testing.T) { <-ch time.Sleep(100 * time.Millisecond) // Send the container a SIGTERM which will cause it to stop. - if err := c.Signal(syscall.SIGTERM); err != nil { + if err := c.Signal(syscall.SIGTERM, false); err != nil { t.Fatalf("error sending signal %v to container: %v", syscall.SIGTERM, err) } // Wait for it to die. diff --git a/runsc/container/multi_container_test.go b/runsc/container/multi_container_test.go index 2867aa3b9..dc938066b 100644 --- a/runsc/container/multi_container_test.go +++ b/runsc/container/multi_container_test.go @@ -16,6 +16,7 @@ package container import ( "io/ioutil" + "math" "os" "path" "path/filepath" @@ -91,11 +92,16 @@ func TestMultiContainerSanity(t *testing.T) { // Check via ps that multiple processes are running. expectedPL := []*control.Process{ {PID: 1, Cmd: "sleep"}, - {PID: 2, Cmd: "sleep"}, } if err := waitForProcessList(containers[0], expectedPL); err != nil { t.Errorf("failed to wait for sleep to start: %v", err) } + expectedPL = []*control.Process{ + {PID: 2, Cmd: "sleep"}, + } + if err := waitForProcessList(containers[1], expectedPL); err != nil { + t.Errorf("failed to wait for sleep to start: %v", err) + } } } @@ -134,10 +140,9 @@ func TestMultiContainerWait(t *testing.T) { // Check via ps that multiple processes are running. expectedPL := []*control.Process{ - {PID: 1, Cmd: "sleep"}, {PID: 2, Cmd: "sleep"}, } - if err := waitForProcessList(containers[0], expectedPL); err != nil { + if err := waitForProcessList(containers[1], expectedPL); err != nil { t.Errorf("failed to wait for sleep to start: %v", err) } @@ -179,7 +184,10 @@ func TestMultiContainerWait(t *testing.T) { // After Wait returns, ensure that the root container is running and // the child has finished. - if err := waitForProcessList(containers[0], expectedPL[:1]); err != nil { + expectedPL = []*control.Process{ + {PID: 1, Cmd: "sleep"}, + } + if err := waitForProcessList(containers[0], expectedPL); err != nil { t.Errorf("failed to wait for %q to start: %v", strings.Join(containers[0].Spec.Process.Args, " "), err) } } @@ -219,17 +227,16 @@ func TestExecWait(t *testing.T) { containers = append(containers, cont) } - // Check via ps that multiple processes are running. + // Check via ps that process is running. expectedPL := []*control.Process{ - {PID: 1, Cmd: "sleep"}, {PID: 2, Cmd: "sleep"}, } - if err := waitForProcessList(containers[0], expectedPL); err != nil { + if err := waitForProcessList(containers[1], expectedPL); err != nil { t.Fatalf("failed to wait for sleep to start: %v", err) } // Wait for the second container to finish. - if err := waitForProcessList(containers[0], expectedPL[:1]); err != nil { + if err := waitForProcessCount(containers[1], 0); err != nil { t.Fatalf("failed to wait for second container to stop: %v", err) } @@ -256,7 +263,10 @@ func TestExecWait(t *testing.T) { } // Wait for the exec'd process to exit. - if err := waitForProcessList(containers[0], expectedPL[:1]); err != nil { + expectedPL = []*control.Process{ + {PID: 1, Cmd: "sleep"}, + } + if err := waitForProcessList(containers[0], expectedPL); err != nil { t.Fatalf("failed to wait for second container to stop: %v", err) } @@ -360,23 +370,25 @@ func TestMultiContainerSignal(t *testing.T) { containers = append(containers, cont) } - // Check via ps that multiple processes are running. + // Check via ps that container 1 process is running. expectedPL := []*control.Process{ - {PID: 1, Cmd: "sleep"}, {PID: 2, Cmd: "sleep"}, } - if err := waitForProcessList(containers[0], expectedPL); err != nil { + if err := waitForProcessList(containers[1], expectedPL); err != nil { t.Errorf("failed to wait for sleep to start: %v", err) } // Kill process 2. - if err := containers[1].Signal(syscall.SIGKILL); err != nil { + if err := containers[1].Signal(syscall.SIGKILL, false); err != nil { t.Errorf("failed to kill process 2: %v", err) } // Make sure process 1 is still running. - if err := waitForProcessList(containers[0], expectedPL[:1]); err != nil { + expectedPL = []*control.Process{ + {PID: 1, Cmd: "sleep"}, + } + if err := waitForProcessList(containers[0], expectedPL); err != nil { t.Errorf("failed to wait for sleep to start: %v", err) } @@ -395,18 +407,18 @@ func TestMultiContainerSignal(t *testing.T) { t.Errorf("error waiting for gofer to exit: %v", err) } // Make sure process 1 is still running. - if err := waitForProcessList(containers[0], expectedPL[:1]); err != nil { + if err := waitForProcessList(containers[0], expectedPL); err != nil { t.Errorf("failed to wait for sleep to start: %v", err) } // Now that process 2 is gone, ensure we get an error trying to // signal it again. - if err := containers[1].Signal(syscall.SIGKILL); err == nil { + if err := containers[1].Signal(syscall.SIGKILL, false); err == nil { t.Errorf("container %q shouldn't exist, but we were able to signal it", containers[1].ID) } // Kill process 1. - if err := containers[0].Signal(syscall.SIGKILL); err != nil { + if err := containers[0].Signal(syscall.SIGKILL, false); err != nil { t.Errorf("failed to kill process 1: %v", err) } @@ -428,7 +440,7 @@ func TestMultiContainerSignal(t *testing.T) { } // The sentry should be gone, so signaling should yield an error. - if err := containers[0].Signal(syscall.SIGKILL); err == nil { + if err := containers[0].Signal(syscall.SIGKILL, false); err == nil { t.Errorf("sandbox %q shouldn't exist, but we were able to signal it", containers[0].Sandbox.ID) } } @@ -453,7 +465,6 @@ func TestMultiContainerDestroy(t *testing.T) { // Setup the containers. var containers []*Container for i, spec := range specs { - conf := testutil.TestConfig() bundleDir, err := testutil.SetupContainerInRoot(rootDir, spec, conf) if err != nil { t.Fatalf("error setting up container: %v", err) @@ -501,3 +512,144 @@ func TestMultiContainerDestroy(t *testing.T) { } } } + +func TestMultiContainerProcesses(t *testing.T) { + // Note: use 'while true' to keep 'sh' process around. Otherwise, shell will + // just execve into 'sleep' and both containers will look the same. + specs, ids := createSpecs( + []string{"sleep", "100"}, + []string{"sh", "-c", "while true; do sleep 100; done"}) + + rootDir, err := testutil.SetupRootDir() + if err != nil { + t.Fatalf("error creating root dir: %v", err) + } + defer os.RemoveAll(rootDir) + + var containers []*Container + for i, spec := range specs { + conf := testutil.TestConfig() + bundleDir, err := testutil.SetupContainerInRoot(rootDir, spec, conf) + if err != nil { + t.Fatalf("error setting up container: %v", err) + } + defer os.RemoveAll(bundleDir) + cont, err := Create(ids[i], spec, conf, bundleDir, "", "") + if err != nil { + t.Fatalf("error creating container: %v", err) + } + defer cont.Destroy() + if err := cont.Start(conf); err != nil { + t.Fatalf("error starting container: %v", err) + } + containers = append(containers, cont) + } + + // Check root's container process list doesn't include other containers. + expectedPL0 := []*control.Process{ + {PID: 1, Cmd: "sleep"}, + } + if err := waitForProcessList(containers[0], expectedPL0); err != nil { + t.Errorf("failed to wait for process to start: %v", err) + } + + // Same for the other container. + expectedPL1 := []*control.Process{ + {PID: 2, Cmd: "sh"}, + {PID: 3, PPID: 2, Cmd: "sleep"}, + } + if err := waitForProcessList(containers[1], expectedPL1); err != nil { + t.Errorf("failed to wait for process to start: %v", err) + } + + // Now exec into the second container and verify it shows up in the container. + args := &control.ExecArgs{ + Filename: "/bin/sleep", + Argv: []string{"/bin/sleep", "100"}, + } + if _, err := containers[1].Execute(args); err != nil { + t.Fatalf("error exec'ing: %v", err) + } + expectedPL1 = append(expectedPL1, &control.Process{PID: 4, Cmd: "sleep"}) + if err := waitForProcessList(containers[1], expectedPL1); err != nil { + t.Errorf("failed to wait for process to start: %v", err) + } + // Root container should remain unchanged. + if err := waitForProcessList(containers[0], expectedPL0); err != nil { + t.Errorf("failed to wait for process to start: %v", err) + } +} + +// TestMultiContainerKillAll checks that all process that belong to a container +// are killed when SIGKILL is sent to *all* processes in that container. +func TestMultiContainerKillAll(t *testing.T) { + app, err := testutil.FindFile("runsc/container/test_app") + if err != nil { + t.Fatal("error finding test_app:", err) + } + + // First container will remain intact while the second container is killed. + specs, ids := createSpecs( + []string{app, "task-tree", "--depth=2", "--width=2"}, + []string{app, "task-tree", "--depth=4", "--width=2"}) + + rootDir, err := testutil.SetupRootDir() + if err != nil { + t.Fatalf("error creating root dir: %v", err) + } + defer os.RemoveAll(rootDir) + + var containers []*Container + for i, spec := range specs { + conf := testutil.TestConfig() + bundleDir, err := testutil.SetupContainerInRoot(rootDir, spec, conf) + if err != nil { + t.Fatalf("error setting up container: %v", err) + } + defer os.RemoveAll(bundleDir) + cont, err := Create(ids[i], spec, conf, bundleDir, "", "") + if err != nil { + t.Fatalf("error creating container: %v", err) + } + defer cont.Destroy() + if err := cont.Start(conf); err != nil { + t.Fatalf("error starting container: %v", err) + } + containers = append(containers, cont) + } + + // Wait until all processes are created. + rootProcCount := int(math.Pow(2, 3) - 1) + if err := waitForProcessCount(containers[0], rootProcCount); err != nil { + t.Fatal(err) + } + procCount := int(math.Pow(2, 5) - 1) + if err := waitForProcessCount(containers[1], procCount); err != nil { + t.Fatal(err) + } + + // Exec more processes to ensure signal works for exec'd processes too. + args := &control.ExecArgs{ + Filename: app, + Argv: []string{app, "task-tree", "--depth=2", "--width=2"}, + } + if _, err := containers[1].Execute(args); err != nil { + t.Fatalf("error exec'ing: %v", err) + } + procCount += 3 + if err := waitForProcessCount(containers[1], procCount); err != nil { + t.Fatal(err) + } + + // Kill'Em All + containers[1].Signal(syscall.SIGKILL, true) + + // Check that all processes are gone. + if err := waitForProcessCount(containers[1], 0); err != nil { + t.Fatal(err) + } + // Check that root container was not affected. + if err := waitForProcessCount(containers[0], rootProcCount); err != nil { + t.Fatal(err) + } +} diff --git a/runsc/container/test_app.go b/runsc/container/test_app.go index 768293cf9..a99eb97c4 100644 --- a/runsc/container/test_app.go +++ b/runsc/container/test_app.go @@ -22,17 +22,20 @@ import ( "log" "net" "os" + "os/exec" "strconv" "time" "flag" "github.com/google/subcommands" + "gvisor.googlesource.com/gvisor/runsc/test/testutil" ) func main() { subcommands.Register(subcommands.HelpCommand(), "") subcommands.Register(subcommands.FlagsCommand(), "") subcommands.Register(new(uds), "") + subcommands.Register(new(taskTree), "") flag.Parse() @@ -114,3 +117,63 @@ func server(listener net.Listener, out *os.File) { fmt.Fprint(out, string(data)+"\n") } } + +type taskTree struct { + depth int + width int +} + +// Name implements subcommands.Command. +func (*taskTree) Name() string { + return "task-tree" +} + +// Synopsis implements subcommands.Command. +func (*taskTree) Synopsis() string { + return "creates a tree of tasks" +} + +// Usage implements subcommands.Command. +func (*taskTree) Usage() string { + return "task-tree " +} + +// SetFlags implements subcommands.Command. +func (c *taskTree) SetFlags(f *flag.FlagSet) { + f.IntVar(&c.depth, "depth", 1, "number of levels to create") + f.IntVar(&c.width, "width", 1, "number of tasks at each level") +} + +// Execute implements subcommands.Command. +func (c *taskTree) Execute(ctx context.Context, f *flag.FlagSet, args ...interface{}) subcommands.ExitStatus { + stop := testutil.StartReaper() + defer stop() + + if c.depth == 0 { + log.Printf("Child sleeping, PID: %d\n", os.Getpid()) + for { + time.Sleep(24 * time.Hour) + } + } + log.Printf("Parent %d sleeping, PID: %d\n", c.depth, os.Getpid()) + + var cmds []*exec.Cmd + for i := 0; i < c.width; i++ { + cmd := exec.Command( + "/proc/self/exe", c.Name(), + "--depth", strconv.Itoa(c.depth-1), + "--width", strconv.Itoa(c.width)) + cmd.Stdout = os.Stdout + cmd.Stderr = os.Stderr + + if err := cmd.Start(); err != nil { + log.Fatal("failed to call self:", err) + } + cmds = append(cmds, cmd) + } + + for _, c := range cmds { + c.Wait() + } + return subcommands.ExitSuccess +} diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index c3d90d5f4..ef85f175f 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -183,10 +183,9 @@ func (s *Sandbox) Processes(cid string) ([]*control.Process, error) { } defer conn.Close() + args := boot.ProcessesArgs{CID: cid} var pl []*control.Process - // TODO: Pass in the container id (cid) here. The sandbox - // should return process info for only that container. - if err := conn.Call(boot.ContainerProcesses, nil, &pl); err != nil { + if err := conn.Call(boot.ContainerProcesses, &args, &pl); err != nil { return nil, fmt.Errorf("error retrieving process data from sandbox: %v", err) } return pl, nil @@ -194,19 +193,17 @@ func (s *Sandbox) Processes(cid string) ([]*control.Process, error) { // Execute runs the specified command in the container. It returns the pid of // the newly created process. -func (s *Sandbox) Execute(cid string, args *control.ExecArgs) (int32, error) { - log.Debugf("Executing new process in container %q in sandbox %q", cid, s.ID) +func (s *Sandbox) Execute(args *control.ExecArgs) (int32, error) { + log.Debugf("Executing new process in container %q in sandbox %q", args.ContainerID, s.ID) conn, err := s.sandboxConnect() if err != nil { return 0, s.connError(err) } defer conn.Close() - rpcArgs := &boot.ExecArgs{ExecArgs: *args, CID: cid} - // Send a message to the sandbox control server to start the container. var pid int32 - if err := conn.Call(boot.ContainerExecuteAsync, rpcArgs, &pid); err != nil { + if err := conn.Call(boot.ContainerExecuteAsync, args, &pid); err != nil { return 0, fmt.Errorf("error executing in sandbox: %v", err) } return pid, nil @@ -575,7 +572,7 @@ func (s *Sandbox) destroy() error { } // Signal sends the signal to a container in the sandbox. -func (s *Sandbox) Signal(cid string, sig syscall.Signal) error { +func (s *Sandbox) Signal(cid string, sig syscall.Signal, all bool) error { log.Debugf("Signal sandbox %q", s.ID) conn, err := s.sandboxConnect() if err != nil { @@ -586,6 +583,7 @@ func (s *Sandbox) Signal(cid string, sig syscall.Signal) error { args := boot.SignalArgs{ CID: cid, Signo: int32(sig), + All: all, } if err := conn.Call(boot.ContainerSignal, &args, nil); err != nil { return fmt.Errorf("err signaling container %q: %v", cid, err) -- cgit v1.2.3 From f1c01ed88666ea81d8f5cef7931153a9951a6e64 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Mon, 1 Oct 2018 22:05:41 -0700 Subject: runsc: Support job control signals in "exec -it". Terminal support in runsc relies on host tty file descriptors that are imported into the sandbox. Application tty ioctls are sent directly to the host fd. However, those host tty ioctls are associated in the host kernel with a host process (in this case runsc), and the host kernel intercepts job control characters like ^C and send signals to the host process. Thus, typing ^C into a "runsc exec" shell will send a SIGINT to the runsc process. This change makes "runsc exec" handle all signals, and forward them into the sandbox via the "ContainerSignal" urpc method. Since the "runsc exec" is associated with a particular container process in the sandbox, the signal must be associated with the same container process. One big difficulty is that the signal should not necessarily be sent to the sandbox process started by "exec", but instead must be sent to the foreground process group for the tty. For example, we may exec "bash", and from bash call "sleep 100". A ^C at this point should SIGINT sleep, not bash. To handle this, tty files inside the sandbox must keep track of their foreground process group, which is set/get via ioctls. When an incoming ContainerSignal urpc comes in, we look up the foreground process group via the tty file. Unfortunately, this means we have to expose and cache the tty file in the Loader. Note that "runsc exec" now handles signals properly, but "runs run" does not. That will come in a later CL, as this one is complex enough already. Example: root@:/usr/local/apache2# sleep 100 ^C root@:/usr/local/apache2# sleep 100 ^Z [1]+ Stopped sleep 100 root@:/usr/local/apache2# fg sleep 100 ^C root@:/usr/local/apache2# PiperOrigin-RevId: 215334554 Change-Id: I53cdce39653027908510a5ba8d08c49f9cf24f39 --- pkg/sentry/control/proc.go | 78 ++++++++++---- pkg/sentry/fs/host/BUILD | 1 + pkg/sentry/fs/host/file.go | 144 +++++-------------------- pkg/sentry/fs/host/inode.go | 18 ++-- pkg/sentry/fs/host/tty.go | 185 ++++++++++++++++++++++++++++++++ pkg/sentry/kernel/sessions.go | 5 + runsc/boot/controller.go | 46 ++++++-- runsc/boot/fds.go | 8 +- runsc/boot/loader.go | 108 ++++++++++++++----- runsc/cmd/exec.go | 13 ++- runsc/container/BUILD | 2 + runsc/container/container.go | 51 ++++++--- runsc/container/container_test.go | 117 ++++++++++++++++++++ runsc/container/multi_container_test.go | 7 +- runsc/sandbox/sandbox.go | 49 ++++++--- runsc/test/integration/exec_test.go | 55 ++++++++++ runsc/test/testutil/BUILD | 1 + runsc/test/testutil/docker.go | 21 ++++ runsc/test/testutil/testutil.go | 36 +++++++ 19 files changed, 732 insertions(+), 213 deletions(-) create mode 100644 pkg/sentry/fs/host/tty.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go index faf1168bb..0ba730c1e 100644 --- a/pkg/sentry/control/proc.go +++ b/pkg/sentry/control/proc.go @@ -78,7 +78,7 @@ type ExecArgs struct { Capabilities *auth.TaskCapabilities // StdioIsPty indicates that FDs 0, 1, and 2 are connected to a host - // pty fd. + // pty FD. StdioIsPty bool // FilePayload determines the files to give to the new process. @@ -90,7 +90,7 @@ type ExecArgs struct { // Exec runs a new task. func (proc *Proc) Exec(args *ExecArgs, waitStatus *uint32) error { - newTG, _, err := proc.execAsync(args) + newTG, _, _, err := proc.execAsync(args) if err != nil { return err } @@ -103,18 +103,27 @@ func (proc *Proc) Exec(args *ExecArgs, waitStatus *uint32) error { // ExecAsync runs a new task, but doesn't wait for it to finish. It is defined // as a function rather than a method to avoid exposing execAsync as an RPC. -func ExecAsync(proc *Proc, args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadID, error) { +func ExecAsync(proc *Proc, args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadID, *host.TTYFileOperations, error) { return proc.execAsync(args) } // execAsync runs a new task, but doesn't wait for it to finish. It returns the -// newly created thread group and its PID. -func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadID, error) { +// newly created thread group and its PID. If the stdio FDs are TTYs, then a +// TTYFileOperations that wraps the TTY is also returned. +func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadID, *host.TTYFileOperations, error) { // Import file descriptors. l := limits.NewLimitSet() fdm := proc.Kernel.NewFDMap() defer fdm.DecRef() + // No matter what happens, we should close all files in the FilePayload + // before returning. Any files that are imported will be duped. + defer func() { + for _, f := range args.FilePayload.Files { + f.Close() + } + }() + creds := auth.NewUserCredentials( args.KUID, args.KGID, @@ -150,31 +159,62 @@ func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadI paths := fs.GetPath(initArgs.Envv) f, err := proc.Kernel.RootMountNamespace().ResolveExecutablePath(ctx, initArgs.WorkingDirectory, initArgs.Argv[0], paths) if err != nil { - return nil, 0, fmt.Errorf("error finding executable %q in PATH %v: %v", initArgs.Argv[0], paths, err) + return nil, 0, nil, fmt.Errorf("error finding executable %q in PATH %v: %v", initArgs.Argv[0], paths, err) } initArgs.Filename = f } mounter := fs.FileOwnerFromContext(ctx) - for appFD, f := range args.FilePayload.Files { - enableIoctl := args.StdioIsPty && appFD <= 2 - // Import the given file FD. This dups the FD as well. - file, err := host.ImportFile(ctx, int(f.Fd()), mounter, enableIoctl) - if err != nil { - return nil, 0, err + var ttyFile *fs.File + for appFD, hostFile := range args.FilePayload.Files { + var appFile *fs.File + + if args.StdioIsPty && appFD < 3 { + // Import the file as a host TTY file. + if ttyFile == nil { + var err error + appFile, err = host.ImportFile(ctx, int(hostFile.Fd()), mounter, true /* isTTY */) + if err != nil { + return nil, 0, nil, err + } + defer appFile.DecRef() + + // Remember this in the TTY file, as we will + // use it for the other stdio FDs. + ttyFile = appFile + } else { + // Re-use the existing TTY file, as all three + // stdio FDs must point to the same fs.File in + // order to share TTY state, specifically the + // foreground process group id. + appFile = ttyFile + } + } else { + // Import the file as a regular host file. + var err error + appFile, err = host.ImportFile(ctx, int(hostFile.Fd()), mounter, false /* isTTY */) + if err != nil { + return nil, 0, nil, err + } + defer appFile.DecRef() } - defer file.DecRef() - - // We're done with this file. - f.Close() - if err := fdm.NewFDAt(kdefs.FD(appFD), file, kernel.FDFlags{}, l); err != nil { - return nil, 0, err + // Add the file to the FD map. + if err := fdm.NewFDAt(kdefs.FD(appFD), appFile, kernel.FDFlags{}, l); err != nil { + return nil, 0, nil, err } } - return proc.Kernel.CreateProcess(initArgs) + tg, tid, err := proc.Kernel.CreateProcess(initArgs) + if err != nil { + return nil, 0, nil, err + } + + if ttyFile == nil { + return tg, tid, nil, nil + } + return tg, tid, ttyFile.FileOperations.(*host.TTYFileOperations), nil } // PsArgs is the set of arguments to ps. diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index d1a6eaf6e..c34f1c26b 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -17,6 +17,7 @@ go_library( "socket.go", "socket_state.go", "socket_unsafe.go", + "tty.go", "util.go", "util_unsafe.go", ], diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index 6f469b5cc..22a5d9f12 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -18,15 +18,12 @@ import ( "fmt" "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/fd" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/secio" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -39,6 +36,7 @@ import ( // // +stateify savable type fileOperations struct { + fsutil.NoIoctl `state:"nosave"` fsutil.NoopRelease `state:"nosave"` // iops are the Inode operations for this file. @@ -49,49 +47,49 @@ type fileOperations struct { // dirCursor is the directory cursor. dirCursor string - - // allowIoctl determines whether ioctls should be passed through to the - // host. - allowIoctl bool } // fileOperations implements fs.FileOperations. var _ fs.FileOperations = (*fileOperations)(nil) // NewFile creates a new File backed by the provided host file descriptor. If -// NewFile succeeds, ownership of the fd is transferred to the returned File. +// NewFile succeeds, ownership of the FD is transferred to the returned File. // // The returned File cannot be saved, since there is no guarantee that the same -// fd will exist or represent the same file at time of restore. If such a +// FD will exist or represent the same file at time of restore. If such a // guarantee does exist, use ImportFile instead. func NewFile(ctx context.Context, fd int, mounter fs.FileOwner) (*fs.File, error) { return newFileFromDonatedFD(ctx, fd, mounter, false, false) } // ImportFile creates a new File backed by the provided host file descriptor. -// Unlike NewFile, the file descriptor used by the File is duped from fd to -// ensure that later changes to fd are not reflected by the fs.File. +// Unlike NewFile, the file descriptor used by the File is duped from FD to +// ensure that later changes to FD are not reflected by the fs.File. // -// If the returned file is saved, it will be restored by re-importing the fd +// If the returned file is saved, it will be restored by re-importing the FD // originally passed to ImportFile. It is the restorer's responsibility to -// ensure that the fd represents the same file. -func ImportFile(ctx context.Context, fd int, mounter fs.FileOwner, allowIoctl bool) (*fs.File, error) { - return newFileFromDonatedFD(ctx, fd, mounter, true, allowIoctl) +// ensure that the FD represents the same file. +func ImportFile(ctx context.Context, fd int, mounter fs.FileOwner, isTTY bool) (*fs.File, error) { + return newFileFromDonatedFD(ctx, fd, mounter, true, isTTY) } -// newFileFromDonatedFD returns an fs.File from a donated fd. If the fd is +// newFileFromDonatedFD returns an fs.File from a donated FD. If the FD is // saveable, then saveable is true. -func newFileFromDonatedFD(ctx context.Context, donated int, mounter fs.FileOwner, saveable, allowIoctl bool) (*fs.File, error) { +func newFileFromDonatedFD(ctx context.Context, donated int, mounter fs.FileOwner, saveable, isTTY bool) (*fs.File, error) { var s syscall.Stat_t if err := syscall.Fstat(donated, &s); err != nil { return nil, err } + flags, err := fileFlagsFromDonatedFD(donated) + if err != nil { + return nil, err + } switch s.Mode & syscall.S_IFMT { case syscall.S_IFSOCK: - flags, err := fileFlagsFromDonatedFD(donated) - if err != nil { - return nil, err + if isTTY { + return nil, fmt.Errorf("cannot import host socket as TTY") } + s, err := newSocket(ctx, donated, saveable) if err != nil { return nil, err @@ -101,10 +99,6 @@ func newFileFromDonatedFD(ctx context.Context, donated int, mounter fs.FileOwner }) return s, nil default: - flags, err := fileFlagsFromDonatedFD(donated) - if err != nil { - return nil, err - } msrc := newMountSource(ctx, "/", mounter, &Filesystem{}, fs.MountSourceFlags{}, false /* dontTranslateOwnership */) inode, err := newInode(ctx, msrc, donated, saveable, true /* donated */) if err != nil { @@ -116,14 +110,18 @@ func newFileFromDonatedFD(ctx context.Context, donated int, mounter fs.FileOwner dirent := fs.NewDirent(inode, name) defer dirent.DecRef() - return newFile(ctx, dirent, flags, iops, allowIoctl), nil + if isTTY { + return newTTYFile(ctx, dirent, flags, iops), nil + } + + return newFile(ctx, dirent, flags, iops), nil } } func fileFlagsFromDonatedFD(donated int) (fs.FileFlags, error) { flags, _, errno := syscall.Syscall(syscall.SYS_FCNTL, uintptr(donated), syscall.F_GETFL, 0) if errno != 0 { - log.Warningf("Failed to get file flags for donated fd %d (errno=%d)", donated, errno) + log.Warningf("Failed to get file flags for donated FD %d (errno=%d)", donated, errno) return fs.FileFlags{}, syscall.EIO } accmode := flags & syscall.O_ACCMODE @@ -138,17 +136,14 @@ func fileFlagsFromDonatedFD(donated int) (fs.FileFlags, error) { } // newFile returns a new fs.File. -func newFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags, iops *inodeOperations, allowIoctl bool) *fs.File { +func newFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags, iops *inodeOperations) *fs.File { if !iops.ReturnsWouldBlock() { // Allow reading/writing at an arbitrary offset for files // that support it. flags.Pread = true flags.Pwrite = true } - return fs.NewFile(ctx, dirent, flags, &fileOperations{ - iops: iops, - allowIoctl: allowIoctl, - }) + return fs.NewFile(ctx, dirent, flags, &fileOperations{iops: iops}) } // EventRegister implements waiter.Waitable.EventRegister. @@ -269,7 +264,7 @@ func (f *fileOperations) Fsync(ctx context.Context, file *fs.File, start int64, func (f *fileOperations) Flush(context.Context, *fs.File) error { // This is a no-op because flushing the resource backing this // file would mean closing it. We can't do that because other - // open files may depend on the backing host fd. + // open files may depend on the backing host FD. return nil } @@ -285,88 +280,3 @@ func (f *fileOperations) ConfigureMMap(ctx context.Context, file *fs.File, opts func (f *fileOperations) Seek(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64) (int64, error) { return fsutil.SeekWithDirCursor(ctx, file, whence, offset, &f.dirCursor) } - -// Ioctl implements fs.FileOperations.Iocotl. -func (f *fileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { - if !f.allowIoctl { - return 0, syserror.ENOTTY - } - // Ignore arg[0]. This is the real FD: - fd := f.iops.fileState.FD() - ioctl := args[1].Uint64() - switch ioctl { - case linux.TCGETS: - termios, err := ioctlGetTermios(fd) - if err != nil { - return 0, err - } - _, err = usermem.CopyObjectOut(ctx, io, args[2].Pointer(), termios, usermem.IOOpts{ - AddressSpaceActive: true, - }) - return 0, err - - case linux.TCSETS, linux.TCSETSW, linux.TCSETSF: - var termios linux.Termios - if _, err := usermem.CopyObjectIn(ctx, io, args[2].Pointer(), &termios, usermem.IOOpts{ - AddressSpaceActive: true, - }); err != nil { - return 0, err - } - err := ioctlSetTermios(fd, ioctl, &termios) - return 0, err - - case linux.TIOCGPGRP: - // Args: pid_t *argp - // When successful, equivalent to *argp = tcgetpgrp(fd). - // Get the process group ID of the foreground process group on - // this terminal. - - t := kernel.TaskFromContext(ctx) - if t == nil { - panic(fmt.Sprintf("cannot get thread group from context %v", ctx)) - } - tid := t.ThreadID() - _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), &tid, usermem.IOOpts{ - AddressSpaceActive: true, - }) - return 0, err - - case linux.TIOCSPGRP: - // Args: const pid_t *argp - // Equivalent to tcsetpgrp(fd, *argp). - // Set the foreground process group ID of this terminal. - - // Not much we can do with this one at the moment, so we just - // lie and pretend everything is great. Bash and Sh seem fine - // with this. - log.Warningf("Ignoring application ioctl(TIOCSPGRP) call") - return 0, nil - - case linux.TIOCGWINSZ: - // Args: struct winsize *argp - // Get window size. - winsize, err := ioctlGetWinsize(fd) - if err != nil { - return 0, err - } - _, err = usermem.CopyObjectOut(ctx, io, args[2].Pointer(), winsize, usermem.IOOpts{ - AddressSpaceActive: true, - }) - return 0, err - - case linux.TIOCSWINSZ: - // Args: const struct winsize *argp - // Set window size. - var winsize linux.Winsize - if _, err := usermem.CopyObjectIn(ctx, io, args[2].Pointer(), &winsize, usermem.IOOpts{ - AddressSpaceActive: true, - }); err != nil { - return 0, err - } - err := ioctlSetWinsize(fd, &winsize) - return 0, err - - default: - return 0, syserror.ENOTTY - } -} diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index e7254fa7d..c2e8ba62f 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -73,7 +73,7 @@ type inodeFileState struct { // Common file system state. mops *superOperations `state:"wait"` - // descriptor is the backing host fd. + // descriptor is the backing host FD. descriptor *descriptor `state:"wait"` // Event queue for blocking operations. @@ -167,7 +167,7 @@ func (i *inodeFileState) unstableAttr(ctx context.Context) (fs.UnstableAttr, err // inodeOperations implements fs.InodeOperations. var _ fs.InodeOperations = (*inodeOperations)(nil) -// newInode returns a new fs.Inode backed by the host fd. +// newInode returns a new fs.Inode backed by the host FD. func newInode(ctx context.Context, msrc *fs.MountSource, fd int, saveable bool, donated bool) (*fs.Inode, error) { // Retrieve metadata. var s syscall.Stat_t @@ -212,8 +212,8 @@ func (i *inodeOperations) Mappable(inode *fs.Inode) memmap.Mappable { return i.cachingInodeOps } -// ReturnsWouldBlock returns true if this host fd can return EWOULDBLOCK -// for operations that would block. +// ReturnsWouldBlock returns true if this host FD can return EWOULDBLOCK for +// operations that would block. func (i *inodeOperations) ReturnsWouldBlock() bool { return i.fileState.descriptor.wouldBlock } @@ -226,7 +226,7 @@ func (i *inodeOperations) Release(context.Context) { // Lookup implements fs.InodeOperations.Lookup. func (i *inodeOperations) Lookup(ctx context.Context, dir *fs.Inode, name string) (*fs.Dirent, error) { - // Get a new fd relative to i at name. + // Get a new FD relative to i at name. fd, err := open(i, name) if err != nil { if err == syserror.ENOENT { @@ -321,7 +321,7 @@ func (i *inodeOperations) BoundEndpoint(inode *fs.Inode, path string) unix.Bound // GetFile implements fs.InodeOperations.GetFile. func (i *inodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { - return newFile(ctx, d, flags, i, false), nil + return newFile(ctx, d, flags, i), nil } // canMap returns true if this fs.Inode can be memory mapped. @@ -362,7 +362,7 @@ func (i *inodeOperations) SetOwner(context.Context, *fs.Inode, fs.FileOwner) err func (i *inodeOperations) SetPermissions(ctx context.Context, inode *fs.Inode, f fs.FilePermissions) bool { // Can we use host kernel metadata caches? if !inode.MountSource.Flags.ForcePageCache || !canMap(inode) { - // Then just change the timestamps on the fd, the host + // Then just change the timestamps on the FD, the host // will synchronize the metadata update with any host // inode and page cache. return syscall.Fchmod(i.fileState.FD(), uint32(f.LinuxMode())) == nil @@ -375,7 +375,7 @@ func (i *inodeOperations) SetPermissions(ctx context.Context, inode *fs.Inode, f func (i *inodeOperations) SetTimestamps(ctx context.Context, inode *fs.Inode, ts fs.TimeSpec) error { // Can we use host kernel metadata caches? if !inode.MountSource.Flags.ForcePageCache || !canMap(inode) { - // Then just change the timestamps on the fd, the host + // Then just change the timestamps on the FD, the host // will synchronize the metadata update with any host // inode and page cache. return setTimestamps(i.fileState.FD(), ts) @@ -388,7 +388,7 @@ func (i *inodeOperations) SetTimestamps(ctx context.Context, inode *fs.Inode, ts func (i *inodeOperations) Truncate(ctx context.Context, inode *fs.Inode, size int64) error { // Is the file not memory-mappable? if !canMap(inode) { - // Then just change the file size on the fd, the host + // Then just change the file size on the FD, the host // will synchronize the metadata update with any host // inode and page cache. return syscall.Ftruncate(i.fileState.FD(), size) diff --git a/pkg/sentry/fs/host/tty.go b/pkg/sentry/fs/host/tty.go new file mode 100644 index 000000000..ad1323610 --- /dev/null +++ b/pkg/sentry/fs/host/tty.go @@ -0,0 +1,185 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package host + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// TTYFileOperations implements fs.FileOperations for a host file descriptor +// that wraps a TTY FD. +// +// +stateify savable +type TTYFileOperations struct { + fileOperations + + // mu protects the fields below. + mu sync.Mutex + + // FGProcessGroup is the foreground process group this TTY. Will be + // nil if not set or if this file has been released. + fgProcessGroup *kernel.ProcessGroup +} + +// newTTYFile returns a new fs.File that wraps a TTY FD. +func newTTYFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags, iops *inodeOperations) *fs.File { + return fs.NewFile(ctx, dirent, flags, &TTYFileOperations{ + fileOperations: fileOperations{iops: iops}, + }) +} + +// ForegroundProcessGroup returns the foreground process for the TTY. This will +// be nil if the foreground process has not been set or if the file has been +// released. +func (t *TTYFileOperations) ForegroundProcessGroup() *kernel.ProcessGroup { + t.mu.Lock() + defer t.mu.Unlock() + return t.fgProcessGroup +} + +// Release implements fs.FileOperations.Release. +func (t *TTYFileOperations) Release() { + t.mu.Lock() + t.fgProcessGroup = nil + t.mu.Unlock() + + t.fileOperations.Release() +} + +// Ioctl implements fs.FileOperations.Ioctl. +func (t *TTYFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { + // Ignore arg[0]. This is the real FD: + fd := t.fileOperations.iops.fileState.FD() + ioctl := args[1].Uint64() + switch ioctl { + case linux.TCGETS: + termios, err := ioctlGetTermios(fd) + if err != nil { + return 0, err + } + _, err = usermem.CopyObjectOut(ctx, io, args[2].Pointer(), termios, usermem.IOOpts{ + AddressSpaceActive: true, + }) + return 0, err + + case linux.TCSETS, linux.TCSETSW, linux.TCSETSF: + var termios linux.Termios + if _, err := usermem.CopyObjectIn(ctx, io, args[2].Pointer(), &termios, usermem.IOOpts{ + AddressSpaceActive: true, + }); err != nil { + return 0, err + } + err := ioctlSetTermios(fd, ioctl, &termios) + return 0, err + + case linux.TIOCGPGRP: + // Args: pid_t *argp + // When successful, equivalent to *argp = tcgetpgrp(fd). + // Get the process group ID of the foreground process group on + // this terminal. + + t.mu.Lock() + defer t.mu.Unlock() + + if t.fgProcessGroup == nil { + // No process group has been set yet. Let's just lie + // and tell it the process group from the current task. + // The app is probably going to set it to something + // else very soon anyways. + t.fgProcessGroup = kernel.TaskFromContext(ctx).ThreadGroup().ProcessGroup() + } + + // Map the ProcessGroup into a ProcessGroupID in the task's PID + // namespace. + pgID := kernel.TaskFromContext(ctx).ThreadGroup().PIDNamespace().IDOfProcessGroup(t.fgProcessGroup) + _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), &pgID, usermem.IOOpts{ + AddressSpaceActive: true, + }) + return 0, err + + case linux.TIOCSPGRP: + // Args: const pid_t *argp + // Equivalent to tcsetpgrp(fd, *argp). + // Set the foreground process group ID of this terminal. + + var pgID kernel.ProcessGroupID + if _, err := usermem.CopyObjectIn(ctx, io, args[2].Pointer(), &pgID, usermem.IOOpts{ + AddressSpaceActive: true, + }); err != nil { + return 0, err + } + + // pgID must be non-negative. + if pgID < 0 { + return 0, syserror.EINVAL + } + + // Process group with pgID must exist in this PID namespace. + task := kernel.TaskFromContext(ctx) + pidns := task.PIDNamespace() + pg := pidns.ProcessGroupWithID(pgID) + if pg == nil { + return 0, syserror.ESRCH + } + + // Process group must be in same session as calling task's + // process group. + curSession := task.ThreadGroup().ProcessGroup().Session() + curSessionID := pidns.IDOfSession(curSession) + if pidns.IDOfSession(pg.Session()) != curSessionID { + return 0, syserror.EPERM + } + + t.mu.Lock() + t.fgProcessGroup = pg + t.mu.Unlock() + return 0, nil + + case linux.TIOCGWINSZ: + // Args: struct winsize *argp + // Get window size. + winsize, err := ioctlGetWinsize(fd) + if err != nil { + return 0, err + } + _, err = usermem.CopyObjectOut(ctx, io, args[2].Pointer(), winsize, usermem.IOOpts{ + AddressSpaceActive: true, + }) + return 0, err + + case linux.TIOCSWINSZ: + // Args: const struct winsize *argp + // Set window size. + var winsize linux.Winsize + if _, err := usermem.CopyObjectIn(ctx, io, args[2].Pointer(), &winsize, usermem.IOOpts{ + AddressSpaceActive: true, + }); err != nil { + return 0, err + } + err := ioctlSetWinsize(fd, &winsize) + return 0, err + + default: + return 0, syserror.ENOTTY + } +} diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index cf4e18805..b44d218d9 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -219,6 +219,11 @@ func (pg *ProcessGroup) handleOrphan() { return } +// Session returns the process group's session without taking a reference. +func (pg *ProcessGroup) Session() *Session { + return pg.session +} + // CreateSession creates a new Session, with the ThreadGroup as the leader. // // EPERM may be returned if either the given ThreadGroup is already a Session diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index 98356e8b7..eaeb9e2d8 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -65,6 +65,10 @@ const ( // ContainerSignal is used to send a signal to a container. ContainerSignal = "containerManager.Signal" + // ContainerSignalProcess is used to send a signal to a particular + // process in a container. + ContainerSignalProcess = "containerManager.SignalProcess" + // ContainerStart is the URPC endpoint for running a non-root container // within a sandbox. ContainerStart = "containerManager.Start" @@ -92,7 +96,7 @@ const ( SandboxStacks = "debug.Stacks" ) -// ControlSocketAddr generates an abstract unix socket name for the given id. +// ControlSocketAddr generates an abstract unix socket name for the given ID. func ControlSocketAddr(id string) string { return fmt.Sprintf("\x00runsc-sandbox.%s", id) } @@ -248,7 +252,7 @@ func (cm *containerManager) Destroy(cid *string, _ *struct{}) error { } // ExecuteAsync starts running a command on a created or running sandbox. It -// returns the pid of the new process. +// returns the PID of the new process. func (cm *containerManager) ExecuteAsync(args *control.ExecArgs, pid *int32) error { log.Debugf("containerManager.ExecuteAsync: %+v", args) tgid, err := cm.l.executeAsync(args) @@ -373,8 +377,12 @@ func (cm *containerManager) Restore(o *RestoreOpts, _ *struct{}) error { // restore the state of multiple containers, nor exec processes. cm.l.sandboxID = o.SandboxID cm.l.mu.Lock() - key := execID{cid: o.SandboxID} - cm.l.processes = map[execID]*kernel.ThreadGroup{key: cm.l.k.GlobalInit()} + eid := execID{cid: o.SandboxID} + cm.l.processes = map[execID]*execProcess{ + eid: &execProcess{ + tg: cm.l.k.GlobalInit(), + }, + } cm.l.mu.Unlock() // Tell the root container to start and wait for the result. @@ -419,7 +427,7 @@ func (cm *containerManager) WaitPID(args *WaitPIDArgs, waitStatus *uint32) error // SignalArgs are arguments to the Signal method. type SignalArgs struct { - // CID is the container id. + // CID is the container ID. CID string // Signo is the signal to send to the process. @@ -430,9 +438,31 @@ type SignalArgs struct { All bool } -// Signal sends a signal to the init process of the container. -// TODO: Send signal to exec process. +// Signal sends a signal to the root process of the container. func (cm *containerManager) Signal(args *SignalArgs, _ *struct{}) error { log.Debugf("containerManager.Signal %q %d, all: %t", args.CID, args.Signo, args.All) - return cm.l.signal(args.CID, args.Signo, args.All) + return cm.l.signalContainer(args.CID, args.Signo, args.All) +} + +// SignalProcessArgs are arguments to the Signal method. +type SignalProcessArgs struct { + // CID is the container ID. + CID string + + // PID is the process ID in the given container that will be signaled. + PID int32 + + // Signo is the signal to send to the process. + Signo int32 + + // SendToForegroundProcess indicates that the signal should be sent to + // the foreground process group in the session that PID belongs to. + // This is only valid if the process is attached to a host TTY. + SendToForegroundProcess bool +} + +// SignalProcess sends a signal to a particular process in the container. +func (cm *containerManager) SignalProcess(args *SignalProcessArgs, _ *struct{}) error { + log.Debugf("containerManager.Signal: %+v", args) + return cm.l.signalProcess(args.CID, args.PID, args.Signo, args.SendToForegroundProcess) } diff --git a/runsc/boot/fds.go b/runsc/boot/fds.go index 92d641b68..a5a6ba8af 100644 --- a/runsc/boot/fds.go +++ b/runsc/boot/fds.go @@ -25,8 +25,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/limits" ) -// createFDMap creates an fd map that contains stdin, stdout, and stderr. If -// console is true, then ioctl calls will be passed through to the host fd. +// createFDMap creates an FD map that contains stdin, stdout, and stderr. If +// console is true, then ioctl calls will be passed through to the host FD. // Upon success, createFDMap dups then closes stdioFDs. func createFDMap(ctx context.Context, k *kernel.Kernel, l *limits.LimitSet, console bool, stdioFDs []int) (*kernel.FDMap, error) { if len(stdioFDs) != 3 { @@ -36,7 +36,7 @@ func createFDMap(ctx context.Context, k *kernel.Kernel, l *limits.LimitSet, cons fdm := k.NewFDMap() defer fdm.DecRef() - // Maps sandbox fd to host fd. + // Maps sandbox FD to host FD. fdMap := map[int]int{ 0: stdioFDs[0], 1: stdioFDs[1], @@ -45,7 +45,7 @@ func createFDMap(ctx context.Context, k *kernel.Kernel, l *limits.LimitSet, cons mounter := fs.FileOwnerFromContext(ctx) for sfd, hfd := range fdMap { - file, err := host.ImportFile(ctx, hfd, mounter, console /* allow ioctls */) + file, err := host.ImportFile(ctx, hfd, mounter, console /* isTTY */) if err != nil { return nil, fmt.Errorf("failed to import fd %d: %v", hfd, err) } diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 9fa9b51a0..766a2e968 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -31,6 +31,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/control" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/host" "gvisor.googlesource.com/gvisor/pkg/sentry/inet" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" @@ -112,7 +113,7 @@ type Loader struct { // have the corresponding pid set. // // processes is guardded by mu. - processes map[execID]*kernel.ThreadGroup + processes map[execID]*execProcess } // execID uniquely identifies a sentry process. @@ -121,6 +122,14 @@ type execID struct { pid kernel.ThreadID } +// execProcess contains the thread group and host TTY of a sentry process. +type execProcess struct { + tg *kernel.ThreadGroup + + // tty will be nil if the process is not attached to a terminal. + tty *host.TTYFileOperations +} + func init() { // Initialize the random number generator. rand.Seed(gtime.Now().UnixNano()) @@ -276,7 +285,7 @@ func New(id string, spec *specs.Spec, conf *Config, controllerFD, deviceFD int, startSignalForwarding: startSignalForwarding, rootProcArgs: procArgs, sandboxID: id, - processes: make(map[execID]*kernel.ThreadGroup), + processes: make(map[execID]*execProcess), } ctrl.manager.l = l return l, nil @@ -330,7 +339,7 @@ func createPlatform(conf *Config, deviceFD int) (platform.Platform, error) { case PlatformKVM: log.Infof("Platform: kvm") if deviceFD < 0 { - return nil, fmt.Errorf("kvm device fd must be provided") + return nil, fmt.Errorf("kvm device FD must be provided") } return kvm.New(os.NewFile(uintptr(deviceFD), "kvm device")) default: @@ -413,8 +422,8 @@ func (l *Loader) run() error { } l.mu.Lock() - key := execID{cid: l.sandboxID} - l.processes[key] = l.k.GlobalInit() + eid := execID{cid: l.sandboxID} + l.processes[eid] = &execProcess{tg: l.k.GlobalInit()} l.mu.Unlock() // Start signal forwarding only after an init process is created. @@ -510,8 +519,8 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config l.mu.Lock() defer l.mu.Unlock() - key := execID{cid: cid} - l.processes[key] = tg + eid := execID{cid: cid} + l.processes[eid] = &execProcess{tg: tg} return nil } @@ -520,7 +529,7 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config // filesystem. func (l *Loader) destroyContainer(cid string) error { // First kill and wait for all processes in the container. - if err := l.signal(cid, int32(linux.SIGKILL), true /*all*/); err != nil { + if err := l.signalContainer(cid, int32(linux.SIGKILL), true /*all*/); err != nil { return fmt.Errorf("failed to SIGKILL all container processes: %v", err) } @@ -549,12 +558,12 @@ func (l *Loader) executeAsync(args *control.ExecArgs) (kernel.ThreadID, error) { // process with the same Root. l.mu.Lock() rootKey := execID{cid: args.ContainerID} - tg, ok := l.processes[rootKey] + ep, ok := l.processes[rootKey] l.mu.Unlock() if !ok { return 0, fmt.Errorf("cannot exec in container %q: no such container", args.ContainerID) } - tg.Leader().WithMuLocked(func(t *kernel.Task) { + ep.tg.Leader().WithMuLocked(func(t *kernel.Task) { args.Root = t.FSContext().RootDirectory() }) if args.Root != nil { @@ -563,7 +572,7 @@ func (l *Loader) executeAsync(args *control.ExecArgs) (kernel.ThreadID, error) { // Start the process. proc := control.Proc{Kernel: l.k} - tg, tgid, err := control.ExecAsync(&proc, args) + tg, tgid, ttyFile, err := control.ExecAsync(&proc, args) if err != nil { return 0, fmt.Errorf("error executing: %+v: %v", args, err) } @@ -573,7 +582,10 @@ func (l *Loader) executeAsync(args *control.ExecArgs) (kernel.ThreadID, error) { l.mu.Lock() defer l.mu.Unlock() eid := execID{cid: args.ContainerID, pid: tgid} - l.processes[eid] = tg + l.processes[eid] = &execProcess{ + tg: tg, + tty: ttyFile, + } log.Debugf("updated processes: %v", l.processes) return tgid, nil @@ -584,8 +596,8 @@ func (l *Loader) waitContainer(cid string, waitStatus *uint32) error { // Don't defer unlock, as doing so would make it impossible for // multiple clients to wait on the same container. l.mu.Lock() - key := execID{cid: cid} - tg, ok := l.processes[key] + eid := execID{cid: cid} + ep, ok := l.processes[eid] l.mu.Unlock() if !ok { return fmt.Errorf("can't find process for container %q in %v", cid, l.processes) @@ -593,7 +605,7 @@ func (l *Loader) waitContainer(cid string, waitStatus *uint32) error { // If the thread either has already exited or exits during waiting, // consider the container exited. - ws := l.wait(tg) + ws := l.wait(ep.tg) *waitStatus = ws return nil } @@ -610,10 +622,10 @@ func (l *Loader) waitPID(tgid kernel.ThreadID, cid string, clearStatus bool, wai // entry in l.processes. l.mu.Lock() eid := execID{cid: cid, pid: tgid} - tg, ok := l.processes[eid] + ep, ok := l.processes[eid] l.mu.Unlock() if ok { - ws := l.wait(tg) + ws := l.wait(ep.tg) *waitStatus = ws if clearStatus { // Remove tg from the cache. @@ -626,8 +638,8 @@ func (l *Loader) waitPID(tgid kernel.ThreadID, cid string, clearStatus bool, wai } // This process wasn't created by runsc exec or start, so just find it - // by pid and hope it hasn't exited yet. - tg = l.k.TaskSet().Root.ThreadGroupWithID(kernel.ThreadID(tgid)) + // by PID and hope it hasn't exited yet. + tg := l.k.TaskSet().Root.ThreadGroupWithID(kernel.ThreadID(tgid)) if tg == nil { return fmt.Errorf("no thread group with ID %d", tgid) } @@ -682,18 +694,66 @@ func newEmptyNetworkStack(conf *Config, clock tcpip.Clock) (inet.Stack, error) { } } -func (l *Loader) signal(cid string, signo int32, all bool) error { +// signalProcess sends a signal to the process with the given PID. If +// sendToFGProcess is true, then the signal will be sent to the foreground +// process group in the same session that PID belongs to. +func (l *Loader) signalProcess(cid string, pid, signo int32, sendToFGProcess bool) error { + si := arch.SignalInfo{Signo: signo} + + if pid <= 0 { + return fmt.Errorf("failed to signal container %q PID %d: PID must be positive", cid, pid) + } + + eid := execID{ + cid: cid, + pid: kernel.ThreadID(pid), + } l.mu.Lock() - key := execID{cid: cid} - tg, ok := l.processes[key] + ep, ok := l.processes[eid] l.mu.Unlock() + if !ok { - return fmt.Errorf("failed to signal container %q: no such container", cid) + return fmt.Errorf("failed to signal container %q PID %d: no such PID", cid, pid) + } + + if !sendToFGProcess { + // Send signal directly to exec process. + return ep.tg.SendSignal(&si) } + // Lookup foreground process group from the TTY for the given process, + // and send the signal to it. + if ep.tty == nil { + return fmt.Errorf("failed to signal foreground process group in container %q PID %d: no TTY attached", cid, pid) + } + pg := ep.tty.ForegroundProcessGroup() + if pg == nil { + // No foreground process group has been set. Signal the + // original thread group. + log.Warningf("No foreground process group for container %q and PID %d. Sending signal directly to PID %d.", cid, pid, pid) + return ep.tg.SendSignal(&si) + } + + // Send the signal. + return pg.Originator().SendSignal(&si) +} + +// signalContainer sends a signal to the root container process, or to all +// processes in the container if all is true. +func (l *Loader) signalContainer(cid string, signo int32, all bool) error { si := arch.SignalInfo{Signo: signo} + + l.mu.Lock() + defer l.mu.Unlock() + + eid := execID{cid: cid} + ep, ok := l.processes[eid] + if !ok { + return fmt.Errorf("failed to signal container %q: no such container", cid) + } + if !all { - return tg.Leader().SendSignal(&si) + return ep.tg.SendSignal(&si) } // Pause the kernel to prevent new processes from being created while diff --git a/runsc/cmd/exec.go b/runsc/cmd/exec.go index 28229dbcf..336edf3f6 100644 --- a/runsc/cmd/exec.go +++ b/runsc/cmd/exec.go @@ -158,6 +158,13 @@ func (ex *Exec) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) Fatalf("error getting processes for container: %v", err) } + if e.StdioIsPty { + // Forward signals sent to this process to the foreground + // process in the sandbox. + stopForwarding := c.ForwardSignals(pid, true /* fgProcess */) + defer stopForwarding() + } + // Write the sandbox-internal pid if required. if ex.internalPidFile != "" { pidStr := []byte(strconv.Itoa(int(pid))) @@ -216,9 +223,9 @@ func (ex *Exec) execAndWait(waitStatus *syscall.WaitStatus) subcommands.ExitStat cmd.Stderr = os.Stderr // If the console control socket file is provided, then create a new - // pty master/slave pair and set the tty on the sandbox process. + // pty master/slave pair and set the TTY on the sandbox process. if ex.consoleSocket != "" { - // Create a new tty pair and send the master on the provided + // Create a new TTY pair and send the master on the provided // socket. tty, err := console.NewWithSocket(ex.consoleSocket) if err != nil { @@ -226,7 +233,7 @@ func (ex *Exec) execAndWait(waitStatus *syscall.WaitStatus) subcommands.ExitStat } defer tty.Close() - // Set stdio to the new tty slave. + // Set stdio to the new TTY slave. cmd.Stdin = tty cmd.Stdout = tty cmd.Stderr = tty diff --git a/runsc/container/BUILD b/runsc/container/BUILD index e68fb1e8e..bf8b9a2ab 100644 --- a/runsc/container/BUILD +++ b/runsc/container/BUILD @@ -50,10 +50,12 @@ go_test( "//pkg/sentry/control", "//pkg/sentry/kernel/auth", "//pkg/unet", + "//pkg/urpc", "//runsc/boot", "//runsc/specutils", "//runsc/test/testutil", "@com_github_cenkalti_backoff//:go_default_library", + "@com_github_kr_pty//:go_default_library", "@com_github_opencontainers_runtime-spec//specs-go:go_default_library", "@org_golang_x_sys//unix:go_default_library", ], diff --git a/runsc/container/container.go b/runsc/container/container.go index be833c03d..4b0037b4e 100644 --- a/runsc/container/container.go +++ b/runsc/container/container.go @@ -22,6 +22,7 @@ import ( "io/ioutil" "os" "os/exec" + "os/signal" "path/filepath" "regexp" "strconv" @@ -107,14 +108,13 @@ type Container struct { Owner string `json:"owner"` // ConsoleSocket is the path to a unix domain socket that will receive - // the console FD. It is only used during create, so we don't need to - // store it in the metadata. - ConsoleSocket string `json:"-"` + // the console FD. + ConsoleSocket string `json:"consoleSocket"` // Status is the current container Status. Status Status `json:"status"` - // GoferPid is the pid of the gofer running along side the sandbox. May + // GoferPid is the PID of the gofer running along side the sandbox. May // be 0 if the gofer has been killed. GoferPid int `json:"goferPid"` @@ -313,12 +313,12 @@ func Create(id string, spec *specs.Spec, conf *boot.Config, bundleDir, consoleSo return nil, err } - // Write the pid file. Containerd considers the create complete after + // Write the PID file. Containerd considers the create complete after // this file is created, so it must be the last thing we do. if pidFile != "" { if err := ioutil.WriteFile(pidFile, []byte(strconv.Itoa(c.Pid())), 0644); err != nil { c.Destroy() - return nil, fmt.Errorf("error writing pid file: %v", err) + return nil, fmt.Errorf("error writing PID file: %v", err) } } @@ -406,7 +406,7 @@ func Run(id string, spec *specs.Spec, conf *boot.Config, bundleDir, consoleSocke return c.Wait() } -// Execute runs the specified command in the container. It returns the pid of +// Execute runs the specified command in the container. It returns the PID of // the newly created process. func (c *Container) Execute(args *control.ExecArgs) (int32, error) { log.Debugf("Execute in container %q, args: %+v", c.ID, args) @@ -429,7 +429,7 @@ func (c *Container) Event() (*boot.Event, error) { // Pid returns the Pid of the sandbox the container is running in, or -1 if the // container is not running. func (c *Container) Pid() int { - if err := c.requireStatus("pid", Created, Running, Paused); err != nil { + if err := c.requireStatus("get PID", Created, Running, Paused); err != nil { return -1 } return c.Sandbox.Pid @@ -449,7 +449,7 @@ func (c *Container) Wait() (syscall.WaitStatus, error) { // WaitRootPID waits for process 'pid' in the sandbox's PID namespace and // returns its WaitStatus. func (c *Container) WaitRootPID(pid int32, clearStatus bool) (syscall.WaitStatus, error) { - log.Debugf("Wait on pid %d in sandbox %q", pid, c.Sandbox.ID) + log.Debugf("Wait on PID %d in sandbox %q", pid, c.Sandbox.ID) if !c.isSandboxRunning() { return 0, fmt.Errorf("container is not running") } @@ -459,7 +459,7 @@ func (c *Container) WaitRootPID(pid int32, clearStatus bool) (syscall.WaitStatus // WaitPID waits for process 'pid' in the container's PID namespace and returns // its WaitStatus. func (c *Container) WaitPID(pid int32, clearStatus bool) (syscall.WaitStatus, error) { - log.Debugf("Wait on pid %d in container %q", pid, c.ID) + log.Debugf("Wait on PID %d in container %q", pid, c.ID) if !c.isSandboxRunning() { return 0, fmt.Errorf("container is not running") } @@ -483,7 +483,30 @@ func (c *Container) Signal(sig syscall.Signal, all bool) error { if !c.isSandboxRunning() { return fmt.Errorf("container is not running") } - return c.Sandbox.Signal(c.ID, sig, all) + return c.Sandbox.SignalContainer(c.ID, sig, all) +} + +// ForwardSignals forwards all signals received by the current process to the +// container process inside the sandbox. It returns a function that will stop +// forwarding signals. +func (c *Container) ForwardSignals(pid int32, fgProcess bool) func() { + log.Debugf("Forwarding all signals to container %q PID %d fgProcess=%t", c.ID, pid, fgProcess) + sigCh := make(chan os.Signal, 1) + signal.Notify(sigCh) + go func() { + for s := range sigCh { + log.Debugf("Forwarding signal %d to container %q PID %d fgProcess=%t", s, c.ID, pid, fgProcess) + if err := c.Sandbox.SignalProcess(c.ID, pid, s.(syscall.Signal), fgProcess); err != nil { + log.Warningf("error forwarding signal %d to container %q: %v", s, c.ID, err) + } + } + log.Debugf("Done forwarding signals to container %q PID %d fgProcess=%t", c.ID, pid, fgProcess) + }() + + return func() { + signal.Stop(sigCh) + close(sigCh) + } } // Checkpoint sends the checkpoint call to the container. @@ -683,9 +706,9 @@ func (c *Container) createGoferProcess(spec *specs.Spec, conf *boot.Config, bund if err != nil { return nil, err } - sandEnds = append(sandEnds, os.NewFile(uintptr(fds[0]), "sandbox io fd")) + sandEnds = append(sandEnds, os.NewFile(uintptr(fds[0]), "sandbox IO FD")) - goferEnd := os.NewFile(uintptr(fds[1]), "gofer io fd") + goferEnd := os.NewFile(uintptr(fds[1]), "gofer IO FD") defer goferEnd.Close() goferEnds = append(goferEnds, goferEnd) @@ -710,7 +733,7 @@ func (c *Container) createGoferProcess(spec *specs.Spec, conf *boot.Config, bund if err := specutils.StartInNS(cmd, nss); err != nil { return nil, err } - log.Infof("Gofer started, pid: %d", cmd.Process.Pid) + log.Infof("Gofer started, PID: %d", cmd.Process.Pid) c.GoferPid = cmd.Process.Pid return sandEnds, nil } diff --git a/runsc/container/container_test.go b/runsc/container/container_test.go index aebfb2878..84b59ffd8 100644 --- a/runsc/container/container_test.go +++ b/runsc/container/container_test.go @@ -31,6 +31,7 @@ import ( "time" "github.com/cenkalti/backoff" + "github.com/kr/pty" specs "github.com/opencontainers/runtime-spec/specs-go" "golang.org/x/sys/unix" "gvisor.googlesource.com/gvisor/pkg/abi/linux" @@ -38,6 +39,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/control" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/unet" + "gvisor.googlesource.com/gvisor/pkg/urpc" "gvisor.googlesource.com/gvisor/runsc/boot" "gvisor.googlesource.com/gvisor/runsc/test/testutil" ) @@ -1577,6 +1579,121 @@ func TestRootNotMount(t *testing.T) { } } +func TestJobControlSignalExec(t *testing.T) { + spec := testutil.NewSpecWithArgs("/bin/sleep", "10000") + conf := testutil.TestConfig() + + rootDir, bundleDir, err := testutil.SetupContainer(spec, conf) + if err != nil { + t.Fatalf("error setting up container: %v", err) + } + defer os.RemoveAll(rootDir) + defer os.RemoveAll(bundleDir) + + // Create and start the container. + c, err := Create(testutil.UniqueContainerID(), spec, conf, bundleDir, "", "") + if err != nil { + t.Fatalf("error creating container: %v", err) + } + defer c.Destroy() + if err := c.Start(conf); err != nil { + t.Fatalf("error starting container: %v", err) + } + + // Create a pty master/slave. The slave will be passed to the exec + // process. + ptyMaster, ptySlave, err := pty.Open() + if err != nil { + t.Fatalf("error opening pty: %v", err) + } + defer ptyMaster.Close() + defer ptySlave.Close() + + // Exec bash and attach a terminal. + args := &control.ExecArgs{ + Filename: "/bin/bash", + // Don't let bash execute from profile or rc files, otherwise + // our PID counts get messed up. + Argv: []string{"/bin/bash", "--noprofile", "--norc"}, + // Pass the pty slave as FD 0, 1, and 2. + FilePayload: urpc.FilePayload{ + Files: []*os.File{ptySlave, ptySlave, ptySlave}, + }, + StdioIsPty: true, + } + + pid, err := c.Execute(args) + if err != nil { + t.Fatalf("error executing: %v", err) + } + if pid != 2 { + t.Fatalf("exec got pid %d, wanted %d", pid, 2) + } + + // Make sure all the processes are running. + expectedPL := []*control.Process{ + // Root container process. + {PID: 1, Cmd: "sleep"}, + // Bash from exec process. + {PID: 2, Cmd: "bash"}, + } + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Execute sleep. + ptyMaster.Write([]byte("sleep 100\n")) + + // Wait for it to start. Sleep's PPID is bash's PID. + expectedPL = append(expectedPL, &control.Process{PID: 3, PPID: 2, Cmd: "sleep"}) + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Send a SIGTERM to the foreground process for the exec PID. Note that + // although we pass in the PID of "bash", it should actually terminate + // "sleep", since that is the foreground process. + if err := c.Sandbox.SignalProcess(c.ID, pid, syscall.SIGTERM, true /* fgProcess */); err != nil { + t.Fatalf("error signaling container: %v", err) + } + + // Sleep process should be gone. + expectedPL = expectedPL[:len(expectedPL)-1] + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Sleep is dead, but it may take more time for bash to notice and + // change the foreground process back to itself. We know it is done + // when bash writes "Terminated" to the pty. + if err := testutil.WaitUntilRead(ptyMaster, "Terminated", nil, 5*time.Second); err != nil { + t.Fatalf("bash did not take over pty: %v", err) + } + + // Send a SIGKILL to the foreground process again. This time "bash" + // should be killed. We use SIGKILL instead of SIGTERM or SIGINT + // because bash ignores those. + if err := c.Sandbox.SignalProcess(c.ID, pid, syscall.SIGKILL, true /* fgProcess */); err != nil { + t.Fatalf("error signaling container: %v", err) + } + expectedPL = expectedPL[:1] + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Make sure the process indicates it was killed by a SIGKILL. + ws, err := c.WaitPID(pid, true) + if err != nil { + t.Errorf("waiting on container failed: %v", err) + } + if !ws.Signaled() { + t.Error("ws.Signaled() got false, want true") + } + if got, want := ws.Signal(), syscall.SIGKILL; got != want { + t.Errorf("ws.Signal() got %v, want %v", got, want) + } +} + // executeSync synchronously executes a new process. func (cont *Container) executeSync(args *control.ExecArgs) (syscall.WaitStatus, error) { pid, err := cont.Execute(args) diff --git a/runsc/container/multi_container_test.go b/runsc/container/multi_container_test.go index e5f7daf60..ab200b75c 100644 --- a/runsc/container/multi_container_test.go +++ b/runsc/container/multi_container_test.go @@ -477,11 +477,12 @@ func TestMultiContainerDestroy(t *testing.T) { } func TestMultiContainerProcesses(t *testing.T) { - // Note: use 'while true' to keep 'sh' process around. Otherwise, shell will - // just execve into 'sleep' and both containers will look the same. + // Note: use curly braces to keep 'sh' process around. Otherwise, shell + // will just execve into 'sleep' and both containers will look the + // same. specs, ids := createSpecs( []string{"sleep", "100"}, - []string{"sh", "-c", "while true; do sleep 100; done"}) + []string{"sh", "-c", "{ sleep 100; }"}) conf := testutil.TestConfig() containers, cleanup, err := startContainers(conf, specs, ids) if err != nil { diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 4111b1a60..e4853af69 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -80,7 +80,7 @@ func Create(id string, spec *specs.Spec, conf *boot.Config, bundleDir, consoleSo // StartRoot starts running the root container process inside the sandbox. func (s *Sandbox) StartRoot(spec *specs.Spec, conf *boot.Config) error { - log.Debugf("Start root sandbox %q, pid: %d", s.ID, s.Pid) + log.Debugf("Start root sandbox %q, PID: %d", s.ID, s.Pid) conn, err := s.sandboxConnect() if err != nil { return err @@ -107,7 +107,7 @@ func (s *Sandbox) Start(spec *specs.Spec, conf *boot.Config, cid string, goferFi defer f.Close() } - log.Debugf("Start non-root container sandbox %q, pid: %d", s.ID, s.Pid) + log.Debugf("Start non-root container sandbox %q, PID: %d", s.ID, s.Pid) sandboxConn, err := s.sandboxConnect() if err != nil { return fmt.Errorf("couldn't connect to sandbox: %v", err) @@ -147,7 +147,7 @@ func (s *Sandbox) Restore(cid string, spec *specs.Spec, conf *boot.Config, f str SandboxID: s.ID, } - // If the platform needs a device fd we must pass it in. + // If the platform needs a device FD we must pass it in. if deviceFile, err := deviceFileForPlatform(conf.Platform); err != nil { return err } else if deviceFile != nil { @@ -192,7 +192,7 @@ func (s *Sandbox) Processes(cid string) ([]*control.Process, error) { return pl, nil } -// Execute runs the specified command in the container. It returns the pid of +// Execute runs the specified command in the container. It returns the PID of // the newly created process. func (s *Sandbox) Execute(args *control.ExecArgs) (int32, error) { log.Debugf("Executing new process in container %q in sandbox %q", args.ContainerID, s.ID) @@ -239,7 +239,7 @@ func (s *Sandbox) sandboxConnect() (*urpc.Client, error) { } func (s *Sandbox) connError(err error) error { - return fmt.Errorf("error connecting to control server at pid %d: %v", s.Pid, err) + return fmt.Errorf("error connecting to control server at PID %d: %v", s.Pid, err) } // createSandboxProcess starts the sandbox as a subprocess by running the "boot" @@ -322,7 +322,7 @@ func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bund nextFD++ } - // If the platform needs a device fd we must pass it in. + // If the platform needs a device FD we must pass it in. if deviceFile, err := deviceFileForPlatform(conf.Platform); err != nil { return err } else if deviceFile != nil { @@ -338,7 +338,7 @@ func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bund cmd.Stderr = os.Stderr // If the console control socket file is provided, then create a new - // pty master/slave pair and set the tty on the sandbox process. + // pty master/slave pair and set the TTY on the sandbox process. if consoleEnabled { // console.NewWithSocket will send the master on the socket, // and return the slave. @@ -461,7 +461,7 @@ func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bund // Add container as the last argument. cmd.Args = append(cmd.Args, s.ID) - // Log the fds we are donating to the sandbox process. + // Log the FDs we are donating to the sandbox process. for i, f := range cmd.ExtraFiles { log.Debugf("Donating FD %d: %q", i+3, f.Name()) } @@ -472,7 +472,7 @@ func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bund return err } s.Pid = cmd.Process.Pid - log.Infof("Sandbox started, pid: %d", s.Pid) + log.Infof("Sandbox started, PID: %d", s.Pid) return nil } @@ -572,9 +572,10 @@ func (s *Sandbox) destroy() error { return nil } -// Signal sends the signal to a container in the sandbox. If all is true and -// signal is SIGKILL, then waits for all processes to exit before returning. -func (s *Sandbox) Signal(cid string, sig syscall.Signal, all bool) error { +// SignalContainer sends the signal to a container in the sandbox. If all is +// true and signal is SIGKILL, then waits for all processes to exit before +// returning. +func (s *Sandbox) SignalContainer(cid string, sig syscall.Signal, all bool) error { log.Debugf("Signal sandbox %q", s.ID) conn, err := s.sandboxConnect() if err != nil { @@ -593,6 +594,30 @@ func (s *Sandbox) Signal(cid string, sig syscall.Signal, all bool) error { return nil } +// SignalProcess sends the signal to a particular process in the container. If +// fgProcess is true, then the signal is sent to the foreground process group +// in the same session that PID belongs to. This is only valid if the process +// is attached to a host TTY. +func (s *Sandbox) SignalProcess(cid string, pid int32, sig syscall.Signal, fgProcess bool) error { + log.Debugf("Signal sandbox %q", s.ID) + conn, err := s.sandboxConnect() + if err != nil { + return err + } + defer conn.Close() + + args := boot.SignalProcessArgs{ + CID: cid, + Signo: int32(sig), + PID: pid, + SendToForegroundProcess: fgProcess, + } + if err := conn.Call(boot.ContainerSignalProcess, &args, nil); err != nil { + return fmt.Errorf("err signaling container %q PID %d: %v", cid, pid, err) + } + return nil +} + // Checkpoint sends the checkpoint call for a container in the sandbox. // The statefile will be written to f. func (s *Sandbox) Checkpoint(cid string, f *os.File) error { diff --git a/runsc/test/integration/exec_test.go b/runsc/test/integration/exec_test.go index 910c36597..ddd088223 100644 --- a/runsc/test/integration/exec_test.go +++ b/runsc/test/integration/exec_test.go @@ -27,6 +27,7 @@ package integration import ( + "syscall" "testing" "time" @@ -60,3 +61,57 @@ func TestExecCapabilities(t *testing.T) { t.Errorf("wrong capabilities, got: %q, want: %q", got, want) } } + +func TestExecJobControl(t *testing.T) { + if err := testutil.Pull("alpine"); err != nil { + t.Fatalf("docker pull failed: %v", err) + } + d := testutil.MakeDocker("exec-test") + + // Start the container. + if _, err := d.Run("alpine", "sleep", "1000"); err != nil { + t.Fatalf("docker run failed: %v", err) + } + defer d.CleanUp() + + // Exec 'sh' with an attached pty. + cmd, ptmx, err := d.ExecWithTerminal("sh") + if err != nil { + t.Fatalf("docker exec failed: %v", err) + } + defer ptmx.Close() + + // Call "sleep 100" in the shell. + if _, err := ptmx.Write([]byte("sleep 100\n")); err != nil { + t.Fatalf("error writing to pty: %v", err) + } + + // Give shell a few seconds to start executing the sleep. + time.Sleep(2 * time.Second) + + // Send a ^C to the pty, which should kill sleep, but not the shell. + // \x03 is ASCII "end of text", which is the same as ^C. + if _, err := ptmx.Write([]byte{'\x03'}); err != nil { + t.Fatalf("error writing to pty: %v", err) + } + + // The shell should still be alive at this point. Sleep should have + // exited with code 2+128=130. We'll exit with 10 plus that number, so + // that we can be sure that the shell did not get signalled. + if _, err := ptmx.Write([]byte("exit $(expr $? + 10)\n")); err != nil { + t.Fatalf("error writing to pty: %v", err) + } + + // Exec process should exit with code 10+130=140. + ps, err := cmd.Process.Wait() + if err != nil { + t.Fatalf("error waiting for exec process: %v", err) + } + ws := ps.Sys().(syscall.WaitStatus) + if !ws.Exited() { + t.Errorf("ws.Exited got false, want true") + } + if got, want := ws.ExitStatus(), 140; got != want { + t.Errorf("ws.ExitedStatus got %d, want %d", got, want) + } +} diff --git a/runsc/test/testutil/BUILD b/runsc/test/testutil/BUILD index ca91e07ff..da2535bfa 100644 --- a/runsc/test/testutil/BUILD +++ b/runsc/test/testutil/BUILD @@ -17,6 +17,7 @@ go_library( "//runsc/boot", "//runsc/specutils", "@com_github_cenkalti_backoff//:go_default_library", + "@com_github_kr_pty//:go_default_library", "@com_github_opencontainers_runtime-spec//specs-go:go_default_library", "@com_github_syndtr_gocapability//capability:go_default_library", ], diff --git a/runsc/test/testutil/docker.go b/runsc/test/testutil/docker.go index 7f5909987..55ca353b8 100644 --- a/runsc/test/testutil/docker.go +++ b/runsc/test/testutil/docker.go @@ -26,6 +26,8 @@ import ( "strconv" "strings" "time" + + "github.com/kr/pty" ) func init() { @@ -131,6 +133,17 @@ func do(args ...string) (string, error) { return string(out), nil } +// doWithPty executes docker command with stdio attached to a pty. +func doWithPty(args ...string) (*exec.Cmd, *os.File, error) { + fmt.Printf("Running with pty: docker %s\n", args) + cmd := exec.Command("docker", args...) + ptmx, err := pty.Start(cmd) + if err != nil { + return nil, nil, fmt.Errorf("error executing docker %s with a pty: %v", args, err) + } + return cmd, ptmx, nil +} + // Pull pulls a docker image. This is used in tests to isolate the // time to pull the image off the network from the time to actually // start the container, to avoid timeouts over slow networks. @@ -197,6 +210,14 @@ func (d *Docker) Exec(args ...string) (string, error) { return do(a...) } +// ExecWithTerminal calls 'docker exec -it' with the arguments provided and +// attaches a pty to stdio. +func (d *Docker) ExecWithTerminal(args ...string) (*exec.Cmd, *os.File, error) { + a := []string{"exec", "-it", d.Name} + a = append(a, args...) + return doWithPty(a...) +} + // Pause calls 'docker pause'. func (d *Docker) Pause() error { if _, err := do("pause", d.Name); err != nil { diff --git a/runsc/test/testutil/testutil.go b/runsc/test/testutil/testutil.go index 07d66e469..cdc7f78c3 100644 --- a/runsc/test/testutil/testutil.go +++ b/runsc/test/testutil/testutil.go @@ -16,6 +16,7 @@ package testutil import ( + "bufio" "context" "encoding/json" "fmt" @@ -27,6 +28,8 @@ import ( "os/signal" "path/filepath" "runtime" + "strings" + "sync/atomic" "syscall" "time" @@ -315,3 +318,36 @@ func RetryEintr(f func() (uintptr, uintptr, error)) (uintptr, uintptr, error) { } } } + +// WaitUntilRead reads from the given reader until the wanted string is found +// or until timeout. +func WaitUntilRead(r io.Reader, want string, split bufio.SplitFunc, timeout time.Duration) error { + sc := bufio.NewScanner(r) + if split != nil { + sc.Split(split) + } + // done must be accessed atomically. A value greater than 0 indicates + // that the read loop can exit. + var done uint32 + doneCh := make(chan struct{}) + go func() { + for sc.Scan() { + t := sc.Text() + if strings.Contains(t, want) { + atomic.StoreUint32(&done, 1) + close(doneCh) + break + } + if atomic.LoadUint32(&done) > 0 { + break + } + } + }() + select { + case <-time.After(timeout): + atomic.StoreUint32(&done, 1) + return fmt.Errorf("timeout waiting to read %q", want) + case <-doneCh: + return nil + } +} -- cgit v1.2.3 From 4fef31f96c289d5e58c3c2997ee38fcb22c0378f Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Wed, 3 Oct 2018 17:02:05 -0700 Subject: Add S/R support for FIOASYNC PiperOrigin-RevId: 215655197 Change-Id: I668b1bc7c29daaf2999f8f759138bcbb09c4de6f --- pkg/sentry/fs/file.go | 10 +++++++++- pkg/sentry/fs/file_state.go | 11 +++++++++++ pkg/sentry/kernel/fasync/fasync.go | 4 +++- pkg/waiter/waiter.go | 2 ++ 4 files changed, 25 insertions(+), 2 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 904827a3e..36794d378 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -85,6 +85,9 @@ type File struct { // async handles O_ASYNC notifications. async FileAsync + // saving indicates that this file is in the process of being saved. + saving bool `state:"nosave"` + // mu is dual-purpose: first, to make read(2) and write(2) thread-safe // in conformity with POSIX, and second, to cancel operations before they // begin in response to interruptions (i.e. signals). @@ -127,10 +130,15 @@ func (f *File) DecRef() { // Release a reference on the Dirent. f.Dirent.DecRef() + // Only unregister if we are currently registered. There is nothing + // to register if f.async is nil (this happens when async mode is + // enabled without setting an owner). Also, we unregister during + // save. f.flagsMu.Lock() - if f.flags.Async && f.async != nil { + if !f.saving && f.flags.Async && f.async != nil { f.async.Unregister(f) } + f.async = nil f.flagsMu.Unlock() }) } diff --git a/pkg/sentry/fs/file_state.go b/pkg/sentry/fs/file_state.go index 3384737ab..f848d1b79 100644 --- a/pkg/sentry/fs/file_state.go +++ b/pkg/sentry/fs/file_state.go @@ -14,7 +14,18 @@ package fs +// beforeSave is invoked by stateify. +func (f *File) beforeSave() { + f.saving = true + if f.flags.Async && f.async != nil { + f.async.Unregister(f) + } +} + // afterLoad is invoked by stateify. func (f *File) afterLoad() { f.mu.Init() + if f.flags.Async && f.async != nil { + f.async.Register(f) + } } diff --git a/pkg/sentry/kernel/fasync/fasync.go b/pkg/sentry/kernel/fasync/fasync.go index 69c7970fa..7d01abe90 100644 --- a/pkg/sentry/kernel/fasync/fasync.go +++ b/pkg/sentry/kernel/fasync/fasync.go @@ -32,8 +32,10 @@ func New() fs.FileAsync { } // FileAsync sends signals when the registered file is ready for IO. +// +// +stateify savable type FileAsync struct { - mu sync.Mutex + mu sync.Mutex `state:"nosave"` e waiter.Entry requester *auth.Credentials diff --git a/pkg/waiter/waiter.go b/pkg/waiter/waiter.go index 9825880ca..832b6a5a9 100644 --- a/pkg/waiter/waiter.go +++ b/pkg/waiter/waiter.go @@ -113,6 +113,8 @@ type EntryCallback interface { // Entry represents a waiter that can be add to the a wait queue. It can // only be in one queue at a time, and is added "intrusively" to the queue with // no extra memory allocations. +// +// +stateify savable type Entry struct { // Context stores any state the waiter may wish to store in the entry // itself, which may be used at wake up time. -- cgit v1.2.3 From beac59b37a8b0ea834904870e5c236d2627947a2 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Wed, 3 Oct 2018 20:21:25 -0700 Subject: Fix panic if FIOASYNC callback is registered and triggered without target PiperOrigin-RevId: 215674589 Change-Id: I4f8871b64c570dc6da448d2fe351cec8a406efeb --- pkg/sentry/kernel/fasync/fasync.go | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/fasync/fasync.go b/pkg/sentry/kernel/fasync/fasync.go index 7d01abe90..f77339cae 100644 --- a/pkg/sentry/kernel/fasync/fasync.go +++ b/pkg/sentry/kernel/fasync/fasync.go @@ -60,6 +60,11 @@ func (a *FileAsync) Callback(e *waiter.Entry) { if tg != nil { t = tg.Leader() } + if t == nil { + // No recipient has been registered. + a.mu.Unlock() + return + } c := t.Credentials() // Logic from sigio_perm in fs/fcntl.c. if a.requester.EffectiveKUID == 0 || -- cgit v1.2.3 From e9e8be661328661b5527f1643727b9a13bbeab48 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Mon, 8 Oct 2018 10:19:27 -0700 Subject: Implement shared futexes. - Shared futex objects on shared mappings are represented by Mappable + offset, analogous to Linux's use of inode + offset. Add type futex.Key, and change the futex.Manager bucket API to use futex.Keys instead of addresses. - Extend the futex.Checker interface to be able to return Keys for memory mappings. It returns Keys rather than just mappings because whether the address or the target of the mapping is used in the Key depends on whether the mapping is MAP_SHARED or MAP_PRIVATE; this matters because using mapping target for a futex on a MAP_PRIVATE mapping causes it to stop working across COW-breaking. - futex.Manager.WaitComplete depends on atomic updates to futex.Waiter.addr to determine when it has locked the right bucket, which is much less straightforward for struct futex.Waiter.key. Switch to an atomically-accessed futex.Waiter.bucket pointer. - futex.Manager.Wake now needs to take a futex.Checker to resolve addresses for shared futexes. CLONE_CHILD_CLEARTID requires the exit path to perform a shared futex wakeup (Linux: kernel/fork.c:mm_release() => sys_futex(tsk->clear_child_tid, FUTEX_WAKE, ...)). This is a problem because futexChecker is in the syscalls/linux package. Move it to kernel. PiperOrigin-RevId: 216207039 Change-Id: I708d68e2d1f47e526d9afd95e7fed410c84afccf --- pkg/sentry/kernel/BUILD | 1 + pkg/sentry/kernel/futex/BUILD | 17 +- pkg/sentry/kernel/futex/futex.go | 382 ++++++++++------ pkg/sentry/kernel/futex/futex_test.go | 765 +++++++++++++++++---------------- pkg/sentry/kernel/kernel.go | 7 + pkg/sentry/kernel/task_clone.go | 2 +- pkg/sentry/kernel/task_context.go | 15 +- pkg/sentry/kernel/task_exit.go | 2 +- pkg/sentry/kernel/task_futex.go | 148 +++++++ pkg/sentry/mm/BUILD | 1 + pkg/sentry/mm/syscalls.go | 35 ++ pkg/sentry/syscalls/linux/sys_futex.go | 135 +----- 12 files changed, 880 insertions(+), 630 deletions(-) create mode 100644 pkg/sentry/kernel/task_futex.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 7eb2bffeb..31ad96612 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -95,6 +95,7 @@ go_library( "task_context.go", "task_exec.go", "task_exit.go", + "task_futex.go", "task_identity.go", "task_list.go", "task_log.go", diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index 0ff5b0a95..e13fcb5ff 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -3,6 +3,17 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +go_template_instance( + name = "atomicptr_bucket", + out = "atomicptr_bucket.go", + package = "futex", + suffix = "Bucket", + template = "//pkg/sync:generic_atomicptr", + types = { + "Value": "bucket", + }, +) + go_template_instance( name = "waiter_list", out = "waiter_list.go", @@ -18,12 +29,16 @@ go_template_instance( go_library( name = "futex", srcs = [ + "atomicptr_bucket.go", "futex.go", "waiter_list.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex", visibility = ["//pkg/sentry:internal"], - deps = ["//pkg/syserror"], + deps = [ + "//pkg/sentry/memmap", + "//pkg/syserror", + ], ) go_test( diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index 4a1f2a0ef..54b1982a0 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -19,11 +19,78 @@ package futex import ( "sync" - "sync/atomic" + "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/syserror" ) +// KeyKind indicates the type of a Key. +type KeyKind int + +const ( + // KindPrivate indicates a private futex (a futex syscall with the + // FUTEX_PRIVATE_FLAG set). + KindPrivate KeyKind = iota + + // KindSharedPrivate indicates a shared futex on a private memory mapping. + // Although KindPrivate and KindSharedPrivate futexes both use memory + // addresses to identify futexes, they do not interoperate (in Linux, the + // two are distinguished by the FUT_OFF_MMSHARED flag, which is used in key + // comparison). + KindSharedPrivate + + // KindSharedMappable indicates a shared futex on a memory mapping other + // than a private anonymous memory mapping. + KindSharedMappable +) + +// Key represents something that a futex waiter may wait on. +type Key struct { + // Kind is the type of the Key. + Kind KeyKind + + // Mappable is the memory-mapped object that is represented by the Key. + // Mappable is always nil if Kind is not KindSharedMappable, and may be nil + // even if it is. + Mappable memmap.Mappable + + // MappingIdentity is the MappingIdentity associated with Mappable. + // MappingIdentity is always nil is Mappable is nil, and may be nil even if + // it isn't. + MappingIdentity memmap.MappingIdentity + + // If Kind is KindPrivate or KindSharedPrivate, Offset is the represented + // memory address. Otherwise, Offset is the represented offset into + // Mappable. + Offset uint64 +} + +func (k *Key) release() { + if k.MappingIdentity != nil { + k.MappingIdentity.DecRef() + } + k.Mappable = nil + k.MappingIdentity = nil +} + +func (k *Key) clone() Key { + if k.MappingIdentity != nil { + k.MappingIdentity.IncRef() + } + return *k +} + +// Preconditions: k.Kind == KindPrivate or KindSharedPrivate. +func (k *Key) addr() uintptr { + return uintptr(k.Offset) +} + +// matches returns true if a wakeup on k2 should wake a waiter waiting on k. +func (k *Key) matches(k2 *Key) bool { + // k.MappingIdentity is ignored; it's only used for reference counting. + return k.Kind == k2.Kind && k.Mappable == k2.Mappable && k.Offset == k2.Offset +} + // Checker abstracts memory accesses. This is useful because the "addresses" // used in this package may not be real addresses (they could be indices of an // array, for example), or they could be mapped via some special mechanism. @@ -41,6 +108,14 @@ type Checker interface { // Note that op is an opaque operation whose behaviour is defined // outside of the futex manager. Op(addr uintptr, op uint32) (bool, error) + + // GetSharedKey returns a Key with kind KindSharedPrivate or + // KindSharedMappable corresponding to the memory mapped at address addr. + // + // If GetSharedKey returns a Key with a non-nil MappingIdentity, a + // reference is held on the MappingIdentity, which must be dropped by the + // caller when the Key is no longer in use. + GetSharedKey(addr uintptr) (Key, error) } // Waiter is the struct which gets enqueued into buckets for wake up routines @@ -53,11 +128,11 @@ type Waiter struct { // synchronization applies). // // - A Waiter is enqueued in a bucket by calling WaitPrepare(). After this, - // waiterEntry, complete, and addr are protected by the bucket.mu ("bucket - // lock") of the containing bucket, and bitmask is immutable. complete and - // addr are additionally mutated using atomic memory operations, ensuring - // that they can be read using atomic memory operations without holding the - // bucket lock. + // waiterEntry, bucket, and key are protected by the bucket.mu ("bucket + // lock") of the containing bucket, and bitmask is immutable. Note that + // since bucket is mutated using atomic memory operations, bucket.Load() + // may be called without holding the bucket lock, although it may change + // racily. See WaitComplete(). // // - A Waiter is only guaranteed to be no longer queued after calling // WaitComplete(). @@ -65,15 +140,15 @@ type Waiter struct { // waiterEntry links Waiter into bucket.waiters. waiterEntry - // complete is 1 if the Waiter was removed from its bucket by a wakeup and - // 0 otherwise. - complete int32 + // bucket is the bucket this waiter is queued in. If bucket is nil, the + // waiter is not waiting and is not in any bucket. + bucket AtomicPtrBucket // C is sent to when the Waiter is woken. C chan struct{} - // addr is the address being waited on. - addr uintptr + // key is what this waiter is waiting on. + key Key // The bitmask we're waiting on. // This is used the case of a FUTEX_WAKE_BITSET. @@ -87,7 +162,14 @@ func NewWaiter() *Waiter { } } +// woken returns true if w has been woken since the last call to WaitPrepare. +func (w *Waiter) woken() bool { + return len(w.C) != 0 +} + // bucket holds a list of waiters for a given address hash. +// +// +stateify savable type bucket struct { // mu protects waiters and contained Waiter state. See comment in Waiter. mu sync.Mutex `state:"nosave"` @@ -99,10 +181,10 @@ type bucket struct { // bucket and returns the number of waiters woken. // // Preconditions: b.mu must be locked. -func (b *bucket) wakeLocked(addr uintptr, bitmask uint32, n int) int { +func (b *bucket) wakeLocked(key *Key, bitmask uint32, n int) int { done := 0 for w := b.waiters.Front(); done < n && w != nil; { - if w.addr != addr || w.bitmask&bitmask == 0 { + if !w.key.matches(key) || w.bitmask&bitmask == 0 { // Not matching. w = w.Next() continue @@ -114,15 +196,15 @@ func (b *bucket) wakeLocked(addr uintptr, bitmask uint32, n int) int { b.waiters.Remove(woke) woke.C <- struct{}{} - // NOTE: The above channel write establishes a write barrier - // according to the memory model, so nothing may be ordered - // around it. Since we've dequeued w and will never touch it - // again, we can safely store 1 to w.complete here and allow - // the WaitComplete() to short-circuit grabbing the bucket - // lock. If they somehow miss the w.complete, we are still - // holding the lock, so we can know that they won't dequeue w, - // assume it's free and have the below operation afterwards. - atomic.StoreInt32(&woke.complete, 1) + // NOTE: The above channel write establishes a write barrier according + // to the memory model, so nothing may be ordered around it. Since + // we've dequeued woke and will never touch it again, we can safely + // store nil to woke.bucket here and allow the WaitComplete() to + // short-circuit grabbing the bucket lock. If they somehow miss the + // store, we are still holding the lock, so we can know that they won't + // dequeue woke, assume it's free and have the below operation + // afterwards. + woke.bucket.Store(nil) done++ } return done @@ -132,10 +214,10 @@ func (b *bucket) wakeLocked(addr uintptr, bitmask uint32, n int) int { // bucket "to". // // Preconditions: b and to must be locked. -func (b *bucket) requeueLocked(to *bucket, addr, naddr uintptr, n int) int { +func (b *bucket) requeueLocked(to *bucket, key, nkey *Key, n int) int { done := 0 for w := b.waiters.Front(); done < n && w != nil; { - if w.addr != addr { + if !w.key.matches(key) { // Not matching. w = w.Next() continue @@ -144,8 +226,10 @@ func (b *bucket) requeueLocked(to *bucket, addr, naddr uintptr, n int) int { requeued := w w = w.Next() // Next iteration. b.waiters.Remove(requeued) - atomic.StoreUintptr(&requeued.addr, naddr) + requeued.key.release() + requeued.key = nkey.clone() to.waiters.PushBack(requeued) + requeued.bucket.Store(to) done++ } return done @@ -158,19 +242,22 @@ const ( bucketCountBits = 10 ) -func checkAddr(addr uintptr) error { +// getKey returns a Key representing address addr in c. +func getKey(c Checker, addr uintptr, private bool) (Key, error) { // Ensure the address is aligned. // It must be a DWORD boundary. if addr&0x3 != 0 { - return syserror.EINVAL + return Key{}, syserror.EINVAL } - - return nil + if private { + return Key{Kind: KindPrivate, Offset: uint64(addr)}, nil + } + return c.GetSharedKey(addr) } // bucketIndexForAddr returns the index into Manager.buckets for addr. func bucketIndexForAddr(addr uintptr) uintptr { - // - The bottom 2 bits of addr must be 0, per checkAddr. + // - The bottom 2 bits of addr must be 0, per getKey. // // - On amd64, the top 16 bits of addr (bits 48-63) must be equal to bit 47 // for a canonical address, and (on all existing platforms) bit 47 must be @@ -199,171 +286,216 @@ func bucketIndexForAddr(addr uintptr) uintptr { // // +stateify savable type Manager struct { - buckets [bucketCount]bucket `state:"zerovalue"` + // privateBuckets holds buckets for KindPrivate and KindSharedPrivate + // futexes. + privateBuckets [bucketCount]bucket `state:"zerovalue"` + + // sharedBucket is the bucket for KindSharedMappable futexes. sharedBucket + // may be shared by multiple Managers. The sharedBucket pointer is + // immutable. + sharedBucket *bucket } // NewManager returns an initialized futex manager. -// N.B. we use virtual address to tag futexes, so it only works for private -// (within a single process) futex. func NewManager() *Manager { - return &Manager{} + return &Manager{ + sharedBucket: &bucket{}, + } } -// lockBucket returns a locked bucket for the given addr. -// -// Preconditions: checkAddr(addr) == nil. -func (m *Manager) lockBucket(addr uintptr) *bucket { - b := &m.buckets[bucketIndexForAddr(addr)] +// Fork returns a new Manager. Shared futex clients using the returned Manager +// may interoperate with those using m. +func (m *Manager) Fork() *Manager { + return &Manager{ + sharedBucket: m.sharedBucket, + } +} + +// lockBucket returns a locked bucket for the given key. +func (m *Manager) lockBucket(k *Key) *bucket { + var b *bucket + if k.Kind == KindSharedMappable { + b = m.sharedBucket + } else { + b = &m.privateBuckets[bucketIndexForAddr(k.addr())] + } b.mu.Lock() return b } -// lockBuckets returns locked buckets for the given addrs. -// -// Preconditions: checkAddr(addr1) == checkAddr(addr2) == nil. -func (m *Manager) lockBuckets(addr1 uintptr, addr2 uintptr) (*bucket, *bucket) { - i1 := bucketIndexForAddr(addr1) - i2 := bucketIndexForAddr(addr2) - b1 := &m.buckets[i1] - b2 := &m.buckets[i2] - - // Ensure that buckets are locked in a consistent order (lowest index - // first) to avoid circular locking. - switch { - case i1 < i2: - b1.mu.Lock() - b2.mu.Lock() - case i2 < i1: - b2.mu.Lock() - b1.mu.Lock() - default: - b1.mu.Lock() +// lockBuckets returns locked buckets for the given keys. +func (m *Manager) lockBuckets(k1, k2 *Key) (*bucket, *bucket) { + // Buckets must be consistently ordered to avoid circular lock + // dependencies. We order buckets in m.privateBuckets by index (lowest + // index first), and all buckets in m.privateBuckets precede + // m.sharedBucket. + + // Handle the common case first: + if k1.Kind != KindSharedMappable && k2.Kind != KindSharedMappable { + i1 := bucketIndexForAddr(k1.addr()) + i2 := bucketIndexForAddr(k2.addr()) + b1 := &m.privateBuckets[i1] + b2 := &m.privateBuckets[i2] + switch { + case i1 < i2: + b1.mu.Lock() + b2.mu.Lock() + case i2 < i1: + b2.mu.Lock() + b1.mu.Lock() + default: + b1.mu.Lock() + } + return b1, b2 } + // At least one of b1 or b2 should be m.sharedBucket. + b1 := m.sharedBucket + b2 := m.sharedBucket + if k1.Kind != KindSharedMappable { + b1 = m.lockBucket(k1) + } else if k2.Kind != KindSharedMappable { + b2 = m.lockBucket(k2) + } + m.sharedBucket.mu.Lock() return b1, b2 } // Wake wakes up to n waiters matching the bitmask on the given addr. // The number of waiters woken is returned. -func (m *Manager) Wake(addr uintptr, bitmask uint32, n int) (int, error) { - if err := checkAddr(addr); err != nil { +func (m *Manager) Wake(c Checker, addr uintptr, private bool, bitmask uint32, n int) (int, error) { + // This function is very hot; avoid defer. + k, err := getKey(c, addr, private) + if err != nil { return 0, err } - b := m.lockBucket(addr) - // This function is very hot; avoid defer. - r := b.wakeLocked(addr, bitmask, n) + b := m.lockBucket(&k) + r := b.wakeLocked(&k, bitmask, n) + b.mu.Unlock() + k.release() return r, nil } -func (m *Manager) doRequeue(c Checker, addr uintptr, val uint32, naddr uintptr, nwake int, nreq int) (int, error) { - if err := checkAddr(addr); err != nil { +func (m *Manager) doRequeue(c Checker, addr, naddr uintptr, private bool, checkval bool, val uint32, nwake int, nreq int) (int, error) { + k1, err := getKey(c, addr, private) + if err != nil { return 0, err } - if err := checkAddr(naddr); err != nil { + defer k1.release() + k2, err := getKey(c, naddr, private) + if err != nil { return 0, err } + defer k2.release() - b1, b2 := m.lockBuckets(addr, naddr) + b1, b2 := m.lockBuckets(&k1, &k2) defer b1.mu.Unlock() if b2 != b1 { defer b2.mu.Unlock() } - // Check our value. - // This only applied for RequeueCmp(). - if c != nil { + if checkval { if err := c.Check(addr, val); err != nil { return 0, err } } // Wake the number required. - done := b1.wakeLocked(addr, ^uint32(0), nwake) + done := b1.wakeLocked(&k1, ^uint32(0), nwake) // Requeue the number required. - b1.requeueLocked(b2, addr, naddr, nreq) + b1.requeueLocked(b2, &k1, &k2, nreq) return done, nil } // Requeue wakes up to nwake waiters on the given addr, and unconditionally // requeues up to nreq waiters on naddr. -func (m *Manager) Requeue(addr uintptr, naddr uintptr, nwake int, nreq int) (int, error) { - return m.doRequeue(nil, addr, 0, naddr, nwake, nreq) +func (m *Manager) Requeue(c Checker, addr, naddr uintptr, private bool, nwake int, nreq int) (int, error) { + return m.doRequeue(c, addr, naddr, private, false, 0, nwake, nreq) } // RequeueCmp atomically checks that the addr contains val (via the Checker), // wakes up to nwake waiters on addr and then unconditionally requeues nreq // waiters on naddr. -func (m *Manager) RequeueCmp(c Checker, addr uintptr, val uint32, naddr uintptr, nwake int, nreq int) (int, error) { - return m.doRequeue(c, addr, val, naddr, nwake, nreq) +func (m *Manager) RequeueCmp(c Checker, addr, naddr uintptr, private bool, val uint32, nwake int, nreq int) (int, error) { + return m.doRequeue(c, addr, naddr, private, true, val, nwake, nreq) } // WakeOp atomically applies op to the memory address addr2, wakes up to nwake1 // waiters unconditionally from addr1, and, based on the original value at addr2 // and a comparison encoded in op, wakes up to nwake2 waiters from addr2. // It returns the total number of waiters woken. -func (m *Manager) WakeOp(c Checker, addr1 uintptr, addr2 uintptr, nwake1 int, nwake2 int, op uint32) (int, error) { - if err := checkAddr(addr1); err != nil { +func (m *Manager) WakeOp(c Checker, addr1, addr2 uintptr, private bool, nwake1 int, nwake2 int, op uint32) (int, error) { + k1, err := getKey(c, addr1, private) + if err != nil { return 0, err } - if err := checkAddr(addr2); err != nil { + defer k1.release() + k2, err := getKey(c, addr2, private) + if err != nil { return 0, err } + defer k2.release() - b1, b2 := m.lockBuckets(addr1, addr2) + b1, b2 := m.lockBuckets(&k1, &k2) + defer b1.mu.Unlock() + if b2 != b1 { + defer b2.mu.Unlock() + } done := 0 cond, err := c.Op(addr2, op) - if err == nil { - // Wake up up to nwake1 entries from the first bucket. - done = b1.wakeLocked(addr1, ^uint32(0), nwake1) - - // Wake up up to nwake2 entries from the second bucket if the - // operation yielded true. - if cond { - done += b2.wakeLocked(addr2, ^uint32(0), nwake2) - } + if err != nil { + return 0, err } - b1.mu.Unlock() - if b2 != b1 { - b2.mu.Unlock() + // Wake up up to nwake1 entries from the first bucket. + done = b1.wakeLocked(&k1, ^uint32(0), nwake1) + + // Wake up up to nwake2 entries from the second bucket if the + // operation yielded true. + if cond { + done += b2.wakeLocked(&k2, ^uint32(0), nwake2) } - return done, err + + return done, nil } // WaitPrepare atomically checks that addr contains val (via the Checker), then // enqueues w to be woken by a send to w.C. If WaitPrepare returns nil, the // Waiter must be subsequently removed by calling WaitComplete, whether or not // a wakeup is received on w.C. -func (m *Manager) WaitPrepare(w *Waiter, c Checker, addr uintptr, val uint32, bitmask uint32) error { - if err := checkAddr(addr); err != nil { +func (m *Manager) WaitPrepare(w *Waiter, c Checker, addr uintptr, private bool, val uint32, bitmask uint32) error { + k, err := getKey(c, addr, private) + if err != nil { return err } + // Ownership of k is transferred to w below. // Prepare the Waiter before taking the bucket lock. - w.complete = 0 select { case <-w.C: default: } - w.addr = addr + w.key = k w.bitmask = bitmask - b := m.lockBucket(addr) + b := m.lockBucket(&k) // This function is very hot; avoid defer. // Perform our atomic check. if err := c.Check(addr, val); err != nil { b.mu.Unlock() + w.key.release() return err } // Add the waiter to the bucket. b.waiters.PushBack(w) + w.bucket.Store(b) b.mu.Unlock() return nil @@ -372,36 +504,36 @@ func (m *Manager) WaitPrepare(w *Waiter, c Checker, addr uintptr, val uint32, bi // WaitComplete must be called when a Waiter previously added by WaitPrepare is // no longer eligible to be woken. func (m *Manager) WaitComplete(w *Waiter) { - // Can we short-circuit acquiring the lock? - // This is the happy path where a notification - // was received and we don't need to dequeue this - // waiter from any list (or take any locks). - if atomic.LoadInt32(&w.complete) != 0 { - return - } - - // Take the bucket lock. Note that without holding the bucket lock, the - // waiter is not guaranteed to stay in that bucket, so after we take the - // bucket lock, we must ensure that the bucket hasn't changed: if it - // happens to have changed, we release the old bucket lock and try again - // with the new bucket; if it hasn't changed, we know it won't change now - // because we hold the lock. - var b *bucket + // Remove w from the bucket it's in. for { - addr := atomic.LoadUintptr(&w.addr) - b = m.lockBucket(addr) - // We still have to use an atomic load here, because if w was racily - // requeued then w.addr is not protected by b.mu. - if addr == atomic.LoadUintptr(&w.addr) { + b := w.bucket.Load() + + // If b is nil, the waiter isn't in any bucket anymore. This can't be + // racy because the waiter can't be concurrently re-queued in another + // bucket. + if b == nil { break } - b.mu.Unlock() - } - // Remove waiter from the bucket. w.complete can only be stored with b.mu - // locked, so this load doesn't need to use sync/atomic. - if w.complete == 0 { + // Take the bucket lock. Note that without holding the bucket lock, the + // waiter is not guaranteed to stay in that bucket, so after we take + // the bucket lock, we must ensure that the bucket hasn't changed: if + // it happens to have changed, we release the old bucket lock and try + // again with the new bucket; if it hasn't changed, we know it won't + // change now because we hold the lock. + b.mu.Lock() + if b != w.bucket.Load() { + b.mu.Unlock() + continue + } + + // Remove w from b. b.waiters.Remove(w) + w.bucket.Store(nil) + b.mu.Unlock() + break } - b.mu.Unlock() + + // Release references held by the waiter. + w.key.release() } diff --git a/pkg/sentry/kernel/futex/futex_test.go b/pkg/sentry/kernel/futex/futex_test.go index 7b81358ec..726c26990 100644 --- a/pkg/sentry/kernel/futex/futex_test.go +++ b/pkg/sentry/kernel/futex/futex_test.go @@ -24,17 +24,13 @@ import ( "unsafe" ) -const ( - testMutexSize = 4 - testMutexLocked uint32 = 1 - testMutexUnlocked uint32 = 0 -) - // testData implements the Checker interface, and allows us to // treat the address passed for futex operations as an index in // a byte slice for testing simplicity. type testData []byte +const sizeofInt32 = 4 + func newTestData(size uint) testData { return make([]byte, size) } @@ -50,451 +46,478 @@ func (t testData) Op(addr uintptr, val uint32) (bool, error) { return val == 0, nil } -// testMutex ties together a testData slice, an address, and a -// futex manager in order to implement the sync.Locker interface. -// Beyond being used as a Locker, this is a simple mechanism for -// changing the underlying values for simpler tests. -type testMutex struct { - a uintptr - d testData - m *Manager -} - -func newTestMutex(addr uintptr, d testData, m *Manager) *testMutex { - return &testMutex{a: addr, d: d, m: m} +func (t testData) GetSharedKey(addr uintptr) (Key, error) { + return Key{ + Kind: KindSharedMappable, + Offset: uint64(addr), + }, nil } -// Lock acquires the testMutex. -// This may wait for it to be available via the futex manager. -func (t *testMutex) Lock() { - for { - // Attempt to grab the lock. - if atomic.CompareAndSwapUint32( - ((*uint32)(unsafe.Pointer(&t.d[t.a]))), - testMutexUnlocked, - testMutexLocked) { - // Lock held. - return - } - - // Wait for it to be "not locked". - w := NewWaiter() - err := t.m.WaitPrepare(w, t.d, t.a, testMutexLocked, ^uint32(0)) - if err == syscall.EAGAIN { - continue - } - if err != nil { - // Should never happen. - panic("WaitPrepare returned unexpected error: " + err.Error()) - } - <-w.C - t.m.WaitComplete(w) +func futexKind(private bool) string { + if private { + return "private" } + return "shared" } -// Unlock releases the testMutex. -// This will notify any waiters via the futex manager. -func (t *testMutex) Unlock() { - // Unlock. - atomic.StoreUint32(((*uint32)(unsafe.Pointer(&t.d[t.a]))), testMutexUnlocked) - - // Notify all waiters. - t.m.Wake(t.a, ^uint32(0), math.MaxInt32) +func newPreparedTestWaiter(t *testing.T, m *Manager, c Checker, addr uintptr, private bool, val uint32, bitmask uint32) *Waiter { + w := NewWaiter() + if err := m.WaitPrepare(w, c, addr, private, val, bitmask); err != nil { + t.Fatalf("WaitPrepare failed: %v", err) + } + return w } func TestFutexWake(t *testing.T) { - m := NewManager() - d := newTestData(testMutexSize) - - // Wait for it to be locked. - // (This won't trigger the wake in testMutex) - w := NewWaiter() - m.WaitPrepare(w, d, 0, testMutexUnlocked, ^uint32(0)) + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(sizeofInt32) + + // Start waiting for wakeup. + w := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(w) + + // Perform a wakeup. + if n, err := m.Wake(d, 0, private, ^uint32(0), 1); err != nil || n != 1 { + t.Errorf("Wake: got (%d, %v), wanted (1, nil)", n, err) + } - // Wake the single thread. - if _, err := m.Wake(0, ^uint32(0), 1); err != nil { - t.Error("wake error:", err) + // Expect the waiter to have been woken. + if !w.woken() { + t.Error("waiter not woken") + } + }) } - - <-w.C - m.WaitComplete(w) } func TestFutexWakeBitmask(t *testing.T) { - m := NewManager() - d := newTestData(testMutexSize) - - // Wait for it to be locked. - // (This won't trigger the wake in testMutex) - w := NewWaiter() - m.WaitPrepare(w, d, 0, testMutexUnlocked, 0x0000ffff) + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(sizeofInt32) + + // Start waiting for wakeup. + w := newPreparedTestWaiter(t, m, d, 0, private, 0, 0x0000ffff) + defer m.WaitComplete(w) + + // Perform a wakeup using the wrong bitmask. + if n, err := m.Wake(d, 0, private, 0xffff0000, 1); err != nil || n != 0 { + t.Errorf("Wake with non-matching bitmask: got (%d, %v), wanted (0, nil)", n, err) + } - // Wake the single thread, not using the bitmask. - if _, err := m.Wake(0, 0xffff0000, 1); err != nil { - t.Error("wake non-matching bitmask error:", err) - } + // Expect the waiter to still be waiting. + if w.woken() { + t.Error("waiter woken unexpectedly") + } - select { - case <-w.C: - t.Error("w is alive?") - default: - } + // Perform a wakeup using the right bitmask. + if n, err := m.Wake(d, 0, private, 0x00000001, 1); err != nil || n != 1 { + t.Errorf("Wake with matching bitmask: got (%d, %v), wanted (1, nil)", n, err) + } - // Now use a matching bitmask. - if _, err := m.Wake(0, 0x00000001, 1); err != nil { - t.Error("wake matching bitmask error:", err) + // Expect that the waiter was woken. + if !w.woken() { + t.Error("waiter not woken") + } + }) } - - <-w.C - m.WaitComplete(w) } func TestFutexWakeTwo(t *testing.T) { - m := NewManager() - d := newTestData(testMutexSize) + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(sizeofInt32) + + // Start three waiters waiting for wakeup. + var ws [3]*Waiter + for i := range ws { + ws[i] = newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(ws[i]) + } - // Wait for it to be locked. - // (This won't trigger the wake in testMutex) - w1 := NewWaiter() - w2 := NewWaiter() - w3 := NewWaiter() - m.WaitPrepare(w1, d, 0, testMutexUnlocked, ^uint32(0)) - m.WaitPrepare(w2, d, 0, testMutexUnlocked, ^uint32(0)) - m.WaitPrepare(w3, d, 0, testMutexUnlocked, ^uint32(0)) - - // Wake exactly two threads. - if _, err := m.Wake(0, ^uint32(0), 2); err != nil { - t.Error("wake error:", err) - } + // Perform two wakeups. + const wakeups = 2 + if n, err := m.Wake(d, 0, private, ^uint32(0), 2); err != nil || n != wakeups { + t.Errorf("Wake: got (%d, %v), wanted (%d, nil)", n, err, wakeups) + } - // Ensure exactly two are alive. - // We don't get guarantees about exactly which two, - // (although we expect them to be w1 and w2). - awake := 0 - for { - select { - case <-w1.C: - awake++ - case <-w2.C: - awake++ - case <-w3.C: - awake++ - default: - if awake != 2 { - t.Error("awake != 2?") - } - - // Success. - return - } + // Expect that exactly two waiters were woken. + // We don't get guarantees about exactly which two, + // (although we expect them to be w1 and w2). + awake := 0 + for i := range ws { + if ws[i].woken() { + awake++ + } + } + if awake != wakeups { + t.Errorf("got %d woken waiters, wanted %d", awake, wakeups) + } + }) } } func TestFutexWakeUnrelated(t *testing.T) { - m := NewManager() - d := newTestData(2 * testMutexSize) - - // Wait for it to be locked. - w1 := NewWaiter() - w2 := NewWaiter() - m.WaitPrepare(w1, d, 0*testMutexSize, testMutexUnlocked, ^uint32(0)) - m.WaitPrepare(w2, d, 1*testMutexSize, testMutexUnlocked, ^uint32(0)) - - // Wake only the second one. - if _, err := m.Wake(1*testMutexSize, ^uint32(0), 2); err != nil { - t.Error("wake error:", err) - } - - // Ensure only r2 is alive. - select { - case <-w1.C: - t.Error("w1 is alive?") - default: - } - <-w2.C -} - -// This function was shamelessly stolen from mutex_test.go. -func HammerMutex(l sync.Locker, loops int, cdone chan bool) { - for i := 0; i < loops; i++ { - l.Lock() - runtime.Gosched() - l.Unlock() - } - cdone <- true -} - -func TestFutexStress(t *testing.T) { - m := NewManager() - d := newTestData(testMutexSize) - tm := newTestMutex(0*testMutexSize, d, m) - c := make(chan bool) - - for i := 0; i < 10; i++ { - go HammerMutex(tm, 1000, c) - } + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(2 * sizeofInt32) + + // Start two waiters waiting for wakeup on different addresses. + w1 := newPreparedTestWaiter(t, m, d, 0*sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w1) + w2 := newPreparedTestWaiter(t, m, d, 1*sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w2) + + // Perform two wakeups on the second address. + if n, err := m.Wake(d, 1*sizeofInt32, private, ^uint32(0), 2); err != nil || n != 1 { + t.Errorf("Wake: got (%d, %v), wanted (1, nil)", n, err) + } - for i := 0; i < 10; i++ { - <-c + // Expect that only the second waiter was woken. + if w1.woken() { + t.Error("w1 woken unexpectedly") + } + if !w2.woken() { + t.Error("w2 not woken") + } + }) } } func TestWakeOpEmpty(t *testing.T) { - m := NewManager() - d := newTestData(8) - - n, err := m.WakeOp(d, 0, 4, 10, 10, 0) - if err != nil { - t.Fatalf("WakeOp failed: %v", err) - } - - if n != 0 { - t.Fatalf("Invalid number of wakes: want 0, got %d", n) + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(2 * sizeofInt32) + + // Perform wakeups with no waiters. + if n, err := m.WakeOp(d, 0, sizeofInt32, private, 10, 10, 0); err != nil || n != 0 { + t.Fatalf("WakeOp: got (%d, %v), wanted (0, nil)", n, err) + } + }) } } func TestWakeOpFirstNonEmpty(t *testing.T) { - m := NewManager() - d := newTestData(8) - - // Add two waiters on address 0. - w1 := NewWaiter() - if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w1) - - w2 := NewWaiter() - if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w2) - - // Wake up all waiters on address 0. - n, err := m.WakeOp(d, 0, 4, 10, 10, 0) - if err != nil { - t.Fatalf("WakeOp failed: %v", err) - } + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address 0. + w1 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(w1) + w2 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(w2) + + // Perform 10 wakeups on address 0. + if n, err := m.WakeOp(d, 0, sizeofInt32, private, 10, 0, 0); err != nil || n != 2 { + t.Errorf("WakeOp: got (%d, %v), wanted (2, nil)", n, err) + } - if n != 2 { - t.Fatalf("Invalid number of wakes: want 2, got %d", n) + // Expect that both waiters were woken. + if !w1.woken() { + t.Error("w1 not woken") + } + if !w2.woken() { + t.Error("w2 not woken") + } + }) } } func TestWakeOpSecondNonEmpty(t *testing.T) { - m := NewManager() - d := newTestData(8) - - // Add two waiters on address 4. - w1 := NewWaiter() - if err := m.WaitPrepare(w1, d, 4, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w1) - - w2 := NewWaiter() - if err := m.WaitPrepare(w2, d, 4, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w2) - - // Wake up all waiters on address 4. - n, err := m.WakeOp(d, 0, 4, 10, 10, 0) - if err != nil { - t.Fatalf("WakeOp failed: %v", err) - } + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address sizeofInt32. + w1 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w1) + w2 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w2) + + // Perform 10 wakeups on address sizeofInt32 (contingent on + // d.Op(0), which should succeed). + if n, err := m.WakeOp(d, 0, sizeofInt32, private, 0, 10, 0); err != nil || n != 2 { + t.Errorf("WakeOp: got (%d, %v), wanted (2, nil)", n, err) + } - if n != 2 { - t.Fatalf("Invalid number of wakes: want 2, got %d", n) + // Expect that both waiters were woken. + if !w1.woken() { + t.Error("w1 not woken") + } + if !w2.woken() { + t.Error("w2 not woken") + } + }) } } func TestWakeOpSecondNonEmptyFailingOp(t *testing.T) { - m := NewManager() - d := newTestData(8) - - // Add two waiters on address 4. - w1 := NewWaiter() - if err := m.WaitPrepare(w1, d, 4, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w1) - - w2 := NewWaiter() - if err := m.WaitPrepare(w2, d, 4, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w2) - - // Wake up all waiters on address 4. - n, err := m.WakeOp(d, 0, 4, 10, 10, 1) - if err != nil { - t.Fatalf("WakeOp failed: %v", err) - } + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address sizeofInt32. + w1 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w1) + w2 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w2) + + // Perform 10 wakeups on address sizeofInt32 (contingent on + // d.Op(1), which should fail). + if n, err := m.WakeOp(d, 0, sizeofInt32, private, 0, 10, 1); err != nil || n != 0 { + t.Errorf("WakeOp: got (%d, %v), wanted (0, nil)", n, err) + } - if n != 0 { - t.Fatalf("Invalid number of wakes: want 0, got %d", n) + // Expect that neither waiter was woken. + if w1.woken() { + t.Error("w1 woken unexpectedly") + } + if w2.woken() { + t.Error("w2 woken unexpectedly") + } + }) } } func TestWakeOpAllNonEmpty(t *testing.T) { - m := NewManager() - d := newTestData(8) + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address 0. + w1 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(w1) + w2 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(w2) + + // Add two waiters on address sizeofInt32. + w3 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w3) + w4 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w4) + + // Perform 10 wakeups on address 0 (unconditionally), and 10 + // wakeups on address sizeofInt32 (contingent on d.Op(0), which + // should succeed). + if n, err := m.WakeOp(d, 0, sizeofInt32, private, 10, 10, 0); err != nil || n != 4 { + t.Errorf("WakeOp: got (%d, %v), wanted (4, nil)", n, err) + } - // Add two waiters on address 0. - w1 := NewWaiter() - if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) + // Expect that all waiters were woken. + if !w1.woken() { + t.Error("w1 not woken") + } + if !w2.woken() { + t.Error("w2 not woken") + } + if !w3.woken() { + t.Error("w3 not woken") + } + if !w4.woken() { + t.Error("w4 not woken") + } + }) } - defer m.WaitComplete(w1) +} - w2 := NewWaiter() - if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w2) +func TestWakeOpAllNonEmptyFailingOp(t *testing.T) { + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add two waiters on address 0. + w1 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(w1) + w2 := newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(w2) + + // Add two waiters on address sizeofInt32. + w3 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w3) + w4 := newPreparedTestWaiter(t, m, d, sizeofInt32, private, 0, ^uint32(0)) + defer m.WaitComplete(w4) + + // Perform 10 wakeups on address 0 (unconditionally), and 10 + // wakeups on address sizeofInt32 (contingent on d.Op(1), which + // should fail). + if n, err := m.WakeOp(d, 0, sizeofInt32, private, 10, 10, 1); err != nil || n != 2 { + t.Errorf("WakeOp: got (%d, %v), wanted (2, nil)", n, err) + } - // Add two waiters on address 4. - w3 := NewWaiter() - if err := m.WaitPrepare(w3, d, 4, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) + // Expect that only the first two waiters were woken. + if !w1.woken() { + t.Error("w1 not woken") + } + if !w2.woken() { + t.Error("w2 not woken") + } + if w3.woken() { + t.Error("w3 woken unexpectedly") + } + if w4.woken() { + t.Error("w4 woken unexpectedly") + } + }) } - defer m.WaitComplete(w3) +} - w4 := NewWaiter() - if err := m.WaitPrepare(w4, d, 4, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w4) +func TestWakeOpSameAddress(t *testing.T) { + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add four waiters on address 0. + var ws [4]*Waiter + for i := range ws { + ws[i] = newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(ws[i]) + } - // Wake up all waiters on both addresses. - n, err := m.WakeOp(d, 0, 4, 10, 10, 0) - if err != nil { - t.Fatalf("WakeOp failed: %v", err) - } + // Perform 1 wakeup on address 0 (unconditionally), and 1 wakeup + // on address 0 (contingent on d.Op(0), which should succeed). + const wakeups = 2 + if n, err := m.WakeOp(d, 0, 0, private, 1, 1, 0); err != nil || n != wakeups { + t.Errorf("WakeOp: got (%d, %v), wanted (%d, nil)", n, err, wakeups) + } - if n != 4 { - t.Fatalf("Invalid number of wakes: want 4, got %d", n) + // Expect that exactly two waiters were woken. + awake := 0 + for i := range ws { + if ws[i].woken() { + awake++ + } + } + if awake != wakeups { + t.Errorf("got %d woken waiters, wanted %d", awake, wakeups) + } + }) } } -func TestWakeOpAllNonEmptyFailingOp(t *testing.T) { - m := NewManager() - d := newTestData(8) - - // Add two waiters on address 0. - w1 := NewWaiter() - if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w1) - - w2 := NewWaiter() - if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w2) +func TestWakeOpSameAddressFailingOp(t *testing.T) { + for _, private := range []bool{false, true} { + t.Run(futexKind(private), func(t *testing.T) { + m := NewManager() + d := newTestData(8) + + // Add four waiters on address 0. + var ws [4]*Waiter + for i := range ws { + ws[i] = newPreparedTestWaiter(t, m, d, 0, private, 0, ^uint32(0)) + defer m.WaitComplete(ws[i]) + } - // Add two waiters on address 4. - w3 := NewWaiter() - if err := m.WaitPrepare(w3, d, 4, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w3) + // Perform 1 wakeup on address 0 (unconditionally), and 1 wakeup + // on address 0 (contingent on d.Op(1), which should fail). + const wakeups = 1 + if n, err := m.WakeOp(d, 0, 0, private, 1, 1, 1); err != nil || n != wakeups { + t.Errorf("WakeOp: got (%d, %v), wanted (%d, nil)", n, err, wakeups) + } - w4 := NewWaiter() - if err := m.WaitPrepare(w4, d, 4, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) + // Expect that exactly one waiter was woken. + awake := 0 + for i := range ws { + if ws[i].woken() { + awake++ + } + } + if awake != wakeups { + t.Errorf("got %d woken waiters, wanted %d", awake, wakeups) + } + }) } - defer m.WaitComplete(w4) +} - // Wake up all waiters on both addresses. - n, err := m.WakeOp(d, 0, 4, 10, 10, 1) - if err != nil { - t.Fatalf("WakeOp failed: %v", err) - } +const ( + testMutexSize = sizeofInt32 + testMutexLocked uint32 = 1 + testMutexUnlocked uint32 = 0 +) - if n != 2 { - t.Fatalf("Invalid number of wakes: want 2, got %d", n) - } +// testMutex ties together a testData slice, an address, and a +// futex manager in order to implement the sync.Locker interface. +// Beyond being used as a Locker, this is a simple mechanism for +// changing the underlying values for simpler tests. +type testMutex struct { + a uintptr + d testData + m *Manager } -func TestWakeOpSameAddress(t *testing.T) { - m := NewManager() - d := newTestData(8) - - // Add four waiters on address 0. - w1 := NewWaiter() - if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w1) +func newTestMutex(addr uintptr, d testData, m *Manager) *testMutex { + return &testMutex{a: addr, d: d, m: m} +} - w2 := NewWaiter() - if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w2) +// Lock acquires the testMutex. +// This may wait for it to be available via the futex manager. +func (t *testMutex) Lock() { + for { + // Attempt to grab the lock. + if atomic.CompareAndSwapUint32( + (*uint32)(unsafe.Pointer(&t.d[t.a])), + testMutexUnlocked, + testMutexLocked) { + // Lock held. + return + } - w3 := NewWaiter() - if err := m.WaitPrepare(w3, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) + // Wait for it to be "not locked". + w := NewWaiter() + err := t.m.WaitPrepare(w, t.d, t.a, true, testMutexLocked, ^uint32(0)) + if err == syscall.EAGAIN { + continue + } + if err != nil { + // Should never happen. + panic("WaitPrepare returned unexpected error: " + err.Error()) + } + <-w.C + t.m.WaitComplete(w) } - defer m.WaitComplete(w3) +} - w4 := NewWaiter() - if err := m.WaitPrepare(w4, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w4) +// Unlock releases the testMutex. +// This will notify any waiters via the futex manager. +func (t *testMutex) Unlock() { + // Unlock. + atomic.StoreUint32((*uint32)(unsafe.Pointer(&t.d[t.a])), testMutexUnlocked) - // Use the same address, with one at most one waiter from each. - n, err := m.WakeOp(d, 0, 0, 1, 1, 0) - if err != nil { - t.Fatalf("WakeOp failed: %v", err) - } + // Notify all waiters. + t.m.Wake(t.d, t.a, true, ^uint32(0), math.MaxInt32) +} - if n != 2 { - t.Fatalf("Invalid number of wakes: want 2, got %d", n) +// This function was shamelessly stolen from mutex_test.go. +func HammerMutex(l sync.Locker, loops int, cdone chan bool) { + for i := 0; i < loops; i++ { + l.Lock() + runtime.Gosched() + l.Unlock() } + cdone <- true } -func TestWakeOpSameAddressFailingOp(t *testing.T) { +func TestMutexStress(t *testing.T) { m := NewManager() - d := newTestData(8) - - // Add four waiters on address 0. - w1 := NewWaiter() - if err := m.WaitPrepare(w1, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w1) - - w2 := NewWaiter() - if err := m.WaitPrepare(w2, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w2) - - w3 := NewWaiter() - if err := m.WaitPrepare(w3, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w3) - - w4 := NewWaiter() - if err := m.WaitPrepare(w4, d, 0, 0, ^uint32(0)); err != nil { - t.Fatalf("WaitPrepare failed: %v", err) - } - defer m.WaitComplete(w4) + d := newTestData(testMutexSize) + tm := newTestMutex(0*testMutexSize, d, m) + c := make(chan bool) - // Use the same address, with one at most one waiter from each. - n, err := m.WakeOp(d, 0, 0, 1, 1, 1) - if err != nil { - t.Fatalf("WakeOp failed: %v", err) + for i := 0; i < 10; i++ { + go HammerMutex(tm, 1000, c) } - if n != 1 { - t.Fatalf("Invalid number of wakes: want 1, got %d", n) + for i := 0; i < 10; i++ { + <-c } } diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 1ace0b501..238fd127b 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -49,6 +49,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/inet" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" @@ -108,6 +109,11 @@ type Kernel struct { // Kernel.CreateProcess can succeed. mounts *fs.MountNamespace + // futexes is the "root" futex.Manager, from which all others are forked. + // This is necessary to ensure that shared futexes are coherent across all + // tasks, including those created by CreateProcess. + futexes *futex.Manager + // globalInit is the thread group whose leader has ID 1 in the root PID // namespace. globalInit is stored separately so that it is accessible even // after all tasks in the thread group have exited, such that ID 1 is no @@ -254,6 +260,7 @@ func (k *Kernel) Init(args InitKernelArgs) error { k.vdso = args.Vdso k.realtimeClock = &timekeeperClock{tk: args.Timekeeper, c: sentrytime.Realtime} k.monotonicClock = &timekeeperClock{tk: args.Timekeeper, c: sentrytime.Monotonic} + k.futexes = futex.NewManager() k.netlinkPorts = port.New() return nil diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 130bd652b..7c469ec46 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -200,7 +200,7 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { ipcns = NewIPCNamespace(userns) } - tc, err := t.tc.Fork(t, !opts.NewAddressSpace) + tc, err := t.tc.Fork(t, t.k, !opts.NewAddressSpace) if err != nil { return 0, nil, err } diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index 9a59cbd33..d2df7e9d1 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -72,7 +72,7 @@ func (tc *TaskContext) release() { // TaskContext shares an address space with the original; otherwise, the copied // TaskContext has an independent address space that is initially a duplicate // of the original's. -func (tc *TaskContext) Fork(ctx context.Context, shareAddressSpace bool) (*TaskContext, error) { +func (tc *TaskContext) Fork(ctx context.Context, k *Kernel, shareAddressSpace bool) (*TaskContext, error) { newTC := &TaskContext{ Arch: tc.Arch.Fork(), st: tc.st, @@ -93,8 +93,7 @@ func (tc *TaskContext) Fork(ctx context.Context, shareAddressSpace bool) (*TaskC return nil, err } newTC.MemoryManager = newMM - // TODO: revisit when shmem is supported. - newTC.fu = futex.NewManager() + newTC.fu = k.futexes.Fork() } return newTC, nil } @@ -116,14 +115,6 @@ func (t *Task) MemoryManager() *mm.MemoryManager { return t.tc.MemoryManager } -// Futex returns t's futex manager. -// -// Preconditions: The caller must be running on the task goroutine, or t.mu -// must be locked. -func (t *Task) Futex() *futex.Manager { - return t.tc.fu -} - // SyscallTable returns t's syscall table. // // Preconditions: The caller must be running on the task goroutine, or t.mu @@ -175,7 +166,7 @@ func (k *Kernel) LoadTaskImage(ctx context.Context, mounts *fs.MountNamespace, r Name: name, Arch: ac, MemoryManager: m, - fu: futex.NewManager(), + fu: k.futexes.Fork(), st: st, }, nil } diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index a1b24e1c6..f5b45fb17 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -247,7 +247,7 @@ func (*runExitMain) execute(t *Task) taskRunState { t.tg.signalHandlers.mu.Unlock() if !signaled { if _, err := t.CopyOut(t.cleartid, ThreadID(0)); err == nil { - t.Futex().Wake(uintptr(t.cleartid), ^uint32(0), 1) + t.Futex().Wake(t.FutexChecker(), uintptr(t.cleartid), false, ^uint32(0), 1) } // If the CopyOut fails, there's nothing we can do. } diff --git a/pkg/sentry/kernel/task_futex.go b/pkg/sentry/kernel/task_futex.go new file mode 100644 index 000000000..62ebbcb0d --- /dev/null +++ b/pkg/sentry/kernel/task_futex.go @@ -0,0 +1,148 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// Futex returns t's futex manager. +// +// Preconditions: The caller must be running on the task goroutine, or t.mu +// must be locked. +func (t *Task) Futex() *futex.Manager { + return t.tc.fu +} + +// FutexChecker returns a futex.Checker that interprets addresses in t's +// address space. +// +// Preconditions: All uses of the returned futex.Checker must be on the task +// goroutine. +func (t *Task) FutexChecker() futex.Checker { + return futexChecker{t} +} + +type futexChecker struct { + t *Task +} + +// Check implements futex.Checker.Check. +func (f futexChecker) Check(addr uintptr, val uint32) error { + // FIXME + in := f.t.CopyScratchBuffer(4) + _, err := f.t.CopyInBytes(usermem.Addr(addr), in) + if err != nil { + return err + } + nval := usermem.ByteOrder.Uint32(in) + if val != nval { + return syserror.EAGAIN + } + return nil +} + +func (f futexChecker) atomicOp(addr uintptr, op func(uint32) uint32) (uint32, error) { + // FIXME + in := f.t.CopyScratchBuffer(4) + _, err := f.t.CopyInBytes(usermem.Addr(addr), in) + if err != nil { + return 0, err + } + o := usermem.ByteOrder.Uint32(in) + mm := f.t.MemoryManager() + for { + n := op(o) + r, err := mm.CompareAndSwapUint32(f.t, usermem.Addr(addr), o, n, usermem.IOOpts{ + AddressSpaceActive: true, + }) + if err != nil { + return 0, err + } + + if r == o { + return o, nil + } + o = r + } +} + +// Op implements futex.Checker.Op, interpreting opIn consistently with Linux. +func (f futexChecker) Op(addr uintptr, opIn uint32) (bool, error) { + op := (opIn >> 28) & 0xf + cmp := (opIn >> 24) & 0xf + opArg := (opIn >> 12) & 0xfff + cmpArg := opIn & 0xfff + + if op&linux.FUTEX_OP_OPARG_SHIFT != 0 { + opArg = 1 << opArg + op &^= linux.FUTEX_OP_OPARG_SHIFT // clear flag + } + + var oldVal uint32 + var err error + switch op { + case linux.FUTEX_OP_SET: + oldVal, err = f.t.MemoryManager().SwapUint32(f.t, usermem.Addr(addr), opArg, usermem.IOOpts{ + AddressSpaceActive: true, + }) + case linux.FUTEX_OP_ADD: + oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { + return a + opArg + }) + case linux.FUTEX_OP_OR: + oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { + return a | opArg + }) + case linux.FUTEX_OP_ANDN: + oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { + return a &^ opArg + }) + case linux.FUTEX_OP_XOR: + oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { + return a ^ opArg + }) + default: + return false, syserror.ENOSYS + } + if err != nil { + return false, err + } + + switch cmp { + case linux.FUTEX_OP_CMP_EQ: + return oldVal == cmpArg, nil + case linux.FUTEX_OP_CMP_NE: + return oldVal != cmpArg, nil + case linux.FUTEX_OP_CMP_LT: + return oldVal < cmpArg, nil + case linux.FUTEX_OP_CMP_LE: + return oldVal <= cmpArg, nil + case linux.FUTEX_OP_CMP_GT: + return oldVal > cmpArg, nil + case linux.FUTEX_OP_CMP_GE: + return oldVal >= cmpArg, nil + default: + return false, syserror.ENOSYS + } +} + +// GetSharedKey implements futex.Checker.GetSharedKey. +func (f futexChecker) GetSharedKey(addr uintptr) (futex.Key, error) { + return f.t.MemoryManager().GetSharedFutexKey(f.t, usermem.Addr(addr)) +} diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index ad9231774..744e73a39 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -106,6 +106,7 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/proc/seqfile", + "//pkg/sentry/kernel/futex", "//pkg/sentry/kernel/shm", "//pkg/sentry/limits", "//pkg/sentry/memmap", diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index 21aeabde8..b0622b0c3 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -20,6 +20,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" @@ -793,6 +794,40 @@ func (mm *MemoryManager) Sync(ctx context.Context, addr usermem.Addr, length uin return nil } +// GetSharedFutexKey is used by kernel.futexChecker.GetSharedKey to implement +// futex.Checker.GetSharedKey. +func (mm *MemoryManager) GetSharedFutexKey(ctx context.Context, addr usermem.Addr) (futex.Key, error) { + ar, ok := addr.ToRange(4) // sizeof(int32) + if !ok { + return futex.Key{}, syserror.EFAULT + } + + mm.mappingMu.RLock() + defer mm.mappingMu.RUnlock() + vseg, _, err := mm.getVMAsLocked(ctx, ar, usermem.Read, false) + if err != nil { + return futex.Key{}, err + } + vma := vseg.ValuePtr() + + if vma.private { + return futex.Key{ + Kind: futex.KindSharedPrivate, + Offset: uint64(addr), + }, nil + } + + if vma.id != nil { + vma.id.IncRef() + } + return futex.Key{ + Kind: futex.KindSharedMappable, + Mappable: vma.mappable, + MappingIdentity: vma.id, + Offset: vseg.mappableOffsetAt(addr), + }, nil +} + // VirtualMemorySize returns the combined length in bytes of all mappings in // mm. func (mm *MemoryManager) VirtualMemorySize() uint64 { diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index 1a0e1f5fb..d35dcecbe 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -21,115 +21,9 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) -// futexChecker is a futex.Checker that uses a Task's MemoryManager. -type futexChecker struct { - t *kernel.Task -} - -// Check checks if the address contains the given value, and returns -// syserror.EAGAIN if it doesn't. See Checker interface in futex package -// for more information. -func (f futexChecker) Check(addr uintptr, val uint32) error { - in := f.t.CopyScratchBuffer(4) - _, err := f.t.CopyInBytes(usermem.Addr(addr), in) - if err != nil { - return err - } - nval := usermem.ByteOrder.Uint32(in) - if val != nval { - return syserror.EAGAIN - } - return nil -} - -func (f futexChecker) atomicOp(addr uintptr, op func(uint32) uint32) (uint32, error) { - in := f.t.CopyScratchBuffer(4) - _, err := f.t.CopyInBytes(usermem.Addr(addr), in) - if err != nil { - return 0, err - } - o := usermem.ByteOrder.Uint32(in) - mm := f.t.MemoryManager() - for { - n := op(o) - r, err := mm.CompareAndSwapUint32(f.t, usermem.Addr(addr), o, n, usermem.IOOpts{ - AddressSpaceActive: true, - }) - if err != nil { - return 0, err - } - - if r == o { - return o, nil - } - o = r - } -} - -// Op performs an operation on addr and returns a result based on the operation. -func (f futexChecker) Op(addr uintptr, opIn uint32) (bool, error) { - op := (opIn >> 28) & 0xf - cmp := (opIn >> 24) & 0xf - opArg := (opIn >> 12) & 0xfff - cmpArg := opIn & 0xfff - - if op&linux.FUTEX_OP_OPARG_SHIFT != 0 { - opArg = 1 << opArg - op &^= linux.FUTEX_OP_OPARG_SHIFT // clear flag - } - - var oldVal uint32 - var err error - switch op { - case linux.FUTEX_OP_SET: - oldVal, err = f.t.MemoryManager().SwapUint32(f.t, usermem.Addr(addr), opArg, usermem.IOOpts{ - AddressSpaceActive: true, - }) - case linux.FUTEX_OP_ADD: - oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { - return a + opArg - }) - case linux.FUTEX_OP_OR: - oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { - return a | opArg - }) - case linux.FUTEX_OP_ANDN: - oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { - return a & ^opArg - }) - case linux.FUTEX_OP_XOR: - oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { - return a ^ opArg - }) - default: - return false, syserror.ENOSYS - } - if err != nil { - return false, err - } - - switch cmp { - case linux.FUTEX_OP_CMP_EQ: - return oldVal == cmpArg, nil - case linux.FUTEX_OP_CMP_NE: - return oldVal != cmpArg, nil - case linux.FUTEX_OP_CMP_LT: - return oldVal < cmpArg, nil - case linux.FUTEX_OP_CMP_LE: - return oldVal <= cmpArg, nil - case linux.FUTEX_OP_CMP_GT: - return oldVal > cmpArg, nil - case linux.FUTEX_OP_CMP_GE: - return oldVal >= cmpArg, nil - default: - return false, syserror.ENOSYS - } -} - // futexWaitRestartBlock encapsulates the state required to restart futex(2) // via restart_syscall(2). // @@ -140,13 +34,14 @@ type futexWaitRestartBlock struct { // addr stored as uint64 since uintptr is not save-able. addr uint64 - val uint32 - mask uint32 + private bool + val uint32 + mask uint32 } // Restart implements kernel.SyscallRestartBlock.Restart. func (f *futexWaitRestartBlock) Restart(t *kernel.Task) (uintptr, error) { - return futexWaitDuration(t, f.duration, false, uintptr(f.addr), f.val, f.mask) + return futexWaitDuration(t, f.duration, false, uintptr(f.addr), f.private, f.val, f.mask) } // futexWaitAbsolute performs a FUTEX_WAIT_BITSET, blocking until the wait is @@ -156,9 +51,9 @@ func (f *futexWaitRestartBlock) Restart(t *kernel.Task) (uintptr, error) { // // If blocking is interrupted, the syscall is restarted with the original // arguments. -func futexWaitAbsolute(t *kernel.Task, clockRealtime bool, ts linux.Timespec, forever bool, addr uintptr, val, mask uint32) (uintptr, error) { +func futexWaitAbsolute(t *kernel.Task, clockRealtime bool, ts linux.Timespec, forever bool, addr uintptr, private bool, val, mask uint32) (uintptr, error) { w := t.FutexWaiter() - err := t.Futex().WaitPrepare(w, futexChecker{t}, addr, val, mask) + err := t.Futex().WaitPrepare(w, t.FutexChecker(), addr, private, val, mask) if err != nil { return 0, err } @@ -192,9 +87,9 @@ func futexWaitAbsolute(t *kernel.Task, clockRealtime bool, ts linux.Timespec, fo // syscall. If forever is true, the syscall is restarted with the original // arguments. If forever is false, duration is a relative timeout and the // syscall is restarted with the remaining timeout. -func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, addr uintptr, val, mask uint32) (uintptr, error) { +func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, addr uintptr, private bool, val, mask uint32) (uintptr, error) { w := t.FutexWaiter() - err := t.Futex().WaitPrepare(w, futexChecker{t}, addr, val, mask) + err := t.Futex().WaitPrepare(w, t.FutexChecker(), addr, private, val, mask) if err != nil { return 0, err } @@ -222,6 +117,7 @@ func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, add t.SetSyscallRestartBlock(&futexWaitRestartBlock{ duration: remaining, addr: uint64(addr), + private: private, val: val, mask: mask, }) @@ -243,6 +139,7 @@ func Futex(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall addr := uintptr(uaddr) naddr := uintptr(uaddr2) cmd := futexOp &^ (linux.FUTEX_PRIVATE_FLAG | linux.FUTEX_CLOCK_REALTIME) + private := (futexOp & linux.FUTEX_PRIVATE_FLAG) != 0 clockRealtime := (futexOp & linux.FUTEX_CLOCK_REALTIME) == linux.FUTEX_CLOCK_REALTIME mask := uint32(val3) @@ -268,7 +165,7 @@ func Futex(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall if !forever { timeoutDur = time.Duration(timespec.ToNsecCapped()) * time.Nanosecond } - n, err := futexWaitDuration(t, timeoutDur, forever, addr, uint32(val), mask) + n, err := futexWaitDuration(t, timeoutDur, forever, addr, private, uint32(val), mask) return n, nil, err case linux.FUTEX_WAIT_BITSET: @@ -277,7 +174,7 @@ func Futex(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall if mask == 0 { return 0, nil, syserror.EINVAL } - n, err := futexWaitAbsolute(t, clockRealtime, timespec, forever, addr, uint32(val), mask) + n, err := futexWaitAbsolute(t, clockRealtime, timespec, forever, addr, private, uint32(val), mask) return n, nil, err default: panic("unreachable") @@ -291,23 +188,23 @@ func Futex(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall if mask == 0 { return 0, nil, syserror.EINVAL } - n, err := t.Futex().Wake(addr, mask, val) + n, err := t.Futex().Wake(t.FutexChecker(), addr, private, mask, val) return uintptr(n), nil, err case linux.FUTEX_REQUEUE: - n, err := t.Futex().Requeue(addr, naddr, val, nreq) + n, err := t.Futex().Requeue(t.FutexChecker(), addr, naddr, private, val, nreq) return uintptr(n), nil, err case linux.FUTEX_CMP_REQUEUE: // 'val3' contains the value to be checked at 'addr' and // 'val' is the number of waiters that should be woken up. nval := uint32(val3) - n, err := t.Futex().RequeueCmp(futexChecker{t}, addr, nval, naddr, val, nreq) + n, err := t.Futex().RequeueCmp(t.FutexChecker(), addr, naddr, private, nval, val, nreq) return uintptr(n), nil, err case linux.FUTEX_WAKE_OP: op := uint32(val3) - n, err := t.Futex().WakeOp(futexChecker{t}, addr, naddr, val, nreq, op) + n, err := t.Futex().WakeOp(t.FutexChecker(), addr, naddr, private, val, nreq, op) return uintptr(n), nil, err case linux.FUTEX_LOCK_PI, linux.FUTEX_UNLOCK_PI, linux.FUTEX_TRYLOCK_PI, linux.FUTEX_WAIT_REQUEUE_PI, linux.FUTEX_CMP_REQUEUE_PI: -- cgit v1.2.3 From b78552d30e0af4122710e01bc86cbde6bb412686 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Wed, 10 Oct 2018 10:41:18 -0700 Subject: When creating a new process group, add it to the session. PiperOrigin-RevId: 216554791 Change-Id: Ia6b7a2e6eaad80a81b2a8f2e3241e93ebc2bda35 --- pkg/sentry/kernel/sessions.go | 3 +++ 1 file changed, 3 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index b44d218d9..a9b4e7647 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -367,6 +367,9 @@ func (tg *ThreadGroup) CreateProcessGroup() error { tg.processGroup.decRefWithParent(oldParentPG) tg.processGroup = pg + // Add the new process group to the session. + pg.session.processGroups.PushBack(pg) + // Ensure this translation is added to all namespaces. for ns := tg.pidns; ns != nil; ns = ns.parent { local := ns.tids[tg.leader] -- cgit v1.2.3 From 167f2401c4abb1ebda1f4536a04d9854e9008e0b Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Mon, 15 Oct 2018 17:47:24 -0700 Subject: Merge host.endpoint into host.ConnectedEndpoint host.endpoint contained duplicated logic from the sockerpair implementation and host.ConnectedEndpoint. Remove host.endpoint in favor of a host.ConnectedEndpoint wrapped in a socketpair end. PiperOrigin-RevId: 217240096 Change-Id: I4a3d51e3fe82bdf30e2d0152458b8499ab4c987c --- pkg/sentry/context/contexttest/contexttest.go | 12 +- pkg/sentry/fs/host/BUILD | 2 + pkg/sentry/fs/host/socket.go | 319 +++++++------------------- pkg/sentry/fs/host/socket_state.go | 19 +- pkg/sentry/kernel/kernel.go | 4 + pkg/sentry/kernel/task.go | 2 + pkg/sentry/uniqueid/BUILD | 5 +- pkg/sentry/uniqueid/context.go | 10 + pkg/tcpip/transport/unix/connectioned.go | 11 + 9 files changed, 134 insertions(+), 250 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/context/contexttest/contexttest.go b/pkg/sentry/context/contexttest/contexttest.go index 193ce3440..b3c6a566b 100644 --- a/pkg/sentry/context/contexttest/contexttest.go +++ b/pkg/sentry/context/contexttest/contexttest.go @@ -53,6 +53,14 @@ type testContext struct { // globalUniqueID tracks incremental unique identifiers for tests. var globalUniqueID uint64 +// globalUniqueIDProvider implements unix.UniqueIDProvider. +type globalUniqueIDProvider struct{} + +// UniqueID implements unix.UniqueIDProvider.UniqueID. +func (*globalUniqueIDProvider) UniqueID() uint64 { + return atomic.AddUint64(&globalUniqueID, 1) +} + // lastInotifyCookie is a monotonically increasing counter for generating unique // inotify cookies. Must be accessed using atomic ops. var lastInotifyCookie uint32 @@ -76,7 +84,9 @@ func (t *testContext) Value(key interface{}) interface{} { case platform.CtxPlatform: return t.platform case uniqueid.CtxGlobalUniqueID: - return atomic.AddUint64(&globalUniqueID, 1) + return (*globalUniqueIDProvider).UniqueID(nil) + case uniqueid.CtxGlobalUniqueIDProvider: + return &globalUniqueIDProvider{} case uniqueid.CtxInotifyCookie: return atomic.AddUint32(&lastInotifyCookie, 1) case ktime.CtxRealtimeClock: diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 6d5640f0a..5ada32ee1 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -42,7 +42,9 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket/control", "//pkg/sentry/socket/unix", + "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", + "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/link/rawfile", diff --git a/pkg/sentry/fs/host/socket.go b/pkg/sentry/fs/host/socket.go index 68ebf6402..577e9e272 100644 --- a/pkg/sentry/fs/host/socket.go +++ b/pkg/sentry/fs/host/socket.go @@ -25,6 +25,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control" unixsocket "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" + "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/tcpip/link/rawfile" @@ -39,88 +41,35 @@ import ( // N.B. 8MB is the default maximum on Linux (2 * sysctl_wmem_max). const maxSendBufferSize = 8 << 20 -// endpoint encapsulates the state needed to represent a host Unix socket. -// -// TODO: Remove/merge with ConnectedEndpoint. -// -// +stateify savable -type endpoint struct { - queue waiter.Queue `state:"zerovalue"` - - // fd is the host fd backing this file. - fd int `state:"nosave"` - - // If srfd >= 0, it is the host fd that fd was imported from. - srfd int `state:"wait"` - - // stype is the type of Unix socket. - stype unix.SockType `state:"nosave"` - - // sndbuf is the size of the send buffer. - sndbuf int `state:"nosave"` -} - -func (e *endpoint) init() error { - family, err := syscall.GetsockoptInt(e.fd, syscall.SOL_SOCKET, syscall.SO_DOMAIN) - if err != nil { - return err - } - - if family != syscall.AF_UNIX { - // We only allow Unix sockets. - return syserror.EINVAL - } - - stype, err := syscall.GetsockoptInt(e.fd, syscall.SOL_SOCKET, syscall.SO_TYPE) - if err != nil { - return err - } - e.stype = unix.SockType(stype) - - e.sndbuf, err = syscall.GetsockoptInt(e.fd, syscall.SOL_SOCKET, syscall.SO_SNDBUF) - if err != nil { - return err - } - if e.sndbuf > maxSendBufferSize { - log.Warningf("Socket send buffer too large: %d", e.sndbuf) - return syserror.EINVAL - } - - if err := syscall.SetNonblock(e.fd, true); err != nil { - return err - } - - return fdnotifier.AddFD(int32(e.fd), &e.queue) -} - -// newEndpoint creates a new host endpoint. -func newEndpoint(fd int, srfd int) (*endpoint, error) { - ep := &endpoint{fd: fd, srfd: srfd} - if err := ep.init(); err != nil { - return nil, err - } - return ep, nil -} - // newSocket allocates a new unix socket with host endpoint. -func newSocket(ctx context.Context, fd int, saveable bool) (*fs.File, error) { - ownedfd := fd +func newSocket(ctx context.Context, orgfd int, saveable bool) (*fs.File, error) { + ownedfd := orgfd srfd := -1 if saveable { var err error - ownedfd, err = syscall.Dup(fd) + ownedfd, err = syscall.Dup(orgfd) if err != nil { return nil, err } - srfd = fd + srfd = orgfd } - ep, err := newEndpoint(ownedfd, srfd) + f := fd.New(ownedfd) + var q waiter.Queue + e, err := NewConnectedEndpoint(f, &q, "" /* path */) if err != nil { if saveable { - syscall.Close(ownedfd) + f.Close() + } else { + f.Release() } - return nil, err + return nil, syserr.TranslateNetstackError(err).ToError() } + + e.srfd = srfd + e.Init() + + ep := unix.NewExternal(e.stype, uniqueid.GlobalProviderFromContext(ctx), &q, e, e) + return unixsocket.New(ctx, ep), nil } @@ -130,139 +79,22 @@ func newSocket(ctx context.Context, fd int, saveable bool) (*fs.File, error) { // // NewSocketWithDirent takes ownership of f on success. func NewSocketWithDirent(ctx context.Context, d *fs.Dirent, f *fd.FD, flags fs.FileFlags) (*fs.File, error) { - ep, err := newEndpoint(f.FD(), -1) + f2 := fd.New(f.FD()) + var q waiter.Queue + e, err := NewConnectedEndpoint(f2, &q, "" /* path */) if err != nil { - return nil, err + f2.Release() + return nil, syserr.TranslateNetstackError(err).ToError() } // Take ownship of the FD. f.Release() - return unixsocket.NewWithDirent(ctx, d, ep, flags), nil -} - -// Close implements unix.Endpoint.Close. -func (e *endpoint) Close() { - fdnotifier.RemoveFD(int32(e.fd)) - syscall.Close(e.fd) - e.fd = -1 -} - -// EventRegister implements waiter.Waitable.EventRegister. -func (e *endpoint) EventRegister(we *waiter.Entry, mask waiter.EventMask) { - e.queue.EventRegister(we, mask) - fdnotifier.UpdateFD(int32(e.fd)) -} - -// EventUnregister implements waiter.Waitable.EventUnregister. -func (e *endpoint) EventUnregister(we *waiter.Entry) { - e.queue.EventUnregister(we) - fdnotifier.UpdateFD(int32(e.fd)) -} - -// Readiness implements unix.Endpoint.Readiness. -func (e *endpoint) Readiness(mask waiter.EventMask) waiter.EventMask { - return fdnotifier.NonBlockingPoll(int32(e.fd), mask) -} - -// Type implements unix.Endpoint.Type. -func (e *endpoint) Type() unix.SockType { - return e.stype -} - -// Connect implements unix.Endpoint.Connect. -func (e *endpoint) Connect(server unix.BoundEndpoint) *tcpip.Error { - return tcpip.ErrInvalidEndpointState -} - -// Bind implements unix.Endpoint.Bind. -func (e *endpoint) Bind(address tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error { - return tcpip.ErrInvalidEndpointState -} - -// Listen implements unix.Endpoint.Listen. -func (e *endpoint) Listen(backlog int) *tcpip.Error { - return tcpip.ErrInvalidEndpointState -} - -// Accept implements unix.Endpoint.Accept. -func (e *endpoint) Accept() (unix.Endpoint, *tcpip.Error) { - return nil, tcpip.ErrInvalidEndpointState -} - -// Shutdown implements unix.Endpoint.Shutdown. -func (e *endpoint) Shutdown(flags tcpip.ShutdownFlags) *tcpip.Error { - return tcpip.ErrInvalidEndpointState -} - -// GetSockOpt implements unix.Endpoint.GetSockOpt. -func (e *endpoint) GetSockOpt(opt interface{}) *tcpip.Error { - switch o := opt.(type) { - case tcpip.ErrorOption: - _, err := syscall.GetsockoptInt(e.fd, syscall.SOL_SOCKET, syscall.SO_ERROR) - return translateError(err) - case *tcpip.PasscredOption: - // We don't support passcred on host sockets. - *o = 0 - return nil - case *tcpip.SendBufferSizeOption: - *o = tcpip.SendBufferSizeOption(e.sndbuf) - return nil - case *tcpip.ReceiveBufferSizeOption: - // N.B. Unix sockets don't use the receive buffer. We'll claim it is - // the same size as the send buffer. - *o = tcpip.ReceiveBufferSizeOption(e.sndbuf) - return nil - case *tcpip.ReuseAddressOption: - v, err := syscall.GetsockoptInt(e.fd, syscall.SOL_SOCKET, syscall.SO_REUSEADDR) - *o = tcpip.ReuseAddressOption(v) - return translateError(err) - case *tcpip.ReceiveQueueSizeOption: - return tcpip.ErrQueueSizeNotSupported - } - return tcpip.ErrInvalidEndpointState -} - -// SetSockOpt implements unix.Endpoint.SetSockOpt. -func (e *endpoint) SetSockOpt(opt interface{}) *tcpip.Error { - return nil -} - -// GetLocalAddress implements unix.Endpoint.GetLocalAddress. -func (e *endpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) { - return tcpip.FullAddress{}, nil -} - -// GetRemoteAddress implements unix.Endpoint.GetRemoteAddress. -func (e *endpoint) GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) { - return tcpip.FullAddress{}, nil -} - -// Passcred returns whether or not the SO_PASSCRED socket option is -// enabled on this end. -func (e *endpoint) Passcred() bool { - // We don't support credential passing for host sockets. - return false -} - -// ConnectedPasscred returns whether or not the SO_PASSCRED socket option -// is enabled on the connected end. -func (e *endpoint) ConnectedPasscred() bool { - // We don't support credential passing for host sockets. - return false -} + e.Init() -// SendMsg implements unix.Endpoint.SendMsg. -func (e *endpoint) SendMsg(data [][]byte, controlMessages unix.ControlMessages, to unix.BoundEndpoint) (uintptr, *tcpip.Error) { - if to != nil { - return 0, tcpip.ErrInvalidEndpointState - } - - // Since stream sockets don't preserve message boundaries, we can write - // only as much of the message as fits in the send buffer. - truncate := e.stype == unix.SockStream + ep := unix.NewExternal(e.stype, uniqueid.GlobalProviderFromContext(ctx), &q, e, e) - return sendMsg(e.fd, data, controlMessages, e.sndbuf, truncate) + return unixsocket.NewWithDirent(ctx, d, ep, flags), nil } func sendMsg(fd int, data [][]byte, controlMessages unix.ControlMessages, maxlen int, truncate bool) (uintptr, *tcpip.Error) { @@ -278,18 +110,6 @@ func sendMsg(fd int, data [][]byte, controlMessages unix.ControlMessages, maxlen return n, translateError(err) } -// RecvMsg implements unix.Endpoint.RecvMsg. -func (e *endpoint) RecvMsg(data [][]byte, creds bool, numRights uintptr, peek bool, addr *tcpip.FullAddress) (uintptr, uintptr, unix.ControlMessages, *tcpip.Error) { - // N.B. Unix sockets don't have a receive buffer, the send buffer - // serves both purposes. - rl, ml, cm, err := recvMsg(e.fd, data, numRights, peek, addr, e.sndbuf) - if rl > 0 && err == tcpip.ErrWouldBlock { - // Message did not fill buffer; that's fine, no need to block. - err = nil - } - return rl, ml, cm, err -} - func recvMsg(fd int, data [][]byte, numRights uintptr, peek bool, addr *tcpip.FullAddress, maxlen int) (uintptr, uintptr, unix.ControlMessages, *tcpip.Error) { var cm unet.ControlMessage if numRights > 0 { @@ -328,42 +148,21 @@ func recvMsg(fd int, data [][]byte, numRights uintptr, peek bool, addr *tcpip.Fu // be called twice because host.ConnectedEndpoint is both a unix.Receiver and // unix.ConnectedEndpoint. func NewConnectedEndpoint(file *fd.FD, queue *waiter.Queue, path string) (*ConnectedEndpoint, *tcpip.Error) { - family, err := syscall.GetsockoptInt(file.FD(), syscall.SOL_SOCKET, syscall.SO_DOMAIN) - if err != nil { - return nil, translateError(err) - } - - if family != syscall.AF_UNIX { - // We only allow Unix sockets. - return nil, tcpip.ErrInvalidEndpointState - } - - stype, err := syscall.GetsockoptInt(file.FD(), syscall.SOL_SOCKET, syscall.SO_TYPE) - if err != nil { - return nil, translateError(err) - } - - sndbuf, err := syscall.GetsockoptInt(file.FD(), syscall.SOL_SOCKET, syscall.SO_SNDBUF) - if err != nil { - return nil, translateError(err) - } - if sndbuf > maxSendBufferSize { - log.Warningf("Socket send buffer too large: %d", sndbuf) - return nil, tcpip.ErrInvalidEndpointState + e := ConnectedEndpoint{ + path: path, + queue: queue, + file: file, + srfd: -1, } - e := &ConnectedEndpoint{ - path: path, - queue: queue, - file: file, - stype: unix.SockType(stype), - sndbuf: sndbuf, + if err := e.init(); err != nil { + return nil, err } // AtomicRefCounters start off with a single reference. We need two. e.ref.IncRef() - return e, nil + return &e, nil } // Init will do initialization required without holding other locks. @@ -376,7 +175,7 @@ func (c *ConnectedEndpoint) Init() { // ConnectedEndpoint is a host FD backed implementation of // unix.ConnectedEndpoint and unix.Receiver. // -// ConnectedEndpoint does not support save/restore for now. +// +stateify savable type ConnectedEndpoint struct { queue *waiter.Queue path string @@ -385,11 +184,11 @@ type ConnectedEndpoint struct { ref refs.AtomicRefCount // mu protects fd, readClosed and writeClosed. - mu sync.RWMutex + mu sync.RWMutex `state:"nosave"` // file is an *fd.FD containing the FD backing this endpoint. It must be // set to nil if it has been closed. - file *fd.FD + file *fd.FD `state:"nosave"` // readClosed is true if the FD has read shutdown or if it has been closed. readClosed bool @@ -398,6 +197,9 @@ type ConnectedEndpoint struct { // closed. writeClosed bool + // If srfd >= 0, it is the host FD that file was imported from. + srfd int `state:"wait"` + // stype is the type of Unix socket. stype unix.SockType @@ -407,7 +209,44 @@ type ConnectedEndpoint struct { // GetSockOpt and message splitting/rejection in SendMsg, but do not // prevent lots of small messages from filling the real send buffer // size on the host. - sndbuf int + sndbuf int `state:"nosave"` +} + +// init performs initialization required for creating new ConnectedEndpoints and +// for restoring them. +func (c *ConnectedEndpoint) init() *tcpip.Error { + family, err := syscall.GetsockoptInt(c.file.FD(), syscall.SOL_SOCKET, syscall.SO_DOMAIN) + if err != nil { + return translateError(err) + } + + if family != syscall.AF_UNIX { + // We only allow Unix sockets. + return tcpip.ErrInvalidEndpointState + } + + stype, err := syscall.GetsockoptInt(c.file.FD(), syscall.SOL_SOCKET, syscall.SO_TYPE) + if err != nil { + return translateError(err) + } + + if err := syscall.SetNonblock(c.file.FD(), true); err != nil { + return translateError(err) + } + + sndbuf, err := syscall.GetsockoptInt(c.file.FD(), syscall.SOL_SOCKET, syscall.SO_SNDBUF) + if err != nil { + return translateError(err) + } + if sndbuf > maxSendBufferSize { + log.Warningf("Socket send buffer too large: %d", sndbuf) + return tcpip.ErrInvalidEndpointState + } + + c.stype = unix.SockType(stype) + c.sndbuf = sndbuf + + return nil } // Send implements unix.ConnectedEndpoint.Send. diff --git a/pkg/sentry/fs/host/socket_state.go b/pkg/sentry/fs/host/socket_state.go index 6acabd55a..7fa500bfb 100644 --- a/pkg/sentry/fs/host/socket_state.go +++ b/pkg/sentry/fs/host/socket_state.go @@ -17,23 +17,26 @@ package host import ( "fmt" "syscall" + + "gvisor.googlesource.com/gvisor/pkg/fd" ) // beforeSave is invoked by stateify. -func (ep *endpoint) beforeSave() { - if ep.srfd < 0 { +func (c *ConnectedEndpoint) beforeSave() { + if c.srfd < 0 { panic("only host file descriptors provided at sentry startup can be saved") } } // afterLoad is invoked by stateify. -func (ep *endpoint) afterLoad() { - fd, err := syscall.Dup(ep.srfd) +func (c *ConnectedEndpoint) afterLoad() { + f, err := syscall.Dup(c.srfd) if err != nil { - panic(fmt.Sprintf("failed to dup restored fd %d: %v", ep.srfd, err)) + panic(fmt.Sprintf("failed to dup restored FD %d: %v", c.srfd, err)) } - ep.fd = fd - if err := ep.init(); err != nil { - panic(fmt.Sprintf("Could not restore host socket fd %d: %v", ep.srfd, err)) + c.file = fd.New(f) + if err := c.init(); err != nil { + panic(fmt.Sprintf("Could not restore host socket FD %d: %v", c.srfd, err)) } + c.Init() } diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 238fd127b..cc664deec 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -591,6 +591,8 @@ func (ctx *createProcessContext) Value(key interface{}) interface{} { return ctx.k case uniqueid.CtxGlobalUniqueID: return ctx.k.UniqueID() + case uniqueid.CtxGlobalUniqueIDProvider: + return ctx.k case uniqueid.CtxInotifyCookie: return ctx.k.GenerateInotifyCookie() default: @@ -1045,6 +1047,8 @@ func (ctx supervisorContext) Value(key interface{}) interface{} { return ctx.k case uniqueid.CtxGlobalUniqueID: return ctx.k.UniqueID() + case uniqueid.CtxGlobalUniqueIDProvider: + return ctx.k case uniqueid.CtxInotifyCookie: return ctx.k.GenerateInotifyCookie() default: diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 07ad1614c..4f0b7fe3f 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -590,6 +590,8 @@ func (t *Task) Value(key interface{}) interface{} { return t.k case uniqueid.CtxGlobalUniqueID: return t.k.UniqueID() + case uniqueid.CtxGlobalUniqueIDProvider: + return t.k case uniqueid.CtxInotifyCookie: return t.k.GenerateInotifyCookie() default: diff --git a/pkg/sentry/uniqueid/BUILD b/pkg/sentry/uniqueid/BUILD index 8eba3609e..ff50b9925 100644 --- a/pkg/sentry/uniqueid/BUILD +++ b/pkg/sentry/uniqueid/BUILD @@ -7,5 +7,8 @@ go_library( srcs = ["context.go"], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid", visibility = ["//pkg/sentry:internal"], - deps = ["//pkg/sentry/context"], + deps = [ + "//pkg/sentry/context", + "//pkg/tcpip/transport/unix", + ], ) diff --git a/pkg/sentry/uniqueid/context.go b/pkg/sentry/uniqueid/context.go index eeb8c4286..541e0611d 100644 --- a/pkg/sentry/uniqueid/context.go +++ b/pkg/sentry/uniqueid/context.go @@ -18,6 +18,7 @@ package uniqueid import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) // contextID is the kernel package's type for context.Context.Value keys. @@ -28,6 +29,10 @@ const ( // unique identifier. CtxGlobalUniqueID contextID = iota + // CtxGlobalUniqueIDProvider is a Context.Value key for a + // system-wide unique identifier generator. + CtxGlobalUniqueIDProvider + // CtxInotifyCookie is a Context.Value key for a unique inotify // event cookie. CtxInotifyCookie @@ -38,6 +43,11 @@ func GlobalFromContext(ctx context.Context) uint64 { return ctx.Value(CtxGlobalUniqueID).(uint64) } +// GlobalProviderFromContext returns a system-wide unique identifier from ctx. +func GlobalProviderFromContext(ctx context.Context) unix.UniqueIDProvider { + return ctx.Value(CtxGlobalUniqueIDProvider).(unix.UniqueIDProvider) +} + // InotifyCookie generates a unique inotify event cookie from ctx. func InotifyCookie(ctx context.Context) uint32 { return ctx.Value(CtxInotifyCookie).(uint32) diff --git a/pkg/tcpip/transport/unix/connectioned.go b/pkg/tcpip/transport/unix/connectioned.go index dd7c03cf1..e319b3bb8 100644 --- a/pkg/tcpip/transport/unix/connectioned.go +++ b/pkg/tcpip/transport/unix/connectioned.go @@ -158,6 +158,17 @@ func NewPair(stype SockType, uid UniqueIDProvider) (Endpoint, Endpoint) { return a, b } +// NewExternal creates a new externally backed Endpoint. It behaves like a +// socketpair. +func NewExternal(stype SockType, uid UniqueIDProvider, queue *waiter.Queue, receiver Receiver, connected ConnectedEndpoint) Endpoint { + return &connectionedEndpoint{ + baseEndpoint: baseEndpoint{Queue: queue, receiver: receiver, connected: connected}, + id: uid.UniqueID(), + idGenerator: uid, + stype: stype, + } +} + // ID implements ConnectingEndpoint.ID. func (e *connectionedEndpoint) ID() uint64 { return e.id -- cgit v1.2.3 From 6cba410df0ea2eabb87bad5074a8a79ed89312b8 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Wed, 17 Oct 2018 11:36:32 -0700 Subject: Move Unix transport out of netstack PiperOrigin-RevId: 217557656 Change-Id: I63d27635b1a6c12877279995d2d9847b6a19da9b --- pkg/sentry/fs/BUILD | 2 +- pkg/sentry/fs/ashmem/BUILD | 1 - pkg/sentry/fs/dirent.go | 4 +- pkg/sentry/fs/fsutil/BUILD | 2 +- pkg/sentry/fs/fsutil/inode.go | 6 +- pkg/sentry/fs/gofer/BUILD | 2 +- pkg/sentry/fs/gofer/path.go | 4 +- pkg/sentry/fs/gofer/session.go | 22 +- pkg/sentry/fs/gofer/socket.go | 25 +- pkg/sentry/fs/host/BUILD | 4 +- pkg/sentry/fs/host/control.go | 8 +- pkg/sentry/fs/host/inode.go | 6 +- pkg/sentry/fs/host/socket.go | 66 +- pkg/sentry/fs/host/socket_test.go | 12 +- pkg/sentry/fs/inode.go | 6 +- pkg/sentry/fs/inode_operations.go | 6 +- pkg/sentry/fs/inode_overlay.go | 6 +- pkg/sentry/fs/ramfs/BUILD | 2 +- pkg/sentry/fs/ramfs/dir.go | 6 +- pkg/sentry/fs/ramfs/ramfs.go | 4 +- pkg/sentry/fs/ramfs/socket.go | 8 +- pkg/sentry/fs/tmpfs/BUILD | 2 +- pkg/sentry/fs/tmpfs/tmpfs.go | 6 +- pkg/sentry/fs/tty/BUILD | 2 +- pkg/sentry/fs/tty/dir.go | 4 +- pkg/sentry/kernel/BUILD | 2 +- pkg/sentry/kernel/abstract_socket_namespace.go | 16 +- pkg/sentry/socket/BUILD | 2 +- pkg/sentry/socket/control/BUILD | 2 +- pkg/sentry/socket/control/control.go | 46 +- pkg/sentry/socket/epsocket/BUILD | 2 +- pkg/sentry/socket/epsocket/epsocket.go | 20 +- pkg/sentry/socket/epsocket/provider.go | 8 +- pkg/sentry/socket/hostinet/BUILD | 2 +- pkg/sentry/socket/hostinet/socket.go | 6 +- pkg/sentry/socket/netlink/BUILD | 2 +- pkg/sentry/socket/netlink/provider.go | 8 +- pkg/sentry/socket/netlink/socket.go | 20 +- pkg/sentry/socket/rpcinet/BUILD | 2 +- pkg/sentry/socket/rpcinet/socket.go | 6 +- pkg/sentry/socket/socket.go | 12 +- pkg/sentry/socket/unix/BUILD | 2 +- pkg/sentry/socket/unix/io.go | 21 +- pkg/sentry/socket/unix/transport/BUILD | 22 + pkg/sentry/socket/unix/transport/connectioned.go | 454 ++++++++++ .../socket/unix/transport/connectioned_state.go | 53 ++ pkg/sentry/socket/unix/transport/connectionless.go | 192 +++++ pkg/sentry/socket/unix/transport/queue/BUILD | 15 + pkg/sentry/socket/unix/transport/queue/queue.go | 227 +++++ pkg/sentry/socket/unix/transport/unix.go | 953 +++++++++++++++++++++ pkg/sentry/socket/unix/unix.go | 62 +- pkg/sentry/syscalls/linux/BUILD | 2 +- pkg/sentry/syscalls/linux/sys_socket.go | 8 +- pkg/sentry/uniqueid/BUILD | 2 +- pkg/sentry/uniqueid/context.go | 6 +- pkg/tcpip/transport/queue/BUILD | 15 - pkg/tcpip/transport/queue/queue.go | 227 ----- pkg/tcpip/transport/unix/BUILD | 22 - pkg/tcpip/transport/unix/connectioned.go | 454 ---------- pkg/tcpip/transport/unix/connectioned_state.go | 53 -- pkg/tcpip/transport/unix/connectionless.go | 192 ----- pkg/tcpip/transport/unix/unix.go | 953 --------------------- 62 files changed, 2154 insertions(+), 2153 deletions(-) create mode 100644 pkg/sentry/socket/unix/transport/BUILD create mode 100644 pkg/sentry/socket/unix/transport/connectioned.go create mode 100644 pkg/sentry/socket/unix/transport/connectioned_state.go create mode 100644 pkg/sentry/socket/unix/transport/connectionless.go create mode 100644 pkg/sentry/socket/unix/transport/queue/BUILD create mode 100644 pkg/sentry/socket/unix/transport/queue/queue.go create mode 100644 pkg/sentry/socket/unix/transport/unix.go delete mode 100644 pkg/tcpip/transport/queue/BUILD delete mode 100644 pkg/tcpip/transport/queue/queue.go delete mode 100644 pkg/tcpip/transport/unix/BUILD delete mode 100644 pkg/tcpip/transport/unix/connectioned.go delete mode 100644 pkg/tcpip/transport/unix/connectioned_state.go delete mode 100644 pkg/tcpip/transport/unix/connectionless.go delete mode 100644 pkg/tcpip/transport/unix/unix.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index a949fffbf..548898aaa 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -59,13 +59,13 @@ go_library( "//pkg/sentry/limits", "//pkg/sentry/memmap", "//pkg/sentry/platform", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/uniqueid", "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/state", "//pkg/syserror", "//pkg/tcpip", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/ashmem/BUILD b/pkg/sentry/fs/ashmem/BUILD index dc893d22f..44ef82e64 100644 --- a/pkg/sentry/fs/ashmem/BUILD +++ b/pkg/sentry/fs/ashmem/BUILD @@ -28,7 +28,6 @@ go_library( "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/syserror", - "//pkg/tcpip/transport/unix", ], ) diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go index a42c03e98..27fea0019 100644 --- a/pkg/sentry/fs/dirent.go +++ b/pkg/sentry/fs/dirent.go @@ -26,9 +26,9 @@ import ( "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) type globalDirentMap struct { @@ -800,7 +800,7 @@ func (d *Dirent) CreateDirectory(ctx context.Context, root *Dirent, name string, } // Bind satisfies the InodeOperations interface; otherwise same as GetFile. -func (d *Dirent) Bind(ctx context.Context, root *Dirent, name string, data unix.BoundEndpoint, perms FilePermissions) (*Dirent, error) { +func (d *Dirent) Bind(ctx context.Context, root *Dirent, name string, data transport.BoundEndpoint, perms FilePermissions) (*Dirent, error) { var childDir *Dirent err := d.genericCreate(ctx, root, name, func() error { var e error diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index 3512bae6f..6834e1272 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -87,11 +87,11 @@ go_library( "//pkg/sentry/memmap", "//pkg/sentry/platform", "//pkg/sentry/safemem", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/state", "//pkg/syserror", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index 3479f2fad..3acc32752 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -19,9 +19,9 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -254,7 +254,7 @@ func (InodeNotDirectory) CreateDirectory(context.Context, *fs.Inode, string, fs. } // Bind implements fs.InodeOperations.Bind. -func (InodeNotDirectory) Bind(context.Context, *fs.Inode, string, unix.BoundEndpoint, fs.FilePermissions) (*fs.Dirent, error) { +func (InodeNotDirectory) Bind(context.Context, *fs.Inode, string, transport.BoundEndpoint, fs.FilePermissions) (*fs.Dirent, error) { return nil, syserror.ENOTDIR } @@ -277,7 +277,7 @@ func (InodeNotDirectory) RemoveDirectory(context.Context, *fs.Inode, string) err type InodeNotSocket struct{} // BoundEndpoint implements fs.InodeOperations.BoundEndpoint. -func (InodeNotSocket) BoundEndpoint(*fs.Inode, string) unix.BoundEndpoint { +func (InodeNotSocket) BoundEndpoint(*fs.Inode, string) transport.BoundEndpoint { return nil } diff --git a/pkg/sentry/fs/gofer/BUILD b/pkg/sentry/fs/gofer/BUILD index cb17339c9..cef01829a 100644 --- a/pkg/sentry/fs/gofer/BUILD +++ b/pkg/sentry/fs/gofer/BUILD @@ -41,10 +41,10 @@ go_library( "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", "//pkg/sentry/safemem", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserror", "//pkg/tcpip", - "//pkg/tcpip/transport/unix", "//pkg/unet", "//pkg/waiter", ], diff --git a/pkg/sentry/fs/gofer/path.go b/pkg/sentry/fs/gofer/path.go index bec9680f8..0bf7881da 100644 --- a/pkg/sentry/fs/gofer/path.go +++ b/pkg/sentry/fs/gofer/path.go @@ -22,8 +22,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/device" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) // Lookup loads an Inode at name into a Dirent based on the session's cache @@ -180,7 +180,7 @@ func (i *inodeOperations) CreateDirectory(ctx context.Context, dir *fs.Inode, s } // Bind implements InodeOperations.Bind. -func (i *inodeOperations) Bind(ctx context.Context, dir *fs.Inode, name string, ep unix.BoundEndpoint, perm fs.FilePermissions) (*fs.Dirent, error) { +func (i *inodeOperations) Bind(ctx context.Context, dir *fs.Inode, name string, ep transport.BoundEndpoint, perm fs.FilePermissions) (*fs.Dirent, error) { if i.session().endpoints == nil { return nil, syscall.EOPNOTSUPP } diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index 49d27ee88..4e2293398 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -24,7 +24,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/device" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/unet" ) @@ -36,23 +36,23 @@ type endpointMaps struct { // direntMap links sockets to their dirents. // It is filled concurrently with the keyMap and is stored upon save. // Before saving, this map is used to populate the pathMap. - direntMap map[unix.BoundEndpoint]*fs.Dirent + direntMap map[transport.BoundEndpoint]*fs.Dirent // keyMap links MultiDeviceKeys (containing inode IDs) to their sockets. // It is not stored during save because the inode ID may change upon restore. - keyMap map[device.MultiDeviceKey]unix.BoundEndpoint `state:"nosave"` + keyMap map[device.MultiDeviceKey]transport.BoundEndpoint `state:"nosave"` // pathMap links the sockets to their paths. // It is filled before saving from the direntMap and is stored upon save. // Upon restore, this map is used to re-populate the keyMap. - pathMap map[unix.BoundEndpoint]string + pathMap map[transport.BoundEndpoint]string } // add adds the endpoint to the maps. // A reference is taken on the dirent argument. // // Precondition: maps must have been locked with 'lock'. -func (e *endpointMaps) add(key device.MultiDeviceKey, d *fs.Dirent, ep unix.BoundEndpoint) { +func (e *endpointMaps) add(key device.MultiDeviceKey, d *fs.Dirent, ep transport.BoundEndpoint) { e.keyMap[key] = ep d.IncRef() e.direntMap[ep] = d @@ -81,7 +81,7 @@ func (e *endpointMaps) lock() func() { // get returns the endpoint mapped to the given key. // // Precondition: maps must have been locked for reading. -func (e *endpointMaps) get(key device.MultiDeviceKey) unix.BoundEndpoint { +func (e *endpointMaps) get(key device.MultiDeviceKey) transport.BoundEndpoint { return e.keyMap[key] } @@ -285,9 +285,9 @@ func Root(ctx context.Context, dev string, filesystem fs.Filesystem, superBlockF // newEndpointMaps creates a new endpointMaps. func newEndpointMaps() *endpointMaps { return &endpointMaps{ - direntMap: make(map[unix.BoundEndpoint]*fs.Dirent), - keyMap: make(map[device.MultiDeviceKey]unix.BoundEndpoint), - pathMap: make(map[unix.BoundEndpoint]string), + direntMap: make(map[transport.BoundEndpoint]*fs.Dirent), + keyMap: make(map[device.MultiDeviceKey]transport.BoundEndpoint), + pathMap: make(map[transport.BoundEndpoint]string), } } @@ -341,7 +341,7 @@ func (s *session) fillPathMap() error { func (s *session) restoreEndpointMaps(ctx context.Context) error { // When restoring, only need to create the keyMap because the dirent and path // maps got stored through the save. - s.endpoints.keyMap = make(map[device.MultiDeviceKey]unix.BoundEndpoint) + s.endpoints.keyMap = make(map[device.MultiDeviceKey]transport.BoundEndpoint) if err := s.fillKeyMap(ctx); err != nil { return fmt.Errorf("failed to insert sockets into endpoint map: %v", err) } @@ -349,6 +349,6 @@ func (s *session) restoreEndpointMaps(ctx context.Context) error { // Re-create pathMap because it can no longer be trusted as socket paths can // change while process continues to run. Empty pathMap will be re-filled upon // next save. - s.endpoints.pathMap = make(map[unix.BoundEndpoint]string) + s.endpoints.pathMap = make(map[transport.BoundEndpoint]string) return nil } diff --git a/pkg/sentry/fs/gofer/socket.go b/pkg/sentry/fs/gofer/socket.go index 0190bc006..d072da624 100644 --- a/pkg/sentry/fs/gofer/socket.go +++ b/pkg/sentry/fs/gofer/socket.go @@ -19,13 +19,13 @@ import ( "gvisor.googlesource.com/gvisor/pkg/p9" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/host" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// BoundEndpoint returns a gofer-backed unix.BoundEndpoint. -func (i *inodeOperations) BoundEndpoint(inode *fs.Inode, path string) unix.BoundEndpoint { +// BoundEndpoint returns a gofer-backed transport.BoundEndpoint. +func (i *inodeOperations) BoundEndpoint(inode *fs.Inode, path string) transport.BoundEndpoint { if !fs.IsSocket(i.fileState.sattr) { return nil } @@ -45,7 +45,7 @@ func (i *inodeOperations) BoundEndpoint(inode *fs.Inode, path string) unix.Bound return &endpoint{inode, i.fileState.file.file, path} } -// endpoint is a Gofer-backed unix.BoundEndpoint. +// endpoint is a Gofer-backed transport.BoundEndpoint. // // An endpoint's lifetime is the time between when InodeOperations.BoundEndpoint() // is called and either BoundEndpoint.BidirectionalConnect or @@ -61,20 +61,20 @@ type endpoint struct { path string } -func unixSockToP9(t unix.SockType) (p9.ConnectFlags, bool) { +func unixSockToP9(t transport.SockType) (p9.ConnectFlags, bool) { switch t { - case unix.SockStream: + case transport.SockStream: return p9.StreamSocket, true - case unix.SockSeqpacket: + case transport.SockSeqpacket: return p9.SeqpacketSocket, true - case unix.SockDgram: + case transport.SockDgram: return p9.DgramSocket, true } return 0, false } // BidirectionalConnect implements ConnectableEndpoint.BidirectionalConnect. -func (e *endpoint) BidirectionalConnect(ce unix.ConnectingEndpoint, returnConnect func(unix.Receiver, unix.ConnectedEndpoint)) *tcpip.Error { +func (e *endpoint) BidirectionalConnect(ce transport.ConnectingEndpoint, returnConnect func(transport.Receiver, transport.ConnectedEndpoint)) *tcpip.Error { cf, ok := unixSockToP9(ce.Type()) if !ok { return tcpip.ErrConnectionRefused @@ -113,8 +113,9 @@ func (e *endpoint) BidirectionalConnect(ce unix.ConnectingEndpoint, returnConnec return nil } -// UnidirectionalConnect implements unix.BoundEndpoint.UnidirectionalConnect. -func (e *endpoint) UnidirectionalConnect() (unix.ConnectedEndpoint, *tcpip.Error) { +// UnidirectionalConnect implements +// transport.BoundEndpoint.UnidirectionalConnect. +func (e *endpoint) UnidirectionalConnect() (transport.ConnectedEndpoint, *tcpip.Error) { hostFile, err := e.file.Connect(p9.DgramSocket) if err != nil { return nil, tcpip.ErrConnectionRefused @@ -134,7 +135,7 @@ func (e *endpoint) UnidirectionalConnect() (unix.ConnectedEndpoint, *tcpip.Error return c, nil } -// Release implements unix.BoundEndpoint.Release. +// Release implements transport.BoundEndpoint.Release. func (e *endpoint) Release() { e.inode.DecRef() } diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 5ada32ee1..4f264a024 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -42,13 +42,13 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket/control", "//pkg/sentry/socket/unix", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/link/rawfile", - "//pkg/tcpip/transport/unix", "//pkg/unet", "//pkg/waiter", "//pkg/waiter/fdnotifier", @@ -72,10 +72,10 @@ go_test( "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/sentry/socket", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/tcpip", - "//pkg/tcpip/transport/unix", "//pkg/waiter", "//pkg/waiter/fdnotifier", ], diff --git a/pkg/sentry/fs/host/control.go b/pkg/sentry/fs/host/control.go index d2b007ab2..d2e34a69d 100644 --- a/pkg/sentry/fs/host/control.go +++ b/pkg/sentry/fs/host/control.go @@ -20,7 +20,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" ) type scmRights struct { @@ -45,13 +45,13 @@ func (c *scmRights) Files(ctx context.Context, max int) control.RightsFiles { return rf } -// Clone implements unix.RightsControlMessage.Clone. -func (c *scmRights) Clone() unix.RightsControlMessage { +// Clone implements transport.RightsControlMessage.Clone. +func (c *scmRights) Clone() transport.RightsControlMessage { // Host rights never need to be cloned. return nil } -// Release implements unix.RightsControlMessage.Release. +// Release implements transport.RightsControlMessage.Release. func (c *scmRights) Release() { for _, fd := range c.fds { syscall.Close(fd) diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index c2e8ba62f..e32497203 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -27,8 +27,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -310,12 +310,12 @@ func (i *inodeOperations) Rename(ctx context.Context, oldParent *fs.Inode, oldNa } // Bind implements fs.InodeOperations.Bind. -func (i *inodeOperations) Bind(ctx context.Context, dir *fs.Inode, name string, data unix.BoundEndpoint, perm fs.FilePermissions) (*fs.Dirent, error) { +func (i *inodeOperations) Bind(ctx context.Context, dir *fs.Inode, name string, data transport.BoundEndpoint, perm fs.FilePermissions) (*fs.Dirent, error) { return nil, syserror.EOPNOTSUPP } // BoundEndpoint implements fs.InodeOperations.BoundEndpoint. -func (i *inodeOperations) BoundEndpoint(inode *fs.Inode, path string) unix.BoundEndpoint { +func (i *inodeOperations) BoundEndpoint(inode *fs.Inode, path string) transport.BoundEndpoint { return nil } diff --git a/pkg/sentry/fs/host/socket.go b/pkg/sentry/fs/host/socket.go index e454b6fe5..0eb267c00 100644 --- a/pkg/sentry/fs/host/socket.go +++ b/pkg/sentry/fs/host/socket.go @@ -25,12 +25,12 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control" unixsocket "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/tcpip/link/rawfile" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/unet" "gvisor.googlesource.com/gvisor/pkg/waiter" "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" @@ -42,7 +42,7 @@ import ( const maxSendBufferSize = 8 << 20 // ConnectedEndpoint is a host FD backed implementation of -// unix.ConnectedEndpoint and unix.Receiver. +// transport.ConnectedEndpoint and transport.Receiver. // // +stateify savable type ConnectedEndpoint struct { @@ -70,7 +70,7 @@ type ConnectedEndpoint struct { srfd int `state:"wait"` // stype is the type of Unix socket. - stype unix.SockType + stype transport.SockType // sndbuf is the size of the send buffer. // @@ -112,7 +112,7 @@ func (c *ConnectedEndpoint) init() *tcpip.Error { return tcpip.ErrInvalidEndpointState } - c.stype = unix.SockType(stype) + c.stype = transport.SockType(stype) c.sndbuf = sndbuf return nil @@ -122,8 +122,8 @@ func (c *ConnectedEndpoint) init() *tcpip.Error { // that will pretend to be bound at a given sentry path. // // The caller is responsible for calling Init(). Additionaly, Release needs to -// be called twice because ConnectedEndpoint is both a unix.Receiver and -// unix.ConnectedEndpoint. +// be called twice because ConnectedEndpoint is both a transport.Receiver and +// transport.ConnectedEndpoint. func NewConnectedEndpoint(file *fd.FD, queue *waiter.Queue, path string) (*ConnectedEndpoint, *tcpip.Error) { e := ConnectedEndpoint{ path: path, @@ -168,7 +168,7 @@ func NewSocketWithDirent(ctx context.Context, d *fs.Dirent, f *fd.FD, flags fs.F e.Init() - ep := unix.NewExternal(e.stype, uniqueid.GlobalProviderFromContext(ctx), &q, e, e) + ep := transport.NewExternal(e.stype, uniqueid.GlobalProviderFromContext(ctx), &q, e, e) return unixsocket.NewWithDirent(ctx, d, ep, flags), nil } @@ -200,13 +200,13 @@ func newSocket(ctx context.Context, orgfd int, saveable bool) (*fs.File, error) e.srfd = srfd e.Init() - ep := unix.NewExternal(e.stype, uniqueid.GlobalProviderFromContext(ctx), &q, e, e) + ep := transport.NewExternal(e.stype, uniqueid.GlobalProviderFromContext(ctx), &q, e, e) return unixsocket.New(ctx, ep), nil } -// Send implements unix.ConnectedEndpoint.Send. -func (c *ConnectedEndpoint) Send(data [][]byte, controlMessages unix.ControlMessages, from tcpip.FullAddress) (uintptr, bool, *tcpip.Error) { +// Send implements transport.ConnectedEndpoint.Send. +func (c *ConnectedEndpoint) Send(data [][]byte, controlMessages transport.ControlMessages, from tcpip.FullAddress) (uintptr, bool, *tcpip.Error) { c.mu.RLock() defer c.mu.RUnlock() if c.writeClosed { @@ -219,7 +219,7 @@ func (c *ConnectedEndpoint) Send(data [][]byte, controlMessages unix.ControlMess // Since stream sockets don't preserve message boundaries, we can write // only as much of the message as fits in the send buffer. - truncate := c.stype == unix.SockStream + truncate := c.stype == transport.SockStream n, totalLen, err := fdWriteVec(c.file.FD(), data, c.sndbuf, truncate) if n < totalLen && err == nil { @@ -239,20 +239,20 @@ func (c *ConnectedEndpoint) Send(data [][]byte, controlMessages unix.ControlMess return n, false, translateError(err) } -// SendNotify implements unix.ConnectedEndpoint.SendNotify. +// SendNotify implements transport.ConnectedEndpoint.SendNotify. func (c *ConnectedEndpoint) SendNotify() {} -// CloseSend implements unix.ConnectedEndpoint.CloseSend. +// CloseSend implements transport.ConnectedEndpoint.CloseSend. func (c *ConnectedEndpoint) CloseSend() { c.mu.Lock() c.writeClosed = true c.mu.Unlock() } -// CloseNotify implements unix.ConnectedEndpoint.CloseNotify. +// CloseNotify implements transport.ConnectedEndpoint.CloseNotify. func (c *ConnectedEndpoint) CloseNotify() {} -// Writable implements unix.ConnectedEndpoint.Writable. +// Writable implements transport.ConnectedEndpoint.Writable. func (c *ConnectedEndpoint) Writable() bool { c.mu.RLock() defer c.mu.RUnlock() @@ -262,18 +262,18 @@ func (c *ConnectedEndpoint) Writable() bool { return fdnotifier.NonBlockingPoll(int32(c.file.FD()), waiter.EventOut)&waiter.EventOut != 0 } -// Passcred implements unix.ConnectedEndpoint.Passcred. +// Passcred implements transport.ConnectedEndpoint.Passcred. func (c *ConnectedEndpoint) Passcred() bool { // We don't support credential passing for host sockets. return false } -// GetLocalAddress implements unix.ConnectedEndpoint.GetLocalAddress. +// GetLocalAddress implements transport.ConnectedEndpoint.GetLocalAddress. func (c *ConnectedEndpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) { return tcpip.FullAddress{Addr: tcpip.Address(c.path)}, nil } -// EventUpdate implements unix.ConnectedEndpoint.EventUpdate. +// EventUpdate implements transport.ConnectedEndpoint.EventUpdate. func (c *ConnectedEndpoint) EventUpdate() { c.mu.RLock() defer c.mu.RUnlock() @@ -282,12 +282,12 @@ func (c *ConnectedEndpoint) EventUpdate() { } } -// Recv implements unix.Receiver.Recv. -func (c *ConnectedEndpoint) Recv(data [][]byte, creds bool, numRights uintptr, peek bool) (uintptr, uintptr, unix.ControlMessages, tcpip.FullAddress, bool, *tcpip.Error) { +// Recv implements transport.Receiver.Recv. +func (c *ConnectedEndpoint) Recv(data [][]byte, creds bool, numRights uintptr, peek bool) (uintptr, uintptr, transport.ControlMessages, tcpip.FullAddress, bool, *tcpip.Error) { c.mu.RLock() defer c.mu.RUnlock() if c.readClosed { - return 0, 0, unix.ControlMessages{}, tcpip.FullAddress{}, false, tcpip.ErrClosedForReceive + return 0, 0, transport.ControlMessages{}, tcpip.FullAddress{}, false, tcpip.ErrClosedForReceive } var cm unet.ControlMessage @@ -305,7 +305,7 @@ func (c *ConnectedEndpoint) Recv(data [][]byte, creds bool, numRights uintptr, p err = nil } if err != nil { - return 0, 0, unix.ControlMessages{}, tcpip.FullAddress{}, false, translateError(err) + return 0, 0, transport.ControlMessages{}, tcpip.FullAddress{}, false, translateError(err) } // There is no need for the callee to call RecvNotify because fdReadVec uses @@ -318,16 +318,16 @@ func (c *ConnectedEndpoint) Recv(data [][]byte, creds bool, numRights uintptr, p // Avoid extra allocations in the case where there isn't any control data. if len(cm) == 0 { - return rl, ml, unix.ControlMessages{}, tcpip.FullAddress{Addr: tcpip.Address(c.path)}, false, nil + return rl, ml, transport.ControlMessages{}, tcpip.FullAddress{Addr: tcpip.Address(c.path)}, false, nil } fds, err := cm.ExtractFDs() if err != nil { - return 0, 0, unix.ControlMessages{}, tcpip.FullAddress{}, false, translateError(err) + return 0, 0, transport.ControlMessages{}, tcpip.FullAddress{}, false, translateError(err) } if len(fds) == 0 { - return rl, ml, unix.ControlMessages{}, tcpip.FullAddress{Addr: tcpip.Address(c.path)}, false, nil + return rl, ml, transport.ControlMessages{}, tcpip.FullAddress{Addr: tcpip.Address(c.path)}, false, nil } return rl, ml, control.New(nil, nil, newSCMRights(fds)), tcpip.FullAddress{Addr: tcpip.Address(c.path)}, false, nil } @@ -339,17 +339,17 @@ func (c *ConnectedEndpoint) close() { c.file = nil } -// RecvNotify implements unix.Receiver.RecvNotify. +// RecvNotify implements transport.Receiver.RecvNotify. func (c *ConnectedEndpoint) RecvNotify() {} -// CloseRecv implements unix.Receiver.CloseRecv. +// CloseRecv implements transport.Receiver.CloseRecv. func (c *ConnectedEndpoint) CloseRecv() { c.mu.Lock() c.readClosed = true c.mu.Unlock() } -// Readable implements unix.Receiver.Readable. +// Readable implements transport.Receiver.Readable. func (c *ConnectedEndpoint) Readable() bool { c.mu.RLock() defer c.mu.RUnlock() @@ -359,33 +359,33 @@ func (c *ConnectedEndpoint) Readable() bool { return fdnotifier.NonBlockingPoll(int32(c.file.FD()), waiter.EventIn)&waiter.EventIn != 0 } -// SendQueuedSize implements unix.Receiver.SendQueuedSize. +// SendQueuedSize implements transport.Receiver.SendQueuedSize. func (c *ConnectedEndpoint) SendQueuedSize() int64 { // SendQueuedSize isn't supported for host sockets because we don't allow the // sentry to call ioctl(2). return -1 } -// RecvQueuedSize implements unix.Receiver.RecvQueuedSize. +// RecvQueuedSize implements transport.Receiver.RecvQueuedSize. func (c *ConnectedEndpoint) RecvQueuedSize() int64 { // RecvQueuedSize isn't supported for host sockets because we don't allow the // sentry to call ioctl(2). return -1 } -// SendMaxQueueSize implements unix.Receiver.SendMaxQueueSize. +// SendMaxQueueSize implements transport.Receiver.SendMaxQueueSize. func (c *ConnectedEndpoint) SendMaxQueueSize() int64 { return int64(c.sndbuf) } -// RecvMaxQueueSize implements unix.Receiver.RecvMaxQueueSize. +// RecvMaxQueueSize implements transport.Receiver.RecvMaxQueueSize. func (c *ConnectedEndpoint) RecvMaxQueueSize() int64 { // N.B. Unix sockets don't use the receive buffer. We'll claim it is // the same size as the send buffer. return int64(c.sndbuf) } -// Release implements unix.ConnectedEndpoint.Release and unix.Receiver.Release. +// Release implements transport.ConnectedEndpoint.Release and transport.Receiver.Release. func (c *ConnectedEndpoint) Release() { c.ref.DecRefWithDestructor(c.close) } diff --git a/pkg/sentry/fs/host/socket_test.go b/pkg/sentry/fs/host/socket_test.go index 8b752737d..1c6f9ddb1 100644 --- a/pkg/sentry/fs/host/socket_test.go +++ b/pkg/sentry/fs/host/socket_test.go @@ -22,20 +22,20 @@ import ( "gvisor.googlesource.com/gvisor/pkg/fd" "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) var ( - // Make sure that ConnectedEndpoint implements unix.ConnectedEndpoint. - _ = unix.ConnectedEndpoint(new(ConnectedEndpoint)) + // Make sure that ConnectedEndpoint implements transport.ConnectedEndpoint. + _ = transport.ConnectedEndpoint(new(ConnectedEndpoint)) - // Make sure that ConnectedEndpoint implements unix.Receiver. - _ = unix.Receiver(new(ConnectedEndpoint)) + // Make sure that ConnectedEndpoint implements transport.Receiver. + _ = transport.Receiver(new(ConnectedEndpoint)) ) func getFl(fd int) (uint32, error) { @@ -199,7 +199,7 @@ func TestListen(t *testing.T) { func TestSend(t *testing.T) { e := ConnectedEndpoint{writeClosed: true} - if _, _, err := e.Send(nil, unix.ControlMessages{}, tcpip.FullAddress{}); err != tcpip.ErrClosedForSend { + if _, _, err := e.Send(nil, transport.ControlMessages{}, tcpip.FullAddress{}); err != tcpip.ErrClosedForSend { t.Errorf("Got %#v.Send() = %v, want = %v", e, err, tcpip.ErrClosedForSend) } } diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index db7240dca..409c81a97 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -22,8 +22,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/lock" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) // Inode is a file system object that can be simultaneously referenced by different @@ -223,7 +223,7 @@ func (i *Inode) Rename(ctx context.Context, oldParent *Dirent, renamed *Dirent, } // Bind calls i.InodeOperations.Bind with i as the directory. -func (i *Inode) Bind(ctx context.Context, name string, data unix.BoundEndpoint, perm FilePermissions) (*Dirent, error) { +func (i *Inode) Bind(ctx context.Context, name string, data transport.BoundEndpoint, perm FilePermissions) (*Dirent, error) { if i.overlay != nil { return overlayBind(ctx, i.overlay, name, data, perm) } @@ -231,7 +231,7 @@ func (i *Inode) Bind(ctx context.Context, name string, data unix.BoundEndpoint, } // BoundEndpoint calls i.InodeOperations.BoundEndpoint with i as the Inode. -func (i *Inode) BoundEndpoint(path string) unix.BoundEndpoint { +func (i *Inode) BoundEndpoint(path string) transport.BoundEndpoint { if i.overlay != nil { return overlayBoundEndpoint(i.overlay, path) } diff --git a/pkg/sentry/fs/inode_operations.go b/pkg/sentry/fs/inode_operations.go index 952f9704d..3ee3de10e 100644 --- a/pkg/sentry/fs/inode_operations.go +++ b/pkg/sentry/fs/inode_operations.go @@ -20,8 +20,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -146,7 +146,7 @@ type InodeOperations interface { // Implementations must ensure that name does not already exist. // // The caller must ensure that this operation is permitted. - Bind(ctx context.Context, dir *Inode, name string, data unix.BoundEndpoint, perm FilePermissions) (*Dirent, error) + Bind(ctx context.Context, dir *Inode, name string, data transport.BoundEndpoint, perm FilePermissions) (*Dirent, error) // BoundEndpoint returns the socket endpoint at path stored in // or generated by an Inode. @@ -160,7 +160,7 @@ type InodeOperations interface { // generally implies that this Inode was created via CreateSocket. // // If there is no socket endpoint available, nil will be returned. - BoundEndpoint(inode *Inode, path string) unix.BoundEndpoint + BoundEndpoint(inode *Inode, path string) transport.BoundEndpoint // GetFile returns a new open File backed by a Dirent and FileFlags. // It may block as long as it is done with ctx. diff --git a/pkg/sentry/fs/inode_overlay.go b/pkg/sentry/fs/inode_overlay.go index e18e095a0..cf698a4da 100644 --- a/pkg/sentry/fs/inode_overlay.go +++ b/pkg/sentry/fs/inode_overlay.go @@ -20,8 +20,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) func overlayHasWhiteout(parent *Inode, name string) bool { @@ -356,7 +356,7 @@ func overlayRename(ctx context.Context, o *overlayEntry, oldParent *Dirent, rena return nil } -func overlayBind(ctx context.Context, o *overlayEntry, name string, data unix.BoundEndpoint, perm FilePermissions) (*Dirent, error) { +func overlayBind(ctx context.Context, o *overlayEntry, name string, data transport.BoundEndpoint, perm FilePermissions) (*Dirent, error) { o.copyMu.RLock() defer o.copyMu.RUnlock() // We do not support doing anything exciting with sockets unless there @@ -383,7 +383,7 @@ func overlayBind(ctx context.Context, o *overlayEntry, name string, data unix.Bo return NewDirent(newOverlayInode(ctx, entry, inode.MountSource), name), nil } -func overlayBoundEndpoint(o *overlayEntry, path string) unix.BoundEndpoint { +func overlayBoundEndpoint(o *overlayEntry, path string) transport.BoundEndpoint { o.copyMu.RLock() defer o.copyMu.RUnlock() diff --git a/pkg/sentry/fs/ramfs/BUILD b/pkg/sentry/fs/ramfs/BUILD index 5230157fe..a93ad6240 100644 --- a/pkg/sentry/fs/ramfs/BUILD +++ b/pkg/sentry/fs/ramfs/BUILD @@ -23,9 +23,9 @@ go_library( "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", "//pkg/sentry/safemem", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserror", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index d8333194b..075e13b01 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -20,9 +20,9 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) // CreateOps represents operations to create different file types. @@ -37,7 +37,7 @@ type CreateOps struct { NewSymlink func(ctx context.Context, dir *fs.Inode, target string) (*fs.Inode, error) // NewBoundEndpoint creates a new socket. - NewBoundEndpoint func(ctx context.Context, dir *fs.Inode, ep unix.BoundEndpoint, perms fs.FilePermissions) (*fs.Inode, error) + NewBoundEndpoint func(ctx context.Context, dir *fs.Inode, ep transport.BoundEndpoint, perms fs.FilePermissions) (*fs.Inode, error) // NewFifo creates a new fifo. NewFifo func(ctx context.Context, dir *fs.Inode, perm fs.FilePermissions) (*fs.Inode, error) @@ -314,7 +314,7 @@ func (d *Dir) CreateDirectory(ctx context.Context, dir *fs.Inode, name string, p } // Bind implements fs.InodeOperations.Bind. -func (d *Dir) Bind(ctx context.Context, dir *fs.Inode, name string, ep unix.BoundEndpoint, perms fs.FilePermissions) (*fs.Dirent, error) { +func (d *Dir) Bind(ctx context.Context, dir *fs.Inode, name string, ep transport.BoundEndpoint, perms fs.FilePermissions) (*fs.Dirent, error) { if d.CreateOps == nil || d.CreateOps.NewBoundEndpoint == nil { return nil, ErrDenied } diff --git a/pkg/sentry/fs/ramfs/ramfs.go b/pkg/sentry/fs/ramfs/ramfs.go index 1028b5f1d..83cbcab23 100644 --- a/pkg/sentry/fs/ramfs/ramfs.go +++ b/pkg/sentry/fs/ramfs/ramfs.go @@ -26,9 +26,9 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -279,7 +279,7 @@ func (*Entry) CreateDirectory(context.Context, *fs.Inode, string, fs.FilePermiss } // Bind is not supported by default. -func (*Entry) Bind(context.Context, *fs.Inode, string, unix.BoundEndpoint, fs.FilePermissions) (*fs.Dirent, error) { +func (*Entry) Bind(context.Context, *fs.Inode, string, transport.BoundEndpoint, fs.FilePermissions) (*fs.Dirent, error) { return nil, ErrInvalidOp } diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index 93427a1ff..9ac00eb18 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -17,7 +17,7 @@ package ramfs import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" ) // Socket represents a socket. @@ -27,17 +27,17 @@ type Socket struct { Entry // ep is the bound endpoint. - ep unix.BoundEndpoint + ep transport.BoundEndpoint } // InitSocket initializes a socket. -func (s *Socket) InitSocket(ctx context.Context, ep unix.BoundEndpoint, owner fs.FileOwner, perms fs.FilePermissions) { +func (s *Socket) InitSocket(ctx context.Context, ep transport.BoundEndpoint, owner fs.FileOwner, perms fs.FilePermissions) { s.InitEntry(ctx, owner, perms) s.ep = ep } // BoundEndpoint returns the socket data. -func (s *Socket) BoundEndpoint(*fs.Inode, string) unix.BoundEndpoint { +func (s *Socket) BoundEndpoint(*fs.Inode, string) transport.BoundEndpoint { // ramfs only supports stored sentry internal sockets. Only gofer sockets // care about the path argument. return s.ep diff --git a/pkg/sentry/fs/tmpfs/BUILD b/pkg/sentry/fs/tmpfs/BUILD index cfe11ab02..277583113 100644 --- a/pkg/sentry/fs/tmpfs/BUILD +++ b/pkg/sentry/fs/tmpfs/BUILD @@ -25,9 +25,9 @@ go_library( "//pkg/sentry/memmap", "//pkg/sentry/platform", "//pkg/sentry/safemem", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 10cb5451d..38be6db46 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -22,9 +22,9 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/pipe" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) var fsInfo = fs.Info{ @@ -104,7 +104,7 @@ func (d *Dir) newCreateOps() *ramfs.CreateOps { NewSymlink: func(ctx context.Context, dir *fs.Inode, target string) (*fs.Inode, error) { return NewSymlink(ctx, target, fs.FileOwnerFromContext(ctx), dir.MountSource), nil }, - NewBoundEndpoint: func(ctx context.Context, dir *fs.Inode, socket unix.BoundEndpoint, perms fs.FilePermissions) (*fs.Inode, error) { + NewBoundEndpoint: func(ctx context.Context, dir *fs.Inode, socket transport.BoundEndpoint, perms fs.FilePermissions) (*fs.Inode, error) { return NewSocket(ctx, socket, fs.FileOwnerFromContext(ctx), perms, dir.MountSource), nil }, NewFifo: func(ctx context.Context, dir *fs.Inode, perms fs.FilePermissions) (*fs.Inode, error) { @@ -160,7 +160,7 @@ type Socket struct { } // NewSocket returns a new socket with the provided permissions. -func NewSocket(ctx context.Context, socket unix.BoundEndpoint, owner fs.FileOwner, perms fs.FilePermissions, msrc *fs.MountSource) *fs.Inode { +func NewSocket(ctx context.Context, socket transport.BoundEndpoint, owner fs.FileOwner, perms fs.FilePermissions, msrc *fs.MountSource) *fs.Inode { s := &Socket{} s.InitSocket(ctx, socket, owner, perms) return fs.NewInode(s, msrc, fs.StableAttr{ diff --git a/pkg/sentry/fs/tty/BUILD b/pkg/sentry/fs/tty/BUILD index 3c446eef4..d4dd20e30 100644 --- a/pkg/sentry/fs/tty/BUILD +++ b/pkg/sentry/fs/tty/BUILD @@ -26,9 +26,9 @@ go_library( "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserror", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index c6f39fce3..7c0c0b0c1 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -26,9 +26,9 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -215,7 +215,7 @@ func (d *dirInodeOperations) RemoveDirectory(ctx context.Context, dir *fs.Inode, } // Bind implements fs.InodeOperations.Bind. -func (d *dirInodeOperations) Bind(ctx context.Context, dir *fs.Inode, name string, data unix.BoundEndpoint, perm fs.FilePermissions) (*fs.Dirent, error) { +func (d *dirInodeOperations) Bind(ctx context.Context, dir *fs.Inode, name string, data transport.BoundEndpoint, perm fs.FilePermissions) (*fs.Dirent, error) { return nil, syserror.EPERM } diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 31ad96612..acc61cb09 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -156,6 +156,7 @@ go_library( "//pkg/sentry/platform", "//pkg/sentry/safemem", "//pkg/sentry/socket/netlink/port", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/time", "//pkg/sentry/uniqueid", "//pkg/sentry/usage", @@ -166,7 +167,6 @@ go_library( "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/stack", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/kernel/abstract_socket_namespace.go b/pkg/sentry/kernel/abstract_socket_namespace.go index d6d1d341d..45088c988 100644 --- a/pkg/sentry/kernel/abstract_socket_namespace.go +++ b/pkg/sentry/kernel/abstract_socket_namespace.go @@ -19,12 +19,12 @@ import ( "syscall" "gvisor.googlesource.com/gvisor/pkg/refs" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" ) // +stateify savable type abstractEndpoint struct { - ep unix.BoundEndpoint + ep transport.BoundEndpoint wr *refs.WeakRef name string ns *AbstractSocketNamespace @@ -56,14 +56,14 @@ func NewAbstractSocketNamespace() *AbstractSocketNamespace { } } -// A boundEndpoint wraps a unix.BoundEndpoint to maintain a reference on its -// backing object. +// A boundEndpoint wraps a transport.BoundEndpoint to maintain a reference on +// its backing object. type boundEndpoint struct { - unix.BoundEndpoint + transport.BoundEndpoint rc refs.RefCounter } -// Release implements unix.BoundEndpoint.Release. +// Release implements transport.BoundEndpoint.Release. func (e *boundEndpoint) Release() { e.rc.DecRef() e.BoundEndpoint.Release() @@ -71,7 +71,7 @@ func (e *boundEndpoint) Release() { // BoundEndpoint retrieves the endpoint bound to the given name. The return // value is nil if no endpoint was bound. -func (a *AbstractSocketNamespace) BoundEndpoint(name string) unix.BoundEndpoint { +func (a *AbstractSocketNamespace) BoundEndpoint(name string) transport.BoundEndpoint { a.mu.Lock() defer a.mu.Unlock() @@ -93,7 +93,7 @@ func (a *AbstractSocketNamespace) BoundEndpoint(name string) unix.BoundEndpoint // // When the last reference managed by rc is dropped, ep may be removed from the // namespace. -func (a *AbstractSocketNamespace) Bind(name string, ep unix.BoundEndpoint, rc refs.RefCounter) error { +func (a *AbstractSocketNamespace) Bind(name string, ep transport.BoundEndpoint, rc refs.RefCounter) error { a.mu.Lock() defer a.mu.Unlock() diff --git a/pkg/sentry/socket/BUILD b/pkg/sentry/socket/BUILD index a320fca0b..3a8044b5f 100644 --- a/pkg/sentry/socket/BUILD +++ b/pkg/sentry/socket/BUILD @@ -16,9 +16,9 @@ go_library( "//pkg/sentry/kernel", "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/time", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/tcpip", - "//pkg/tcpip/transport/unix", ], ) diff --git a/pkg/sentry/socket/control/BUILD b/pkg/sentry/socket/control/BUILD index c4874fdfb..d3a63f15f 100644 --- a/pkg/sentry/socket/control/BUILD +++ b/pkg/sentry/socket/control/BUILD @@ -18,8 +18,8 @@ go_library( "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/kdefs", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserror", - "//pkg/tcpip/transport/unix", ], ) diff --git a/pkg/sentry/socket/control/control.go b/pkg/sentry/socket/control/control.go index c31182e69..db97e95f2 100644 --- a/pkg/sentry/socket/control/control.go +++ b/pkg/sentry/socket/control/control.go @@ -24,16 +24,16 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) const maxInt = int(^uint(0) >> 1) // SCMCredentials represents a SCM_CREDENTIALS socket control message. type SCMCredentials interface { - unix.CredentialsControlMessage + transport.CredentialsControlMessage // Credentials returns properly namespaced values for the sender's pid, uid // and gid. @@ -42,7 +42,7 @@ type SCMCredentials interface { // SCMRights represents a SCM_RIGHTS socket control message. type SCMRights interface { - unix.RightsControlMessage + transport.RightsControlMessage // Files returns up to max RightsFiles. Files(ctx context.Context, max int) RightsFiles @@ -81,8 +81,8 @@ func (fs *RightsFiles) Files(ctx context.Context, max int) RightsFiles { return rf } -// Clone implements unix.RightsControlMessage.Clone. -func (fs *RightsFiles) Clone() unix.RightsControlMessage { +// Clone implements transport.RightsControlMessage.Clone. +func (fs *RightsFiles) Clone() transport.RightsControlMessage { nfs := append(RightsFiles(nil), *fs...) for _, nf := range nfs { nf.IncRef() @@ -90,7 +90,7 @@ func (fs *RightsFiles) Clone() unix.RightsControlMessage { return &nfs } -// Release implements unix.RightsControlMessage.Release. +// Release implements transport.RightsControlMessage.Release. func (fs *RightsFiles) Release() { for _, f := range *fs { f.DecRef() @@ -156,8 +156,8 @@ func NewSCMCredentials(t *kernel.Task, cred linux.ControlMessageCredentials) (SC return &scmCredentials{t, kuid, kgid}, nil } -// Equals implements unix.CredentialsControlMessage.Equals. -func (c *scmCredentials) Equals(oc unix.CredentialsControlMessage) bool { +// Equals implements transport.CredentialsControlMessage.Equals. +func (c *scmCredentials) Equals(oc transport.CredentialsControlMessage) bool { if oc, _ := oc.(*scmCredentials); oc != nil && *c == *oc { return true } @@ -301,7 +301,7 @@ func PackTimestamp(t *kernel.Task, timestamp int64, buf []byte) []byte { } // Parse parses a raw socket control message into portable objects. -func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (unix.ControlMessages, error) { +func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (transport.ControlMessages, error) { var ( fds linux.ControlMessageRights @@ -311,20 +311,20 @@ func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (unix.Contr for i := 0; i < len(buf); { if i+linux.SizeOfControlMessageHeader > len(buf) { - return unix.ControlMessages{}, syserror.EINVAL + return transport.ControlMessages{}, syserror.EINVAL } var h linux.ControlMessageHeader binary.Unmarshal(buf[i:i+linux.SizeOfControlMessageHeader], usermem.ByteOrder, &h) if h.Length < uint64(linux.SizeOfControlMessageHeader) { - return unix.ControlMessages{}, syserror.EINVAL + return transport.ControlMessages{}, syserror.EINVAL } if h.Length > uint64(len(buf)-i) { - return unix.ControlMessages{}, syserror.EINVAL + return transport.ControlMessages{}, syserror.EINVAL } if h.Level != linux.SOL_SOCKET { - return unix.ControlMessages{}, syserror.EINVAL + return transport.ControlMessages{}, syserror.EINVAL } i += linux.SizeOfControlMessageHeader @@ -340,7 +340,7 @@ func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (unix.Contr numRights := rightsSize / linux.SizeOfControlMessageRight if len(fds)+numRights > linux.SCM_MAX_FD { - return unix.ControlMessages{}, syserror.EINVAL + return transport.ControlMessages{}, syserror.EINVAL } for j := i; j < i+rightsSize; j += linux.SizeOfControlMessageRight { @@ -351,7 +351,7 @@ func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (unix.Contr case linux.SCM_CREDENTIALS: if length < linux.SizeOfControlMessageCredentials { - return unix.ControlMessages{}, syserror.EINVAL + return transport.ControlMessages{}, syserror.EINVAL } binary.Unmarshal(buf[i:i+linux.SizeOfControlMessageCredentials], usermem.ByteOrder, &creds) @@ -360,7 +360,7 @@ func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (unix.Contr default: // Unknown message type. - return unix.ControlMessages{}, syserror.EINVAL + return transport.ControlMessages{}, syserror.EINVAL } } @@ -368,7 +368,7 @@ func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (unix.Contr if haveCreds { var err error if credentials, err = NewSCMCredentials(t, creds); err != nil { - return unix.ControlMessages{}, err + return transport.ControlMessages{}, err } } else { credentials = makeCreds(t, socketOrEndpoint) @@ -378,22 +378,22 @@ func Parse(t *kernel.Task, socketOrEndpoint interface{}, buf []byte) (unix.Contr if len(fds) > 0 { var err error if rights, err = NewSCMRights(t, fds); err != nil { - return unix.ControlMessages{}, err + return transport.ControlMessages{}, err } } if credentials == nil && rights == nil { - return unix.ControlMessages{}, nil + return transport.ControlMessages{}, nil } - return unix.ControlMessages{Credentials: credentials, Rights: rights}, nil + return transport.ControlMessages{Credentials: credentials, Rights: rights}, nil } func makeCreds(t *kernel.Task, socketOrEndpoint interface{}) SCMCredentials { if t == nil || socketOrEndpoint == nil { return nil } - if cr, ok := socketOrEndpoint.(unix.Credentialer); ok && (cr.Passcred() || cr.ConnectedPasscred()) { + if cr, ok := socketOrEndpoint.(transport.Credentialer); ok && (cr.Passcred() || cr.ConnectedPasscred()) { tcred := t.Credentials() return &scmCredentials{t, tcred.EffectiveKUID, tcred.EffectiveKGID} } @@ -401,8 +401,8 @@ func makeCreds(t *kernel.Task, socketOrEndpoint interface{}) SCMCredentials { } // New creates default control messages if needed. -func New(t *kernel.Task, socketOrEndpoint interface{}, rights SCMRights) unix.ControlMessages { - return unix.ControlMessages{ +func New(t *kernel.Task, socketOrEndpoint interface{}, rights SCMRights) transport.ControlMessages { + return transport.ControlMessages{ Credentials: makeCreds(t, socketOrEndpoint), Rights: rights, } diff --git a/pkg/sentry/socket/epsocket/BUILD b/pkg/sentry/socket/epsocket/BUILD index 7f9ea9edc..dbabc931c 100644 --- a/pkg/sentry/socket/epsocket/BUILD +++ b/pkg/sentry/socket/epsocket/BUILD @@ -31,6 +31,7 @@ go_library( "//pkg/sentry/kernel/time", "//pkg/sentry/safemem", "//pkg/sentry/socket", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/syserror", @@ -42,7 +43,6 @@ go_library( "//pkg/tcpip/stack", "//pkg/tcpip/transport/tcp", "//pkg/tcpip/transport/udp", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index a44679f0b..47c575e7b 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -44,13 +44,13 @@ import ( ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" "gvisor.googlesource.com/gvisor/pkg/tcpip/stack" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -108,26 +108,26 @@ func htons(v uint16) uint16 { } // commonEndpoint represents the intersection of a tcpip.Endpoint and a -// unix.Endpoint. +// transport.Endpoint. type commonEndpoint interface { // GetLocalAddress implements tcpip.Endpoint.GetLocalAddress and - // unix.Endpoint.GetLocalAddress. + // transport.Endpoint.GetLocalAddress. GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) // GetRemoteAddress implements tcpip.Endpoint.GetRemoteAddress and - // unix.Endpoint.GetRemoteAddress. + // transport.Endpoint.GetRemoteAddress. GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) // Readiness implements tcpip.Endpoint.Readiness and - // unix.Endpoint.Readiness. + // transport.Endpoint.Readiness. Readiness(mask waiter.EventMask) waiter.EventMask // SetSockOpt implements tcpip.Endpoint.SetSockOpt and - // unix.Endpoint.SetSockOpt. + // transport.Endpoint.SetSockOpt. SetSockOpt(interface{}) *tcpip.Error // GetSockOpt implements tcpip.Endpoint.GetSockOpt and - // unix.Endpoint.GetSockOpt. + // transport.Endpoint.GetSockOpt. GetSockOpt(interface{}) *tcpip.Error } @@ -146,7 +146,7 @@ type SocketOperations struct { family int Endpoint tcpip.Endpoint - skType unix.SockType + skType transport.SockType // readMu protects access to readView, control, and sender. readMu sync.Mutex `state:"nosave"` @@ -156,7 +156,7 @@ type SocketOperations struct { } // New creates a new endpoint socket. -func New(t *kernel.Task, family int, skType unix.SockType, queue *waiter.Queue, endpoint tcpip.Endpoint) *fs.File { +func New(t *kernel.Task, family int, skType transport.SockType, queue *waiter.Queue, endpoint tcpip.Endpoint) *fs.File { dirent := socket.NewDirent(t, epsocketDevice) defer dirent.DecRef() return fs.NewFile(t, dirent, fs.FileFlags{Read: true, Write: true}, &SocketOperations{ @@ -502,7 +502,7 @@ func (s *SocketOperations) GetSockOpt(t *kernel.Task, level, name, outLen int) ( // GetSockOpt can be used to implement the linux syscall getsockopt(2) for // sockets backed by a commonEndpoint. -func GetSockOpt(t *kernel.Task, s socket.Socket, ep commonEndpoint, family int, skType unix.SockType, level, name, outLen int) (interface{}, *syserr.Error) { +func GetSockOpt(t *kernel.Task, s socket.Socket, ep commonEndpoint, family int, skType transport.SockType, level, name, outLen int) (interface{}, *syserr.Error) { switch level { case linux.SOL_SOCKET: switch name { diff --git a/pkg/sentry/socket/epsocket/provider.go b/pkg/sentry/socket/epsocket/provider.go index 6c1e3b6b9..dbc232d26 100644 --- a/pkg/sentry/socket/epsocket/provider.go +++ b/pkg/sentry/socket/epsocket/provider.go @@ -21,6 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/tcpip/header" @@ -28,7 +29,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/tcpip/network/ipv6" "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/tcp" "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/udp" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -40,7 +40,7 @@ type provider struct { // GetTransportProtocol figures out transport protocol. Currently only TCP, // UDP, and ICMP are supported. -func GetTransportProtocol(stype unix.SockType, protocol int) (tcpip.TransportProtocolNumber, *syserr.Error) { +func GetTransportProtocol(stype transport.SockType, protocol int) (tcpip.TransportProtocolNumber, *syserr.Error) { switch stype { case linux.SOCK_STREAM: if protocol != 0 && protocol != syscall.IPPROTO_TCP { @@ -62,7 +62,7 @@ func GetTransportProtocol(stype unix.SockType, protocol int) (tcpip.TransportPro } // Socket creates a new socket object for the AF_INET or AF_INET6 family. -func (p *provider) Socket(t *kernel.Task, stype unix.SockType, protocol int) (*fs.File, *syserr.Error) { +func (p *provider) Socket(t *kernel.Task, stype transport.SockType, protocol int) (*fs.File, *syserr.Error) { // Fail right away if we don't have a stack. stack := t.NetworkContext() if stack == nil { @@ -92,7 +92,7 @@ func (p *provider) Socket(t *kernel.Task, stype unix.SockType, protocol int) (*f } // Pair just returns nil sockets (not supported). -func (*provider) Pair(*kernel.Task, unix.SockType, int) (*fs.File, *fs.File, *syserr.Error) { +func (*provider) Pair(*kernel.Task, transport.SockType, int) (*fs.File, *fs.File, *syserr.Error) { return nil, nil, nil } diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD index d623718b3..c30220a46 100644 --- a/pkg/sentry/socket/hostinet/BUILD +++ b/pkg/sentry/socket/hostinet/BUILD @@ -29,10 +29,10 @@ go_library( "//pkg/sentry/kernel/time", "//pkg/sentry/safemem", "//pkg/sentry/socket", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/syserror", - "//pkg/tcpip/transport/unix", "//pkg/waiter", "//pkg/waiter/fdnotifier", ], diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index d0f3054dc..e82624b44 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -27,10 +27,10 @@ import ( ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) @@ -511,7 +511,7 @@ type socketProvider struct { } // Socket implements socket.Provider.Socket. -func (p *socketProvider) Socket(t *kernel.Task, stypeflags unix.SockType, protocol int) (*fs.File, *syserr.Error) { +func (p *socketProvider) Socket(t *kernel.Task, stypeflags transport.SockType, protocol int) (*fs.File, *syserr.Error) { // Check that we are using the host network stack. stack := t.NetworkContext() if stack == nil { @@ -553,7 +553,7 @@ func (p *socketProvider) Socket(t *kernel.Task, stypeflags unix.SockType, protoc } // Pair implements socket.Provider.Pair. -func (p *socketProvider) Pair(t *kernel.Task, stype unix.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) { +func (p *socketProvider) Pair(t *kernel.Task, stype transport.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) { // Not supported by AF_INET/AF_INET6. return nil, nil, nil } diff --git a/pkg/sentry/socket/netlink/BUILD b/pkg/sentry/socket/netlink/BUILD index b852165f7..cff922cb8 100644 --- a/pkg/sentry/socket/netlink/BUILD +++ b/pkg/sentry/socket/netlink/BUILD @@ -25,11 +25,11 @@ go_library( "//pkg/sentry/socket", "//pkg/sentry/socket/netlink/port", "//pkg/sentry/socket/unix", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/socket/netlink/provider.go b/pkg/sentry/socket/netlink/provider.go index e874216f4..5d0a04a07 100644 --- a/pkg/sentry/socket/netlink/provider.go +++ b/pkg/sentry/socket/netlink/provider.go @@ -22,8 +22,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/syserr" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) // Protocol is the implementation of a netlink socket protocol. @@ -66,10 +66,10 @@ type socketProvider struct { } // Socket implements socket.Provider.Socket. -func (*socketProvider) Socket(t *kernel.Task, stype unix.SockType, protocol int) (*fs.File, *syserr.Error) { +func (*socketProvider) Socket(t *kernel.Task, stype transport.SockType, protocol int) (*fs.File, *syserr.Error) { // Netlink sockets must be specified as datagram or raw, but they // behave the same regardless of type. - if stype != unix.SockDgram && stype != unix.SockRaw { + if stype != transport.SockDgram && stype != transport.SockRaw { return nil, syserr.ErrSocketNotSupported } @@ -94,7 +94,7 @@ func (*socketProvider) Socket(t *kernel.Task, stype unix.SockType, protocol int) } // Pair implements socket.Provider.Pair by returning an error. -func (*socketProvider) Pair(*kernel.Task, unix.SockType, int) (*fs.File, *fs.File, *syserr.Error) { +func (*socketProvider) Pair(*kernel.Task, transport.SockType, int) (*fs.File, *fs.File, *syserr.Error) { // Netlink sockets never supports creating socket pairs. return nil, nil, syserr.ErrNotSupported } diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index f3b2c7256..0c03997f2 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -31,12 +31,12 @@ import ( ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/port" - sunix "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -80,11 +80,11 @@ type Socket struct { // ep is a datagram unix endpoint used to buffer messages sent from the // kernel to userspace. RecvMsg reads messages from this endpoint. - ep unix.Endpoint + ep transport.Endpoint // connection is the kernel's connection to ep, used to write messages // sent to userspace. - connection unix.ConnectedEndpoint + connection transport.ConnectedEndpoint // mu protects the fields below. mu sync.Mutex `state:"nosave"` @@ -105,7 +105,7 @@ var _ socket.Socket = (*Socket)(nil) // NewSocket creates a new Socket. func NewSocket(t *kernel.Task, protocol Protocol) (*Socket, *syserr.Error) { // Datagram endpoint used to buffer kernel -> user messages. - ep := unix.NewConnectionless() + ep := transport.NewConnectionless() // Bind the endpoint for good measure so we can connect to it. The // bound address will never be exposed. @@ -115,7 +115,7 @@ func NewSocket(t *kernel.Task, protocol Protocol) (*Socket, *syserr.Error) { } // Create a connection from which the kernel can write messages. - connection, terr := ep.(unix.BoundEndpoint).UnidirectionalConnect() + connection, terr := ep.(transport.BoundEndpoint).UnidirectionalConnect() if terr != nil { ep.Close() return nil, syserr.TranslateNetstackError(terr) @@ -368,7 +368,7 @@ func (s *Socket) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, have trunc := flags&linux.MSG_TRUNC != 0 - r := sunix.EndpointReader{ + r := unix.EndpointReader{ Endpoint: s.ep, Peek: flags&linux.MSG_PEEK != 0, } @@ -408,7 +408,7 @@ func (s *Socket) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, _ if dst.NumBytes() == 0 { return 0, nil } - return dst.CopyOutFrom(ctx, &sunix.EndpointReader{ + return dst.CopyOutFrom(ctx, &unix.EndpointReader{ Endpoint: s.ep, }) } @@ -424,7 +424,7 @@ func (s *Socket) sendResponse(ctx context.Context, ms *MessageSet) *syserr.Error if len(bufs) > 0 { // RecvMsg never receives the address, so we don't need to send // one. - _, notify, terr := s.connection.Send(bufs, unix.ControlMessages{}, tcpip.FullAddress{}) + _, notify, terr := s.connection.Send(bufs, transport.ControlMessages{}, tcpip.FullAddress{}) // If the buffer is full, we simply drop messages, just like // Linux. if terr != nil && terr != tcpip.ErrWouldBlock { @@ -448,7 +448,7 @@ func (s *Socket) sendResponse(ctx context.Context, ms *MessageSet) *syserr.Error PortID: uint32(ms.PortID), }) - _, notify, terr := s.connection.Send([][]byte{m.Finalize()}, unix.ControlMessages{}, tcpip.FullAddress{}) + _, notify, terr := s.connection.Send([][]byte{m.Finalize()}, transport.ControlMessages{}, tcpip.FullAddress{}) if terr != nil && terr != tcpip.ErrWouldBlock { return syserr.TranslateNetstackError(terr) } diff --git a/pkg/sentry/socket/rpcinet/BUILD b/pkg/sentry/socket/rpcinet/BUILD index 288199779..3ea433360 100644 --- a/pkg/sentry/socket/rpcinet/BUILD +++ b/pkg/sentry/socket/rpcinet/BUILD @@ -31,12 +31,12 @@ go_library( "//pkg/sentry/socket/hostinet", "//pkg/sentry/socket/rpcinet/conn", "//pkg/sentry/socket/rpcinet/notifier", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/buffer", - "//pkg/tcpip/transport/unix", "//pkg/unet", "//pkg/waiter", ], diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index 72fa1ca8f..c7e761d54 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -31,12 +31,12 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/socket/rpcinet/conn" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/rpcinet/notifier" pb "gvisor.googlesource.com/gvisor/pkg/sentry/socket/rpcinet/syscall_rpc_go_proto" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -763,7 +763,7 @@ type socketProvider struct { } // Socket implements socket.Provider.Socket. -func (p *socketProvider) Socket(t *kernel.Task, stypeflags unix.SockType, protocol int) (*fs.File, *syserr.Error) { +func (p *socketProvider) Socket(t *kernel.Task, stypeflags transport.SockType, protocol int) (*fs.File, *syserr.Error) { // Check that we are using the RPC network stack. stack := t.NetworkContext() if stack == nil { @@ -803,7 +803,7 @@ func (p *socketProvider) Socket(t *kernel.Task, stypeflags unix.SockType, protoc } // Pair implements socket.Provider.Pair. -func (p *socketProvider) Pair(t *kernel.Task, stype unix.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) { +func (p *socketProvider) Pair(t *kernel.Task, stype transport.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) { // Not supported by AF_INET/AF_INET6. return nil, nil, nil } diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index 54fe64595..31f8d42d7 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -29,16 +29,16 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) // ControlMessages represents the union of unix control messages and tcpip // control messages. type ControlMessages struct { - Unix unix.ControlMessages + Unix transport.ControlMessages IP tcpip.ControlMessages } @@ -109,12 +109,12 @@ type Provider interface { // If a nil Socket _and_ a nil error is returned, it means that the // protocol is not supported. A non-nil error should only be returned // if the protocol is supported, but an error occurs during creation. - Socket(t *kernel.Task, stype unix.SockType, protocol int) (*fs.File, *syserr.Error) + Socket(t *kernel.Task, stype transport.SockType, protocol int) (*fs.File, *syserr.Error) // Pair creates a pair of connected sockets. // // See Socket for error information. - Pair(t *kernel.Task, stype unix.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) + Pair(t *kernel.Task, stype transport.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) } // families holds a map of all known address families and their providers. @@ -128,7 +128,7 @@ func RegisterProvider(family int, provider Provider) { } // New creates a new socket with the given family, type and protocol. -func New(t *kernel.Task, family int, stype unix.SockType, protocol int) (*fs.File, *syserr.Error) { +func New(t *kernel.Task, family int, stype transport.SockType, protocol int) (*fs.File, *syserr.Error) { for _, p := range families[family] { s, err := p.Socket(t, stype, protocol) if err != nil { @@ -144,7 +144,7 @@ func New(t *kernel.Task, family int, stype unix.SockType, protocol int) (*fs.Fil // Pair creates a new connected socket pair with the given family, type and // protocol. -func Pair(t *kernel.Task, family int, stype unix.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) { +func Pair(t *kernel.Task, family int, stype transport.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) { providers, ok := families[family] if !ok { return nil, nil, syserr.ErrAddressFamilyNotSupported diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index 9fe681e9a..a12fa93db 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -26,11 +26,11 @@ go_library( "//pkg/sentry/socket", "//pkg/sentry/socket/control", "//pkg/sentry/socket/epsocket", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/socket/unix/io.go b/pkg/sentry/socket/unix/io.go index 0ca2e35d0..06333e14b 100644 --- a/pkg/sentry/socket/unix/io.go +++ b/pkg/sentry/socket/unix/io.go @@ -16,23 +16,23 @@ package unix import ( "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) -// EndpointWriter implements safemem.Writer that writes to a unix.Endpoint. +// EndpointWriter implements safemem.Writer that writes to a transport.Endpoint. // // EndpointWriter is not thread-safe. type EndpointWriter struct { - // Endpoint is the unix.Endpoint to write to. - Endpoint unix.Endpoint + // Endpoint is the transport.Endpoint to write to. + Endpoint transport.Endpoint // Control is the control messages to send. - Control unix.ControlMessages + Control transport.ControlMessages // To is the endpoint to send to. May be nil. - To unix.BoundEndpoint + To transport.BoundEndpoint } // WriteFromBlocks implements safemem.Writer.WriteFromBlocks. @@ -46,12 +46,13 @@ func (w *EndpointWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error) }}.WriteFromBlocks(srcs) } -// EndpointReader implements safemem.Reader that reads from a unix.Endpoint. +// EndpointReader implements safemem.Reader that reads from a +// transport.Endpoint. // // EndpointReader is not thread-safe. type EndpointReader struct { - // Endpoint is the unix.Endpoint to read from. - Endpoint unix.Endpoint + // Endpoint is the transport.Endpoint to read from. + Endpoint transport.Endpoint // Creds indicates if credential control messages are requested. Creds bool @@ -71,7 +72,7 @@ type EndpointReader struct { From *tcpip.FullAddress // Control contains the received control messages. - Control unix.ControlMessages + Control transport.ControlMessages } // ReadToBlocks implements safemem.Reader.ReadToBlocks. diff --git a/pkg/sentry/socket/unix/transport/BUILD b/pkg/sentry/socket/unix/transport/BUILD new file mode 100644 index 000000000..04ef0d438 --- /dev/null +++ b/pkg/sentry/socket/unix/transport/BUILD @@ -0,0 +1,22 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("//tools/go_stateify:defs.bzl", "go_library") + +go_library( + name = "transport", + srcs = [ + "connectioned.go", + "connectioned_state.go", + "connectionless.go", + "unix.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport", + visibility = ["//:sandbox"], + deps = [ + "//pkg/ilist", + "//pkg/sentry/socket/unix/transport/queue", + "//pkg/tcpip", + "//pkg/tcpip/buffer", + "//pkg/waiter", + ], +) diff --git a/pkg/sentry/socket/unix/transport/connectioned.go b/pkg/sentry/socket/unix/transport/connectioned.go new file mode 100644 index 000000000..f09935765 --- /dev/null +++ b/pkg/sentry/socket/unix/transport/connectioned.go @@ -0,0 +1,454 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package transport + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport/queue" + "gvisor.googlesource.com/gvisor/pkg/tcpip" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// UniqueIDProvider generates a sequence of unique identifiers useful for, +// among other things, lock ordering. +type UniqueIDProvider interface { + // UniqueID returns a new unique identifier. + UniqueID() uint64 +} + +// A ConnectingEndpoint is a connectioned unix endpoint that is attempting to +// establish a bidirectional connection with a BoundEndpoint. +type ConnectingEndpoint interface { + // ID returns the endpoint's globally unique identifier. This identifier + // must be used to determine locking order if more than one endpoint is + // to be locked in the same codepath. The endpoint with the smaller + // identifier must be locked before endpoints with larger identifiers. + ID() uint64 + + // Passcred implements socket.Credentialer.Passcred. + Passcred() bool + + // Type returns the socket type, typically either SockStream or + // SockSeqpacket. The connection attempt must be aborted if this + // value doesn't match the ConnectableEndpoint's type. + Type() SockType + + // GetLocalAddress returns the bound path. + GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) + + // Locker protects the following methods. While locked, only the holder of + // the lock can change the return value of the protected methods. + sync.Locker + + // Connected returns true iff the ConnectingEndpoint is in the connected + // state. ConnectingEndpoints can only be connected to a single endpoint, + // so the connection attempt must be aborted if this returns true. + Connected() bool + + // Listening returns true iff the ConnectingEndpoint is in the listening + // state. ConnectingEndpoints cannot make connections while listening, so + // the connection attempt must be aborted if this returns true. + Listening() bool + + // WaiterQueue returns a pointer to the endpoint's waiter queue. + WaiterQueue() *waiter.Queue +} + +// connectionedEndpoint is a Unix-domain connected or connectable endpoint and implements +// ConnectingEndpoint, ConnectableEndpoint and tcpip.Endpoint. +// +// connectionedEndpoints must be in connected state in order to transfer data. +// +// This implementation includes STREAM and SEQPACKET Unix sockets created with +// socket(2), accept(2) or socketpair(2) and dgram unix sockets created with +// socketpair(2). See unix_connectionless.go for the implementation of DGRAM +// Unix sockets created with socket(2). +// +// The state is much simpler than a TCP endpoint, so it is not encoded +// explicitly. Instead we enforce the following invariants: +// +// receiver != nil, connected != nil => connected. +// path != "" && acceptedChan == nil => bound, not listening. +// path != "" && acceptedChan != nil => bound and listening. +// +// Only one of these will be true at any moment. +// +// +stateify savable +type connectionedEndpoint struct { + baseEndpoint + + // id is the unique endpoint identifier. This is used exclusively for + // lock ordering within connect. + id uint64 + + // idGenerator is used to generate new unique endpoint identifiers. + idGenerator UniqueIDProvider + + // stype is used by connecting sockets to ensure that they are the + // same type. The value is typically either tcpip.SockSeqpacket or + // tcpip.SockStream. + stype SockType + + // acceptedChan is per the TCP endpoint implementation. Note that the + // sockets in this channel are _already in the connected state_, and + // have another associated connectionedEndpoint. + // + // If nil, then no listen call has been made. + acceptedChan chan *connectionedEndpoint `state:".([]*connectionedEndpoint)"` +} + +// NewConnectioned creates a new unbound connectionedEndpoint. +func NewConnectioned(stype SockType, uid UniqueIDProvider) Endpoint { + return &connectionedEndpoint{ + baseEndpoint: baseEndpoint{Queue: &waiter.Queue{}}, + id: uid.UniqueID(), + idGenerator: uid, + stype: stype, + } +} + +// NewPair allocates a new pair of connected unix-domain connectionedEndpoints. +func NewPair(stype SockType, uid UniqueIDProvider) (Endpoint, Endpoint) { + a := &connectionedEndpoint{ + baseEndpoint: baseEndpoint{Queue: &waiter.Queue{}}, + id: uid.UniqueID(), + idGenerator: uid, + stype: stype, + } + b := &connectionedEndpoint{ + baseEndpoint: baseEndpoint{Queue: &waiter.Queue{}}, + id: uid.UniqueID(), + idGenerator: uid, + stype: stype, + } + + q1 := queue.New(a.Queue, b.Queue, initialLimit) + q2 := queue.New(b.Queue, a.Queue, initialLimit) + + if stype == SockStream { + a.receiver = &streamQueueReceiver{queueReceiver: queueReceiver{q1}} + b.receiver = &streamQueueReceiver{queueReceiver: queueReceiver{q2}} + } else { + a.receiver = &queueReceiver{q1} + b.receiver = &queueReceiver{q2} + } + + a.connected = &connectedEndpoint{ + endpoint: b, + writeQueue: q2, + } + b.connected = &connectedEndpoint{ + endpoint: a, + writeQueue: q1, + } + + return a, b +} + +// NewExternal creates a new externally backed Endpoint. It behaves like a +// socketpair. +func NewExternal(stype SockType, uid UniqueIDProvider, queue *waiter.Queue, receiver Receiver, connected ConnectedEndpoint) Endpoint { + return &connectionedEndpoint{ + baseEndpoint: baseEndpoint{Queue: queue, receiver: receiver, connected: connected}, + id: uid.UniqueID(), + idGenerator: uid, + stype: stype, + } +} + +// ID implements ConnectingEndpoint.ID. +func (e *connectionedEndpoint) ID() uint64 { + return e.id +} + +// Type implements ConnectingEndpoint.Type and Endpoint.Type. +func (e *connectionedEndpoint) Type() SockType { + return e.stype +} + +// WaiterQueue implements ConnectingEndpoint.WaiterQueue. +func (e *connectionedEndpoint) WaiterQueue() *waiter.Queue { + return e.Queue +} + +// isBound returns true iff the connectionedEndpoint is bound (but not +// listening). +func (e *connectionedEndpoint) isBound() bool { + return e.path != "" && e.acceptedChan == nil +} + +// Listening implements ConnectingEndpoint.Listening. +func (e *connectionedEndpoint) Listening() bool { + return e.acceptedChan != nil +} + +// Close puts the connectionedEndpoint in a closed state and frees all +// resources associated with it. +// +// The socket will be a fresh state after a call to close and may be reused. +// That is, close may be used to "unbind" or "disconnect" the socket in error +// paths. +func (e *connectionedEndpoint) Close() { + e.Lock() + var c ConnectedEndpoint + var r Receiver + switch { + case e.Connected(): + e.connected.CloseSend() + e.receiver.CloseRecv() + c = e.connected + r = e.receiver + e.connected = nil + e.receiver = nil + case e.isBound(): + e.path = "" + case e.Listening(): + close(e.acceptedChan) + for n := range e.acceptedChan { + n.Close() + } + e.acceptedChan = nil + e.path = "" + } + e.Unlock() + if c != nil { + c.CloseNotify() + c.Release() + } + if r != nil { + r.CloseNotify() + r.Release() + } +} + +// BidirectionalConnect implements BoundEndpoint.BidirectionalConnect. +func (e *connectionedEndpoint) BidirectionalConnect(ce ConnectingEndpoint, returnConnect func(Receiver, ConnectedEndpoint)) *tcpip.Error { + if ce.Type() != e.stype { + return tcpip.ErrConnectionRefused + } + + // Check if ce is e to avoid a deadlock. + if ce, ok := ce.(*connectionedEndpoint); ok && ce == e { + return tcpip.ErrInvalidEndpointState + } + + // Do a dance to safely acquire locks on both endpoints. + if e.id < ce.ID() { + e.Lock() + ce.Lock() + } else { + ce.Lock() + e.Lock() + } + + // Check connecting state. + if ce.Connected() { + e.Unlock() + ce.Unlock() + return tcpip.ErrAlreadyConnected + } + if ce.Listening() { + e.Unlock() + ce.Unlock() + return tcpip.ErrInvalidEndpointState + } + + // Check bound state. + if !e.Listening() { + e.Unlock() + ce.Unlock() + return tcpip.ErrConnectionRefused + } + + // Create a newly bound connectionedEndpoint. + ne := &connectionedEndpoint{ + baseEndpoint: baseEndpoint{ + path: e.path, + Queue: &waiter.Queue{}, + }, + id: e.idGenerator.UniqueID(), + idGenerator: e.idGenerator, + stype: e.stype, + } + readQueue := queue.New(ce.WaiterQueue(), ne.Queue, initialLimit) + writeQueue := queue.New(ne.Queue, ce.WaiterQueue(), initialLimit) + ne.connected = &connectedEndpoint{ + endpoint: ce, + writeQueue: readQueue, + } + if e.stype == SockStream { + ne.receiver = &streamQueueReceiver{queueReceiver: queueReceiver{readQueue: writeQueue}} + } else { + ne.receiver = &queueReceiver{readQueue: writeQueue} + } + + select { + case e.acceptedChan <- ne: + // Commit state. + connected := &connectedEndpoint{ + endpoint: ne, + writeQueue: writeQueue, + } + if e.stype == SockStream { + returnConnect(&streamQueueReceiver{queueReceiver: queueReceiver{readQueue: readQueue}}, connected) + } else { + returnConnect(&queueReceiver{readQueue: readQueue}, connected) + } + + // Notify can deadlock if we are holding these locks. + e.Unlock() + ce.Unlock() + + // Notify on both ends. + e.Notify(waiter.EventIn) + ce.WaiterQueue().Notify(waiter.EventOut) + + return nil + default: + // Busy; return ECONNREFUSED per spec. + ne.Close() + e.Unlock() + ce.Unlock() + return tcpip.ErrConnectionRefused + } +} + +// UnidirectionalConnect implements BoundEndpoint.UnidirectionalConnect. +func (e *connectionedEndpoint) UnidirectionalConnect() (ConnectedEndpoint, *tcpip.Error) { + return nil, tcpip.ErrConnectionRefused +} + +// Connect attempts to directly connect to another Endpoint. +// Implements Endpoint.Connect. +func (e *connectionedEndpoint) Connect(server BoundEndpoint) *tcpip.Error { + returnConnect := func(r Receiver, ce ConnectedEndpoint) { + e.receiver = r + e.connected = ce + } + + return server.BidirectionalConnect(e, returnConnect) +} + +// Listen starts listening on the connection. +func (e *connectionedEndpoint) Listen(backlog int) *tcpip.Error { + e.Lock() + defer e.Unlock() + if e.Listening() { + // Adjust the size of the channel iff we can fix existing + // pending connections into the new one. + if len(e.acceptedChan) > backlog { + return tcpip.ErrInvalidEndpointState + } + origChan := e.acceptedChan + e.acceptedChan = make(chan *connectionedEndpoint, backlog) + close(origChan) + for ep := range origChan { + e.acceptedChan <- ep + } + return nil + } + if !e.isBound() { + return tcpip.ErrInvalidEndpointState + } + + // Normal case. + e.acceptedChan = make(chan *connectionedEndpoint, backlog) + return nil +} + +// Accept accepts a new connection. +func (e *connectionedEndpoint) Accept() (Endpoint, *tcpip.Error) { + e.Lock() + defer e.Unlock() + + if !e.Listening() { + return nil, tcpip.ErrInvalidEndpointState + } + + select { + case ne := <-e.acceptedChan: + return ne, nil + + default: + // Nothing left. + return nil, tcpip.ErrWouldBlock + } +} + +// Bind binds the connection. +// +// For Unix connectionedEndpoints, this _only sets the address associated with +// the socket_. Work associated with sockets in the filesystem or finding those +// sockets must be done by a higher level. +// +// Bind will fail only if the socket is connected, bound or the passed address +// is invalid (the empty string). +func (e *connectionedEndpoint) Bind(addr tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error { + e.Lock() + defer e.Unlock() + if e.isBound() || e.Listening() { + return tcpip.ErrAlreadyBound + } + if addr.Addr == "" { + // The empty string is not permitted. + return tcpip.ErrBadLocalAddress + } + if commit != nil { + if err := commit(); err != nil { + return err + } + } + + // Save the bound address. + e.path = string(addr.Addr) + return nil +} + +// SendMsg writes data and a control message to the endpoint's peer. +// This method does not block if the data cannot be written. +func (e *connectionedEndpoint) SendMsg(data [][]byte, c ControlMessages, to BoundEndpoint) (uintptr, *tcpip.Error) { + // Stream sockets do not support specifying the endpoint. Seqpacket + // sockets ignore the passed endpoint. + if e.stype == SockStream && to != nil { + return 0, tcpip.ErrNotSupported + } + return e.baseEndpoint.SendMsg(data, c, to) +} + +// Readiness returns the current readiness of the connectionedEndpoint. For +// example, if waiter.EventIn is set, the connectionedEndpoint is immediately +// readable. +func (e *connectionedEndpoint) Readiness(mask waiter.EventMask) waiter.EventMask { + e.Lock() + defer e.Unlock() + + ready := waiter.EventMask(0) + switch { + case e.Connected(): + if mask&waiter.EventIn != 0 && e.receiver.Readable() { + ready |= waiter.EventIn + } + if mask&waiter.EventOut != 0 && e.connected.Writable() { + ready |= waiter.EventOut + } + case e.Listening(): + if mask&waiter.EventIn != 0 && len(e.acceptedChan) > 0 { + ready |= waiter.EventIn + } + } + + return ready +} diff --git a/pkg/sentry/socket/unix/transport/connectioned_state.go b/pkg/sentry/socket/unix/transport/connectioned_state.go new file mode 100644 index 000000000..7e6c73dcc --- /dev/null +++ b/pkg/sentry/socket/unix/transport/connectioned_state.go @@ -0,0 +1,53 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package transport + +// saveAcceptedChan is invoked by stateify. +func (e *connectionedEndpoint) saveAcceptedChan() []*connectionedEndpoint { + // If acceptedChan is nil (i.e. we are not listening) then we will save nil. + // Otherwise we create a (possibly empty) slice of the values in acceptedChan and + // save that. + var acceptedSlice []*connectionedEndpoint + if e.acceptedChan != nil { + // Swap out acceptedChan with a new empty channel of the same capacity. + saveChan := e.acceptedChan + e.acceptedChan = make(chan *connectionedEndpoint, cap(saveChan)) + + // Create a new slice with the same len and capacity as the channel. + acceptedSlice = make([]*connectionedEndpoint, len(saveChan), cap(saveChan)) + // Drain acceptedChan into saveSlice, and fill up the new acceptChan at the + // same time. + for i := range acceptedSlice { + ep := <-saveChan + acceptedSlice[i] = ep + e.acceptedChan <- ep + } + close(saveChan) + } + return acceptedSlice +} + +// loadAcceptedChan is invoked by stateify. +func (e *connectionedEndpoint) loadAcceptedChan(acceptedSlice []*connectionedEndpoint) { + // If acceptedSlice is nil, then acceptedChan should also be nil. + if acceptedSlice != nil { + // Otherwise, create a new channel with the same capacity as acceptedSlice. + e.acceptedChan = make(chan *connectionedEndpoint, cap(acceptedSlice)) + // Seed the channel with values from acceptedSlice. + for _, ep := range acceptedSlice { + e.acceptedChan <- ep + } + } +} diff --git a/pkg/sentry/socket/unix/transport/connectionless.go b/pkg/sentry/socket/unix/transport/connectionless.go new file mode 100644 index 000000000..fb2728010 --- /dev/null +++ b/pkg/sentry/socket/unix/transport/connectionless.go @@ -0,0 +1,192 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package transport + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport/queue" + "gvisor.googlesource.com/gvisor/pkg/tcpip" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// connectionlessEndpoint is a unix endpoint for unix sockets that support operating in +// a conectionless fashon. +// +// Specifically, this means datagram unix sockets not created with +// socketpair(2). +// +// +stateify savable +type connectionlessEndpoint struct { + baseEndpoint +} + +// NewConnectionless creates a new unbound dgram endpoint. +func NewConnectionless() Endpoint { + ep := &connectionlessEndpoint{baseEndpoint{Queue: &waiter.Queue{}}} + ep.receiver = &queueReceiver{readQueue: queue.New(&waiter.Queue{}, ep.Queue, initialLimit)} + return ep +} + +// isBound returns true iff the endpoint is bound. +func (e *connectionlessEndpoint) isBound() bool { + return e.path != "" +} + +// Close puts the endpoint in a closed state and frees all resources associated +// with it. +// +// The socket will be a fresh state after a call to close and may be reused. +// That is, close may be used to "unbind" or "disconnect" the socket in error +// paths. +func (e *connectionlessEndpoint) Close() { + e.Lock() + var r Receiver + if e.Connected() { + e.receiver.CloseRecv() + r = e.receiver + e.receiver = nil + + e.connected.Release() + e.connected = nil + } + if e.isBound() { + e.path = "" + } + e.Unlock() + if r != nil { + r.CloseNotify() + r.Release() + } +} + +// BidirectionalConnect implements BoundEndpoint.BidirectionalConnect. +func (e *connectionlessEndpoint) BidirectionalConnect(ce ConnectingEndpoint, returnConnect func(Receiver, ConnectedEndpoint)) *tcpip.Error { + return tcpip.ErrConnectionRefused +} + +// UnidirectionalConnect implements BoundEndpoint.UnidirectionalConnect. +func (e *connectionlessEndpoint) UnidirectionalConnect() (ConnectedEndpoint, *tcpip.Error) { + e.Lock() + r := e.receiver + e.Unlock() + if r == nil { + return nil, tcpip.ErrConnectionRefused + } + return &connectedEndpoint{ + endpoint: e, + writeQueue: r.(*queueReceiver).readQueue, + }, nil +} + +// SendMsg writes data and a control message to the specified endpoint. +// This method does not block if the data cannot be written. +func (e *connectionlessEndpoint) SendMsg(data [][]byte, c ControlMessages, to BoundEndpoint) (uintptr, *tcpip.Error) { + if to == nil { + return e.baseEndpoint.SendMsg(data, c, nil) + } + + connected, err := to.UnidirectionalConnect() + if err != nil { + return 0, tcpip.ErrInvalidEndpointState + } + defer connected.Release() + + e.Lock() + n, notify, err := connected.Send(data, c, tcpip.FullAddress{Addr: tcpip.Address(e.path)}) + e.Unlock() + + if notify { + connected.SendNotify() + } + + return n, err +} + +// Type implements Endpoint.Type. +func (e *connectionlessEndpoint) Type() SockType { + return SockDgram +} + +// Connect attempts to connect directly to server. +func (e *connectionlessEndpoint) Connect(server BoundEndpoint) *tcpip.Error { + connected, err := server.UnidirectionalConnect() + if err != nil { + return err + } + + e.Lock() + e.connected = connected + e.Unlock() + + return nil +} + +// Listen starts listening on the connection. +func (e *connectionlessEndpoint) Listen(int) *tcpip.Error { + return tcpip.ErrNotSupported +} + +// Accept accepts a new connection. +func (e *connectionlessEndpoint) Accept() (Endpoint, *tcpip.Error) { + return nil, tcpip.ErrNotSupported +} + +// Bind binds the connection. +// +// For Unix endpoints, this _only sets the address associated with the socket_. +// Work associated with sockets in the filesystem or finding those sockets must +// be done by a higher level. +// +// Bind will fail only if the socket is connected, bound or the passed address +// is invalid (the empty string). +func (e *connectionlessEndpoint) Bind(addr tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error { + e.Lock() + defer e.Unlock() + if e.isBound() { + return tcpip.ErrAlreadyBound + } + if addr.Addr == "" { + // The empty string is not permitted. + return tcpip.ErrBadLocalAddress + } + if commit != nil { + if err := commit(); err != nil { + return err + } + } + + // Save the bound address. + e.path = string(addr.Addr) + return nil +} + +// Readiness returns the current readiness of the endpoint. For example, if +// waiter.EventIn is set, the endpoint is immediately readable. +func (e *connectionlessEndpoint) Readiness(mask waiter.EventMask) waiter.EventMask { + e.Lock() + defer e.Unlock() + + ready := waiter.EventMask(0) + if mask&waiter.EventIn != 0 && e.receiver.Readable() { + ready |= waiter.EventIn + } + + if e.Connected() { + if mask&waiter.EventOut != 0 && e.connected.Writable() { + ready |= waiter.EventOut + } + } + + return ready +} diff --git a/pkg/sentry/socket/unix/transport/queue/BUILD b/pkg/sentry/socket/unix/transport/queue/BUILD new file mode 100644 index 000000000..d914ecc23 --- /dev/null +++ b/pkg/sentry/socket/unix/transport/queue/BUILD @@ -0,0 +1,15 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("//tools/go_stateify:defs.bzl", "go_library") + +go_library( + name = "queue", + srcs = ["queue.go"], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport/queue", + visibility = ["//:sandbox"], + deps = [ + "//pkg/ilist", + "//pkg/tcpip", + "//pkg/waiter", + ], +) diff --git a/pkg/sentry/socket/unix/transport/queue/queue.go b/pkg/sentry/socket/unix/transport/queue/queue.go new file mode 100644 index 000000000..b3d2ea68b --- /dev/null +++ b/pkg/sentry/socket/unix/transport/queue/queue.go @@ -0,0 +1,227 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package queue provides the implementation of buffer queue +// and interface of queue entry with Length method. +package queue + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/ilist" + "gvisor.googlesource.com/gvisor/pkg/tcpip" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// Entry implements Linker interface and has additional required methods. +type Entry interface { + ilist.Linker + + // Length returns the number of bytes stored in the entry. + Length() int64 + + // Release releases any resources held by the entry. + Release() + + // Peek returns a copy of the entry. It must be Released separately. + Peek() Entry + + // Truncate reduces the number of bytes stored in the entry to n bytes. + // + // Preconditions: n <= Length(). + Truncate(n int64) +} + +// Queue is a buffer queue. +// +// +stateify savable +type Queue struct { + ReaderQueue *waiter.Queue + WriterQueue *waiter.Queue + + mu sync.Mutex `state:"nosave"` + closed bool + used int64 + limit int64 + dataList ilist.List +} + +// New allocates and initializes a new queue. +func New(ReaderQueue *waiter.Queue, WriterQueue *waiter.Queue, limit int64) *Queue { + return &Queue{ReaderQueue: ReaderQueue, WriterQueue: WriterQueue, limit: limit} +} + +// Close closes q for reading and writing. It is immediately not writable and +// will become unreadable when no more data is pending. +// +// Both the read and write queues must be notified after closing: +// q.ReaderQueue.Notify(waiter.EventIn) +// q.WriterQueue.Notify(waiter.EventOut) +func (q *Queue) Close() { + q.mu.Lock() + q.closed = true + q.mu.Unlock() +} + +// Reset empties the queue and Releases all of the Entries. +// +// Both the read and write queues must be notified after resetting: +// q.ReaderQueue.Notify(waiter.EventIn) +// q.WriterQueue.Notify(waiter.EventOut) +func (q *Queue) Reset() { + q.mu.Lock() + for cur := q.dataList.Front(); cur != nil; cur = cur.Next() { + cur.(Entry).Release() + } + q.dataList.Reset() + q.used = 0 + q.mu.Unlock() +} + +// IsReadable determines if q is currently readable. +func (q *Queue) IsReadable() bool { + q.mu.Lock() + defer q.mu.Unlock() + + return q.closed || q.dataList.Front() != nil +} + +// bufWritable returns true if there is space for writing. +// +// N.B. Linux only considers a unix socket "writable" if >75% of the buffer is +// free. +// +// See net/unix/af_unix.c:unix_writeable. +func (q *Queue) bufWritable() bool { + return 4*q.used < q.limit +} + +// IsWritable determines if q is currently writable. +func (q *Queue) IsWritable() bool { + q.mu.Lock() + defer q.mu.Unlock() + + return q.closed || q.bufWritable() +} + +// Enqueue adds an entry to the data queue if room is available. +// +// If truncate is true, Enqueue may truncate the message beforing enqueuing it. +// Otherwise, the entire message must fit. If n < e.Length(), err indicates why. +// +// If notify is true, ReaderQueue.Notify must be called: +// q.ReaderQueue.Notify(waiter.EventIn) +func (q *Queue) Enqueue(e Entry, truncate bool) (l int64, notify bool, err *tcpip.Error) { + q.mu.Lock() + + if q.closed { + q.mu.Unlock() + return 0, false, tcpip.ErrClosedForSend + } + + free := q.limit - q.used + + l = e.Length() + + if l > free && truncate { + if free == 0 { + // Message can't fit right now. + q.mu.Unlock() + return 0, false, tcpip.ErrWouldBlock + } + + e.Truncate(free) + l = e.Length() + err = tcpip.ErrWouldBlock + } + + if l > q.limit { + // Message is too big to ever fit. + q.mu.Unlock() + return 0, false, tcpip.ErrMessageTooLong + } + + if l > free { + // Message can't fit right now. + q.mu.Unlock() + return 0, false, tcpip.ErrWouldBlock + } + + notify = q.dataList.Front() == nil + q.used += l + q.dataList.PushBack(e) + + q.mu.Unlock() + + return l, notify, err +} + +// Dequeue removes the first entry in the data queue, if one exists. +// +// If notify is true, WriterQueue.Notify must be called: +// q.WriterQueue.Notify(waiter.EventOut) +func (q *Queue) Dequeue() (e Entry, notify bool, err *tcpip.Error) { + q.mu.Lock() + + if q.dataList.Front() == nil { + err := tcpip.ErrWouldBlock + if q.closed { + err = tcpip.ErrClosedForReceive + } + q.mu.Unlock() + + return nil, false, err + } + + notify = !q.bufWritable() + + e = q.dataList.Front().(Entry) + q.dataList.Remove(e) + q.used -= e.Length() + + notify = notify && q.bufWritable() + + q.mu.Unlock() + + return e, notify, nil +} + +// Peek returns the first entry in the data queue, if one exists. +func (q *Queue) Peek() (Entry, *tcpip.Error) { + q.mu.Lock() + defer q.mu.Unlock() + + if q.dataList.Front() == nil { + err := tcpip.ErrWouldBlock + if q.closed { + err = tcpip.ErrClosedForReceive + } + return nil, err + } + + return q.dataList.Front().(Entry).Peek(), nil +} + +// QueuedSize returns the number of bytes currently in the queue, that is, the +// number of readable bytes. +func (q *Queue) QueuedSize() int64 { + q.mu.Lock() + defer q.mu.Unlock() + return q.used +} + +// MaxQueueSize returns the maximum number of bytes storable in the queue. +func (q *Queue) MaxQueueSize() int64 { + return q.limit +} diff --git a/pkg/sentry/socket/unix/transport/unix.go b/pkg/sentry/socket/unix/transport/unix.go new file mode 100644 index 000000000..577aa87d5 --- /dev/null +++ b/pkg/sentry/socket/unix/transport/unix.go @@ -0,0 +1,953 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package transport contains the implementation of Unix endpoints. +package transport + +import ( + "sync" + "sync/atomic" + + "gvisor.googlesource.com/gvisor/pkg/ilist" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport/queue" + "gvisor.googlesource.com/gvisor/pkg/tcpip" + "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// initialLimit is the starting limit for the socket buffers. +const initialLimit = 16 * 1024 + +// A SockType is a type (as opposed to family) of sockets. These are enumerated +// in the syscall package as syscall.SOCK_* constants. +type SockType int + +const ( + // SockStream corresponds to syscall.SOCK_STREAM. + SockStream SockType = 1 + // SockDgram corresponds to syscall.SOCK_DGRAM. + SockDgram SockType = 2 + // SockRaw corresponds to syscall.SOCK_RAW. + SockRaw SockType = 3 + // SockSeqpacket corresponds to syscall.SOCK_SEQPACKET. + SockSeqpacket SockType = 5 +) + +// A RightsControlMessage is a control message containing FDs. +type RightsControlMessage interface { + // Clone returns a copy of the RightsControlMessage. + Clone() RightsControlMessage + + // Release releases any resources owned by the RightsControlMessage. + Release() +} + +// A CredentialsControlMessage is a control message containing Unix credentials. +type CredentialsControlMessage interface { + // Equals returns true iff the two messages are equal. + Equals(CredentialsControlMessage) bool +} + +// A ControlMessages represents a collection of socket control messages. +// +// +stateify savable +type ControlMessages struct { + // Rights is a control message containing FDs. + Rights RightsControlMessage + + // Credentials is a control message containing Unix credentials. + Credentials CredentialsControlMessage +} + +// Empty returns true iff the ControlMessages does not contain either +// credentials or rights. +func (c *ControlMessages) Empty() bool { + return c.Rights == nil && c.Credentials == nil +} + +// Clone clones both the credentials and the rights. +func (c *ControlMessages) Clone() ControlMessages { + cm := ControlMessages{} + if c.Rights != nil { + cm.Rights = c.Rights.Clone() + } + cm.Credentials = c.Credentials + return cm +} + +// Release releases both the credentials and the rights. +func (c *ControlMessages) Release() { + if c.Rights != nil { + c.Rights.Release() + } + *c = ControlMessages{} +} + +// Endpoint is the interface implemented by Unix transport protocol +// implementations that expose functionality like sendmsg, recvmsg, connect, +// etc. to Unix socket implementations. +type Endpoint interface { + Credentialer + waiter.Waitable + + // Close puts the endpoint in a closed state and frees all resources + // associated with it. + Close() + + // RecvMsg reads data and a control message from the endpoint. This method + // does not block if there is no data pending. + // + // creds indicates if credential control messages are requested by the + // caller. This is useful for determining if control messages can be + // coalesced. creds is a hint and can be safely ignored by the + // implementation if no coalescing is possible. It is fine to return + // credential control messages when none were requested or to not return + // credential control messages when they were requested. + // + // numRights is the number of SCM_RIGHTS FDs requested by the caller. This + // is useful if one must allocate a buffer to receive a SCM_RIGHTS message + // or determine if control messages can be coalesced. numRights is a hint + // and can be safely ignored by the implementation if the number of + // available SCM_RIGHTS FDs is known and no coalescing is possible. It is + // fine for the returned number of SCM_RIGHTS FDs to be either higher or + // lower than the requested number. + // + // If peek is true, no data should be consumed from the Endpoint. Any and + // all data returned from a peek should be available in the next call to + // RecvMsg. + // + // recvLen is the number of bytes copied into data. + // + // msgLen is the length of the read message consumed for datagram Endpoints. + // msgLen is always the same as recvLen for stream Endpoints. + RecvMsg(data [][]byte, creds bool, numRights uintptr, peek bool, addr *tcpip.FullAddress) (recvLen, msgLen uintptr, cm ControlMessages, err *tcpip.Error) + + // SendMsg writes data and a control message to the endpoint's peer. + // This method does not block if the data cannot be written. + // + // SendMsg does not take ownership of any of its arguments on error. + SendMsg([][]byte, ControlMessages, BoundEndpoint) (uintptr, *tcpip.Error) + + // Connect connects this endpoint directly to another. + // + // This should be called on the client endpoint, and the (bound) + // endpoint passed in as a parameter. + // + // The error codes are the same as Connect. + Connect(server BoundEndpoint) *tcpip.Error + + // Shutdown closes the read and/or write end of the endpoint connection + // to its peer. + Shutdown(flags tcpip.ShutdownFlags) *tcpip.Error + + // Listen puts the endpoint in "listen" mode, which allows it to accept + // new connections. + Listen(backlog int) *tcpip.Error + + // Accept returns a new endpoint if a peer has established a connection + // to an endpoint previously set to listen mode. This method does not + // block if no new connections are available. + // + // The returned Queue is the wait queue for the newly created endpoint. + Accept() (Endpoint, *tcpip.Error) + + // Bind binds the endpoint to a specific local address and port. + // Specifying a NIC is optional. + // + // An optional commit function will be executed atomically with respect + // to binding the endpoint. If this returns an error, the bind will not + // occur and the error will be propagated back to the caller. + Bind(address tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error + + // Type return the socket type, typically either SockStream, SockDgram + // or SockSeqpacket. + Type() SockType + + // GetLocalAddress returns the address to which the endpoint is bound. + GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) + + // GetRemoteAddress returns the address to which the endpoint is + // connected. + GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) + + // SetSockOpt sets a socket option. opt should be one of the tcpip.*Option + // types. + SetSockOpt(opt interface{}) *tcpip.Error + + // GetSockOpt gets a socket option. opt should be a pointer to one of the + // tcpip.*Option types. + GetSockOpt(opt interface{}) *tcpip.Error +} + +// A Credentialer is a socket or endpoint that supports the SO_PASSCRED socket +// option. +type Credentialer interface { + // Passcred returns whether or not the SO_PASSCRED socket option is + // enabled on this end. + Passcred() bool + + // ConnectedPasscred returns whether or not the SO_PASSCRED socket option + // is enabled on the connected end. + ConnectedPasscred() bool +} + +// A BoundEndpoint is a unix endpoint that can be connected to. +type BoundEndpoint interface { + // BidirectionalConnect establishes a bi-directional connection between two + // unix endpoints in an all-or-nothing manner. If an error occurs during + // connecting, the state of neither endpoint should be modified. + // + // In order for an endpoint to establish such a bidirectional connection + // with a BoundEndpoint, the endpoint calls the BidirectionalConnect method + // on the BoundEndpoint and sends a representation of itself (the + // ConnectingEndpoint) and a callback (returnConnect) to receive the + // connection information (Receiver and ConnectedEndpoint) upon a + // successful connect. The callback should only be called on a successful + // connect. + // + // For a connection attempt to be successful, the ConnectingEndpoint must + // be unconnected and not listening and the BoundEndpoint whose + // BidirectionalConnect method is being called must be listening. + // + // This method will return tcpip.ErrConnectionRefused on endpoints with a + // type that isn't SockStream or SockSeqpacket. + BidirectionalConnect(ep ConnectingEndpoint, returnConnect func(Receiver, ConnectedEndpoint)) *tcpip.Error + + // UnidirectionalConnect establishes a write-only connection to a unix + // endpoint. + // + // An endpoint which calls UnidirectionalConnect and supports it itself must + // not hold its own lock when calling UnidirectionalConnect. + // + // This method will return tcpip.ErrConnectionRefused on a non-SockDgram + // endpoint. + UnidirectionalConnect() (ConnectedEndpoint, *tcpip.Error) + + // Release releases any resources held by the BoundEndpoint. It must be + // called before dropping all references to a BoundEndpoint returned by a + // function. + Release() +} + +// message represents a message passed over a Unix domain socket. +// +// +stateify savable +type message struct { + ilist.Entry + + // Data is the Message payload. + Data buffer.View + + // Control is auxiliary control message data that goes along with the + // data. + Control ControlMessages + + // Address is the bound address of the endpoint that sent the message. + // + // If the endpoint that sent the message is not bound, the Address is + // the empty string. + Address tcpip.FullAddress +} + +// Length returns number of bytes stored in the message. +func (m *message) Length() int64 { + return int64(len(m.Data)) +} + +// Release releases any resources held by the message. +func (m *message) Release() { + m.Control.Release() +} + +// Peek returns a copy of the message. +func (m *message) Peek() queue.Entry { + return &message{Data: m.Data, Control: m.Control.Clone(), Address: m.Address} +} + +// Truncate reduces the length of the message payload to n bytes. +// +// Preconditions: n <= m.Length(). +func (m *message) Truncate(n int64) { + m.Data.CapLength(int(n)) +} + +// A Receiver can be used to receive Messages. +type Receiver interface { + // Recv receives a single message. This method does not block. + // + // See Endpoint.RecvMsg for documentation on shared arguments. + // + // notify indicates if RecvNotify should be called. + Recv(data [][]byte, creds bool, numRights uintptr, peek bool) (recvLen, msgLen uintptr, cm ControlMessages, source tcpip.FullAddress, notify bool, err *tcpip.Error) + + // RecvNotify notifies the Receiver of a successful Recv. This must not be + // called while holding any endpoint locks. + RecvNotify() + + // CloseRecv prevents the receiving of additional Messages. + // + // After CloseRecv is called, CloseNotify must also be called. + CloseRecv() + + // CloseNotify notifies the Receiver of recv being closed. This must not be + // called while holding any endpoint locks. + CloseNotify() + + // Readable returns if messages should be attempted to be received. This + // includes when read has been shutdown. + Readable() bool + + // RecvQueuedSize returns the total amount of data currently receivable. + // RecvQueuedSize should return -1 if the operation isn't supported. + RecvQueuedSize() int64 + + // RecvMaxQueueSize returns maximum value for RecvQueuedSize. + // RecvMaxQueueSize should return -1 if the operation isn't supported. + RecvMaxQueueSize() int64 + + // Release releases any resources owned by the Receiver. It should be + // called before droping all references to a Receiver. + Release() +} + +// queueReceiver implements Receiver for datagram sockets. +// +// +stateify savable +type queueReceiver struct { + readQueue *queue.Queue +} + +// Recv implements Receiver.Recv. +func (q *queueReceiver) Recv(data [][]byte, creds bool, numRights uintptr, peek bool) (uintptr, uintptr, ControlMessages, tcpip.FullAddress, bool, *tcpip.Error) { + var m queue.Entry + var notify bool + var err *tcpip.Error + if peek { + m, err = q.readQueue.Peek() + } else { + m, notify, err = q.readQueue.Dequeue() + } + if err != nil { + return 0, 0, ControlMessages{}, tcpip.FullAddress{}, false, err + } + msg := m.(*message) + src := []byte(msg.Data) + var copied uintptr + for i := 0; i < len(data) && len(src) > 0; i++ { + n := copy(data[i], src) + copied += uintptr(n) + src = src[n:] + } + return copied, uintptr(len(msg.Data)), msg.Control, msg.Address, notify, nil +} + +// RecvNotify implements Receiver.RecvNotify. +func (q *queueReceiver) RecvNotify() { + q.readQueue.WriterQueue.Notify(waiter.EventOut) +} + +// CloseNotify implements Receiver.CloseNotify. +func (q *queueReceiver) CloseNotify() { + q.readQueue.ReaderQueue.Notify(waiter.EventIn) + q.readQueue.WriterQueue.Notify(waiter.EventOut) +} + +// CloseRecv implements Receiver.CloseRecv. +func (q *queueReceiver) CloseRecv() { + q.readQueue.Close() +} + +// Readable implements Receiver.Readable. +func (q *queueReceiver) Readable() bool { + return q.readQueue.IsReadable() +} + +// RecvQueuedSize implements Receiver.RecvQueuedSize. +func (q *queueReceiver) RecvQueuedSize() int64 { + return q.readQueue.QueuedSize() +} + +// RecvMaxQueueSize implements Receiver.RecvMaxQueueSize. +func (q *queueReceiver) RecvMaxQueueSize() int64 { + return q.readQueue.MaxQueueSize() +} + +// Release implements Receiver.Release. +func (*queueReceiver) Release() {} + +// streamQueueReceiver implements Receiver for stream sockets. +// +// +stateify savable +type streamQueueReceiver struct { + queueReceiver + + mu sync.Mutex `state:"nosave"` + buffer []byte + control ControlMessages + addr tcpip.FullAddress +} + +func vecCopy(data [][]byte, buf []byte) (uintptr, [][]byte, []byte) { + var copied uintptr + for len(data) > 0 && len(buf) > 0 { + n := copy(data[0], buf) + copied += uintptr(n) + buf = buf[n:] + data[0] = data[0][n:] + if len(data[0]) == 0 { + data = data[1:] + } + } + return copied, data, buf +} + +// Readable implements Receiver.Readable. +func (q *streamQueueReceiver) Readable() bool { + q.mu.Lock() + bl := len(q.buffer) + r := q.readQueue.IsReadable() + q.mu.Unlock() + // We're readable if we have data in our buffer or if the queue receiver is + // readable. + return bl > 0 || r +} + +// RecvQueuedSize implements Receiver.RecvQueuedSize. +func (q *streamQueueReceiver) RecvQueuedSize() int64 { + q.mu.Lock() + bl := len(q.buffer) + qs := q.readQueue.QueuedSize() + q.mu.Unlock() + return int64(bl) + qs +} + +// RecvMaxQueueSize implements Receiver.RecvMaxQueueSize. +func (q *streamQueueReceiver) RecvMaxQueueSize() int64 { + // The RecvMaxQueueSize() is the readQueue's MaxQueueSize() plus the largest + // message we can buffer which is also the largest message we can receive. + return 2 * q.readQueue.MaxQueueSize() +} + +// Recv implements Receiver.Recv. +func (q *streamQueueReceiver) Recv(data [][]byte, wantCreds bool, numRights uintptr, peek bool) (uintptr, uintptr, ControlMessages, tcpip.FullAddress, bool, *tcpip.Error) { + q.mu.Lock() + defer q.mu.Unlock() + + var notify bool + + // If we have no data in the endpoint, we need to get some. + if len(q.buffer) == 0 { + // Load the next message into a buffer, even if we are peeking. Peeking + // won't consume the message, so it will be still available to be read + // the next time Recv() is called. + m, n, err := q.readQueue.Dequeue() + if err != nil { + return 0, 0, ControlMessages{}, tcpip.FullAddress{}, false, err + } + notify = n + msg := m.(*message) + q.buffer = []byte(msg.Data) + q.control = msg.Control + q.addr = msg.Address + } + + var copied uintptr + if peek { + // Don't consume control message if we are peeking. + c := q.control.Clone() + + // Don't consume data since we are peeking. + copied, data, _ = vecCopy(data, q.buffer) + + return copied, copied, c, q.addr, notify, nil + } + + // Consume data and control message since we are not peeking. + copied, data, q.buffer = vecCopy(data, q.buffer) + + // Save the original state of q.control. + c := q.control + + // Remove rights from q.control and leave behind just the creds. + q.control.Rights = nil + if !wantCreds { + c.Credentials = nil + } + + if c.Rights != nil && numRights == 0 { + c.Rights.Release() + c.Rights = nil + } + + haveRights := c.Rights != nil + + // If we have more capacity for data and haven't received any usable + // rights. + // + // Linux never coalesces rights control messages. + for !haveRights && len(data) > 0 { + // Get a message from the readQueue. + m, n, err := q.readQueue.Dequeue() + if err != nil { + // We already got some data, so ignore this error. This will + // manifest as a short read to the user, which is what Linux + // does. + break + } + notify = notify || n + msg := m.(*message) + q.buffer = []byte(msg.Data) + q.control = msg.Control + q.addr = msg.Address + + if wantCreds { + if (q.control.Credentials == nil) != (c.Credentials == nil) { + // One message has credentials, the other does not. + break + } + + if q.control.Credentials != nil && c.Credentials != nil && !q.control.Credentials.Equals(c.Credentials) { + // Both messages have credentials, but they don't match. + break + } + } + + if numRights != 0 && c.Rights != nil && q.control.Rights != nil { + // Both messages have rights. + break + } + + var cpd uintptr + cpd, data, q.buffer = vecCopy(data, q.buffer) + copied += cpd + + if cpd == 0 { + // data was actually full. + break + } + + if q.control.Rights != nil { + // Consume rights. + if numRights == 0 { + q.control.Rights.Release() + } else { + c.Rights = q.control.Rights + haveRights = true + } + q.control.Rights = nil + } + } + return copied, copied, c, q.addr, notify, nil +} + +// A ConnectedEndpoint is an Endpoint that can be used to send Messages. +type ConnectedEndpoint interface { + // Passcred implements Endpoint.Passcred. + Passcred() bool + + // GetLocalAddress implements Endpoint.GetLocalAddress. + GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) + + // Send sends a single message. This method does not block. + // + // notify indicates if SendNotify should be called. + // + // tcpip.ErrWouldBlock can be returned along with a partial write if + // the caller should block to send the rest of the data. + Send(data [][]byte, controlMessages ControlMessages, from tcpip.FullAddress) (n uintptr, notify bool, err *tcpip.Error) + + // SendNotify notifies the ConnectedEndpoint of a successful Send. This + // must not be called while holding any endpoint locks. + SendNotify() + + // CloseSend prevents the sending of additional Messages. + // + // After CloseSend is call, CloseNotify must also be called. + CloseSend() + + // CloseNotify notifies the ConnectedEndpoint of send being closed. This + // must not be called while holding any endpoint locks. + CloseNotify() + + // Writable returns if messages should be attempted to be sent. This + // includes when write has been shutdown. + Writable() bool + + // EventUpdate lets the ConnectedEndpoint know that event registrations + // have changed. + EventUpdate() + + // SendQueuedSize returns the total amount of data currently queued for + // sending. SendQueuedSize should return -1 if the operation isn't + // supported. + SendQueuedSize() int64 + + // SendMaxQueueSize returns maximum value for SendQueuedSize. + // SendMaxQueueSize should return -1 if the operation isn't supported. + SendMaxQueueSize() int64 + + // Release releases any resources owned by the ConnectedEndpoint. It should + // be called before droping all references to a ConnectedEndpoint. + Release() +} + +// +stateify savable +type connectedEndpoint struct { + // endpoint represents the subset of the Endpoint functionality needed by + // the connectedEndpoint. It is implemented by both connectionedEndpoint + // and connectionlessEndpoint and allows the use of types which don't + // fully implement Endpoint. + endpoint interface { + // Passcred implements Endpoint.Passcred. + Passcred() bool + + // GetLocalAddress implements Endpoint.GetLocalAddress. + GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) + + // Type implements Endpoint.Type. + Type() SockType + } + + writeQueue *queue.Queue +} + +// Passcred implements ConnectedEndpoint.Passcred. +func (e *connectedEndpoint) Passcred() bool { + return e.endpoint.Passcred() +} + +// GetLocalAddress implements ConnectedEndpoint.GetLocalAddress. +func (e *connectedEndpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) { + return e.endpoint.GetLocalAddress() +} + +// Send implements ConnectedEndpoint.Send. +func (e *connectedEndpoint) Send(data [][]byte, controlMessages ControlMessages, from tcpip.FullAddress) (uintptr, bool, *tcpip.Error) { + var l int64 + for _, d := range data { + l += int64(len(d)) + } + + truncate := false + if e.endpoint.Type() == SockStream { + // Since stream sockets don't preserve message boundaries, we + // can write only as much of the message as fits in the queue. + truncate = true + + // Discard empty stream packets. Since stream sockets don't + // preserve message boundaries, sending zero bytes is a no-op. + // In Linux, the receiver actually uses a zero-length receive + // as an indication that the stream was closed. + if l == 0 { + controlMessages.Release() + return 0, false, nil + } + } + + v := make([]byte, 0, l) + for _, d := range data { + v = append(v, d...) + } + + l, notify, err := e.writeQueue.Enqueue(&message{Data: buffer.View(v), Control: controlMessages, Address: from}, truncate) + return uintptr(l), notify, err +} + +// SendNotify implements ConnectedEndpoint.SendNotify. +func (e *connectedEndpoint) SendNotify() { + e.writeQueue.ReaderQueue.Notify(waiter.EventIn) +} + +// CloseNotify implements ConnectedEndpoint.CloseNotify. +func (e *connectedEndpoint) CloseNotify() { + e.writeQueue.ReaderQueue.Notify(waiter.EventIn) + e.writeQueue.WriterQueue.Notify(waiter.EventOut) +} + +// CloseSend implements ConnectedEndpoint.CloseSend. +func (e *connectedEndpoint) CloseSend() { + e.writeQueue.Close() +} + +// Writable implements ConnectedEndpoint.Writable. +func (e *connectedEndpoint) Writable() bool { + return e.writeQueue.IsWritable() +} + +// EventUpdate implements ConnectedEndpoint.EventUpdate. +func (*connectedEndpoint) EventUpdate() {} + +// SendQueuedSize implements ConnectedEndpoint.SendQueuedSize. +func (e *connectedEndpoint) SendQueuedSize() int64 { + return e.writeQueue.QueuedSize() +} + +// SendMaxQueueSize implements ConnectedEndpoint.SendMaxQueueSize. +func (e *connectedEndpoint) SendMaxQueueSize() int64 { + return e.writeQueue.MaxQueueSize() +} + +// Release implements ConnectedEndpoint.Release. +func (*connectedEndpoint) Release() {} + +// baseEndpoint is an embeddable unix endpoint base used in both the connected and connectionless +// unix domain socket Endpoint implementations. +// +// Not to be used on its own. +// +// +stateify savable +type baseEndpoint struct { + *waiter.Queue + + // passcred specifies whether SCM_CREDENTIALS socket control messages are + // enabled on this endpoint. Must be accessed atomically. + passcred int32 + + // Mutex protects the below fields. + sync.Mutex `state:"nosave"` + + // receiver allows Messages to be received. + receiver Receiver + + // connected allows messages to be sent and state information about the + // connected endpoint to be read. + connected ConnectedEndpoint + + // path is not empty if the endpoint has been bound, + // or may be used if the endpoint is connected. + path string +} + +// EventRegister implements waiter.Waitable.EventRegister. +func (e *baseEndpoint) EventRegister(we *waiter.Entry, mask waiter.EventMask) { + e.Queue.EventRegister(we, mask) + e.Lock() + if e.connected != nil { + e.connected.EventUpdate() + } + e.Unlock() +} + +// EventUnregister implements waiter.Waitable.EventUnregister. +func (e *baseEndpoint) EventUnregister(we *waiter.Entry) { + e.Queue.EventUnregister(we) + e.Lock() + if e.connected != nil { + e.connected.EventUpdate() + } + e.Unlock() +} + +// Passcred implements Credentialer.Passcred. +func (e *baseEndpoint) Passcred() bool { + return atomic.LoadInt32(&e.passcred) != 0 +} + +// ConnectedPasscred implements Credentialer.ConnectedPasscred. +func (e *baseEndpoint) ConnectedPasscred() bool { + e.Lock() + defer e.Unlock() + return e.connected != nil && e.connected.Passcred() +} + +func (e *baseEndpoint) setPasscred(pc bool) { + if pc { + atomic.StoreInt32(&e.passcred, 1) + } else { + atomic.StoreInt32(&e.passcred, 0) + } +} + +// Connected implements ConnectingEndpoint.Connected. +func (e *baseEndpoint) Connected() bool { + return e.receiver != nil && e.connected != nil +} + +// RecvMsg reads data and a control message from the endpoint. +func (e *baseEndpoint) RecvMsg(data [][]byte, creds bool, numRights uintptr, peek bool, addr *tcpip.FullAddress) (uintptr, uintptr, ControlMessages, *tcpip.Error) { + e.Lock() + + if e.receiver == nil { + e.Unlock() + return 0, 0, ControlMessages{}, tcpip.ErrNotConnected + } + + recvLen, msgLen, cms, a, notify, err := e.receiver.Recv(data, creds, numRights, peek) + e.Unlock() + if err != nil { + return 0, 0, ControlMessages{}, err + } + + if notify { + e.receiver.RecvNotify() + } + + if addr != nil { + *addr = a + } + return recvLen, msgLen, cms, nil +} + +// SendMsg writes data and a control message to the endpoint's peer. +// This method does not block if the data cannot be written. +func (e *baseEndpoint) SendMsg(data [][]byte, c ControlMessages, to BoundEndpoint) (uintptr, *tcpip.Error) { + e.Lock() + if !e.Connected() { + e.Unlock() + return 0, tcpip.ErrNotConnected + } + if to != nil { + e.Unlock() + return 0, tcpip.ErrAlreadyConnected + } + + n, notify, err := e.connected.Send(data, c, tcpip.FullAddress{Addr: tcpip.Address(e.path)}) + e.Unlock() + + if notify { + e.connected.SendNotify() + } + + return n, err +} + +// SetSockOpt sets a socket option. Currently not supported. +func (e *baseEndpoint) SetSockOpt(opt interface{}) *tcpip.Error { + switch v := opt.(type) { + case tcpip.PasscredOption: + e.setPasscred(v != 0) + return nil + } + return nil +} + +// GetSockOpt implements tcpip.Endpoint.GetSockOpt. +func (e *baseEndpoint) GetSockOpt(opt interface{}) *tcpip.Error { + switch o := opt.(type) { + case tcpip.ErrorOption: + return nil + case *tcpip.SendQueueSizeOption: + e.Lock() + if !e.Connected() { + e.Unlock() + return tcpip.ErrNotConnected + } + qs := tcpip.SendQueueSizeOption(e.connected.SendQueuedSize()) + e.Unlock() + if qs < 0 { + return tcpip.ErrQueueSizeNotSupported + } + *o = qs + return nil + case *tcpip.ReceiveQueueSizeOption: + e.Lock() + if !e.Connected() { + e.Unlock() + return tcpip.ErrNotConnected + } + qs := tcpip.ReceiveQueueSizeOption(e.receiver.RecvQueuedSize()) + e.Unlock() + if qs < 0 { + return tcpip.ErrQueueSizeNotSupported + } + *o = qs + return nil + case *tcpip.PasscredOption: + if e.Passcred() { + *o = tcpip.PasscredOption(1) + } else { + *o = tcpip.PasscredOption(0) + } + return nil + case *tcpip.SendBufferSizeOption: + e.Lock() + if !e.Connected() { + e.Unlock() + return tcpip.ErrNotConnected + } + qs := tcpip.SendBufferSizeOption(e.connected.SendMaxQueueSize()) + e.Unlock() + if qs < 0 { + return tcpip.ErrQueueSizeNotSupported + } + *o = qs + return nil + case *tcpip.ReceiveBufferSizeOption: + e.Lock() + if e.receiver == nil { + e.Unlock() + return tcpip.ErrNotConnected + } + qs := tcpip.ReceiveBufferSizeOption(e.receiver.RecvMaxQueueSize()) + e.Unlock() + if qs < 0 { + return tcpip.ErrQueueSizeNotSupported + } + *o = qs + return nil + } + return tcpip.ErrUnknownProtocolOption +} + +// Shutdown closes the read and/or write end of the endpoint connection to its +// peer. +func (e *baseEndpoint) Shutdown(flags tcpip.ShutdownFlags) *tcpip.Error { + e.Lock() + if !e.Connected() { + e.Unlock() + return tcpip.ErrNotConnected + } + + if flags&tcpip.ShutdownRead != 0 { + e.receiver.CloseRecv() + } + + if flags&tcpip.ShutdownWrite != 0 { + e.connected.CloseSend() + } + + e.Unlock() + + if flags&tcpip.ShutdownRead != 0 { + e.receiver.CloseNotify() + } + + if flags&tcpip.ShutdownWrite != 0 { + e.connected.CloseNotify() + } + + return nil +} + +// GetLocalAddress returns the bound path. +func (e *baseEndpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) { + e.Lock() + defer e.Unlock() + return tcpip.FullAddress{Addr: tcpip.Address(e.path)}, nil +} + +// GetRemoteAddress returns the local address of the connected endpoint (if +// available). +func (e *baseEndpoint) GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) { + e.Lock() + c := e.connected + e.Unlock() + if c != nil { + return c.GetLocalAddress() + } + return tcpip.FullAddress{}, tcpip.ErrNotConnected +} + +// Release implements BoundEndpoint.Release. +func (*baseEndpoint) Release() {} diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index e30378e60..668363864 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -32,16 +32,16 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/socket" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/epsocket" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// SocketOperations is a Unix socket. It is similar to an epsocket, except it is backed -// by a unix.Endpoint instead of a tcpip.Endpoint. +// SocketOperations is a Unix socket. It is similar to an epsocket, except it +// is backed by a transport.Endpoint instead of a tcpip.Endpoint. // // +stateify savable type SocketOperations struct { @@ -52,18 +52,18 @@ type SocketOperations struct { fsutil.NoFsync `state:"nosave"` fsutil.NoopFlush `state:"nosave"` fsutil.NoMMap `state:"nosave"` - ep unix.Endpoint + ep transport.Endpoint } // New creates a new unix socket. -func New(ctx context.Context, endpoint unix.Endpoint) *fs.File { +func New(ctx context.Context, endpoint transport.Endpoint) *fs.File { dirent := socket.NewDirent(ctx, unixSocketDevice) defer dirent.DecRef() return NewWithDirent(ctx, dirent, endpoint, fs.FileFlags{Read: true, Write: true}) } // NewWithDirent creates a new unix socket using an existing dirent. -func NewWithDirent(ctx context.Context, d *fs.Dirent, ep unix.Endpoint, flags fs.FileFlags) *fs.File { +func NewWithDirent(ctx context.Context, d *fs.Dirent, ep transport.Endpoint, flags fs.FileFlags) *fs.File { return fs.NewFile(ctx, d, flags, &SocketOperations{ ep: ep, }) @@ -83,8 +83,8 @@ func (s *SocketOperations) Release() { s.DecRef() } -// Endpoint extracts the unix.Endpoint. -func (s *SocketOperations) Endpoint() unix.Endpoint { +// Endpoint extracts the transport.Endpoint. +func (s *SocketOperations) Endpoint() transport.Endpoint { return s.ep } @@ -110,7 +110,7 @@ func extractPath(sockaddr []byte) (string, *syserr.Error) { } // GetPeerName implements the linux syscall getpeername(2) for sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) GetPeerName(t *kernel.Task) (interface{}, uint32, *syserr.Error) { addr, err := s.ep.GetRemoteAddress() if err != nil { @@ -122,7 +122,7 @@ func (s *SocketOperations) GetPeerName(t *kernel.Task) (interface{}, uint32, *sy } // GetSockName implements the linux syscall getsockname(2) for sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) GetSockName(t *kernel.Task) (interface{}, uint32, *syserr.Error) { addr, err := s.ep.GetLocalAddress() if err != nil { @@ -139,20 +139,20 @@ func (s *SocketOperations) Ioctl(ctx context.Context, io usermem.IO, args arch.S } // GetSockOpt implements the linux syscall getsockopt(2) for sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) GetSockOpt(t *kernel.Task, level, name, outLen int) (interface{}, *syserr.Error) { return epsocket.GetSockOpt(t, s, s.ep, linux.AF_UNIX, s.ep.Type(), level, name, outLen) } // Listen implements the linux syscall listen(2) for sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) Listen(t *kernel.Task, backlog int) *syserr.Error { return syserr.TranslateNetstackError(s.ep.Listen(backlog)) } // blockingAccept implements a blocking version of accept(2), that is, if no // connections are ready to be accept, it will block until one becomes ready. -func (s *SocketOperations) blockingAccept(t *kernel.Task) (unix.Endpoint, *syserr.Error) { +func (s *SocketOperations) blockingAccept(t *kernel.Task) (transport.Endpoint, *syserr.Error) { // Register for notifications. e, ch := waiter.NewChannelEntry(nil) s.EventRegister(&e, waiter.EventIn) @@ -172,7 +172,7 @@ func (s *SocketOperations) blockingAccept(t *kernel.Task) (unix.Endpoint, *syser } // Accept implements the linux syscall accept(2) for sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) Accept(t *kernel.Task, peerRequested bool, flags int, blocking bool) (kdefs.FD, interface{}, uint32, *syserr.Error) { // Issue the accept request to get the new endpoint. ep, err := s.ep.Accept() @@ -226,7 +226,7 @@ func (s *SocketOperations) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { return e } - bep, ok := s.ep.(unix.BoundEndpoint) + bep, ok := s.ep.(transport.BoundEndpoint) if !ok { // This socket can't be bound. return syserr.ErrInvalidArgument @@ -287,10 +287,10 @@ func (s *SocketOperations) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { })) } -// extractEndpoint retrieves the unix.BoundEndpoint associated with a Unix -// socket path. The Release must be called on the unix.BoundEndpoint when the -// caller is done with it. -func extractEndpoint(t *kernel.Task, sockaddr []byte) (unix.BoundEndpoint, *syserr.Error) { +// extractEndpoint retrieves the transport.BoundEndpoint associated with a Unix +// socket path. The Release must be called on the transport.BoundEndpoint when +// the caller is done with it. +func extractEndpoint(t *kernel.Task, sockaddr []byte) (transport.BoundEndpoint, *syserr.Error) { path, err := extractPath(sockaddr) if err != nil { return nil, err @@ -362,7 +362,7 @@ func (s *SocketOperations) Write(ctx context.Context, _ *fs.File, src usermem.IO } // SendMsg implements the linux syscall sendmsg(2) for unix sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to []byte, flags int, controlMessages socket.ControlMessages) (int, *syserr.Error) { w := EndpointWriter{ Endpoint: s.ep, @@ -408,12 +408,12 @@ func (s *SocketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to [] return int(total), syserr.FromError(err) } -// Passcred implements unix.Credentialer.Passcred. +// Passcred implements transport.Credentialer.Passcred. func (s *SocketOperations) Passcred() bool { return s.ep.Passcred() } -// ConnectedPasscred implements unix.Credentialer.ConnectedPasscred. +// ConnectedPasscred implements transport.Credentialer.ConnectedPasscred. func (s *SocketOperations) ConnectedPasscred() bool { return s.ep.ConnectedPasscred() } @@ -434,13 +434,13 @@ func (s *SocketOperations) EventUnregister(e *waiter.Entry) { } // SetSockOpt implements the linux syscall setsockopt(2) for sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) SetSockOpt(t *kernel.Task, level int, name int, optVal []byte) *syserr.Error { return epsocket.SetSockOpt(t, s, s.ep, level, name, optVal) } // Shutdown implements the linux syscall shutdown(2) for sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) Shutdown(t *kernel.Task, how int) *syserr.Error { f, err := epsocket.ConvertShutdown(how) if err != nil { @@ -465,7 +465,7 @@ func (s *SocketOperations) Read(ctx context.Context, _ *fs.File, dst usermem.IOS } // RecvMsg implements the linux syscall recvmsg(2) for sockets backed by -// a unix.Endpoint. +// a transport.Endpoint. func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (n int, senderAddr interface{}, senderAddrLen uint32, controlMessages socket.ControlMessages, err *syserr.Error) { trunc := flags&linux.MSG_TRUNC != 0 peek := flags&linux.MSG_PEEK != 0 @@ -539,19 +539,19 @@ func (s *SocketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags type provider struct{} // Socket returns a new unix domain socket. -func (*provider) Socket(t *kernel.Task, stype unix.SockType, protocol int) (*fs.File, *syserr.Error) { +func (*provider) Socket(t *kernel.Task, stype transport.SockType, protocol int) (*fs.File, *syserr.Error) { // Check arguments. if protocol != 0 { return nil, syserr.ErrInvalidArgument } // Create the endpoint and socket. - var ep unix.Endpoint + var ep transport.Endpoint switch stype { case linux.SOCK_DGRAM: - ep = unix.NewConnectionless() + ep = transport.NewConnectionless() case linux.SOCK_STREAM, linux.SOCK_SEQPACKET: - ep = unix.NewConnectioned(stype, t.Kernel()) + ep = transport.NewConnectioned(stype, t.Kernel()) default: return nil, syserr.ErrInvalidArgument } @@ -560,7 +560,7 @@ func (*provider) Socket(t *kernel.Task, stype unix.SockType, protocol int) (*fs. } // Pair creates a new pair of AF_UNIX connected sockets. -func (*provider) Pair(t *kernel.Task, stype unix.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) { +func (*provider) Pair(t *kernel.Task, stype transport.SockType, protocol int) (*fs.File, *fs.File, *syserr.Error) { // Check arguments. if protocol != 0 { return nil, nil, syserr.ErrInvalidArgument @@ -573,7 +573,7 @@ func (*provider) Pair(t *kernel.Task, stype unix.SockType, protocol int) (*fs.Fi } // Create the endpoints and sockets. - ep1, ep2 := unix.NewPair(stype, t.Kernel()) + ep1, ep2 := transport.NewPair(stype, t.Kernel()) s1 := New(t, ep1) s2 := New(t, ep2) diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index bbdfad9da..7621bfdbd 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -79,11 +79,11 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket", "//pkg/sentry/socket/control", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/syscalls", "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/syserror", - "//pkg/tcpip/transport/unix", "//pkg/waiter", ], ) diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go index 867fec468..5fa5ddce6 100644 --- a/pkg/sentry/syscalls/linux/sys_socket.go +++ b/pkg/sentry/syscalls/linux/sys_socket.go @@ -27,9 +27,9 @@ import ( ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/control" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" ) // minListenBacklog is the minimum reasonable backlog for listening sockets. @@ -180,7 +180,7 @@ func Socket(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal } // Create the new socket. - s, e := socket.New(t, domain, unix.SockType(stype&0xf), protocol) + s, e := socket.New(t, domain, transport.SockType(stype&0xf), protocol) if e != nil { return 0, nil, e.ToError() } @@ -219,7 +219,7 @@ func SocketPair(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sy } // Create the socket pair. - s1, s2, e := socket.Pair(t, domain, unix.SockType(stype&0xf), protocol) + s1, s2, e := socket.Pair(t, domain, transport.SockType(stype&0xf), protocol) if e != nil { return 0, nil, e.ToError() } @@ -750,7 +750,7 @@ func recvSingleMsg(t *kernel.Task, s socket.Socket, msgPtr usermem.Addr, flags i controlData := make([]byte, 0, msg.ControlLen) - if cr, ok := s.(unix.Credentialer); ok && cr.Passcred() { + if cr, ok := s.(transport.Credentialer); ok && cr.Passcred() { creds, _ := cms.Unix.Credentials.(control.SCMCredentials) controlData = control.PackCredentials(t, creds, controlData) } diff --git a/pkg/sentry/uniqueid/BUILD b/pkg/sentry/uniqueid/BUILD index ff50b9925..68b82af47 100644 --- a/pkg/sentry/uniqueid/BUILD +++ b/pkg/sentry/uniqueid/BUILD @@ -9,6 +9,6 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/sentry/context", - "//pkg/tcpip/transport/unix", + "//pkg/sentry/socket/unix/transport", ], ) diff --git a/pkg/sentry/uniqueid/context.go b/pkg/sentry/uniqueid/context.go index 541e0611d..e48fabc2d 100644 --- a/pkg/sentry/uniqueid/context.go +++ b/pkg/sentry/uniqueid/context.go @@ -18,7 +18,7 @@ package uniqueid import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" ) // contextID is the kernel package's type for context.Context.Value keys. @@ -44,8 +44,8 @@ func GlobalFromContext(ctx context.Context) uint64 { } // GlobalProviderFromContext returns a system-wide unique identifier from ctx. -func GlobalProviderFromContext(ctx context.Context) unix.UniqueIDProvider { - return ctx.Value(CtxGlobalUniqueIDProvider).(unix.UniqueIDProvider) +func GlobalProviderFromContext(ctx context.Context) transport.UniqueIDProvider { + return ctx.Value(CtxGlobalUniqueIDProvider).(transport.UniqueIDProvider) } // InotifyCookie generates a unique inotify event cookie from ctx. diff --git a/pkg/tcpip/transport/queue/BUILD b/pkg/tcpip/transport/queue/BUILD deleted file mode 100644 index 6dcec312e..000000000 --- a/pkg/tcpip/transport/queue/BUILD +++ /dev/null @@ -1,15 +0,0 @@ -package(licenses = ["notice"]) # Apache 2.0 - -load("//tools/go_stateify:defs.bzl", "go_library") - -go_library( - name = "queue", - srcs = ["queue.go"], - importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/queue", - visibility = ["//:sandbox"], - deps = [ - "//pkg/ilist", - "//pkg/tcpip", - "//pkg/waiter", - ], -) diff --git a/pkg/tcpip/transport/queue/queue.go b/pkg/tcpip/transport/queue/queue.go deleted file mode 100644 index b3d2ea68b..000000000 --- a/pkg/tcpip/transport/queue/queue.go +++ /dev/null @@ -1,227 +0,0 @@ -// Copyright 2018 Google Inc. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package queue provides the implementation of buffer queue -// and interface of queue entry with Length method. -package queue - -import ( - "sync" - - "gvisor.googlesource.com/gvisor/pkg/ilist" - "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/waiter" -) - -// Entry implements Linker interface and has additional required methods. -type Entry interface { - ilist.Linker - - // Length returns the number of bytes stored in the entry. - Length() int64 - - // Release releases any resources held by the entry. - Release() - - // Peek returns a copy of the entry. It must be Released separately. - Peek() Entry - - // Truncate reduces the number of bytes stored in the entry to n bytes. - // - // Preconditions: n <= Length(). - Truncate(n int64) -} - -// Queue is a buffer queue. -// -// +stateify savable -type Queue struct { - ReaderQueue *waiter.Queue - WriterQueue *waiter.Queue - - mu sync.Mutex `state:"nosave"` - closed bool - used int64 - limit int64 - dataList ilist.List -} - -// New allocates and initializes a new queue. -func New(ReaderQueue *waiter.Queue, WriterQueue *waiter.Queue, limit int64) *Queue { - return &Queue{ReaderQueue: ReaderQueue, WriterQueue: WriterQueue, limit: limit} -} - -// Close closes q for reading and writing. It is immediately not writable and -// will become unreadable when no more data is pending. -// -// Both the read and write queues must be notified after closing: -// q.ReaderQueue.Notify(waiter.EventIn) -// q.WriterQueue.Notify(waiter.EventOut) -func (q *Queue) Close() { - q.mu.Lock() - q.closed = true - q.mu.Unlock() -} - -// Reset empties the queue and Releases all of the Entries. -// -// Both the read and write queues must be notified after resetting: -// q.ReaderQueue.Notify(waiter.EventIn) -// q.WriterQueue.Notify(waiter.EventOut) -func (q *Queue) Reset() { - q.mu.Lock() - for cur := q.dataList.Front(); cur != nil; cur = cur.Next() { - cur.(Entry).Release() - } - q.dataList.Reset() - q.used = 0 - q.mu.Unlock() -} - -// IsReadable determines if q is currently readable. -func (q *Queue) IsReadable() bool { - q.mu.Lock() - defer q.mu.Unlock() - - return q.closed || q.dataList.Front() != nil -} - -// bufWritable returns true if there is space for writing. -// -// N.B. Linux only considers a unix socket "writable" if >75% of the buffer is -// free. -// -// See net/unix/af_unix.c:unix_writeable. -func (q *Queue) bufWritable() bool { - return 4*q.used < q.limit -} - -// IsWritable determines if q is currently writable. -func (q *Queue) IsWritable() bool { - q.mu.Lock() - defer q.mu.Unlock() - - return q.closed || q.bufWritable() -} - -// Enqueue adds an entry to the data queue if room is available. -// -// If truncate is true, Enqueue may truncate the message beforing enqueuing it. -// Otherwise, the entire message must fit. If n < e.Length(), err indicates why. -// -// If notify is true, ReaderQueue.Notify must be called: -// q.ReaderQueue.Notify(waiter.EventIn) -func (q *Queue) Enqueue(e Entry, truncate bool) (l int64, notify bool, err *tcpip.Error) { - q.mu.Lock() - - if q.closed { - q.mu.Unlock() - return 0, false, tcpip.ErrClosedForSend - } - - free := q.limit - q.used - - l = e.Length() - - if l > free && truncate { - if free == 0 { - // Message can't fit right now. - q.mu.Unlock() - return 0, false, tcpip.ErrWouldBlock - } - - e.Truncate(free) - l = e.Length() - err = tcpip.ErrWouldBlock - } - - if l > q.limit { - // Message is too big to ever fit. - q.mu.Unlock() - return 0, false, tcpip.ErrMessageTooLong - } - - if l > free { - // Message can't fit right now. - q.mu.Unlock() - return 0, false, tcpip.ErrWouldBlock - } - - notify = q.dataList.Front() == nil - q.used += l - q.dataList.PushBack(e) - - q.mu.Unlock() - - return l, notify, err -} - -// Dequeue removes the first entry in the data queue, if one exists. -// -// If notify is true, WriterQueue.Notify must be called: -// q.WriterQueue.Notify(waiter.EventOut) -func (q *Queue) Dequeue() (e Entry, notify bool, err *tcpip.Error) { - q.mu.Lock() - - if q.dataList.Front() == nil { - err := tcpip.ErrWouldBlock - if q.closed { - err = tcpip.ErrClosedForReceive - } - q.mu.Unlock() - - return nil, false, err - } - - notify = !q.bufWritable() - - e = q.dataList.Front().(Entry) - q.dataList.Remove(e) - q.used -= e.Length() - - notify = notify && q.bufWritable() - - q.mu.Unlock() - - return e, notify, nil -} - -// Peek returns the first entry in the data queue, if one exists. -func (q *Queue) Peek() (Entry, *tcpip.Error) { - q.mu.Lock() - defer q.mu.Unlock() - - if q.dataList.Front() == nil { - err := tcpip.ErrWouldBlock - if q.closed { - err = tcpip.ErrClosedForReceive - } - return nil, err - } - - return q.dataList.Front().(Entry).Peek(), nil -} - -// QueuedSize returns the number of bytes currently in the queue, that is, the -// number of readable bytes. -func (q *Queue) QueuedSize() int64 { - q.mu.Lock() - defer q.mu.Unlock() - return q.used -} - -// MaxQueueSize returns the maximum number of bytes storable in the queue. -func (q *Queue) MaxQueueSize() int64 { - return q.limit -} diff --git a/pkg/tcpip/transport/unix/BUILD b/pkg/tcpip/transport/unix/BUILD deleted file mode 100644 index dae0bd079..000000000 --- a/pkg/tcpip/transport/unix/BUILD +++ /dev/null @@ -1,22 +0,0 @@ -package(licenses = ["notice"]) # Apache 2.0 - -load("//tools/go_stateify:defs.bzl", "go_library") - -go_library( - name = "unix", - srcs = [ - "connectioned.go", - "connectioned_state.go", - "connectionless.go", - "unix.go", - ], - importpath = "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/unix", - visibility = ["//:sandbox"], - deps = [ - "//pkg/ilist", - "//pkg/tcpip", - "//pkg/tcpip/buffer", - "//pkg/tcpip/transport/queue", - "//pkg/waiter", - ], -) diff --git a/pkg/tcpip/transport/unix/connectioned.go b/pkg/tcpip/transport/unix/connectioned.go deleted file mode 100644 index e319b3bb8..000000000 --- a/pkg/tcpip/transport/unix/connectioned.go +++ /dev/null @@ -1,454 +0,0 @@ -// Copyright 2018 Google Inc. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package unix - -import ( - "sync" - - "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/queue" - "gvisor.googlesource.com/gvisor/pkg/waiter" -) - -// UniqueIDProvider generates a sequence of unique identifiers useful for, -// among other things, lock ordering. -type UniqueIDProvider interface { - // UniqueID returns a new unique identifier. - UniqueID() uint64 -} - -// A ConnectingEndpoint is a connectioned unix endpoint that is attempting to -// establish a bidirectional connection with a BoundEndpoint. -type ConnectingEndpoint interface { - // ID returns the endpoint's globally unique identifier. This identifier - // must be used to determine locking order if more than one endpoint is - // to be locked in the same codepath. The endpoint with the smaller - // identifier must be locked before endpoints with larger identifiers. - ID() uint64 - - // Passcred implements socket.Credentialer.Passcred. - Passcred() bool - - // Type returns the socket type, typically either SockStream or - // SockSeqpacket. The connection attempt must be aborted if this - // value doesn't match the ConnectableEndpoint's type. - Type() SockType - - // GetLocalAddress returns the bound path. - GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) - - // Locker protects the following methods. While locked, only the holder of - // the lock can change the return value of the protected methods. - sync.Locker - - // Connected returns true iff the ConnectingEndpoint is in the connected - // state. ConnectingEndpoints can only be connected to a single endpoint, - // so the connection attempt must be aborted if this returns true. - Connected() bool - - // Listening returns true iff the ConnectingEndpoint is in the listening - // state. ConnectingEndpoints cannot make connections while listening, so - // the connection attempt must be aborted if this returns true. - Listening() bool - - // WaiterQueue returns a pointer to the endpoint's waiter queue. - WaiterQueue() *waiter.Queue -} - -// connectionedEndpoint is a Unix-domain connected or connectable endpoint and implements -// ConnectingEndpoint, ConnectableEndpoint and tcpip.Endpoint. -// -// connectionedEndpoints must be in connected state in order to transfer data. -// -// This implementation includes STREAM and SEQPACKET Unix sockets created with -// socket(2), accept(2) or socketpair(2) and dgram unix sockets created with -// socketpair(2). See unix_connectionless.go for the implementation of DGRAM -// Unix sockets created with socket(2). -// -// The state is much simpler than a TCP endpoint, so it is not encoded -// explicitly. Instead we enforce the following invariants: -// -// receiver != nil, connected != nil => connected. -// path != "" && acceptedChan == nil => bound, not listening. -// path != "" && acceptedChan != nil => bound and listening. -// -// Only one of these will be true at any moment. -// -// +stateify savable -type connectionedEndpoint struct { - baseEndpoint - - // id is the unique endpoint identifier. This is used exclusively for - // lock ordering within connect. - id uint64 - - // idGenerator is used to generate new unique endpoint identifiers. - idGenerator UniqueIDProvider - - // stype is used by connecting sockets to ensure that they are the - // same type. The value is typically either tcpip.SockSeqpacket or - // tcpip.SockStream. - stype SockType - - // acceptedChan is per the TCP endpoint implementation. Note that the - // sockets in this channel are _already in the connected state_, and - // have another associated connectionedEndpoint. - // - // If nil, then no listen call has been made. - acceptedChan chan *connectionedEndpoint `state:".([]*connectionedEndpoint)"` -} - -// NewConnectioned creates a new unbound connectionedEndpoint. -func NewConnectioned(stype SockType, uid UniqueIDProvider) Endpoint { - return &connectionedEndpoint{ - baseEndpoint: baseEndpoint{Queue: &waiter.Queue{}}, - id: uid.UniqueID(), - idGenerator: uid, - stype: stype, - } -} - -// NewPair allocates a new pair of connected unix-domain connectionedEndpoints. -func NewPair(stype SockType, uid UniqueIDProvider) (Endpoint, Endpoint) { - a := &connectionedEndpoint{ - baseEndpoint: baseEndpoint{Queue: &waiter.Queue{}}, - id: uid.UniqueID(), - idGenerator: uid, - stype: stype, - } - b := &connectionedEndpoint{ - baseEndpoint: baseEndpoint{Queue: &waiter.Queue{}}, - id: uid.UniqueID(), - idGenerator: uid, - stype: stype, - } - - q1 := queue.New(a.Queue, b.Queue, initialLimit) - q2 := queue.New(b.Queue, a.Queue, initialLimit) - - if stype == SockStream { - a.receiver = &streamQueueReceiver{queueReceiver: queueReceiver{q1}} - b.receiver = &streamQueueReceiver{queueReceiver: queueReceiver{q2}} - } else { - a.receiver = &queueReceiver{q1} - b.receiver = &queueReceiver{q2} - } - - a.connected = &connectedEndpoint{ - endpoint: b, - writeQueue: q2, - } - b.connected = &connectedEndpoint{ - endpoint: a, - writeQueue: q1, - } - - return a, b -} - -// NewExternal creates a new externally backed Endpoint. It behaves like a -// socketpair. -func NewExternal(stype SockType, uid UniqueIDProvider, queue *waiter.Queue, receiver Receiver, connected ConnectedEndpoint) Endpoint { - return &connectionedEndpoint{ - baseEndpoint: baseEndpoint{Queue: queue, receiver: receiver, connected: connected}, - id: uid.UniqueID(), - idGenerator: uid, - stype: stype, - } -} - -// ID implements ConnectingEndpoint.ID. -func (e *connectionedEndpoint) ID() uint64 { - return e.id -} - -// Type implements ConnectingEndpoint.Type and Endpoint.Type. -func (e *connectionedEndpoint) Type() SockType { - return e.stype -} - -// WaiterQueue implements ConnectingEndpoint.WaiterQueue. -func (e *connectionedEndpoint) WaiterQueue() *waiter.Queue { - return e.Queue -} - -// isBound returns true iff the connectionedEndpoint is bound (but not -// listening). -func (e *connectionedEndpoint) isBound() bool { - return e.path != "" && e.acceptedChan == nil -} - -// Listening implements ConnectingEndpoint.Listening. -func (e *connectionedEndpoint) Listening() bool { - return e.acceptedChan != nil -} - -// Close puts the connectionedEndpoint in a closed state and frees all -// resources associated with it. -// -// The socket will be a fresh state after a call to close and may be reused. -// That is, close may be used to "unbind" or "disconnect" the socket in error -// paths. -func (e *connectionedEndpoint) Close() { - e.Lock() - var c ConnectedEndpoint - var r Receiver - switch { - case e.Connected(): - e.connected.CloseSend() - e.receiver.CloseRecv() - c = e.connected - r = e.receiver - e.connected = nil - e.receiver = nil - case e.isBound(): - e.path = "" - case e.Listening(): - close(e.acceptedChan) - for n := range e.acceptedChan { - n.Close() - } - e.acceptedChan = nil - e.path = "" - } - e.Unlock() - if c != nil { - c.CloseNotify() - c.Release() - } - if r != nil { - r.CloseNotify() - r.Release() - } -} - -// BidirectionalConnect implements BoundEndpoint.BidirectionalConnect. -func (e *connectionedEndpoint) BidirectionalConnect(ce ConnectingEndpoint, returnConnect func(Receiver, ConnectedEndpoint)) *tcpip.Error { - if ce.Type() != e.stype { - return tcpip.ErrConnectionRefused - } - - // Check if ce is e to avoid a deadlock. - if ce, ok := ce.(*connectionedEndpoint); ok && ce == e { - return tcpip.ErrInvalidEndpointState - } - - // Do a dance to safely acquire locks on both endpoints. - if e.id < ce.ID() { - e.Lock() - ce.Lock() - } else { - ce.Lock() - e.Lock() - } - - // Check connecting state. - if ce.Connected() { - e.Unlock() - ce.Unlock() - return tcpip.ErrAlreadyConnected - } - if ce.Listening() { - e.Unlock() - ce.Unlock() - return tcpip.ErrInvalidEndpointState - } - - // Check bound state. - if !e.Listening() { - e.Unlock() - ce.Unlock() - return tcpip.ErrConnectionRefused - } - - // Create a newly bound connectionedEndpoint. - ne := &connectionedEndpoint{ - baseEndpoint: baseEndpoint{ - path: e.path, - Queue: &waiter.Queue{}, - }, - id: e.idGenerator.UniqueID(), - idGenerator: e.idGenerator, - stype: e.stype, - } - readQueue := queue.New(ce.WaiterQueue(), ne.Queue, initialLimit) - writeQueue := queue.New(ne.Queue, ce.WaiterQueue(), initialLimit) - ne.connected = &connectedEndpoint{ - endpoint: ce, - writeQueue: readQueue, - } - if e.stype == SockStream { - ne.receiver = &streamQueueReceiver{queueReceiver: queueReceiver{readQueue: writeQueue}} - } else { - ne.receiver = &queueReceiver{readQueue: writeQueue} - } - - select { - case e.acceptedChan <- ne: - // Commit state. - connected := &connectedEndpoint{ - endpoint: ne, - writeQueue: writeQueue, - } - if e.stype == SockStream { - returnConnect(&streamQueueReceiver{queueReceiver: queueReceiver{readQueue: readQueue}}, connected) - } else { - returnConnect(&queueReceiver{readQueue: readQueue}, connected) - } - - // Notify can deadlock if we are holding these locks. - e.Unlock() - ce.Unlock() - - // Notify on both ends. - e.Notify(waiter.EventIn) - ce.WaiterQueue().Notify(waiter.EventOut) - - return nil - default: - // Busy; return ECONNREFUSED per spec. - ne.Close() - e.Unlock() - ce.Unlock() - return tcpip.ErrConnectionRefused - } -} - -// UnidirectionalConnect implements BoundEndpoint.UnidirectionalConnect. -func (e *connectionedEndpoint) UnidirectionalConnect() (ConnectedEndpoint, *tcpip.Error) { - return nil, tcpip.ErrConnectionRefused -} - -// Connect attempts to directly connect to another Endpoint. -// Implements Endpoint.Connect. -func (e *connectionedEndpoint) Connect(server BoundEndpoint) *tcpip.Error { - returnConnect := func(r Receiver, ce ConnectedEndpoint) { - e.receiver = r - e.connected = ce - } - - return server.BidirectionalConnect(e, returnConnect) -} - -// Listen starts listening on the connection. -func (e *connectionedEndpoint) Listen(backlog int) *tcpip.Error { - e.Lock() - defer e.Unlock() - if e.Listening() { - // Adjust the size of the channel iff we can fix existing - // pending connections into the new one. - if len(e.acceptedChan) > backlog { - return tcpip.ErrInvalidEndpointState - } - origChan := e.acceptedChan - e.acceptedChan = make(chan *connectionedEndpoint, backlog) - close(origChan) - for ep := range origChan { - e.acceptedChan <- ep - } - return nil - } - if !e.isBound() { - return tcpip.ErrInvalidEndpointState - } - - // Normal case. - e.acceptedChan = make(chan *connectionedEndpoint, backlog) - return nil -} - -// Accept accepts a new connection. -func (e *connectionedEndpoint) Accept() (Endpoint, *tcpip.Error) { - e.Lock() - defer e.Unlock() - - if !e.Listening() { - return nil, tcpip.ErrInvalidEndpointState - } - - select { - case ne := <-e.acceptedChan: - return ne, nil - - default: - // Nothing left. - return nil, tcpip.ErrWouldBlock - } -} - -// Bind binds the connection. -// -// For Unix connectionedEndpoints, this _only sets the address associated with -// the socket_. Work associated with sockets in the filesystem or finding those -// sockets must be done by a higher level. -// -// Bind will fail only if the socket is connected, bound or the passed address -// is invalid (the empty string). -func (e *connectionedEndpoint) Bind(addr tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error { - e.Lock() - defer e.Unlock() - if e.isBound() || e.Listening() { - return tcpip.ErrAlreadyBound - } - if addr.Addr == "" { - // The empty string is not permitted. - return tcpip.ErrBadLocalAddress - } - if commit != nil { - if err := commit(); err != nil { - return err - } - } - - // Save the bound address. - e.path = string(addr.Addr) - return nil -} - -// SendMsg writes data and a control message to the endpoint's peer. -// This method does not block if the data cannot be written. -func (e *connectionedEndpoint) SendMsg(data [][]byte, c ControlMessages, to BoundEndpoint) (uintptr, *tcpip.Error) { - // Stream sockets do not support specifying the endpoint. Seqpacket - // sockets ignore the passed endpoint. - if e.stype == SockStream && to != nil { - return 0, tcpip.ErrNotSupported - } - return e.baseEndpoint.SendMsg(data, c, to) -} - -// Readiness returns the current readiness of the connectionedEndpoint. For -// example, if waiter.EventIn is set, the connectionedEndpoint is immediately -// readable. -func (e *connectionedEndpoint) Readiness(mask waiter.EventMask) waiter.EventMask { - e.Lock() - defer e.Unlock() - - ready := waiter.EventMask(0) - switch { - case e.Connected(): - if mask&waiter.EventIn != 0 && e.receiver.Readable() { - ready |= waiter.EventIn - } - if mask&waiter.EventOut != 0 && e.connected.Writable() { - ready |= waiter.EventOut - } - case e.Listening(): - if mask&waiter.EventIn != 0 && len(e.acceptedChan) > 0 { - ready |= waiter.EventIn - } - } - - return ready -} diff --git a/pkg/tcpip/transport/unix/connectioned_state.go b/pkg/tcpip/transport/unix/connectioned_state.go deleted file mode 100644 index 39e0ca2d6..000000000 --- a/pkg/tcpip/transport/unix/connectioned_state.go +++ /dev/null @@ -1,53 +0,0 @@ -// Copyright 2018 Google Inc. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package unix - -// saveAcceptedChan is invoked by stateify. -func (e *connectionedEndpoint) saveAcceptedChan() []*connectionedEndpoint { - // If acceptedChan is nil (i.e. we are not listening) then we will save nil. - // Otherwise we create a (possibly empty) slice of the values in acceptedChan and - // save that. - var acceptedSlice []*connectionedEndpoint - if e.acceptedChan != nil { - // Swap out acceptedChan with a new empty channel of the same capacity. - saveChan := e.acceptedChan - e.acceptedChan = make(chan *connectionedEndpoint, cap(saveChan)) - - // Create a new slice with the same len and capacity as the channel. - acceptedSlice = make([]*connectionedEndpoint, len(saveChan), cap(saveChan)) - // Drain acceptedChan into saveSlice, and fill up the new acceptChan at the - // same time. - for i := range acceptedSlice { - ep := <-saveChan - acceptedSlice[i] = ep - e.acceptedChan <- ep - } - close(saveChan) - } - return acceptedSlice -} - -// loadAcceptedChan is invoked by stateify. -func (e *connectionedEndpoint) loadAcceptedChan(acceptedSlice []*connectionedEndpoint) { - // If acceptedSlice is nil, then acceptedChan should also be nil. - if acceptedSlice != nil { - // Otherwise, create a new channel with the same capacity as acceptedSlice. - e.acceptedChan = make(chan *connectionedEndpoint, cap(acceptedSlice)) - // Seed the channel with values from acceptedSlice. - for _, ep := range acceptedSlice { - e.acceptedChan <- ep - } - } -} diff --git a/pkg/tcpip/transport/unix/connectionless.go b/pkg/tcpip/transport/unix/connectionless.go deleted file mode 100644 index ae93c61d7..000000000 --- a/pkg/tcpip/transport/unix/connectionless.go +++ /dev/null @@ -1,192 +0,0 @@ -// Copyright 2018 Google Inc. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package unix - -import ( - "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/queue" - "gvisor.googlesource.com/gvisor/pkg/waiter" -) - -// connectionlessEndpoint is a unix endpoint for unix sockets that support operating in -// a conectionless fashon. -// -// Specifically, this means datagram unix sockets not created with -// socketpair(2). -// -// +stateify savable -type connectionlessEndpoint struct { - baseEndpoint -} - -// NewConnectionless creates a new unbound dgram endpoint. -func NewConnectionless() Endpoint { - ep := &connectionlessEndpoint{baseEndpoint{Queue: &waiter.Queue{}}} - ep.receiver = &queueReceiver{readQueue: queue.New(&waiter.Queue{}, ep.Queue, initialLimit)} - return ep -} - -// isBound returns true iff the endpoint is bound. -func (e *connectionlessEndpoint) isBound() bool { - return e.path != "" -} - -// Close puts the endpoint in a closed state and frees all resources associated -// with it. -// -// The socket will be a fresh state after a call to close and may be reused. -// That is, close may be used to "unbind" or "disconnect" the socket in error -// paths. -func (e *connectionlessEndpoint) Close() { - e.Lock() - var r Receiver - if e.Connected() { - e.receiver.CloseRecv() - r = e.receiver - e.receiver = nil - - e.connected.Release() - e.connected = nil - } - if e.isBound() { - e.path = "" - } - e.Unlock() - if r != nil { - r.CloseNotify() - r.Release() - } -} - -// BidirectionalConnect implements BoundEndpoint.BidirectionalConnect. -func (e *connectionlessEndpoint) BidirectionalConnect(ce ConnectingEndpoint, returnConnect func(Receiver, ConnectedEndpoint)) *tcpip.Error { - return tcpip.ErrConnectionRefused -} - -// UnidirectionalConnect implements BoundEndpoint.UnidirectionalConnect. -func (e *connectionlessEndpoint) UnidirectionalConnect() (ConnectedEndpoint, *tcpip.Error) { - e.Lock() - r := e.receiver - e.Unlock() - if r == nil { - return nil, tcpip.ErrConnectionRefused - } - return &connectedEndpoint{ - endpoint: e, - writeQueue: r.(*queueReceiver).readQueue, - }, nil -} - -// SendMsg writes data and a control message to the specified endpoint. -// This method does not block if the data cannot be written. -func (e *connectionlessEndpoint) SendMsg(data [][]byte, c ControlMessages, to BoundEndpoint) (uintptr, *tcpip.Error) { - if to == nil { - return e.baseEndpoint.SendMsg(data, c, nil) - } - - connected, err := to.UnidirectionalConnect() - if err != nil { - return 0, tcpip.ErrInvalidEndpointState - } - defer connected.Release() - - e.Lock() - n, notify, err := connected.Send(data, c, tcpip.FullAddress{Addr: tcpip.Address(e.path)}) - e.Unlock() - - if notify { - connected.SendNotify() - } - - return n, err -} - -// Type implements Endpoint.Type. -func (e *connectionlessEndpoint) Type() SockType { - return SockDgram -} - -// Connect attempts to connect directly to server. -func (e *connectionlessEndpoint) Connect(server BoundEndpoint) *tcpip.Error { - connected, err := server.UnidirectionalConnect() - if err != nil { - return err - } - - e.Lock() - e.connected = connected - e.Unlock() - - return nil -} - -// Listen starts listening on the connection. -func (e *connectionlessEndpoint) Listen(int) *tcpip.Error { - return tcpip.ErrNotSupported -} - -// Accept accepts a new connection. -func (e *connectionlessEndpoint) Accept() (Endpoint, *tcpip.Error) { - return nil, tcpip.ErrNotSupported -} - -// Bind binds the connection. -// -// For Unix endpoints, this _only sets the address associated with the socket_. -// Work associated with sockets in the filesystem or finding those sockets must -// be done by a higher level. -// -// Bind will fail only if the socket is connected, bound or the passed address -// is invalid (the empty string). -func (e *connectionlessEndpoint) Bind(addr tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error { - e.Lock() - defer e.Unlock() - if e.isBound() { - return tcpip.ErrAlreadyBound - } - if addr.Addr == "" { - // The empty string is not permitted. - return tcpip.ErrBadLocalAddress - } - if commit != nil { - if err := commit(); err != nil { - return err - } - } - - // Save the bound address. - e.path = string(addr.Addr) - return nil -} - -// Readiness returns the current readiness of the endpoint. For example, if -// waiter.EventIn is set, the endpoint is immediately readable. -func (e *connectionlessEndpoint) Readiness(mask waiter.EventMask) waiter.EventMask { - e.Lock() - defer e.Unlock() - - ready := waiter.EventMask(0) - if mask&waiter.EventIn != 0 && e.receiver.Readable() { - ready |= waiter.EventIn - } - - if e.Connected() { - if mask&waiter.EventOut != 0 && e.connected.Writable() { - ready |= waiter.EventOut - } - } - - return ready -} diff --git a/pkg/tcpip/transport/unix/unix.go b/pkg/tcpip/transport/unix/unix.go deleted file mode 100644 index 1bca4b0b4..000000000 --- a/pkg/tcpip/transport/unix/unix.go +++ /dev/null @@ -1,953 +0,0 @@ -// Copyright 2018 Google Inc. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package unix contains the implementation of Unix endpoints. -package unix - -import ( - "sync" - "sync/atomic" - - "gvisor.googlesource.com/gvisor/pkg/ilist" - "gvisor.googlesource.com/gvisor/pkg/tcpip" - "gvisor.googlesource.com/gvisor/pkg/tcpip/buffer" - "gvisor.googlesource.com/gvisor/pkg/tcpip/transport/queue" - "gvisor.googlesource.com/gvisor/pkg/waiter" -) - -// initialLimit is the starting limit for the socket buffers. -const initialLimit = 16 * 1024 - -// A SockType is a type (as opposed to family) of sockets. These are enumerated -// in the syscall package as syscall.SOCK_* constants. -type SockType int - -const ( - // SockStream corresponds to syscall.SOCK_STREAM. - SockStream SockType = 1 - // SockDgram corresponds to syscall.SOCK_DGRAM. - SockDgram SockType = 2 - // SockRaw corresponds to syscall.SOCK_RAW. - SockRaw SockType = 3 - // SockSeqpacket corresponds to syscall.SOCK_SEQPACKET. - SockSeqpacket SockType = 5 -) - -// A RightsControlMessage is a control message containing FDs. -type RightsControlMessage interface { - // Clone returns a copy of the RightsControlMessage. - Clone() RightsControlMessage - - // Release releases any resources owned by the RightsControlMessage. - Release() -} - -// A CredentialsControlMessage is a control message containing Unix credentials. -type CredentialsControlMessage interface { - // Equals returns true iff the two messages are equal. - Equals(CredentialsControlMessage) bool -} - -// A ControlMessages represents a collection of socket control messages. -// -// +stateify savable -type ControlMessages struct { - // Rights is a control message containing FDs. - Rights RightsControlMessage - - // Credentials is a control message containing Unix credentials. - Credentials CredentialsControlMessage -} - -// Empty returns true iff the ControlMessages does not contain either -// credentials or rights. -func (c *ControlMessages) Empty() bool { - return c.Rights == nil && c.Credentials == nil -} - -// Clone clones both the credentials and the rights. -func (c *ControlMessages) Clone() ControlMessages { - cm := ControlMessages{} - if c.Rights != nil { - cm.Rights = c.Rights.Clone() - } - cm.Credentials = c.Credentials - return cm -} - -// Release releases both the credentials and the rights. -func (c *ControlMessages) Release() { - if c.Rights != nil { - c.Rights.Release() - } - *c = ControlMessages{} -} - -// Endpoint is the interface implemented by Unix transport protocol -// implementations that expose functionality like sendmsg, recvmsg, connect, -// etc. to Unix socket implementations. -type Endpoint interface { - Credentialer - waiter.Waitable - - // Close puts the endpoint in a closed state and frees all resources - // associated with it. - Close() - - // RecvMsg reads data and a control message from the endpoint. This method - // does not block if there is no data pending. - // - // creds indicates if credential control messages are requested by the - // caller. This is useful for determining if control messages can be - // coalesced. creds is a hint and can be safely ignored by the - // implementation if no coalescing is possible. It is fine to return - // credential control messages when none were requested or to not return - // credential control messages when they were requested. - // - // numRights is the number of SCM_RIGHTS FDs requested by the caller. This - // is useful if one must allocate a buffer to receive a SCM_RIGHTS message - // or determine if control messages can be coalesced. numRights is a hint - // and can be safely ignored by the implementation if the number of - // available SCM_RIGHTS FDs is known and no coalescing is possible. It is - // fine for the returned number of SCM_RIGHTS FDs to be either higher or - // lower than the requested number. - // - // If peek is true, no data should be consumed from the Endpoint. Any and - // all data returned from a peek should be available in the next call to - // RecvMsg. - // - // recvLen is the number of bytes copied into data. - // - // msgLen is the length of the read message consumed for datagram Endpoints. - // msgLen is always the same as recvLen for stream Endpoints. - RecvMsg(data [][]byte, creds bool, numRights uintptr, peek bool, addr *tcpip.FullAddress) (recvLen, msgLen uintptr, cm ControlMessages, err *tcpip.Error) - - // SendMsg writes data and a control message to the endpoint's peer. - // This method does not block if the data cannot be written. - // - // SendMsg does not take ownership of any of its arguments on error. - SendMsg([][]byte, ControlMessages, BoundEndpoint) (uintptr, *tcpip.Error) - - // Connect connects this endpoint directly to another. - // - // This should be called on the client endpoint, and the (bound) - // endpoint passed in as a parameter. - // - // The error codes are the same as Connect. - Connect(server BoundEndpoint) *tcpip.Error - - // Shutdown closes the read and/or write end of the endpoint connection - // to its peer. - Shutdown(flags tcpip.ShutdownFlags) *tcpip.Error - - // Listen puts the endpoint in "listen" mode, which allows it to accept - // new connections. - Listen(backlog int) *tcpip.Error - - // Accept returns a new endpoint if a peer has established a connection - // to an endpoint previously set to listen mode. This method does not - // block if no new connections are available. - // - // The returned Queue is the wait queue for the newly created endpoint. - Accept() (Endpoint, *tcpip.Error) - - // Bind binds the endpoint to a specific local address and port. - // Specifying a NIC is optional. - // - // An optional commit function will be executed atomically with respect - // to binding the endpoint. If this returns an error, the bind will not - // occur and the error will be propagated back to the caller. - Bind(address tcpip.FullAddress, commit func() *tcpip.Error) *tcpip.Error - - // Type return the socket type, typically either SockStream, SockDgram - // or SockSeqpacket. - Type() SockType - - // GetLocalAddress returns the address to which the endpoint is bound. - GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) - - // GetRemoteAddress returns the address to which the endpoint is - // connected. - GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) - - // SetSockOpt sets a socket option. opt should be one of the tcpip.*Option - // types. - SetSockOpt(opt interface{}) *tcpip.Error - - // GetSockOpt gets a socket option. opt should be a pointer to one of the - // tcpip.*Option types. - GetSockOpt(opt interface{}) *tcpip.Error -} - -// A Credentialer is a socket or endpoint that supports the SO_PASSCRED socket -// option. -type Credentialer interface { - // Passcred returns whether or not the SO_PASSCRED socket option is - // enabled on this end. - Passcred() bool - - // ConnectedPasscred returns whether or not the SO_PASSCRED socket option - // is enabled on the connected end. - ConnectedPasscred() bool -} - -// A BoundEndpoint is a unix endpoint that can be connected to. -type BoundEndpoint interface { - // BidirectionalConnect establishes a bi-directional connection between two - // unix endpoints in an all-or-nothing manner. If an error occurs during - // connecting, the state of neither endpoint should be modified. - // - // In order for an endpoint to establish such a bidirectional connection - // with a BoundEndpoint, the endpoint calls the BidirectionalConnect method - // on the BoundEndpoint and sends a representation of itself (the - // ConnectingEndpoint) and a callback (returnConnect) to receive the - // connection information (Receiver and ConnectedEndpoint) upon a - // successful connect. The callback should only be called on a successful - // connect. - // - // For a connection attempt to be successful, the ConnectingEndpoint must - // be unconnected and not listening and the BoundEndpoint whose - // BidirectionalConnect method is being called must be listening. - // - // This method will return tcpip.ErrConnectionRefused on endpoints with a - // type that isn't SockStream or SockSeqpacket. - BidirectionalConnect(ep ConnectingEndpoint, returnConnect func(Receiver, ConnectedEndpoint)) *tcpip.Error - - // UnidirectionalConnect establishes a write-only connection to a unix - // endpoint. - // - // An endpoint which calls UnidirectionalConnect and supports it itself must - // not hold its own lock when calling UnidirectionalConnect. - // - // This method will return tcpip.ErrConnectionRefused on a non-SockDgram - // endpoint. - UnidirectionalConnect() (ConnectedEndpoint, *tcpip.Error) - - // Release releases any resources held by the BoundEndpoint. It must be - // called before dropping all references to a BoundEndpoint returned by a - // function. - Release() -} - -// message represents a message passed over a Unix domain socket. -// -// +stateify savable -type message struct { - ilist.Entry - - // Data is the Message payload. - Data buffer.View - - // Control is auxiliary control message data that goes along with the - // data. - Control ControlMessages - - // Address is the bound address of the endpoint that sent the message. - // - // If the endpoint that sent the message is not bound, the Address is - // the empty string. - Address tcpip.FullAddress -} - -// Length returns number of bytes stored in the message. -func (m *message) Length() int64 { - return int64(len(m.Data)) -} - -// Release releases any resources held by the message. -func (m *message) Release() { - m.Control.Release() -} - -// Peek returns a copy of the message. -func (m *message) Peek() queue.Entry { - return &message{Data: m.Data, Control: m.Control.Clone(), Address: m.Address} -} - -// Truncate reduces the length of the message payload to n bytes. -// -// Preconditions: n <= m.Length(). -func (m *message) Truncate(n int64) { - m.Data.CapLength(int(n)) -} - -// A Receiver can be used to receive Messages. -type Receiver interface { - // Recv receives a single message. This method does not block. - // - // See Endpoint.RecvMsg for documentation on shared arguments. - // - // notify indicates if RecvNotify should be called. - Recv(data [][]byte, creds bool, numRights uintptr, peek bool) (recvLen, msgLen uintptr, cm ControlMessages, source tcpip.FullAddress, notify bool, err *tcpip.Error) - - // RecvNotify notifies the Receiver of a successful Recv. This must not be - // called while holding any endpoint locks. - RecvNotify() - - // CloseRecv prevents the receiving of additional Messages. - // - // After CloseRecv is called, CloseNotify must also be called. - CloseRecv() - - // CloseNotify notifies the Receiver of recv being closed. This must not be - // called while holding any endpoint locks. - CloseNotify() - - // Readable returns if messages should be attempted to be received. This - // includes when read has been shutdown. - Readable() bool - - // RecvQueuedSize returns the total amount of data currently receivable. - // RecvQueuedSize should return -1 if the operation isn't supported. - RecvQueuedSize() int64 - - // RecvMaxQueueSize returns maximum value for RecvQueuedSize. - // RecvMaxQueueSize should return -1 if the operation isn't supported. - RecvMaxQueueSize() int64 - - // Release releases any resources owned by the Receiver. It should be - // called before droping all references to a Receiver. - Release() -} - -// queueReceiver implements Receiver for datagram sockets. -// -// +stateify savable -type queueReceiver struct { - readQueue *queue.Queue -} - -// Recv implements Receiver.Recv. -func (q *queueReceiver) Recv(data [][]byte, creds bool, numRights uintptr, peek bool) (uintptr, uintptr, ControlMessages, tcpip.FullAddress, bool, *tcpip.Error) { - var m queue.Entry - var notify bool - var err *tcpip.Error - if peek { - m, err = q.readQueue.Peek() - } else { - m, notify, err = q.readQueue.Dequeue() - } - if err != nil { - return 0, 0, ControlMessages{}, tcpip.FullAddress{}, false, err - } - msg := m.(*message) - src := []byte(msg.Data) - var copied uintptr - for i := 0; i < len(data) && len(src) > 0; i++ { - n := copy(data[i], src) - copied += uintptr(n) - src = src[n:] - } - return copied, uintptr(len(msg.Data)), msg.Control, msg.Address, notify, nil -} - -// RecvNotify implements Receiver.RecvNotify. -func (q *queueReceiver) RecvNotify() { - q.readQueue.WriterQueue.Notify(waiter.EventOut) -} - -// CloseNotify implements Receiver.CloseNotify. -func (q *queueReceiver) CloseNotify() { - q.readQueue.ReaderQueue.Notify(waiter.EventIn) - q.readQueue.WriterQueue.Notify(waiter.EventOut) -} - -// CloseRecv implements Receiver.CloseRecv. -func (q *queueReceiver) CloseRecv() { - q.readQueue.Close() -} - -// Readable implements Receiver.Readable. -func (q *queueReceiver) Readable() bool { - return q.readQueue.IsReadable() -} - -// RecvQueuedSize implements Receiver.RecvQueuedSize. -func (q *queueReceiver) RecvQueuedSize() int64 { - return q.readQueue.QueuedSize() -} - -// RecvMaxQueueSize implements Receiver.RecvMaxQueueSize. -func (q *queueReceiver) RecvMaxQueueSize() int64 { - return q.readQueue.MaxQueueSize() -} - -// Release implements Receiver.Release. -func (*queueReceiver) Release() {} - -// streamQueueReceiver implements Receiver for stream sockets. -// -// +stateify savable -type streamQueueReceiver struct { - queueReceiver - - mu sync.Mutex `state:"nosave"` - buffer []byte - control ControlMessages - addr tcpip.FullAddress -} - -func vecCopy(data [][]byte, buf []byte) (uintptr, [][]byte, []byte) { - var copied uintptr - for len(data) > 0 && len(buf) > 0 { - n := copy(data[0], buf) - copied += uintptr(n) - buf = buf[n:] - data[0] = data[0][n:] - if len(data[0]) == 0 { - data = data[1:] - } - } - return copied, data, buf -} - -// Readable implements Receiver.Readable. -func (q *streamQueueReceiver) Readable() bool { - q.mu.Lock() - bl := len(q.buffer) - r := q.readQueue.IsReadable() - q.mu.Unlock() - // We're readable if we have data in our buffer or if the queue receiver is - // readable. - return bl > 0 || r -} - -// RecvQueuedSize implements Receiver.RecvQueuedSize. -func (q *streamQueueReceiver) RecvQueuedSize() int64 { - q.mu.Lock() - bl := len(q.buffer) - qs := q.readQueue.QueuedSize() - q.mu.Unlock() - return int64(bl) + qs -} - -// RecvMaxQueueSize implements Receiver.RecvMaxQueueSize. -func (q *streamQueueReceiver) RecvMaxQueueSize() int64 { - // The RecvMaxQueueSize() is the readQueue's MaxQueueSize() plus the largest - // message we can buffer which is also the largest message we can receive. - return 2 * q.readQueue.MaxQueueSize() -} - -// Recv implements Receiver.Recv. -func (q *streamQueueReceiver) Recv(data [][]byte, wantCreds bool, numRights uintptr, peek bool) (uintptr, uintptr, ControlMessages, tcpip.FullAddress, bool, *tcpip.Error) { - q.mu.Lock() - defer q.mu.Unlock() - - var notify bool - - // If we have no data in the endpoint, we need to get some. - if len(q.buffer) == 0 { - // Load the next message into a buffer, even if we are peeking. Peeking - // won't consume the message, so it will be still available to be read - // the next time Recv() is called. - m, n, err := q.readQueue.Dequeue() - if err != nil { - return 0, 0, ControlMessages{}, tcpip.FullAddress{}, false, err - } - notify = n - msg := m.(*message) - q.buffer = []byte(msg.Data) - q.control = msg.Control - q.addr = msg.Address - } - - var copied uintptr - if peek { - // Don't consume control message if we are peeking. - c := q.control.Clone() - - // Don't consume data since we are peeking. - copied, data, _ = vecCopy(data, q.buffer) - - return copied, copied, c, q.addr, notify, nil - } - - // Consume data and control message since we are not peeking. - copied, data, q.buffer = vecCopy(data, q.buffer) - - // Save the original state of q.control. - c := q.control - - // Remove rights from q.control and leave behind just the creds. - q.control.Rights = nil - if !wantCreds { - c.Credentials = nil - } - - if c.Rights != nil && numRights == 0 { - c.Rights.Release() - c.Rights = nil - } - - haveRights := c.Rights != nil - - // If we have more capacity for data and haven't received any usable - // rights. - // - // Linux never coalesces rights control messages. - for !haveRights && len(data) > 0 { - // Get a message from the readQueue. - m, n, err := q.readQueue.Dequeue() - if err != nil { - // We already got some data, so ignore this error. This will - // manifest as a short read to the user, which is what Linux - // does. - break - } - notify = notify || n - msg := m.(*message) - q.buffer = []byte(msg.Data) - q.control = msg.Control - q.addr = msg.Address - - if wantCreds { - if (q.control.Credentials == nil) != (c.Credentials == nil) { - // One message has credentials, the other does not. - break - } - - if q.control.Credentials != nil && c.Credentials != nil && !q.control.Credentials.Equals(c.Credentials) { - // Both messages have credentials, but they don't match. - break - } - } - - if numRights != 0 && c.Rights != nil && q.control.Rights != nil { - // Both messages have rights. - break - } - - var cpd uintptr - cpd, data, q.buffer = vecCopy(data, q.buffer) - copied += cpd - - if cpd == 0 { - // data was actually full. - break - } - - if q.control.Rights != nil { - // Consume rights. - if numRights == 0 { - q.control.Rights.Release() - } else { - c.Rights = q.control.Rights - haveRights = true - } - q.control.Rights = nil - } - } - return copied, copied, c, q.addr, notify, nil -} - -// A ConnectedEndpoint is an Endpoint that can be used to send Messages. -type ConnectedEndpoint interface { - // Passcred implements Endpoint.Passcred. - Passcred() bool - - // GetLocalAddress implements Endpoint.GetLocalAddress. - GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) - - // Send sends a single message. This method does not block. - // - // notify indicates if SendNotify should be called. - // - // tcpip.ErrWouldBlock can be returned along with a partial write if - // the caller should block to send the rest of the data. - Send(data [][]byte, controlMessages ControlMessages, from tcpip.FullAddress) (n uintptr, notify bool, err *tcpip.Error) - - // SendNotify notifies the ConnectedEndpoint of a successful Send. This - // must not be called while holding any endpoint locks. - SendNotify() - - // CloseSend prevents the sending of additional Messages. - // - // After CloseSend is call, CloseNotify must also be called. - CloseSend() - - // CloseNotify notifies the ConnectedEndpoint of send being closed. This - // must not be called while holding any endpoint locks. - CloseNotify() - - // Writable returns if messages should be attempted to be sent. This - // includes when write has been shutdown. - Writable() bool - - // EventUpdate lets the ConnectedEndpoint know that event registrations - // have changed. - EventUpdate() - - // SendQueuedSize returns the total amount of data currently queued for - // sending. SendQueuedSize should return -1 if the operation isn't - // supported. - SendQueuedSize() int64 - - // SendMaxQueueSize returns maximum value for SendQueuedSize. - // SendMaxQueueSize should return -1 if the operation isn't supported. - SendMaxQueueSize() int64 - - // Release releases any resources owned by the ConnectedEndpoint. It should - // be called before droping all references to a ConnectedEndpoint. - Release() -} - -// +stateify savable -type connectedEndpoint struct { - // endpoint represents the subset of the Endpoint functionality needed by - // the connectedEndpoint. It is implemented by both connectionedEndpoint - // and connectionlessEndpoint and allows the use of types which don't - // fully implement Endpoint. - endpoint interface { - // Passcred implements Endpoint.Passcred. - Passcred() bool - - // GetLocalAddress implements Endpoint.GetLocalAddress. - GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) - - // Type implements Endpoint.Type. - Type() SockType - } - - writeQueue *queue.Queue -} - -// Passcred implements ConnectedEndpoint.Passcred. -func (e *connectedEndpoint) Passcred() bool { - return e.endpoint.Passcred() -} - -// GetLocalAddress implements ConnectedEndpoint.GetLocalAddress. -func (e *connectedEndpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) { - return e.endpoint.GetLocalAddress() -} - -// Send implements ConnectedEndpoint.Send. -func (e *connectedEndpoint) Send(data [][]byte, controlMessages ControlMessages, from tcpip.FullAddress) (uintptr, bool, *tcpip.Error) { - var l int64 - for _, d := range data { - l += int64(len(d)) - } - - truncate := false - if e.endpoint.Type() == SockStream { - // Since stream sockets don't preserve message boundaries, we - // can write only as much of the message as fits in the queue. - truncate = true - - // Discard empty stream packets. Since stream sockets don't - // preserve message boundaries, sending zero bytes is a no-op. - // In Linux, the receiver actually uses a zero-length receive - // as an indication that the stream was closed. - if l == 0 { - controlMessages.Release() - return 0, false, nil - } - } - - v := make([]byte, 0, l) - for _, d := range data { - v = append(v, d...) - } - - l, notify, err := e.writeQueue.Enqueue(&message{Data: buffer.View(v), Control: controlMessages, Address: from}, truncate) - return uintptr(l), notify, err -} - -// SendNotify implements ConnectedEndpoint.SendNotify. -func (e *connectedEndpoint) SendNotify() { - e.writeQueue.ReaderQueue.Notify(waiter.EventIn) -} - -// CloseNotify implements ConnectedEndpoint.CloseNotify. -func (e *connectedEndpoint) CloseNotify() { - e.writeQueue.ReaderQueue.Notify(waiter.EventIn) - e.writeQueue.WriterQueue.Notify(waiter.EventOut) -} - -// CloseSend implements ConnectedEndpoint.CloseSend. -func (e *connectedEndpoint) CloseSend() { - e.writeQueue.Close() -} - -// Writable implements ConnectedEndpoint.Writable. -func (e *connectedEndpoint) Writable() bool { - return e.writeQueue.IsWritable() -} - -// EventUpdate implements ConnectedEndpoint.EventUpdate. -func (*connectedEndpoint) EventUpdate() {} - -// SendQueuedSize implements ConnectedEndpoint.SendQueuedSize. -func (e *connectedEndpoint) SendQueuedSize() int64 { - return e.writeQueue.QueuedSize() -} - -// SendMaxQueueSize implements ConnectedEndpoint.SendMaxQueueSize. -func (e *connectedEndpoint) SendMaxQueueSize() int64 { - return e.writeQueue.MaxQueueSize() -} - -// Release implements ConnectedEndpoint.Release. -func (*connectedEndpoint) Release() {} - -// baseEndpoint is an embeddable unix endpoint base used in both the connected and connectionless -// unix domain socket Endpoint implementations. -// -// Not to be used on its own. -// -// +stateify savable -type baseEndpoint struct { - *waiter.Queue - - // passcred specifies whether SCM_CREDENTIALS socket control messages are - // enabled on this endpoint. Must be accessed atomically. - passcred int32 - - // Mutex protects the below fields. - sync.Mutex `state:"nosave"` - - // receiver allows Messages to be received. - receiver Receiver - - // connected allows messages to be sent and state information about the - // connected endpoint to be read. - connected ConnectedEndpoint - - // path is not empty if the endpoint has been bound, - // or may be used if the endpoint is connected. - path string -} - -// EventRegister implements waiter.Waitable.EventRegister. -func (e *baseEndpoint) EventRegister(we *waiter.Entry, mask waiter.EventMask) { - e.Queue.EventRegister(we, mask) - e.Lock() - if e.connected != nil { - e.connected.EventUpdate() - } - e.Unlock() -} - -// EventUnregister implements waiter.Waitable.EventUnregister. -func (e *baseEndpoint) EventUnregister(we *waiter.Entry) { - e.Queue.EventUnregister(we) - e.Lock() - if e.connected != nil { - e.connected.EventUpdate() - } - e.Unlock() -} - -// Passcred implements Credentialer.Passcred. -func (e *baseEndpoint) Passcred() bool { - return atomic.LoadInt32(&e.passcred) != 0 -} - -// ConnectedPasscred implements Credentialer.ConnectedPasscred. -func (e *baseEndpoint) ConnectedPasscred() bool { - e.Lock() - defer e.Unlock() - return e.connected != nil && e.connected.Passcred() -} - -func (e *baseEndpoint) setPasscred(pc bool) { - if pc { - atomic.StoreInt32(&e.passcred, 1) - } else { - atomic.StoreInt32(&e.passcred, 0) - } -} - -// Connected implements ConnectingEndpoint.Connected. -func (e *baseEndpoint) Connected() bool { - return e.receiver != nil && e.connected != nil -} - -// RecvMsg reads data and a control message from the endpoint. -func (e *baseEndpoint) RecvMsg(data [][]byte, creds bool, numRights uintptr, peek bool, addr *tcpip.FullAddress) (uintptr, uintptr, ControlMessages, *tcpip.Error) { - e.Lock() - - if e.receiver == nil { - e.Unlock() - return 0, 0, ControlMessages{}, tcpip.ErrNotConnected - } - - recvLen, msgLen, cms, a, notify, err := e.receiver.Recv(data, creds, numRights, peek) - e.Unlock() - if err != nil { - return 0, 0, ControlMessages{}, err - } - - if notify { - e.receiver.RecvNotify() - } - - if addr != nil { - *addr = a - } - return recvLen, msgLen, cms, nil -} - -// SendMsg writes data and a control message to the endpoint's peer. -// This method does not block if the data cannot be written. -func (e *baseEndpoint) SendMsg(data [][]byte, c ControlMessages, to BoundEndpoint) (uintptr, *tcpip.Error) { - e.Lock() - if !e.Connected() { - e.Unlock() - return 0, tcpip.ErrNotConnected - } - if to != nil { - e.Unlock() - return 0, tcpip.ErrAlreadyConnected - } - - n, notify, err := e.connected.Send(data, c, tcpip.FullAddress{Addr: tcpip.Address(e.path)}) - e.Unlock() - - if notify { - e.connected.SendNotify() - } - - return n, err -} - -// SetSockOpt sets a socket option. Currently not supported. -func (e *baseEndpoint) SetSockOpt(opt interface{}) *tcpip.Error { - switch v := opt.(type) { - case tcpip.PasscredOption: - e.setPasscred(v != 0) - return nil - } - return nil -} - -// GetSockOpt implements tcpip.Endpoint.GetSockOpt. -func (e *baseEndpoint) GetSockOpt(opt interface{}) *tcpip.Error { - switch o := opt.(type) { - case tcpip.ErrorOption: - return nil - case *tcpip.SendQueueSizeOption: - e.Lock() - if !e.Connected() { - e.Unlock() - return tcpip.ErrNotConnected - } - qs := tcpip.SendQueueSizeOption(e.connected.SendQueuedSize()) - e.Unlock() - if qs < 0 { - return tcpip.ErrQueueSizeNotSupported - } - *o = qs - return nil - case *tcpip.ReceiveQueueSizeOption: - e.Lock() - if !e.Connected() { - e.Unlock() - return tcpip.ErrNotConnected - } - qs := tcpip.ReceiveQueueSizeOption(e.receiver.RecvQueuedSize()) - e.Unlock() - if qs < 0 { - return tcpip.ErrQueueSizeNotSupported - } - *o = qs - return nil - case *tcpip.PasscredOption: - if e.Passcred() { - *o = tcpip.PasscredOption(1) - } else { - *o = tcpip.PasscredOption(0) - } - return nil - case *tcpip.SendBufferSizeOption: - e.Lock() - if !e.Connected() { - e.Unlock() - return tcpip.ErrNotConnected - } - qs := tcpip.SendBufferSizeOption(e.connected.SendMaxQueueSize()) - e.Unlock() - if qs < 0 { - return tcpip.ErrQueueSizeNotSupported - } - *o = qs - return nil - case *tcpip.ReceiveBufferSizeOption: - e.Lock() - if e.receiver == nil { - e.Unlock() - return tcpip.ErrNotConnected - } - qs := tcpip.ReceiveBufferSizeOption(e.receiver.RecvMaxQueueSize()) - e.Unlock() - if qs < 0 { - return tcpip.ErrQueueSizeNotSupported - } - *o = qs - return nil - } - return tcpip.ErrUnknownProtocolOption -} - -// Shutdown closes the read and/or write end of the endpoint connection to its -// peer. -func (e *baseEndpoint) Shutdown(flags tcpip.ShutdownFlags) *tcpip.Error { - e.Lock() - if !e.Connected() { - e.Unlock() - return tcpip.ErrNotConnected - } - - if flags&tcpip.ShutdownRead != 0 { - e.receiver.CloseRecv() - } - - if flags&tcpip.ShutdownWrite != 0 { - e.connected.CloseSend() - } - - e.Unlock() - - if flags&tcpip.ShutdownRead != 0 { - e.receiver.CloseNotify() - } - - if flags&tcpip.ShutdownWrite != 0 { - e.connected.CloseNotify() - } - - return nil -} - -// GetLocalAddress returns the bound path. -func (e *baseEndpoint) GetLocalAddress() (tcpip.FullAddress, *tcpip.Error) { - e.Lock() - defer e.Unlock() - return tcpip.FullAddress{Addr: tcpip.Address(e.path)}, nil -} - -// GetRemoteAddress returns the local address of the connected endpoint (if -// available). -func (e *baseEndpoint) GetRemoteAddress() (tcpip.FullAddress, *tcpip.Error) { - e.Lock() - c := e.connected - e.Unlock() - if c != nil { - return c.GetLocalAddress() - } - return tcpip.FullAddress{}, tcpip.ErrNotConnected -} - -// Release implements BoundEndpoint.Release. -func (*baseEndpoint) Release() {} -- cgit v1.2.3 From 578fe5a50dcf8e104b6bce3802987b0f8c069ade Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Wed, 17 Oct 2018 11:51:43 -0700 Subject: Fix PTRACE_GETREGSET write size The existing logic is backwards and writes iov_len == 0 for a full write. PiperOrigin-RevId: 217560377 Change-Id: I5a39c31bf0ba9063a8495993bfef58dc8ab7c5fa --- pkg/sentry/kernel/ptrace.go | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index e21a25ae6..9fe28f435 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -921,7 +921,13 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { if err != nil { return err } - ar.End -= usermem.Addr(n) + + // Update iovecs to represent the range of the written register set. + end, ok := ar.Start.AddLength(uint64(n)) + if !ok { + panic(fmt.Sprintf("%#x + %#x overflows. Invalid reg size > %#x", ar.Start, n, ar.Length())) + } + ar.End = end return t.CopyOutIovecs(data, usermem.AddrRangeSeqOf(ar)) case linux.PTRACE_SETREGS: -- cgit v1.2.3 From 4e6f0892c96c374b1abcf5c39b75ba52d98c97f8 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Wed, 17 Oct 2018 12:27:58 -0700 Subject: runsc: Support job control signals for the root container. Now containers run with "docker run -it" support control characters like ^C and ^Z. This required refactoring our signal handling a bit. Signals delivered to the "runsc boot" process are turned into loader.Signal calls with the appropriate delivery mode. Previously they were always sent directly to PID 1. PiperOrigin-RevId: 217566770 Change-Id: I5b7220d9a0f2b591a56335479454a200c6de8732 --- pkg/sentry/kernel/kernel.go | 27 +- pkg/sentry/sighandling/BUILD | 6 +- pkg/sentry/sighandling/sighandling.go | 25 +- runsc/boot/controller.go | 58 ++-- runsc/boot/fds.go | 45 ++- runsc/boot/loader.go | 221 +++++++------- runsc/container/BUILD | 1 + runsc/container/console_test.go | 452 +++++++++++++++++++++++++++++ runsc/container/container_test.go | 203 ------------- runsc/sandbox/sandbox.go | 24 +- runsc/test/integration/exec_test.go | 2 +- runsc/test/integration/integration_test.go | 48 +++ runsc/test/testutil/docker.go | 39 +++ 13 files changed, 776 insertions(+), 375 deletions(-) create mode 100644 runsc/container/console_test.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index cc664deec..84afdb530 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -839,17 +839,40 @@ func (k *Kernel) SendContainerSignal(cid string, info *arch.SignalInfo) error { k.tasks.mu.RLock() defer k.tasks.mu.RUnlock() + var lastErr error for t := range k.tasks.Root.tids { if t == t.tg.leader && t.ContainerID() == cid { t.tg.signalHandlers.mu.Lock() defer t.tg.signalHandlers.mu.Unlock() infoCopy := *info if err := t.sendSignalLocked(&infoCopy, true /*group*/); err != nil { - return err + lastErr = err } } } - return nil + return lastErr +} + +// SendProcessGroupSignal sends a signal to all processes inside the process +// group. It is analagous to kernel/signal.c:kill_pgrp. +func (k *Kernel) SendProcessGroupSignal(pg *ProcessGroup, info *arch.SignalInfo) error { + k.extMu.Lock() + defer k.extMu.Unlock() + k.tasks.mu.RLock() + defer k.tasks.mu.RUnlock() + + var lastErr error + for t := range k.tasks.Root.tids { + if t == t.tg.leader && t.tg.ProcessGroup() == pg { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + infoCopy := *info + if err := t.sendSignalLocked(&infoCopy, true /*group*/); err != nil { + lastErr = err + } + } + } + return lastErr } // FeatureSet returns the FeatureSet. diff --git a/pkg/sentry/sighandling/BUILD b/pkg/sentry/sighandling/BUILD index f480f0735..751176747 100644 --- a/pkg/sentry/sighandling/BUILD +++ b/pkg/sentry/sighandling/BUILD @@ -10,9 +10,5 @@ go_library( ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/sighandling", visibility = ["//pkg/sentry:internal"], - deps = [ - "//pkg/abi/linux", - "//pkg/sentry/arch", - "//pkg/sentry/kernel", - ], + deps = ["//pkg/abi/linux"], ) diff --git a/pkg/sentry/sighandling/sighandling.go b/pkg/sentry/sighandling/sighandling.go index 0946ab075..29bcf55ab 100644 --- a/pkg/sentry/sighandling/sighandling.go +++ b/pkg/sentry/sighandling/sighandling.go @@ -23,18 +23,17 @@ import ( "syscall" "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" ) // numSignals is the number of normal (non-realtime) signals on Linux. const numSignals = 32 -// forwardSignals listens for incoming signals and delivers them to k. +// handleSignals listens for incoming signals and calls the given handler +// function. // // It starts when the start channel is closed, stops when the stop channel // is closed, and closes done once it will no longer deliver signals to k. -func forwardSignals(k *kernel.Kernel, sigchans []chan os.Signal, start, stop, done chan struct{}) { +func handleSignals(sigchans []chan os.Signal, handler func(linux.Signal), start, stop, done chan struct{}) { // Build a select case. sc := []reflect.SelectCase{{Dir: reflect.SelectRecv, Chan: reflect.ValueOf(start)}} for _, sigchan := range sigchans { @@ -98,18 +97,19 @@ func forwardSignals(k *kernel.Kernel, sigchans []chan os.Signal, start, stop, do } } - k.SendExternalSignal(&arch.SignalInfo{Signo: int32(signal)}, "sentry") + // Pass the signal to the handler. + handler(signal) } } -// PrepareForwarding ensures that synchronous signals are forwarded to k and -// returns a callback that starts signal delivery, which itself returns a -// callback that stops signal forwarding. +// PrepareHandler ensures that synchronous signals are passed to the given +// handler function and returns a callback that starts signal delivery, which +// itself returns a callback that stops signal handling. // // Note that this function permanently takes over signal handling. After the // stop callback, signals revert to the default Go runtime behavior, which // cannot be overridden with external calls to signal.Notify. -func PrepareForwarding(k *kernel.Kernel, skipSignal syscall.Signal) func() func() { +func PrepareHandler(handler func(linux.Signal)) func() func() { start := make(chan struct{}) stop := make(chan struct{}) done := make(chan struct{}) @@ -125,15 +125,10 @@ func PrepareForwarding(k *kernel.Kernel, skipSignal syscall.Signal) func() func( for sig := 1; sig <= numSignals+1; sig++ { sigchan := make(chan os.Signal, 1) sigchans = append(sigchans, sigchan) - - if syscall.Signal(sig) == skipSignal { - continue - } - signal.Notify(sigchan, syscall.Signal(sig)) } // Start up our listener. - go forwardSignals(k, sigchans, start, stop, done) // S/R-SAFE: synchronized by Kernel.extMu. + go handleSignals(sigchans, handler, start, stop, done) // S/R-SAFE: synchronized by Kernel.extMu. return func() func() { close(start) diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index eaeb9e2d8..bee82f344 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -425,6 +425,26 @@ func (cm *containerManager) WaitPID(args *WaitPIDArgs, waitStatus *uint32) error return cm.l.waitPID(kernel.ThreadID(args.PID), args.CID, args.ClearStatus, waitStatus) } +// SignalDeliveryMode enumerates different signal delivery modes. +type SignalDeliveryMode int + +const ( + // DeliverToProcess delivers the signal to the container process with + // the specified PID. If PID is 0, then the container init process is + // signaled. + DeliverToProcess SignalDeliveryMode = iota + + // DeliverToAllProcesses delivers the signal to all processes in the + // container. PID must be 0. + DeliverToAllProcesses + + // DeliverToForegroundProcessGroup delivers the signal to the + // foreground process group in the same TTY session as the specified + // process. If PID is 0, then the signal is delivered to the foreground + // process group for the TTY for the init process. + DeliverToForegroundProcessGroup +) + // SignalArgs are arguments to the Signal method. type SignalArgs struct { // CID is the container ID. @@ -433,36 +453,20 @@ type SignalArgs struct { // Signo is the signal to send to the process. Signo int32 - // All is set when signal should be sent to all processes in the container. - // When false, the signal is sent to the root container process only. - All bool -} - -// Signal sends a signal to the root process of the container. -func (cm *containerManager) Signal(args *SignalArgs, _ *struct{}) error { - log.Debugf("containerManager.Signal %q %d, all: %t", args.CID, args.Signo, args.All) - return cm.l.signalContainer(args.CID, args.Signo, args.All) -} - -// SignalProcessArgs are arguments to the Signal method. -type SignalProcessArgs struct { - // CID is the container ID. - CID string - // PID is the process ID in the given container that will be signaled. + // If 0, the root container will be signalled. PID int32 - // Signo is the signal to send to the process. - Signo int32 - - // SendToForegroundProcess indicates that the signal should be sent to - // the foreground process group in the session that PID belongs to. - // This is only valid if the process is attached to a host TTY. - SendToForegroundProcess bool + // Mode is the signal delivery mode. + Mode SignalDeliveryMode } -// SignalProcess sends a signal to a particular process in the container. -func (cm *containerManager) SignalProcess(args *SignalProcessArgs, _ *struct{}) error { - log.Debugf("containerManager.Signal: %+v", args) - return cm.l.signalProcess(args.CID, args.PID, args.Signo, args.SendToForegroundProcess) +// Signal sends a signal to one or more processes in a container. If args.PID +// is 0, then the container init process is used. Depending on the +// args.SignalDeliveryMode option, the signal may be sent directly to the +// indicated process, to all processes in the container, or to the foreground +// process group. +func (cm *containerManager) Signal(args *SignalArgs, _ *struct{}) error { + log.Debugf("containerManager.Signal %+v", args) + return cm.l.signal(args.CID, args.PID, args.Signo, args.Mode) } diff --git a/runsc/boot/fds.go b/runsc/boot/fds.go index a5a6ba8af..9416e3a5c 100644 --- a/runsc/boot/fds.go +++ b/runsc/boot/fds.go @@ -35,6 +35,7 @@ func createFDMap(ctx context.Context, k *kernel.Kernel, l *limits.LimitSet, cons fdm := k.NewFDMap() defer fdm.DecRef() + mounter := fs.FileOwnerFromContext(ctx) // Maps sandbox FD to host FD. fdMap := map[int]int{ @@ -42,16 +43,44 @@ func createFDMap(ctx context.Context, k *kernel.Kernel, l *limits.LimitSet, cons 1: stdioFDs[1], 2: stdioFDs[2], } - mounter := fs.FileOwnerFromContext(ctx) - for sfd, hfd := range fdMap { - file, err := host.ImportFile(ctx, hfd, mounter, console /* isTTY */) - if err != nil { - return nil, fmt.Errorf("failed to import fd %d: %v", hfd, err) + var ttyFile *fs.File + for appFD, hostFD := range fdMap { + var appFile *fs.File + + if console && appFD < 3 { + // Import the file as a host TTY file. + if ttyFile == nil { + var err error + appFile, err = host.ImportFile(ctx, hostFD, mounter, true /* isTTY */) + if err != nil { + return nil, err + } + defer appFile.DecRef() + + // Remember this in the TTY file, as we will + // use it for the other stdio FDs. + ttyFile = appFile + } else { + // Re-use the existing TTY file, as all three + // stdio FDs must point to the same fs.File in + // order to share TTY state, specifically the + // foreground process group id. + appFile = ttyFile + } + } else { + // Import the file as a regular host file. + var err error + appFile, err = host.ImportFile(ctx, hostFD, mounter, false /* isTTY */) + if err != nil { + return nil, err + } + defer appFile.DecRef() } - defer file.DecRef() - if err := fdm.NewFDAt(kdefs.FD(sfd), file, kernel.FDFlags{}, l); err != nil { - return nil, fmt.Errorf("failed to add imported fd %d to FDMap: %v", hfd, err) + + // Add the file to the FD map. + if err := fdm.NewFDAt(kdefs.FD(appFD), appFile, kernel.FDFlags{}, l); err != nil { + return nil, err } } diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index fa169d090..c79b95bde 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -19,7 +19,6 @@ import ( "fmt" mrand "math/rand" "os" - "os/signal" "runtime" "sync" "sync/atomic" @@ -110,7 +109,7 @@ type Loader struct { // mu guards processes. mu sync.Mutex - // processes maps containers root process and invocation of exec. Root + // processes maps containers init process and invocation of exec. Root // processes are keyed with container ID and pid=0, while exec invocations // have the corresponding pid set. // @@ -291,28 +290,9 @@ func New(args Args) (*Loader, error) { return nil, fmt.Errorf("error creating control server: %v", err) } - // We don't care about child signals; some platforms can generate a - // tremendous number of useless ones (I'm looking at you, ptrace). - if err := sighandling.IgnoreChildStop(); err != nil { - return nil, fmt.Errorf("failed to ignore child stop signals: %v", err) - } - // Ensure that signals received are forwarded to the emulated kernel. - ps := syscall.Signal(args.Conf.PanicSignal) - startSignalForwarding := sighandling.PrepareForwarding(k, ps) - if args.Conf.PanicSignal != -1 { - // Panics if the sentry receives 'Config.PanicSignal'. - panicChan := make(chan os.Signal, 1) - signal.Notify(panicChan, ps) - go func() { // S/R-SAFE: causes sentry panic. - <-panicChan - panic("Signal-induced panic") - }() - log.Infof("Panic signal set to %v(%d)", ps, args.Conf.PanicSignal) - } - procArgs, err := newProcess(args.ID, args.Spec, creds, k) if err != nil { - return nil, fmt.Errorf("failed to create root process: %v", err) + return nil, fmt.Errorf("failed to create init process for root container: %v", err) } if err := initCompatLogs(args.UserLogFD); err != nil { @@ -320,19 +300,47 @@ func New(args Args) (*Loader, error) { } l := &Loader{ - k: k, - ctrl: ctrl, - conf: args.Conf, - console: args.Console, - watchdog: watchdog, - spec: args.Spec, - goferFDs: args.GoferFDs, - stdioFDs: args.StdioFDs, - startSignalForwarding: startSignalForwarding, - rootProcArgs: procArgs, - sandboxID: args.ID, - processes: make(map[execID]*execProcess), + k: k, + ctrl: ctrl, + conf: args.Conf, + console: args.Console, + watchdog: watchdog, + spec: args.Spec, + goferFDs: args.GoferFDs, + stdioFDs: args.StdioFDs, + rootProcArgs: procArgs, + sandboxID: args.ID, + processes: make(map[execID]*execProcess), } + + // We don't care about child signals; some platforms can generate a + // tremendous number of useless ones (I'm looking at you, ptrace). + if err := sighandling.IgnoreChildStop(); err != nil { + return nil, fmt.Errorf("failed to ignore child stop signals: %v", err) + } + + // Handle signals by forwarding them to the root container process + // (except for panic signal, which should cause a panic). + l.startSignalForwarding = sighandling.PrepareHandler(func(sig linux.Signal) { + // Panic signal should cause a panic. + if args.Conf.PanicSignal != -1 && sig == linux.Signal(args.Conf.PanicSignal) { + panic("Signal-induced panic") + } + + // Otherwise forward to root container. + deliveryMode := DeliverToProcess + if args.Console { + // Since we are running with a console, we should + // forward the signal to the foreground process group + // so that job control signals like ^C can be handled + // properly. + deliveryMode = DeliverToForegroundProcessGroup + } + if err := l.signal(args.ID, 0, int32(sig), deliveryMode); err != nil { + log.Warningf("error sending signal %v to container %q: %v", sig, args.ID, err) + } + }) + ctrl.manager.l = l return l, nil } @@ -467,9 +475,15 @@ func (l *Loader) run() error { l.rootProcArgs.FDMap.DecRef() } - l.mu.Lock() eid := execID{cid: l.sandboxID} - l.processes[eid] = &execProcess{tg: l.k.GlobalInit()} + ep := execProcess{tg: l.k.GlobalInit()} + if l.console { + ttyFile := l.rootProcArgs.FDMap.GetFile(0) + defer ttyFile.DecRef() + ep.tty = ttyFile.FileOperations.(*host.TTYFileOperations) + } + l.mu.Lock() + l.processes[eid] = &ep l.mu.Unlock() // Start signal forwarding only after an init process is created. @@ -572,7 +586,7 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config // filesystem. func (l *Loader) destroyContainer(cid string) error { // First kill and wait for all processes in the container. - if err := l.signalContainer(cid, int32(linux.SIGKILL), true /*all*/); err != nil { + if err := l.signal(cid, 0, int32(linux.SIGKILL), DeliverToAllProcesses); err != nil { return fmt.Errorf("failed to SIGKILL all container processes: %v", err) } @@ -634,7 +648,7 @@ func (l *Loader) executeAsync(args *control.ExecArgs) (kernel.ThreadID, error) { return tgid, nil } -// waitContainer waits for the root process of a container to exit. +// waitContainer waits for the init process of a container to exit. func (l *Loader) waitContainer(cid string, waitStatus *uint32) error { // Don't defer unlock, as doing so would make it impossible for // multiple clients to wait on the same container. @@ -740,11 +754,12 @@ func newEmptyNetworkStack(conf *Config, clock tcpip.Clock) (inet.Stack, error) { } } -// signalProcess sends a signal to the process with the given PID. If -// sendToFGProcess is true, then the signal will be sent to the foreground -// process group in the same session that PID belongs to. -func (l *Loader) signalProcess(cid string, pid, signo int32, sendToFGProcess bool) error { - if pid <= 0 { +// signal sends a signal to one or more processes in a container. If PID is 0, +// then the container init process is used. Depending on the SignalDeliveryMode +// option, the signal may be sent directly to the indicated process, to all +// processes in the container, or to the foreground process group. +func (l *Loader) signal(cid string, pid, signo int32, mode SignalDeliveryMode) error { + if pid < 0 { return fmt.Errorf("failed to signal container %q PID %d: PID must be positive", cid, pid) } @@ -756,10 +771,16 @@ func (l *Loader) signalProcess(cid string, pid, signo int32, sendToFGProcess boo ep, ok := l.processes[eid] l.mu.Unlock() - // The caller may be signaling a process not started directly via exec. - // In this case, find the process in the container's PID namespace and - // signal it. - if !ok { + switch mode { + case DeliverToProcess: + if ok { + // Send signal directly to the identified process. + return ep.tg.SendSignal(&arch.SignalInfo{Signo: signo}) + } + + // The caller may be signaling a process not started directly via exec. + // In this case, find the process in the container's PID namespace and + // signal it. ep, ok := l.processes[execID{cid: cid}] if !ok { return fmt.Errorf("no container with ID: %q", cid) @@ -772,74 +793,60 @@ func (l *Loader) signalProcess(cid string, pid, signo int32, sendToFGProcess boo return fmt.Errorf("process %d is part of a different container: %q", pid, tg.Leader().ContainerID()) } return tg.SendSignal(&arch.SignalInfo{Signo: signo}) - } - - if !sendToFGProcess { - // Send signal directly to exec process. - return ep.tg.SendSignal(&arch.SignalInfo{Signo: signo}) - } - // Lookup foreground process group from the TTY for the given process, - // and send the signal to it. - if ep.tty == nil { - return fmt.Errorf("failed to signal foreground process group in container %q PID %d: no TTY attached", cid, pid) - } - pg := ep.tty.ForegroundProcessGroup() - if pg == nil { - // No foreground process group has been set. Signal the - // original thread group. - log.Warningf("No foreground process group for container %q and PID %d. Sending signal directly to PID %d.", cid, pid, pid) - return ep.tg.SendSignal(&arch.SignalInfo{Signo: signo}) - } + case DeliverToForegroundProcessGroup: + if !ok { + return fmt.Errorf("failed to signal foreground process group for container %q PID %d: no such PID", cid, pid) + } - // Send the signal to all processes in the process group. - var lastErr error - for _, tg := range l.k.TaskSet().Root.ThreadGroups() { - if tg.ProcessGroup() != pg { - continue + // Lookup foreground process group from the TTY for the given process, + // and send the signal to it. + if ep.tty == nil { + return fmt.Errorf("failed to signal foreground process group in container %q PID %d: no TTY attached", cid, pid) } - if err := tg.SendSignal(&arch.SignalInfo{Signo: signo}); err != nil { - lastErr = err + pg := ep.tty.ForegroundProcessGroup() + if pg == nil { + // No foreground process group has been set. Signal the + // original thread group. + log.Warningf("No foreground process group for container %q and PID %d. Sending signal directly to PID %d.", cid, pid, pid) + return ep.tg.SendSignal(&arch.SignalInfo{Signo: signo}) + } + // Send the signal to all processes in the process group. + var lastErr error + for _, tg := range l.k.TaskSet().Root.ThreadGroups() { + if tg.ProcessGroup() != pg { + continue + } + if err := tg.SendSignal(&arch.SignalInfo{Signo: signo}); err != nil { + lastErr = err + } + } + return lastErr + case DeliverToAllProcesses: + if !ok { + return fmt.Errorf("failed to signal all processes in container %q PID %d: no such PID", cid, pid) } - } - return lastErr -} - -// signalContainer sends a signal to the root container process, or to all -// processes in the container if all is true. -func (l *Loader) signalContainer(cid string, signo int32, all bool) error { - si := arch.SignalInfo{Signo: signo} - - l.mu.Lock() - defer l.mu.Unlock() - - eid := execID{cid: cid} - ep, ok := l.processes[eid] - if !ok { - return fmt.Errorf("failed to signal container %q: no such container", cid) - } - - if !all { - return ep.tg.SendSignal(&si) - } - // Pause the kernel to prevent new processes from being created while - // the signal is delivered. This prevents process leaks when SIGKILL is - // sent to the entire container. - l.k.Pause() - if err := l.k.SendContainerSignal(cid, &si); err != nil { + // Pause the kernel to prevent new processes from being created while + // the signal is delivered. This prevents process leaks when SIGKILL is + // sent to the entire container. + l.k.Pause() + if err := l.k.SendContainerSignal(cid, &arch.SignalInfo{Signo: signo}); err != nil { + l.k.Unpause() + return err + } l.k.Unpause() - return err - } - l.k.Unpause() - // If killing all processes, wait for them to exit. - if all && linux.Signal(signo) == linux.SIGKILL { - for _, t := range l.k.TaskSet().Root.Tasks() { - if t.ContainerID() == cid { - t.ThreadGroup().WaitExited() + // If SIGKILLing all processes, wait for them to exit. + if linux.Signal(signo) == linux.SIGKILL { + for _, t := range l.k.TaskSet().Root.Tasks() { + if t.ContainerID() == cid { + t.ThreadGroup().WaitExited() + } } } + return nil + default: + panic(fmt.Sprintf("unknown signal signal delivery mode %v", mode)) } - return nil } diff --git a/runsc/container/BUILD b/runsc/container/BUILD index 60f1d3033..f4c6f1525 100644 --- a/runsc/container/BUILD +++ b/runsc/container/BUILD @@ -30,6 +30,7 @@ go_test( name = "container_test", size = "medium", srcs = [ + "console_test.go", "container_test.go", "fs_test.go", "multi_container_test.go", diff --git a/runsc/container/console_test.go b/runsc/container/console_test.go new file mode 100644 index 000000000..82adcbb7d --- /dev/null +++ b/runsc/container/console_test.go @@ -0,0 +1,452 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package container + +import ( + "bytes" + "fmt" + "io" + "os" + "path/filepath" + "sync" + "syscall" + "testing" + "time" + + "github.com/kr/pty" + "golang.org/x/sys/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/control" + "gvisor.googlesource.com/gvisor/pkg/unet" + "gvisor.googlesource.com/gvisor/pkg/urpc" + "gvisor.googlesource.com/gvisor/runsc/test/testutil" +) + +// createConsoleSocket creates a socket that will receive a console fd from the +// sandbox. If no error occurs, it returns the server socket and a cleanup +// function. +func createConsoleSocket(socketPath string) (*unet.ServerSocket, func() error, error) { + cwd, err := os.Getwd() + if err != nil { + return nil, nil, fmt.Errorf("error getting cwd: %v", err) + } + // We use a relative path to avoid overflowing the unix path length + // limit (108 chars). + socketRelPath, err := filepath.Rel(cwd, socketPath) + if err != nil { + return nil, nil, fmt.Errorf("error getting relative path for %q from cwd %q: %v", socketPath, cwd, err) + } + if len(socketRelPath) > len(socketPath) { + socketRelPath = socketPath + } + srv, err := unet.BindAndListen(socketRelPath, false) + if err != nil { + return nil, nil, fmt.Errorf("error binding and listening to socket %q: %v", socketPath, err) + } + + cleanup := func() error { + if err := srv.Close(); err != nil { + return fmt.Errorf("error closing socket %q: %v", socketRelPath, err) + } + if err := os.Remove(socketPath); err != nil { + return fmt.Errorf("error removing socket %q: %v", socketRelPath, err) + } + return nil + } + + return srv, cleanup, nil +} + +// receiveConsolePTY accepts a connection on the server socket and reads fds. +// It fails if more than one FD is received, or if the FD is not a PTY. It +// returns the PTY master file. +func receiveConsolePTY(srv *unet.ServerSocket) (*os.File, error) { + sock, err := srv.Accept() + if err != nil { + return nil, fmt.Errorf("error accepting socket connection: %v", err) + } + + // Allow 3 fds to be received. We only expect 1. + r := sock.Reader(true /* blocking */) + r.EnableFDs(1) + + // The socket is closed right after sending the FD, so EOF is + // an allowed error. + b := [][]byte{{}} + if _, err := r.ReadVec(b); err != nil && err != io.EOF { + return nil, fmt.Errorf("error reading from socket connection: %v", err) + } + + // We should have gotten a control message. + fds, err := r.ExtractFDs() + if err != nil { + return nil, fmt.Errorf("error extracting fds from socket connection: %v", err) + } + if len(fds) != 1 { + return nil, fmt.Errorf("got %d fds from socket, wanted 1", len(fds)) + } + + // Verify that the fd is a terminal. + if _, err := unix.IoctlGetTermios(fds[0], unix.TCGETS); err != nil { + return nil, fmt.Errorf("fd is not a terminal (ioctl TGGETS got %v)", err) + } + + return os.NewFile(uintptr(fds[0]), "pty_master"), nil +} + +// Test that an pty FD is sent over the console socket if one is provided. +func TestConsoleSocket(t *testing.T) { + for _, conf := range configs(all...) { + t.Logf("Running test with conf: %+v", conf) + spec := testutil.NewSpecWithArgs("true") + rootDir, bundleDir, err := testutil.SetupContainer(spec, conf) + if err != nil { + t.Fatalf("error setting up container: %v", err) + } + defer os.RemoveAll(rootDir) + defer os.RemoveAll(bundleDir) + + socketPath := filepath.Join(bundleDir, "socket") + srv, cleanup, err := createConsoleSocket(socketPath) + if err != nil { + t.Fatalf("error creating socket at %q: %v", socketPath, err) + } + defer cleanup() + + // Create the container and pass the socket name. + id := testutil.UniqueContainerID() + c, err := Create(id, spec, conf, bundleDir, socketPath, "", "") + if err != nil { + t.Fatalf("error creating container: %v", err) + } + defer c.Destroy() + + // Make sure we get a console PTY. + ptyMaster, err := receiveConsolePTY(srv) + if err != nil { + t.Fatalf("error receiving console FD: %v", err) + } + ptyMaster.Close() + } +} + +// Test that job control signals work on a console created with "exec -ti". +func TestJobControlSignalExec(t *testing.T) { + spec := testutil.NewSpecWithArgs("/bin/sleep", "10000") + conf := testutil.TestConfig() + + rootDir, bundleDir, err := testutil.SetupContainer(spec, conf) + if err != nil { + t.Fatalf("error setting up container: %v", err) + } + defer os.RemoveAll(rootDir) + defer os.RemoveAll(bundleDir) + + // Create and start the container. + c, err := Create(testutil.UniqueContainerID(), spec, conf, bundleDir, "", "", "") + if err != nil { + t.Fatalf("error creating container: %v", err) + } + defer c.Destroy() + if err := c.Start(conf); err != nil { + t.Fatalf("error starting container: %v", err) + } + + // Create a pty master/slave. The slave will be passed to the exec + // process. + ptyMaster, ptySlave, err := pty.Open() + if err != nil { + t.Fatalf("error opening pty: %v", err) + } + defer ptyMaster.Close() + defer ptySlave.Close() + + // Exec bash and attach a terminal. + args := &control.ExecArgs{ + Filename: "/bin/bash", + // Don't let bash execute from profile or rc files, otherwise + // our PID counts get messed up. + Argv: []string{"/bin/bash", "--noprofile", "--norc"}, + // Pass the pty slave as FD 0, 1, and 2. + FilePayload: urpc.FilePayload{ + Files: []*os.File{ptySlave, ptySlave, ptySlave}, + }, + StdioIsPty: true, + } + + pid, err := c.Execute(args) + if err != nil { + t.Fatalf("error executing: %v", err) + } + if pid != 2 { + t.Fatalf("exec got pid %d, wanted %d", pid, 2) + } + + // Make sure all the processes are running. + expectedPL := []*control.Process{ + // Root container process. + {PID: 1, Cmd: "sleep"}, + // Bash from exec process. + {PID: 2, Cmd: "bash"}, + } + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Execute sleep. + ptyMaster.Write([]byte("sleep 100\n")) + + // Wait for it to start. Sleep's PPID is bash's PID. + expectedPL = append(expectedPL, &control.Process{PID: 3, PPID: 2, Cmd: "sleep"}) + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Send a SIGTERM to the foreground process for the exec PID. Note that + // although we pass in the PID of "bash", it should actually terminate + // "sleep", since that is the foreground process. + if err := c.Sandbox.SignalProcess(c.ID, pid, syscall.SIGTERM, true /* fgProcess */); err != nil { + t.Fatalf("error signaling container: %v", err) + } + + // Sleep process should be gone. + expectedPL = expectedPL[:len(expectedPL)-1] + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Sleep is dead, but it may take more time for bash to notice and + // change the foreground process back to itself. We know it is done + // when bash writes "Terminated" to the pty. + if err := testutil.WaitUntilRead(ptyMaster, "Terminated", nil, 5*time.Second); err != nil { + t.Fatalf("bash did not take over pty: %v", err) + } + + // Send a SIGKILL to the foreground process again. This time "bash" + // should be killed. We use SIGKILL instead of SIGTERM or SIGINT + // because bash ignores those. + if err := c.Sandbox.SignalProcess(c.ID, pid, syscall.SIGKILL, true /* fgProcess */); err != nil { + t.Fatalf("error signaling container: %v", err) + } + expectedPL = expectedPL[:1] + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Make sure the process indicates it was killed by a SIGKILL. + ws, err := c.WaitPID(pid, true) + if err != nil { + t.Errorf("waiting on container failed: %v", err) + } + if !ws.Signaled() { + t.Error("ws.Signaled() got false, want true") + } + if got, want := ws.Signal(), syscall.SIGKILL; got != want { + t.Errorf("ws.Signal() got %v, want %v", got, want) + } +} + +// Test that job control signals work on a console created with "run -ti". +func TestJobControlSignalRootContainer(t *testing.T) { + conf := testutil.TestConfig() + // Don't let bash execute from profile or rc files, otherwise our PID + // counts get messed up. + spec := testutil.NewSpecWithArgs("/bin/bash", "--noprofile", "--norc") + spec.Process.Terminal = true + + rootDir, bundleDir, err := testutil.SetupContainer(spec, conf) + if err != nil { + t.Fatalf("error setting up container: %v", err) + } + defer os.RemoveAll(rootDir) + defer os.RemoveAll(bundleDir) + + socketPath := filepath.Join(bundleDir, "socket") + srv, cleanup, err := createConsoleSocket(socketPath) + if err != nil { + t.Fatalf("error creating socket at %q: %v", socketPath, err) + } + defer cleanup() + + // Create the container and pass the socket name. + id := testutil.UniqueContainerID() + c, err := Create(id, spec, conf, bundleDir, socketPath, "", "") + if err != nil { + t.Fatalf("error creating container: %v", err) + } + defer c.Destroy() + + // Get the PTY master. + ptyMaster, err := receiveConsolePTY(srv) + if err != nil { + t.Fatalf("error receiving console FD: %v", err) + } + defer ptyMaster.Close() + + // Bash output as well as sandbox output will be written to the PTY + // file. Writes after a certain point will block unless we drain the + // PTY, so we must continually copy from it. + // + // We log the output to stdout for debugabilitly, and also to a buffer, + // since we wait on particular output from bash below. We use a custom + // blockingBuffer which is thread-safe and also blocks on Read calls, + // which makes this a suitable Reader for WaitUntilRead. + ptyBuf := newBlockingBuffer() + tee := io.TeeReader(ptyMaster, ptyBuf) + go io.Copy(os.Stdout, tee) + + // Start the container. + if err := c.Start(conf); err != nil { + t.Fatalf("error starting container: %v", err) + } + + // Start waiting for the container to exit in a goroutine. We do this + // very early, otherwise it might exit before we have a chance to call + // Wait. + var ( + ws syscall.WaitStatus + wg sync.WaitGroup + ) + wg.Add(1) + go func() { + var err error + ws, err = c.Wait() + if err != nil { + t.Errorf("error waiting on container: %v", err) + } + wg.Done() + }() + + // Wait for bash to start. + expectedPL := []*control.Process{ + {PID: 1, Cmd: "bash"}, + } + if err := waitForProcessList(c, expectedPL); err != nil { + t.Fatal(err) + } + + // Execute sleep via the terminal. + ptyMaster.Write([]byte("sleep 100\n")) + + // Wait for sleep to start. + expectedPL = append(expectedPL, &control.Process{PID: 2, PPID: 1, Cmd: "sleep"}) + if err := waitForProcessList(c, expectedPL); err != nil { + t.Fatal(err) + } + + // Reset the pty buffer, so there is less output for us to scan later. + ptyBuf.Reset() + + // Send a SIGTERM to the foreground process. We pass PID=0, indicating + // that the root process should be killed. However, by setting + // fgProcess=true, the signal should actually be sent to sleep. + if err := c.Sandbox.SignalProcess(c.ID, 0 /* PID */, syscall.SIGTERM, true /* fgProcess */); err != nil { + t.Fatalf("error signaling container: %v", err) + } + + // Sleep process should be gone. + expectedPL = expectedPL[:len(expectedPL)-1] + if err := waitForProcessList(c, expectedPL); err != nil { + t.Error(err) + } + + // Sleep is dead, but it may take more time for bash to notice and + // change the foreground process back to itself. We know it is done + // when bash writes "Terminated" to the pty. + if err := testutil.WaitUntilRead(ptyBuf, "Terminated", nil, 5*time.Second); err != nil { + t.Fatalf("bash did not take over pty: %v", err) + } + + // Send a SIGKILL to the foreground process again. This time "bash" + // should be killed. We use SIGKILL instead of SIGTERM or SIGINT + // because bash ignores those. + if err := c.Sandbox.SignalProcess(c.ID, 0 /* PID */, syscall.SIGKILL, true /* fgProcess */); err != nil { + t.Fatalf("error signaling container: %v", err) + } + + // Wait for the sandbox to exit. It should exit with a SIGKILL status. + wg.Wait() + if !ws.Signaled() { + t.Error("ws.Signaled() got false, want true") + } + if got, want := ws.Signal(), syscall.SIGKILL; got != want { + t.Errorf("ws.Signal() got %v, want %v", got, want) + } +} + +// blockingBuffer is a thread-safe buffer that blocks when reading if the +// buffer is empty. It implements io.ReadWriter. +type blockingBuffer struct { + // A send to readCh indicates that a previously empty buffer now has + // data for reading. + readCh chan struct{} + + // mu protects buf. + mu sync.Mutex + buf bytes.Buffer +} + +func newBlockingBuffer() *blockingBuffer { + return &blockingBuffer{ + readCh: make(chan struct{}, 1), + } +} + +// Write implements Writer.Write. +func (bb *blockingBuffer) Write(p []byte) (int, error) { + bb.mu.Lock() + defer bb.mu.Unlock() + l := bb.buf.Len() + n, err := bb.buf.Write(p) + if l == 0 && n > 0 { + // New data! + bb.readCh <- struct{}{} + } + return n, err +} + +// Read implements Reader.Read. It will block until data is available. +func (bb *blockingBuffer) Read(p []byte) (int, error) { + for { + bb.mu.Lock() + n, err := bb.buf.Read(p) + if n > 0 || err != io.EOF { + if bb.buf.Len() == 0 { + // Reset the readCh. + select { + case <-bb.readCh: + default: + } + } + bb.mu.Unlock() + return n, err + } + bb.mu.Unlock() + + // Wait for new data. + <-bb.readCh + } +} + +// Reset resets the buffer. +func (bb *blockingBuffer) Reset() { + bb.mu.Lock() + defer bb.mu.Unlock() + bb.buf.Reset() + // Reset the readCh. + select { + case <-bb.readCh: + default: + } +} diff --git a/runsc/container/container_test.go b/runsc/container/container_test.go index d9cd38c0a..e2bb7d8ec 100644 --- a/runsc/container/container_test.go +++ b/runsc/container/container_test.go @@ -17,7 +17,6 @@ package container import ( "bytes" "fmt" - "io" "io/ioutil" "os" "path" @@ -31,15 +30,11 @@ import ( "time" "github.com/cenkalti/backoff" - "github.com/kr/pty" specs "github.com/opencontainers/runtime-spec/specs-go" - "golang.org/x/sys/unix" "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/control" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" - "gvisor.googlesource.com/gvisor/pkg/unet" - "gvisor.googlesource.com/gvisor/pkg/urpc" "gvisor.googlesource.com/gvisor/runsc/boot" "gvisor.googlesource.com/gvisor/runsc/test/testutil" ) @@ -1151,89 +1146,6 @@ func TestCapabilities(t *testing.T) { } } -// Test that an tty FD is sent over the console socket if one is provided. -func TestConsoleSocket(t *testing.T) { - for _, conf := range configs(all...) { - t.Logf("Running test with conf: %+v", conf) - spec := testutil.NewSpecWithArgs("true") - rootDir, bundleDir, err := testutil.SetupContainer(spec, conf) - if err != nil { - t.Fatalf("error setting up container: %v", err) - } - defer os.RemoveAll(rootDir) - defer os.RemoveAll(bundleDir) - - // Create a named socket and start listening. We use a relative path - // to avoid overflowing the unix path length limit (108 chars). - socketPath := filepath.Join(bundleDir, "socket") - cwd, err := os.Getwd() - if err != nil { - t.Fatalf("error getting cwd: %v", err) - } - socketRelPath, err := filepath.Rel(cwd, socketPath) - if err != nil { - t.Fatalf("error getting relative path for %q from cwd %q: %v", socketPath, cwd, err) - } - if len(socketRelPath) > len(socketPath) { - socketRelPath = socketPath - } - srv, err := unet.BindAndListen(socketRelPath, false) - if err != nil { - t.Fatalf("error binding and listening to socket %q: %v", socketPath, err) - } - defer os.Remove(socketPath) - - // Create the container and pass the socket name. - id := testutil.UniqueContainerID() - c, err := Create(id, spec, conf, bundleDir, socketRelPath, "", "") - if err != nil { - t.Fatalf("error creating container: %v", err) - } - c.Destroy() - - // Open the othe end of the socket. - sock, err := srv.Accept() - if err != nil { - t.Fatalf("error accepting socket connection: %v", err) - } - - // Allow 3 fds to be received. We only expect 1. - r := sock.Reader(true /* blocking */) - r.EnableFDs(1) - - // The socket is closed right after sending the FD, so EOF is - // an allowed error. - b := [][]byte{{}} - if _, err := r.ReadVec(b); err != nil && err != io.EOF { - t.Fatalf("error reading from socket connection: %v", err) - } - - // We should have gotten a control message. - fds, err := r.ExtractFDs() - if err != nil { - t.Fatalf("error extracting fds from socket connection: %v", err) - } - if len(fds) != 1 { - t.Fatalf("got %d fds from socket, wanted 1", len(fds)) - } - - // Verify that the fd is a terminal. - if _, err := unix.IoctlGetTermios(fds[0], unix.TCGETS); err != nil { - t.Errorf("fd is not a terminal (ioctl TGGETS got %v)", err) - } - - // Shut it down. - if err := c.Destroy(); err != nil { - t.Fatalf("error destroying container: %v", err) - } - - // Close socket. - if err := srv.Close(); err != nil { - t.Fatalf("error destroying container: %v", err) - } - } -} - // TestRunNonRoot checks that sandbox can be configured when running as // non-privileged user. func TestRunNonRoot(t *testing.T) { @@ -1626,121 +1538,6 @@ func TestRootNotMount(t *testing.T) { } } -func TestJobControlSignalExec(t *testing.T) { - spec := testutil.NewSpecWithArgs("/bin/sleep", "10000") - conf := testutil.TestConfig() - - rootDir, bundleDir, err := testutil.SetupContainer(spec, conf) - if err != nil { - t.Fatalf("error setting up container: %v", err) - } - defer os.RemoveAll(rootDir) - defer os.RemoveAll(bundleDir) - - // Create and start the container. - c, err := Create(testutil.UniqueContainerID(), spec, conf, bundleDir, "", "", "") - if err != nil { - t.Fatalf("error creating container: %v", err) - } - defer c.Destroy() - if err := c.Start(conf); err != nil { - t.Fatalf("error starting container: %v", err) - } - - // Create a pty master/slave. The slave will be passed to the exec - // process. - ptyMaster, ptySlave, err := pty.Open() - if err != nil { - t.Fatalf("error opening pty: %v", err) - } - defer ptyMaster.Close() - defer ptySlave.Close() - - // Exec bash and attach a terminal. - args := &control.ExecArgs{ - Filename: "/bin/bash", - // Don't let bash execute from profile or rc files, otherwise - // our PID counts get messed up. - Argv: []string{"/bin/bash", "--noprofile", "--norc"}, - // Pass the pty slave as FD 0, 1, and 2. - FilePayload: urpc.FilePayload{ - Files: []*os.File{ptySlave, ptySlave, ptySlave}, - }, - StdioIsPty: true, - } - - pid, err := c.Execute(args) - if err != nil { - t.Fatalf("error executing: %v", err) - } - if pid != 2 { - t.Fatalf("exec got pid %d, wanted %d", pid, 2) - } - - // Make sure all the processes are running. - expectedPL := []*control.Process{ - // Root container process. - {PID: 1, Cmd: "sleep"}, - // Bash from exec process. - {PID: 2, Cmd: "bash"}, - } - if err := waitForProcessList(c, expectedPL); err != nil { - t.Error(err) - } - - // Execute sleep. - ptyMaster.Write([]byte("sleep 100\n")) - - // Wait for it to start. Sleep's PPID is bash's PID. - expectedPL = append(expectedPL, &control.Process{PID: 3, PPID: 2, Cmd: "sleep"}) - if err := waitForProcessList(c, expectedPL); err != nil { - t.Error(err) - } - - // Send a SIGTERM to the foreground process for the exec PID. Note that - // although we pass in the PID of "bash", it should actually terminate - // "sleep", since that is the foreground process. - if err := c.Sandbox.SignalProcess(c.ID, pid, syscall.SIGTERM, true /* fgProcess */); err != nil { - t.Fatalf("error signaling container: %v", err) - } - - // Sleep process should be gone. - expectedPL = expectedPL[:len(expectedPL)-1] - if err := waitForProcessList(c, expectedPL); err != nil { - t.Error(err) - } - - // Sleep is dead, but it may take more time for bash to notice and - // change the foreground process back to itself. We know it is done - // when bash writes "Terminated" to the pty. - if err := testutil.WaitUntilRead(ptyMaster, "Terminated", nil, 5*time.Second); err != nil { - t.Fatalf("bash did not take over pty: %v", err) - } - - // Send a SIGKILL to the foreground process again. This time "bash" - // should be killed. We use SIGKILL instead of SIGTERM or SIGINT - // because bash ignores those. - if err := c.Sandbox.SignalProcess(c.ID, pid, syscall.SIGKILL, true /* fgProcess */); err != nil { - t.Fatalf("error signaling container: %v", err) - } - expectedPL = expectedPL[:1] - if err := waitForProcessList(c, expectedPL); err != nil { - t.Error(err) - } - - // Make sure the process indicates it was killed by a SIGKILL. - ws, err := c.WaitPID(pid, true) - if err != nil { - t.Errorf("waiting on container failed: %v", err) - } - if !ws.Signaled() { - t.Error("ws.Signaled() got false, want true") - } - if got, want := ws.Signal(), syscall.SIGKILL; got != want { - t.Errorf("ws.Signal() got %v, want %v", got, want) - } -} - func TestUserLog(t *testing.T) { app, err := testutil.FindFile("runsc/container/test_app") if err != nil { diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 713b326a6..6dc8cf7f0 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -696,10 +696,15 @@ func (s *Sandbox) SignalContainer(cid string, sig syscall.Signal, all bool) erro } defer conn.Close() + mode := boot.DeliverToProcess + if all { + mode = boot.DeliverToAllProcesses + } + args := boot.SignalArgs{ CID: cid, Signo: int32(sig), - All: all, + Mode: mode, } if err := conn.Call(boot.ContainerSignal, &args, nil); err != nil { return fmt.Errorf("err signaling container %q: %v", cid, err) @@ -719,13 +724,18 @@ func (s *Sandbox) SignalProcess(cid string, pid int32, sig syscall.Signal, fgPro } defer conn.Close() - args := boot.SignalProcessArgs{ - CID: cid, - Signo: int32(sig), - PID: pid, - SendToForegroundProcess: fgProcess, + mode := boot.DeliverToProcess + if fgProcess { + mode = boot.DeliverToForegroundProcessGroup + } + + args := boot.SignalArgs{ + CID: cid, + Signo: int32(sig), + PID: pid, + Mode: mode, } - if err := conn.Call(boot.ContainerSignalProcess, &args, nil); err != nil { + if err := conn.Call(boot.ContainerSignal, &args, nil); err != nil { return fmt.Errorf("err signaling container %q PID %d: %v", cid, pid, err) } return nil diff --git a/runsc/test/integration/exec_test.go b/runsc/test/integration/exec_test.go index d08140ad3..3cac674d0 100644 --- a/runsc/test/integration/exec_test.go +++ b/runsc/test/integration/exec_test.go @@ -66,7 +66,7 @@ func TestExecJobControl(t *testing.T) { if err := testutil.Pull("alpine"); err != nil { t.Fatalf("docker pull failed: %v", err) } - d := testutil.MakeDocker("exec-test") + d := testutil.MakeDocker("exec-job-control-test") // Start the container. if err := d.Run("alpine", "sleep", "1000"); err != nil { diff --git a/runsc/test/integration/integration_test.go b/runsc/test/integration/integration_test.go index b7d07309d..536bb17e0 100644 --- a/runsc/test/integration/integration_test.go +++ b/runsc/test/integration/integration_test.go @@ -28,6 +28,7 @@ import ( "os" "strconv" "strings" + "syscall" "testing" "time" @@ -231,6 +232,53 @@ func TestNumCPU(t *testing.T) { } } +// TestJobControl tests that job control characters are handled properly. +func TestJobControl(t *testing.T) { + if err := testutil.Pull("alpine"); err != nil { + t.Fatalf("docker pull failed: %v", err) + } + d := testutil.MakeDocker("job-control-test") + + // Start the container with an attached PTY. + _, ptmx, err := d.RunWithPty("alpine", "sh") + if err != nil { + t.Fatalf("docker run failed: %v", err) + } + defer ptmx.Close() + defer d.CleanUp() + + // Call "sleep 100" in the shell. + if _, err := ptmx.Write([]byte("sleep 100\n")); err != nil { + t.Fatalf("error writing to pty: %v", err) + } + + // Give shell a few seconds to start executing the sleep. + time.Sleep(2 * time.Second) + + // Send a ^C to the pty, which should kill sleep, but not the shell. + // \x03 is ASCII "end of text", which is the same as ^C. + if _, err := ptmx.Write([]byte{'\x03'}); err != nil { + t.Fatalf("error writing to pty: %v", err) + } + + // The shell should still be alive at this point. Sleep should have + // exited with code 2+128=130. We'll exit with 10 plus that number, so + // that we can be sure that the shell did not get signalled. + if _, err := ptmx.Write([]byte("exit $(expr $? + 10)\n")); err != nil { + t.Fatalf("error writing to pty: %v", err) + } + + // Wait for the container to exit. + got, err := d.Wait(5 * time.Second) + if err != nil { + t.Fatalf("error getting exit code: %v", err) + } + // Container should exit with code 10+130=140. + if want := syscall.WaitStatus(140); got != want { + t.Errorf("container exited with code %d want %d", got, want) + } +} + func TestMain(m *testing.M) { testutil.EnsureSupportedDockerVersion() os.Exit(m.Run()) diff --git a/runsc/test/testutil/docker.go b/runsc/test/testutil/docker.go index 8a51d3eed..4e48817cf 100644 --- a/runsc/test/testutil/docker.go +++ b/runsc/test/testutil/docker.go @@ -25,6 +25,7 @@ import ( "regexp" "strconv" "strings" + "syscall" "time" "github.com/kr/pty" @@ -198,6 +199,13 @@ func (d *Docker) Run(args ...string) error { return err } +// RunWithPty is like Run but with an attached pty. +func (d *Docker) RunWithPty(args ...string) (*exec.Cmd, *os.File, error) { + a := []string{"run", "--runtime", d.Runtime, "--name", d.Name, "-it"} + a = append(a, args...) + return doWithPty(a...) +} + // RunFg calls 'docker run' with the arguments provided in the foreground. It // blocks until the container exits and returns the output. func (d *Docker) RunFg(args ...string) (string, error) { @@ -307,6 +315,37 @@ func (d *Docker) ID() (string, error) { return strings.TrimSpace(string(out)), nil } +// Wait waits for container to exit, up to the given timeout. Returns error if +// wait fails or timeout is hit. Returns the application return code otherwise. +// Note that the application may have failed even if err == nil, always check +// the exit code. +func (d *Docker) Wait(timeout time.Duration) (syscall.WaitStatus, error) { + timeoutChan := time.After(timeout) + waitChan := make(chan (syscall.WaitStatus)) + errChan := make(chan (error)) + + go func() { + out, err := do("wait", d.Name) + if err != nil { + errChan <- fmt.Errorf("error waiting for container %q: %v", d.Name, err) + } + exit, err := strconv.Atoi(strings.TrimSuffix(string(out), "\n")) + if err != nil { + errChan <- fmt.Errorf("error parsing exit code %q: %v", out, err) + } + waitChan <- syscall.WaitStatus(uint32(exit)) + }() + + select { + case ws := <-waitChan: + return ws, nil + case err := <-errChan: + return syscall.WaitStatus(1), err + case <-timeoutChan: + return syscall.WaitStatus(1), fmt.Errorf("timeout waiting for container %q", d.Name) + } +} + // WaitForOutput calls 'docker logs' to retrieve containers output and searches // for the given pattern. func (d *Docker) WaitForOutput(pattern string, timeout time.Duration) (string, error) { -- cgit v1.2.3 From b2a88ff4713325fca736f6a3bf200be02d2d72a7 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Wed, 17 Oct 2018 15:48:55 -0700 Subject: Check thread group CPU timers in the CPU clock ticker. This reduces the number of goroutines and runtime timers when ITIMER_VIRTUAL or ITIMER_PROF are enabled, or when RLIMIT_CPU is set. This also ensures that thread group CPU timers only advance if running tasks are observed at the time the CPU clock advances, mostly eliminating the possibility that a CPU timer expiration observes no running tasks and falls back to the group leader. PiperOrigin-RevId: 217603396 Change-Id: Ia24ce934d5574334857d9afb5ad8ca0b6a6e65f4 --- pkg/sentry/kernel/BUILD | 1 - pkg/sentry/kernel/kernel.go | 46 ++--- pkg/sentry/kernel/task_acct.go | 97 ++++++++- pkg/sentry/kernel/task_clone.go | 2 +- pkg/sentry/kernel/task_exit.go | 3 - pkg/sentry/kernel/task_run.go | 4 - pkg/sentry/kernel/task_sched.go | 344 ++++++++++++++++++++++++++++++-- pkg/sentry/kernel/task_signals.go | 66 ------ pkg/sentry/kernel/thread_group.go | 73 +++++-- pkg/sentry/kernel/threads.go | 6 +- pkg/sentry/kernel/time/time.go | 35 +++- pkg/sentry/kernel/timekeeper.go | 26 +++ pkg/sentry/kernel/timer.go | 290 --------------------------- pkg/sentry/syscalls/linux/sys_rlimit.go | 2 +- pkg/sentry/syscalls/linux/sys_timer.go | 61 ++---- 15 files changed, 551 insertions(+), 505 deletions(-) delete mode 100644 pkg/sentry/kernel/timer.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index acc61cb09..e2fb61ba6 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -111,7 +111,6 @@ go_library( "threads.go", "timekeeper.go", "timekeeper_state.go", - "timer.go", "uts_namespace.go", "vdso.go", "version.go", diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 84afdb530..5d6856f3c 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -20,7 +20,7 @@ // // Kernel.extMu // ThreadGroup.timerMu -// ktime.Timer.mu (for IntervalTimer) +// ktime.Timer.mu (for kernelCPUClockTicker and IntervalTimer) // TaskSet.mu // SignalHandlers.mu // Task.mu @@ -617,7 +617,7 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, ThreadID, return nil, 0, fmt.Errorf("no kernel MountNamespace") } - tg := NewThreadGroup(k.tasks.Root, NewSignalHandlers(), linux.SIGCHLD, args.Limits, k.monotonicClock) + tg := k.newThreadGroup(k.tasks.Root, NewSignalHandlers(), linux.SIGCHLD, args.Limits, k.monotonicClock) ctx := args.NewContext(k) // Grab the root directory. @@ -705,7 +705,7 @@ func (k *Kernel) Start() error { } k.started = true - k.cpuClockTicker = ktime.NewTimer(k.monotonicClock, kernelCPUClockListener{k}) + k.cpuClockTicker = ktime.NewTimer(k.monotonicClock, newKernelCPUClockTicker(k)) k.cpuClockTicker.Swap(ktime.Setting{ Enabled: true, Period: linux.ClockTick, @@ -741,14 +741,13 @@ func (k *Kernel) pauseTimeLocked() { // mutex, while holding the Timer mutex.) for t := range k.tasks.Root.tids { if t == t.tg.leader { - t.tg.tm.pause() - } - // This means we'll iterate ThreadGroups and FDMaps shared by multiple - // tasks repeatedly, but ktime.Timer.Pause is idempotent so this is - // harmless. - for _, it := range t.tg.timers { - it.PauseTimer() + t.tg.itimerRealTimer.Pause() + for _, it := range t.tg.timers { + it.PauseTimer() + } } + // This means we'll iterate FDMaps shared by multiple tasks repeatedly, + // but ktime.Timer.Pause is idempotent so this is harmless. if fdm := t.fds; fdm != nil { for _, desc := range fdm.files { if tfd, ok := desc.file.FileOperations.(*timerfd.TimerOperations); ok { @@ -774,10 +773,10 @@ func (k *Kernel) resumeTimeLocked() { k.timekeeper.ResumeUpdates() for t := range k.tasks.Root.tids { if t == t.tg.leader { - t.tg.tm.resume() - } - for _, it := range t.tg.timers { - it.ResumeTimer() + t.tg.itimerRealTimer.Resume() + for _, it := range t.tg.timers { + it.ResumeTimer() + } } if fdm := t.fds; fdm != nil { for _, desc := range fdm.files { @@ -1078,22 +1077,3 @@ func (ctx supervisorContext) Value(key interface{}) interface{} { return nil } } - -type kernelCPUClockListener struct { - k *Kernel -} - -// Notify implements ktime.TimerListener.Notify. -func (l kernelCPUClockListener) Notify(exp uint64) { - // Only increment cpuClock by 1 regardless of the number of expirations. - // This approximately compensates for cases where thread throttling or bad - // Go runtime scheduling prevents the cpuClockTicker goroutine, and - // presumably task goroutines as well, from executing for a long period of - // time. It's also necessary to prevent CPU clocks from seeing large - // discontinuous jumps. - atomic.AddUint64(&l.k.cpuClock, 1) -} - -// Destroy implements ktime.TimerListener.Destroy. -func (l kernelCPUClockListener) Destroy() { -} diff --git a/pkg/sentry/kernel/task_acct.go b/pkg/sentry/kernel/task_acct.go index ce12cdb64..d2052921e 100644 --- a/pkg/sentry/kernel/task_acct.go +++ b/pkg/sentry/kernel/task_acct.go @@ -21,8 +21,99 @@ import ( ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" + "gvisor.googlesource.com/gvisor/pkg/syserror" ) +// Getitimer implements getitimer(2). +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) Getitimer(id int32) (linux.ItimerVal, error) { + var tm ktime.Time + var s ktime.Setting + switch id { + case linux.ITIMER_REAL: + tm, s = t.tg.itimerRealTimer.Get() + case linux.ITIMER_VIRTUAL: + tm = t.tg.UserCPUClock().Now() + t.tg.signalHandlers.mu.Lock() + s, _ = t.tg.itimerVirtSetting.At(tm) + t.tg.signalHandlers.mu.Unlock() + case linux.ITIMER_PROF: + tm = t.tg.CPUClock().Now() + t.tg.signalHandlers.mu.Lock() + s, _ = t.tg.itimerProfSetting.At(tm) + t.tg.signalHandlers.mu.Unlock() + default: + return linux.ItimerVal{}, syserror.EINVAL + } + val, iv := ktime.SpecFromSetting(tm, s) + return linux.ItimerVal{ + Value: linux.DurationToTimeval(val), + Interval: linux.DurationToTimeval(iv), + }, nil +} + +// Setitimer implements setitimer(2). +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) Setitimer(id int32, newitv linux.ItimerVal) (linux.ItimerVal, error) { + var tm ktime.Time + var olds ktime.Setting + switch id { + case linux.ITIMER_REAL: + news, err := ktime.SettingFromSpec(newitv.Value.ToDuration(), newitv.Interval.ToDuration(), t.tg.itimerRealTimer.Clock()) + if err != nil { + return linux.ItimerVal{}, err + } + tm, olds = t.tg.itimerRealTimer.Swap(news) + case linux.ITIMER_VIRTUAL: + c := t.tg.UserCPUClock() + var err error + t.k.cpuClockTicker.Atomically(func() { + tm = c.Now() + var news ktime.Setting + news, err = ktime.SettingFromSpecAt(newitv.Value.ToDuration(), newitv.Interval.ToDuration(), tm) + if err != nil { + return + } + t.tg.signalHandlers.mu.Lock() + olds = t.tg.itimerVirtSetting + t.tg.itimerVirtSetting = news + t.tg.updateCPUTimersEnabledLocked() + t.tg.signalHandlers.mu.Unlock() + }) + if err != nil { + return linux.ItimerVal{}, err + } + case linux.ITIMER_PROF: + c := t.tg.CPUClock() + var err error + t.k.cpuClockTicker.Atomically(func() { + tm = c.Now() + var news ktime.Setting + news, err = ktime.SettingFromSpecAt(newitv.Value.ToDuration(), newitv.Interval.ToDuration(), tm) + if err != nil { + return + } + t.tg.signalHandlers.mu.Lock() + olds = t.tg.itimerProfSetting + t.tg.itimerProfSetting = news + t.tg.updateCPUTimersEnabledLocked() + t.tg.signalHandlers.mu.Unlock() + }) + if err != nil { + return linux.ItimerVal{}, err + } + default: + return linux.ItimerVal{}, syserror.EINVAL + } + oldval, oldiv := ktime.SpecFromSetting(tm, olds) + return linux.ItimerVal{ + Value: linux.DurationToTimeval(oldval), + Interval: linux.DurationToTimeval(oldiv), + }, nil +} + // IOUsage returns the io usage of the thread. func (t *Task) IOUsage() *usage.IO { return t.ioUsage @@ -56,12 +147,6 @@ func (t *Task) SetName(name string) { t.Debugf("Set thread name to %q", name) } -// SetCPUTimer is used by setrlimit(RLIMIT_CPU) to enforce the hard and soft -// limits on CPU time used by this process. -func (tg *ThreadGroup) SetCPUTimer(l *limits.Limit) { - tg.Timer().applyCPULimits(*l) -} - // Limits implements context.Context.Limits. func (t *Task) Limits() *limits.LimitSet { return t.ThreadGroup().Limits() diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 7c469ec46..de3aef40d 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -241,7 +241,7 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { if opts.NewSignalHandlers { sh = sh.Fork() } - tg = NewThreadGroup(pidns, sh, opts.TerminationSignal, tg.limits.GetCopy(), t.k.monotonicClock) + tg = t.k.newThreadGroup(pidns, sh, opts.TerminationSignal, tg.limits.GetCopy(), t.k.monotonicClock) } cfg := &TaskConfig{ diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index f5b45fb17..65969ca9b 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -675,9 +675,6 @@ func (t *Task) exitNotifyLocked(fromPtraceDetach bool) { t.tg.ioUsage.Accumulate(t.ioUsage) t.tg.signalHandlers.mu.Lock() t.tg.tasks.Remove(t) - if t.tg.lastTimerSignalTask == t { - t.tg.lastTimerSignalTask = nil - } t.tg.tasksCount-- tc := t.tg.tasksCount t.tg.signalHandlers.mu.Unlock() diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 8dd0ef6ea..49ac933b7 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -70,10 +70,6 @@ func (t *Task) run(threadID uintptr) { // Platform.CooperativelySharesAddressSpace() == true, we give up the // AddressSpace before the task goroutine finishes executing. - // Ensure that thread group timers for execution time reflect that this - // task now exists. - t.tg.tm.kick() - // If this is a newly-started task, it should check for participation in // group stops. If this is a task resuming after restore, it was // interrupted by saving. In either case, the task is initially diff --git a/pkg/sentry/kernel/task_sched.go b/pkg/sentry/kernel/task_sched.go index 49141ab74..19dcc963a 100644 --- a/pkg/sentry/kernel/task_sched.go +++ b/pkg/sentry/kernel/task_sched.go @@ -18,12 +18,15 @@ package kernel import ( "fmt" + "math/rand" "sync/atomic" "time" "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/hostcpu" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -84,6 +87,33 @@ type TaskGoroutineSchedInfo struct { SysTicks uint64 } +// userTicksAt returns the extrapolated value of ts.UserTicks after +// Kernel.CPUClockNow() indicates a time of now. +// +// Preconditions: now <= Kernel.CPUClockNow(). (Since Kernel.cpuClock is +// monotonic, this is satisfied if now is the result of a previous call to +// Kernel.CPUClockNow().) This requirement exists because otherwise a racing +// change to t.gosched can cause userTicksAt to adjust stats by too much, +// making the observed stats non-monotonic. +func (ts *TaskGoroutineSchedInfo) userTicksAt(now uint64) uint64 { + if ts.Timestamp < now && ts.State == TaskGoroutineRunningApp { + // Update stats to reflect execution since the last update. + return ts.UserTicks + (now - ts.Timestamp) + } + return ts.UserTicks +} + +// sysTicksAt returns the extrapolated value of ts.SysTicks after +// Kernel.CPUClockNow() indicates a time of now. +// +// Preconditions: As for userTicksAt. +func (ts *TaskGoroutineSchedInfo) sysTicksAt(now uint64) uint64 { + if ts.Timestamp < now && ts.State == TaskGoroutineRunningSys { + return ts.SysTicks + (now - ts.Timestamp) + } + return ts.SysTicks +} + // Preconditions: The caller must be running on the task goroutine. func (t *Task) accountTaskGoroutineEnter(state TaskGoroutineState) { now := t.k.CPUClockNow() @@ -127,26 +157,12 @@ func (t *Task) CPUStats() usage.CPUStats { return t.cpuStatsAt(t.k.CPUClockNow()) } -// Preconditions: now <= Kernel.CPUClockNow(). (Since Kernel.cpuClock is -// monotonic, this is satisfied if now is the result of a previous call to -// Kernel.CPUClockNow().) This requirement exists because otherwise a racing -// change to t.gosched can cause cpuStatsAt to adjust stats by too much, making -// the returned stats non-monotonic. +// Preconditions: As for TaskGoroutineSchedInfo.userTicksAt. func (t *Task) cpuStatsAt(now uint64) usage.CPUStats { tsched := t.TaskGoroutineSchedInfo() - if tsched.Timestamp < now { - // Update stats to reflect execution since the last update to - // t.gosched. - switch tsched.State { - case TaskGoroutineRunningSys: - tsched.SysTicks += now - tsched.Timestamp - case TaskGoroutineRunningApp: - tsched.UserTicks += now - tsched.Timestamp - } - } return usage.CPUStats{ - UserTime: time.Duration(tsched.UserTicks * uint64(linux.ClockTick)), - SysTime: time.Duration(tsched.SysTicks * uint64(linux.ClockTick)), + UserTime: time.Duration(tsched.userTicksAt(now) * uint64(linux.ClockTick)), + SysTime: time.Duration(tsched.sysTicksAt(now) * uint64(linux.ClockTick)), VoluntarySwitches: atomic.LoadUint64(&t.yieldCount), } } @@ -162,9 +178,14 @@ func (tg *ThreadGroup) CPUStats() usage.CPUStats { // ThreadGroup has ever executed anyway. return usage.CPUStats{} } - now := tg.leader.k.CPUClockNow() + return tg.cpuStatsAtLocked(tg.leader.k.CPUClockNow()) +} + +// Preconditions: As for TaskGoroutineSchedInfo.userTicksAt. The TaskSet mutex +// must be locked. +func (tg *ThreadGroup) cpuStatsAtLocked(now uint64) usage.CPUStats { stats := tg.exitedCPUStats - // Account for active tasks. + // Account for live tasks. for t := tg.tasks.Front(); t != nil; t = t.Next() { stats.Accumulate(t.cpuStatsAt(now)) } @@ -182,6 +203,291 @@ func (tg *ThreadGroup) JoinedChildCPUStats() usage.CPUStats { return tg.childCPUStats } +// taskClock is a ktime.Clock that measures the time that a task has spent +// executing. taskClock is primarily used to implement CLOCK_THREAD_CPUTIME_ID. +// +// +stateify savable +type taskClock struct { + t *Task + + // If includeSys is true, the taskClock includes both time spent executing + // application code as well as time spent in the sentry. Otherwise, the + // taskClock includes only time spent executing application code. + includeSys bool + + // Implements waiter.Waitable. TimeUntil wouldn't change its estimation + // based on either of the clock events, so there's no event to be + // notified for. + ktime.NoClockEvents `state:"nosave"` + + // Implements ktime.Clock.WallTimeUntil. + // + // As an upper bound, a task's clock cannot advance faster than CPU + // time. It would have to execute at a rate of more than 1 task-second + // per 1 CPU-second, which isn't possible. + ktime.WallRateClock `state:"nosave"` +} + +// UserCPUClock returns a clock measuring the CPU time the task has spent +// executing application code. +func (t *Task) UserCPUClock() ktime.Clock { + return &taskClock{t: t, includeSys: false} +} + +// CPUClock returns a clock measuring the CPU time the task has spent executing +// application and "kernel" code. +func (t *Task) CPUClock() ktime.Clock { + return &taskClock{t: t, includeSys: true} +} + +// Now implements ktime.Clock.Now. +func (tc *taskClock) Now() ktime.Time { + stats := tc.t.CPUStats() + if tc.includeSys { + return ktime.FromNanoseconds((stats.UserTime + stats.SysTime).Nanoseconds()) + } + return ktime.FromNanoseconds(stats.UserTime.Nanoseconds()) +} + +// tgClock is a ktime.Clock that measures the time a thread group has spent +// executing. tgClock is primarily used to implement CLOCK_PROCESS_CPUTIME_ID. +// +// +stateify savable +type tgClock struct { + tg *ThreadGroup + + // If includeSys is true, the tgClock includes both time spent executing + // application code as well as time spent in the sentry. Otherwise, the + // tgClock includes only time spent executing application code. + includeSys bool + + // Implements waiter.Waitable. + ktime.ClockEventsQueue `state:"nosave"` +} + +// Now implements ktime.Clock.Now. +func (tgc *tgClock) Now() ktime.Time { + stats := tgc.tg.CPUStats() + if tgc.includeSys { + return ktime.FromNanoseconds((stats.UserTime + stats.SysTime).Nanoseconds()) + } + return ktime.FromNanoseconds(stats.UserTime.Nanoseconds()) +} + +// WallTimeUntil implements ktime.Clock.WallTimeUntil. +func (tgc *tgClock) WallTimeUntil(t, now ktime.Time) time.Duration { + // Thread group CPU time should not exceed wall time * live tasks, since + // task goroutines exit after the transition to TaskExitZombie in + // runExitNotify. + tgc.tg.pidns.owner.mu.RLock() + n := tgc.tg.liveTasks + tgc.tg.pidns.owner.mu.RUnlock() + if n == 0 { + if t.Before(now) { + return 0 + } + // The timer tick raced with thread group exit, after which no more + // tasks can enter the thread group. So tgc.Now() will never advance + // again. Return a large delay; the timer should be stopped long before + // it comes again anyway. + return time.Hour + } + // This is a lower bound on the amount of time that can elapse before an + // associated timer expires, so returning this value tends to result in a + // sequence of closely-spaced ticks just before timer expiry. To avoid + // this, round up to the nearest ClockTick; CPU usage measurements are + // limited to this resolution anyway. + remaining := time.Duration(t.Sub(now).Nanoseconds()/int64(n)) * time.Nanosecond + return ((remaining + (linux.ClockTick - time.Nanosecond)) / linux.ClockTick) * linux.ClockTick +} + +// UserCPUClock returns a ktime.Clock that measures the time that a thread +// group has spent executing. +func (tg *ThreadGroup) UserCPUClock() ktime.Clock { + return &tgClock{tg: tg, includeSys: false} +} + +// CPUClock returns a ktime.Clock that measures the time that a thread group +// has spent executing, including sentry time. +func (tg *ThreadGroup) CPUClock() ktime.Clock { + return &tgClock{tg: tg, includeSys: true} +} + +type kernelCPUClockTicker struct { + k *Kernel + + // These are essentially kernelCPUClockTicker.Notify local variables that + // are cached between calls to reduce allocations. + rng *rand.Rand + tgs []*ThreadGroup +} + +func newKernelCPUClockTicker(k *Kernel) *kernelCPUClockTicker { + return &kernelCPUClockTicker{ + k: k, + rng: rand.New(rand.NewSource(rand.Int63())), + } +} + +// Notify implements ktime.TimerListener.Notify. +func (ticker *kernelCPUClockTicker) Notify(exp uint64) { + // Only increment cpuClock by 1 regardless of the number of expirations. + // This approximately compensates for cases where thread throttling or bad + // Go runtime scheduling prevents the kernelCPUClockTicker goroutine, and + // presumably task goroutines as well, from executing for a long period of + // time. It's also necessary to prevent CPU clocks from seeing large + // discontinuous jumps. + now := atomic.AddUint64(&ticker.k.cpuClock, 1) + + // Check thread group CPU timers. + tgs := ticker.k.tasks.Root.ThreadGroupsAppend(ticker.tgs) + for _, tg := range tgs { + if atomic.LoadUint32(&tg.cpuTimersEnabled) == 0 { + continue + } + + ticker.k.tasks.mu.RLock() + if tg.leader == nil { + // No tasks have ever run in this thread group. + ticker.k.tasks.mu.RUnlock() + continue + } + // Accumulate thread group CPU stats, and randomly select running tasks + // using reservoir sampling to receive CPU timer signals. + var virtReceiver *Task + nrVirtCandidates := 0 + var profReceiver *Task + nrProfCandidates := 0 + tgUserTime := tg.exitedCPUStats.UserTime + tgSysTime := tg.exitedCPUStats.SysTime + for t := tg.tasks.Front(); t != nil; t = t.Next() { + tsched := t.TaskGoroutineSchedInfo() + tgUserTime += time.Duration(tsched.userTicksAt(now) * uint64(linux.ClockTick)) + tgSysTime += time.Duration(tsched.sysTicksAt(now) * uint64(linux.ClockTick)) + switch tsched.State { + case TaskGoroutineRunningApp: + // Considered by ITIMER_VIRT, ITIMER_PROF, and RLIMIT_CPU + // timers. + nrVirtCandidates++ + if int(randInt31n(ticker.rng, int32(nrVirtCandidates))) == 0 { + virtReceiver = t + } + fallthrough + case TaskGoroutineRunningSys: + // Considered by ITIMER_PROF and RLIMIT_CPU timers. + nrProfCandidates++ + if int(randInt31n(ticker.rng, int32(nrProfCandidates))) == 0 { + profReceiver = t + } + } + } + tgVirtNow := ktime.FromNanoseconds(tgUserTime.Nanoseconds()) + tgProfNow := ktime.FromNanoseconds((tgUserTime + tgSysTime).Nanoseconds()) + + // All of the following are standard (not real-time) signals, which are + // automatically deduplicated, so we ignore the number of expirations. + tg.signalHandlers.mu.Lock() + // It should only be possible for these timers to advance if we found + // at least one running task. + if virtReceiver != nil { + // ITIMER_VIRTUAL + newItimerVirtSetting, exp := tg.itimerVirtSetting.At(tgVirtNow) + tg.itimerVirtSetting = newItimerVirtSetting + if exp != 0 { + virtReceiver.sendSignalLocked(sigPriv(linux.SIGVTALRM), true) + } + } + if profReceiver != nil { + // ITIMER_PROF + newItimerProfSetting, exp := tg.itimerProfSetting.At(tgProfNow) + tg.itimerProfSetting = newItimerProfSetting + if exp != 0 { + profReceiver.sendSignalLocked(sigPriv(linux.SIGPROF), true) + } + // RLIMIT_CPU soft limit + newRlimitCPUSoftSetting, exp := tg.rlimitCPUSoftSetting.At(tgProfNow) + tg.rlimitCPUSoftSetting = newRlimitCPUSoftSetting + if exp != 0 { + profReceiver.sendSignalLocked(sigPriv(linux.SIGXCPU), true) + } + // RLIMIT_CPU hard limit + rlimitCPUMax := tg.limits.Get(limits.CPU).Max + if rlimitCPUMax != limits.Infinity && !tgProfNow.Before(ktime.FromSeconds(int64(rlimitCPUMax))) { + profReceiver.sendSignalLocked(sigPriv(linux.SIGKILL), true) + } + } + tg.signalHandlers.mu.Unlock() + + ticker.k.tasks.mu.RUnlock() + } + + // Retain tgs between calls to Notify to reduce allocations. + for i := range tgs { + tgs[i] = nil + } + ticker.tgs = tgs[:0] +} + +// Destroy implements ktime.TimerListener.Destroy. +func (ticker *kernelCPUClockTicker) Destroy() { +} + +// randInt31n returns a random integer in [0, n). +// +// randInt31n is equivalent to math/rand.Rand.int31n(), which is unexported. +// See that function for details. +func randInt31n(rng *rand.Rand, n int32) int32 { + v := rng.Uint32() + prod := uint64(v) * uint64(n) + low := uint32(prod) + if low < uint32(n) { + thresh := uint32(-n) % uint32(n) + for low < thresh { + v = rng.Uint32() + prod = uint64(v) * uint64(n) + low = uint32(prod) + } + } + return int32(prod >> 32) +} + +// NotifyRlimitCPUUpdated is called by setrlimit. +// +// Preconditions: The caller must be running on the task goroutine. +func (t *Task) NotifyRlimitCPUUpdated() { + t.k.cpuClockTicker.Atomically(func() { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + rlimitCPU := t.tg.limits.Get(limits.CPU) + t.tg.rlimitCPUSoftSetting = ktime.Setting{ + Enabled: rlimitCPU.Cur != limits.Infinity, + Next: ktime.FromNanoseconds((time.Duration(rlimitCPU.Cur) * time.Second).Nanoseconds()), + Period: time.Second, + } + if rlimitCPU.Max != limits.Infinity { + // Check if tg is already over the hard limit. + tgcpu := t.tg.cpuStatsAtLocked(t.k.CPUClockNow()) + tgProfNow := ktime.FromNanoseconds((tgcpu.UserTime + tgcpu.SysTime).Nanoseconds()) + if !tgProfNow.Before(ktime.FromSeconds(int64(rlimitCPU.Max))) { + t.sendSignalLocked(sigPriv(linux.SIGKILL), true) + } + } + t.tg.updateCPUTimersEnabledLocked() + }) +} + +// Preconditions: The signal mutex must be locked. +func (tg *ThreadGroup) updateCPUTimersEnabledLocked() { + rlimitCPU := tg.limits.Get(limits.CPU) + if tg.itimerVirtSetting.Enabled || tg.itimerProfSetting.Enabled || tg.rlimitCPUSoftSetting.Enabled || rlimitCPU.Max != limits.Infinity { + atomic.StoreUint32(&tg.cpuTimersEnabled, 1) + } else { + atomic.StoreUint32(&tg.cpuTimersEnabled, 0) + } +} + // StateStatus returns a string representation of the task's current state, // appropriate for /proc/[pid]/status. func (t *Task) StateStatus() string { diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index afb010f60..e2925a708 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -359,72 +359,6 @@ func (tg *ThreadGroup) SendSignal(info *arch.SignalInfo) error { return tg.leader.sendSignalLocked(info, true /* group */) } -// Preconditions: The TaskSet mutex must be locked. -func (t *Task) onCPULocked(includeSys bool) bool { - // Task is exiting. - if t.exitState != TaskExitNone { - return false - } - - switch t.TaskGoroutineSchedInfo().State { - case TaskGoroutineRunningSys: - return includeSys - case TaskGoroutineRunningApp: - return true - default: - return false - } -} - -// SendTimerSignal mimics the process timer signal delivery behavior in linux: -// signals are delivered to the thread that triggers the timer expiration (see -// kernel/time/posix-cpu-timers.c:check_process_timers(). This -// means -// 1) the thread is running on cpu at the time. -// 2) a thread runs more frequently will get more of those signals. -// -// We approximate this behavior by selecting a running task in a round-robin -// fashion. Statistically, a thread running more often should have a higher -// probability to be selected. -func (tg *ThreadGroup) SendTimerSignal(info *arch.SignalInfo, includeSys bool) error { - tg.pidns.owner.mu.RLock() - defer tg.pidns.owner.mu.RUnlock() - tg.signalHandlers.mu.Lock() - defer tg.signalHandlers.mu.Unlock() - - // Find the next running threads. - var t *Task - if tg.lastTimerSignalTask == nil { - t = tg.tasks.Front() - } else { - t = tg.lastTimerSignalTask.Next() - } - - // Iterate from lastTimerSignalTask.Next() to the last task in the task list. - for t != nil { - if t.onCPULocked(includeSys) { - tg.lastTimerSignalTask = t - return t.sendSignalLocked(info, true /* group */) - } - t = t.Next() - } - - // t is nil when we reach here. If lastTimerSignalTask is not nil, iterate - // from Front to lastTimerSignalTask. - if tg.lastTimerSignalTask != nil { - for t := tg.tasks.Front(); t != tg.lastTimerSignalTask.Next(); t = t.Next() { - if t.onCPULocked(includeSys) { - tg.lastTimerSignalTask = t - return t.sendSignalLocked(info, true /* group */) - } - } - } - - // No running threads? Just try the leader. - tg.lastTimerSignalTask = tg.leader - return tg.leader.sendSignalLocked(info, true /* group */) -} - func (t *Task) sendSignalLocked(info *arch.SignalInfo, group bool) error { return t.sendSignalTimerLocked(info, group, nil) } diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index 13dce08ce..dfff7b52d 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -19,6 +19,7 @@ import ( "sync/atomic" "gvisor.googlesource.com/gvisor/pkg/abi/linux" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" ) @@ -59,12 +60,6 @@ type ThreadGroup struct { // pendingSignals is protected by the signal mutex. pendingSignals pendingSignals - // lastTimerSignalTask records the last task we deliver a process timer signal to. - // Please see SendTimerSignal for more details. - // - // lastTimerSignalTask is protected by the signal mutex. - lastTimerSignalTask *Task - // groupStopPhase indicates the state of a group stop in progress on the // thread group, if any. // @@ -152,14 +147,39 @@ type ThreadGroup struct { // restarted by Task.Start. liveGoroutines sync.WaitGroup `state:"nosave"` - // tm contains process timers. TimerManager fields are immutable. - tm TimerManager + timerMu sync.Mutex `state:"nosave"` + + // itimerRealTimer implements ITIMER_REAL for the thread group. + itimerRealTimer *ktime.Timer + + // itimerVirtSetting is the ITIMER_VIRTUAL setting for the thread group. + // + // itimerVirtSetting is protected by the signal mutex. + itimerVirtSetting ktime.Setting + + // itimerProfSetting is the ITIMER_PROF setting for the thread group. + // + // itimerProfSetting is protected by the signal mutex. + itimerProfSetting ktime.Setting + + // rlimitCPUSoftSetting is the setting for RLIMIT_CPU soft limit + // notifications for the thread group. + // + // rlimitCPUSoftSetting is protected by the signal mutex. + rlimitCPUSoftSetting ktime.Setting + + // cpuTimersEnabled is non-zero if itimerVirtSetting.Enabled is true, + // itimerProfSetting.Enabled is true, rlimitCPUSoftSetting.Enabled is true, + // or limits.Get(CPU) is finite. + // + // cpuTimersEnabled is protected by the signal mutex. cpuTimersEnabled is + // accessed using atomic memory operations. + cpuTimersEnabled uint32 // timers is the thread group's POSIX interval timers. nextTimerID is the // TimerID at which allocation should begin searching for an unused ID. // // timers and nextTimerID are protected by timerMu. - timerMu sync.Mutex `state:"nosave"` timers map[linux.TimerID]*IntervalTimer nextTimerID linux.TimerID @@ -211,11 +231,11 @@ type ThreadGroup struct { rscr atomic.Value `state:".(*RSEQCriticalRegion)"` } -// NewThreadGroup returns a new, empty thread group in PID namespace ns. The +// newThreadGroup returns a new, empty thread group in PID namespace ns. The // thread group leader will send its parent terminationSignal when it exits. // The new thread group isn't visible to the system until a task has been // created inside of it by a successful call to TaskSet.NewTask. -func NewThreadGroup(ns *PIDNamespace, sh *SignalHandlers, terminationSignal linux.Signal, limits *limits.LimitSet, monotonicClock *timekeeperClock) *ThreadGroup { +func (k *Kernel) newThreadGroup(ns *PIDNamespace, sh *SignalHandlers, terminationSignal linux.Signal, limits *limits.LimitSet, monotonicClock *timekeeperClock) *ThreadGroup { tg := &ThreadGroup{ threadGroupNode: threadGroupNode{ pidns: ns, @@ -225,7 +245,7 @@ func NewThreadGroup(ns *PIDNamespace, sh *SignalHandlers, terminationSignal linu ioUsage: &usage.IO{}, limits: limits, } - tg.tm = newTimerManager(tg, monotonicClock) + tg.itimerRealTimer = ktime.NewTimer(k.monotonicClock, &itimerRealListener{tg: tg}) tg.timers = make(map[linux.TimerID]*IntervalTimer) tg.rscr.Store(&RSEQCriticalRegion{}) return tg @@ -249,11 +269,6 @@ func (tg *ThreadGroup) SignalHandlers() *SignalHandlers { return tg.signalHandlers } -// Timer returns tg's timers. -func (tg *ThreadGroup) Timer() *TimerManager { - return &tg.tm -} - // Limits returns tg's limits. func (tg *ThreadGroup) Limits() *limits.LimitSet { return tg.limits @@ -261,11 +276,9 @@ func (tg *ThreadGroup) Limits() *limits.LimitSet { // release releases the thread group's resources. func (tg *ThreadGroup) release() { - // These must be done without holding the TaskSet or signal mutexes since - // timers send signals with Timer.mu locked. - - tg.tm.destroy() - + // Timers must be destroyed without holding the TaskSet or signal mutexes + // since timers send signals with Timer.mu locked. + tg.itimerRealTimer.Destroy() var its []*IntervalTimer tg.pidns.owner.mu.Lock() tg.signalHandlers.mu.Lock() @@ -292,3 +305,19 @@ func (tg *ThreadGroup) forEachChildThreadGroupLocked(fn func(*ThreadGroup)) { } } } + +// itimerRealListener implements ktime.Listener for ITIMER_REAL expirations. +// +// +stateify savable +type itimerRealListener struct { + tg *ThreadGroup +} + +// Notify implements ktime.TimerListener.Notify. +func (l *itimerRealListener) Notify(exp uint64) { + l.tg.SendSignal(sigPriv(linux.SIGALRM)) +} + +// Destroy implements ktime.TimerListener.Destroy. +func (l *itimerRealListener) Destroy() { +} diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index 3d5713106..4e3d19e97 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -243,9 +243,13 @@ func (ns *PIDNamespace) Tasks() []*Task { // ThreadGroups returns a snapshot of the thread groups in ns. func (ns *PIDNamespace) ThreadGroups() []*ThreadGroup { + return ns.ThreadGroupsAppend(nil) +} + +// ThreadGroupsAppend appends a snapshot of the thread groups in ns to tgs. +func (ns *PIDNamespace) ThreadGroupsAppend(tgs []*ThreadGroup) []*ThreadGroup { ns.owner.mu.RLock() defer ns.owner.mu.RUnlock() - var tgs []*ThreadGroup for t := range ns.tids { if t == t.tg.leader { tgs = append(tgs, t.tg) diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go index 1f6fed007..52e0dfba1 100644 --- a/pkg/sentry/kernel/time/time.go +++ b/pkg/sentry/kernel/time/time.go @@ -307,6 +307,12 @@ type Setting struct { // SettingFromSpec converts a (value, interval) pair to a Setting based on a // reading from c. value is interpreted as a time relative to c.Now(). func SettingFromSpec(value time.Duration, interval time.Duration, c Clock) (Setting, error) { + return SettingFromSpecAt(value, interval, c.Now()) +} + +// SettingFromSpecAt converts a (value, interval) pair to a Setting. value is +// interpreted as a time relative to now. +func SettingFromSpecAt(value time.Duration, interval time.Duration, now Time) (Setting, error) { if value < 0 { return Setting{}, syserror.EINVAL } @@ -315,7 +321,7 @@ func SettingFromSpec(value time.Duration, interval time.Duration, c Clock) (Sett } return Setting{ Enabled: true, - Next: c.Now().Add(value), + Next: now.Add(value), Period: interval, }, nil } @@ -365,14 +371,14 @@ func ItimerspecFromSetting(now Time, s Setting) linux.Itimerspec { } } -// advancedTo returns an updated Setting and a number of expirations after -// the associated Clock indicates a time of now. +// At returns an updated Setting and a number of expirations after the +// associated Clock indicates a time of now. // -// Settings may be created by successive calls to advancedTo with decreasing +// Settings may be created by successive calls to At with decreasing // values of now (i.e. time may appear to go backward). Supporting this is // required to support non-monotonic clocks, as well as allowing // Timer.clock.Now() to be called without holding Timer.mu. -func (s Setting) advancedTo(now Time) (Setting, uint64) { +func (s Setting) At(now Time) (Setting, uint64) { if !s.Enabled { return s, 0 } @@ -519,7 +525,7 @@ func (t *Timer) Tick() { if t.paused { return } - s, exp := t.setting.advancedTo(now) + s, exp := t.setting.At(now) t.setting = s if exp > 0 { t.listener.Notify(exp) @@ -574,7 +580,7 @@ func (t *Timer) Get() (Time, Setting) { if t.paused { panic(fmt.Sprintf("Timer.Get called on paused Timer %p", t)) } - s, exp := t.setting.advancedTo(now) + s, exp := t.setting.At(now) t.setting = s if exp > 0 { t.listener.Notify(exp) @@ -607,14 +613,14 @@ func (t *Timer) SwapAnd(s Setting, f func()) (Time, Setting) { if t.paused { panic(fmt.Sprintf("Timer.SwapAnd called on paused Timer %p", t)) } - oldS, oldExp := t.setting.advancedTo(now) + oldS, oldExp := t.setting.At(now) if oldExp > 0 { t.listener.Notify(oldExp) } if f != nil { f() } - newS, newExp := s.advancedTo(now) + newS, newExp := s.At(now) t.setting = newS if newExp > 0 { t.listener.Notify(newExp) @@ -623,6 +629,17 @@ func (t *Timer) SwapAnd(s Setting, f func()) (Time, Setting) { return now, oldS } +// Atomically invokes f atomically with respect to expirations of t; that is, t +// cannot generate expirations while f is being called. +// +// Preconditions: f cannot call any Timer methods since it is called with the +// Timer mutex locked. +func (t *Timer) Atomically(f func()) { + t.mu.Lock() + defer t.mu.Unlock() + f() +} + // Preconditions: t.mu must be locked. func (t *Timer) resetKickerLocked(now Time) { if t.setting.Enabled { diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go index df5dbe128..2167f3efe 100644 --- a/pkg/sentry/kernel/timekeeper.go +++ b/pkg/sentry/kernel/timekeeper.go @@ -15,6 +15,7 @@ package kernel import ( + "fmt" "sync" "time" @@ -277,3 +278,28 @@ func (t *Timekeeper) GetTime(c sentrytime.ClockID) (int64, error) { func (t *Timekeeper) BootTime() ktime.Time { return t.bootTime } + +// timekeeperClock is a ktime.Clock that reads time from a +// kernel.Timekeeper-managed clock. +// +// +stateify savable +type timekeeperClock struct { + tk *Timekeeper + c sentrytime.ClockID + + // Implements ktime.Clock.WallTimeUntil. + ktime.WallRateClock `state:"nosave"` + + // Implements waiter.Waitable. (We have no ability to detect + // discontinuities from external changes to CLOCK_REALTIME). + ktime.NoClockEvents `state:"nosave"` +} + +// Now implements ktime.Clock.Now. +func (tc *timekeeperClock) Now() ktime.Time { + now, err := tc.tk.GetTime(tc.c) + if err != nil { + panic(fmt.Sprintf("timekeeperClock(ClockID=%v)).Now: %v", tc.c, err)) + } + return ktime.FromNanoseconds(now) +} diff --git a/pkg/sentry/kernel/timer.go b/pkg/sentry/kernel/timer.go deleted file mode 100644 index 534d03d0f..000000000 --- a/pkg/sentry/kernel/timer.go +++ /dev/null @@ -1,290 +0,0 @@ -// Copyright 2018 Google Inc. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package kernel - -import ( - "fmt" - "time" - - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/limits" - sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" -) - -// timekeeperClock is a ktime.Clock that reads time from a -// kernel.Timekeeper-managed clock. -// -// +stateify savable -type timekeeperClock struct { - tk *Timekeeper - c sentrytime.ClockID - - // Implements ktime.Clock.WallTimeUntil. - ktime.WallRateClock `state:"nosave"` - - // Implements waiter.Waitable. (We have no ability to detect - // discontinuities from external changes to CLOCK_REALTIME). - ktime.NoClockEvents `state:"nosave"` -} - -// Now implements ktime.Clock.Now. -func (tc *timekeeperClock) Now() ktime.Time { - now, err := tc.tk.GetTime(tc.c) - if err != nil { - panic(fmt.Sprintf("timekeeperClock(ClockID=%v)).Now: %v", tc.c, err)) - } - return ktime.FromNanoseconds(now) -} - -// tgClock is a ktime.Clock that measures the time a thread group has spent -// executing. -// -// +stateify savable -type tgClock struct { - tg *ThreadGroup - - // If includeSys is true, the tgClock includes both time spent executing - // application code as well as time spent in the sentry. Otherwise, the - // tgClock includes only time spent executing application code. - includeSys bool - - // Implements waiter.Waitable. - ktime.ClockEventsQueue `state:"nosave"` -} - -// UserCPUClock returns a ktime.Clock that measures the time that a thread -// group has spent executing. -func (tg *ThreadGroup) UserCPUClock() ktime.Clock { - return tg.tm.virtClock -} - -// CPUClock returns a ktime.Clock that measures the time that a thread group -// has spent executing, including sentry time. -func (tg *ThreadGroup) CPUClock() ktime.Clock { - return tg.tm.profClock -} - -// Now implements ktime.Clock.Now. -func (tgc *tgClock) Now() ktime.Time { - stats := tgc.tg.CPUStats() - if tgc.includeSys { - return ktime.FromNanoseconds((stats.UserTime + stats.SysTime).Nanoseconds()) - } - return ktime.FromNanoseconds(stats.UserTime.Nanoseconds()) -} - -// WallTimeUntil implements ktime.Clock.WallTimeUntil. -func (tgc *tgClock) WallTimeUntil(t, now ktime.Time) time.Duration { - // The assumption here is that the time spent in this process (not matter - // virtual or prof) should not exceed wall time * active tasks, since - // Task.exitThreadGroup stops accounting as it transitions to - // TaskExitInitiated. - tgc.tg.pidns.owner.mu.RLock() - n := tgc.tg.activeTasks - tgc.tg.pidns.owner.mu.RUnlock() - if n == 0 { - if t.Before(now) { - return 0 - } - // The timer tick raced with thread group exit, after which no more - // tasks can enter the thread group. So tgc.Now() will never advance - // again. Return a large delay; the timer should be stopped long before - // it comes again anyway. - return time.Hour - } - // This is a lower bound on the amount of time that can elapse before an - // associated timer expires, so returning this value tends to result in a - // sequence of closely-spaced ticks just before timer expiry. To avoid - // this, round up to the nearest ClockTick; CPU usage measurements are - // limited to this resolution anyway. - remaining := time.Duration(int64(t.Sub(now))/int64(n)) * time.Nanosecond - return ((remaining + (linux.ClockTick - time.Nanosecond)) / linux.ClockTick) * linux.ClockTick -} - -// taskClock is a ktime.Clock that measures the time that a task has spent -// executing. -type taskClock struct { - t *Task - - // If includeSys is true, the taskClock includes both time spent executing - // application code as well as time spent in the sentry. Otherwise, the - // taskClock includes only time spent executing application code. - includeSys bool - - // Implements waiter.Waitable. TimeUntil wouldn't change its estimation - // based on either of the clock events, so there's no event to be - // notified for. - ktime.NoClockEvents `state:"nosave"` - - // Implements ktime.Clock.WallTimeUntil. - // - // As an upper bound, a task's clock cannot advance faster than CPU - // time. It would have to execute at a rate of more than 1 task-second - // per 1 CPU-second, which isn't possible. - ktime.WallRateClock `state:"nosave"` -} - -// UserCPUClock returns a clock measuring the CPU time the task has spent -// executing application code. -func (t *Task) UserCPUClock() ktime.Clock { - return &taskClock{t: t, includeSys: false} -} - -// CPUClock returns a clock measuring the CPU time the task has spent executing -// application and "kernel" code. -func (t *Task) CPUClock() ktime.Clock { - return &taskClock{t: t, includeSys: true} -} - -// Now implements ktime.Clock.Now. -func (tc *taskClock) Now() ktime.Time { - stats := tc.t.CPUStats() - if tc.includeSys { - return ktime.FromNanoseconds((stats.UserTime + stats.SysTime).Nanoseconds()) - } - return ktime.FromNanoseconds(stats.UserTime.Nanoseconds()) -} - -// signalNotifier is a ktime.Listener that sends signals to a ThreadGroup. -// -// +stateify savable -type signalNotifier struct { - tg *ThreadGroup - signal linux.Signal - realTimer bool - includeSys bool -} - -// Notify implements ktime.TimerListener.Notify. -func (s *signalNotifier) Notify(exp uint64) { - // Since all signals sent using a signalNotifier are standard (not - // real-time) signals, we can ignore the number of expirations and send - // only a single signal. - if s.realTimer { - // real timer signal sent to leader. See kernel/time/itimer.c:it_real_fn - s.tg.SendSignal(sigPriv(s.signal)) - } else { - s.tg.SendTimerSignal(sigPriv(s.signal), s.includeSys) - } -} - -// Destroy implements ktime.TimerListener.Destroy. -func (s *signalNotifier) Destroy() {} - -// TimerManager is a collection of supported process cpu timers. -// -// +stateify savable -type TimerManager struct { - // Clocks used to drive thread group execution time timers. - virtClock *tgClock - profClock *tgClock - - RealTimer *ktime.Timer - VirtualTimer *ktime.Timer - ProfTimer *ktime.Timer - SoftLimitTimer *ktime.Timer - HardLimitTimer *ktime.Timer -} - -// newTimerManager returns a new instance of TimerManager. -func newTimerManager(tg *ThreadGroup, monotonicClock ktime.Clock) TimerManager { - virtClock := &tgClock{tg: tg, includeSys: false} - profClock := &tgClock{tg: tg, includeSys: true} - tm := TimerManager{ - virtClock: virtClock, - profClock: profClock, - RealTimer: ktime.NewTimer(monotonicClock, &signalNotifier{ - tg: tg, - signal: linux.SIGALRM, - realTimer: true, - includeSys: false, - }), - VirtualTimer: ktime.NewTimer(virtClock, &signalNotifier{ - tg: tg, - signal: linux.SIGVTALRM, - realTimer: false, - includeSys: false, - }), - ProfTimer: ktime.NewTimer(profClock, &signalNotifier{ - tg: tg, - signal: linux.SIGPROF, - realTimer: false, - includeSys: true, - }), - SoftLimitTimer: ktime.NewTimer(profClock, &signalNotifier{ - tg: tg, - signal: linux.SIGXCPU, - realTimer: false, - includeSys: true, - }), - HardLimitTimer: ktime.NewTimer(profClock, &signalNotifier{ - tg: tg, - signal: linux.SIGKILL, - realTimer: false, - includeSys: true, - }), - } - tm.applyCPULimits(tg.Limits().Get(limits.CPU)) - return tm -} - -// Save saves this TimerManger. - -// destroy destroys all timers. -func (tm *TimerManager) destroy() { - tm.RealTimer.Destroy() - tm.VirtualTimer.Destroy() - tm.ProfTimer.Destroy() - tm.SoftLimitTimer.Destroy() - tm.HardLimitTimer.Destroy() -} - -func (tm *TimerManager) applyCPULimits(l limits.Limit) { - tm.SoftLimitTimer.Swap(ktime.Setting{ - Enabled: l.Cur != limits.Infinity, - Next: ktime.FromNanoseconds((time.Duration(l.Cur) * time.Second).Nanoseconds()), - Period: time.Second, - }) - tm.HardLimitTimer.Swap(ktime.Setting{ - Enabled: l.Max != limits.Infinity, - Next: ktime.FromNanoseconds((time.Duration(l.Max) * time.Second).Nanoseconds()), - }) -} - -// kick is called when the number of threads in the thread group associated -// with tm increases. -func (tm *TimerManager) kick() { - tm.virtClock.Notify(ktime.ClockEventRateIncrease) - tm.profClock.Notify(ktime.ClockEventRateIncrease) -} - -// pause is to pause the timers and stop timer signal delivery. -func (tm *TimerManager) pause() { - tm.RealTimer.Pause() - tm.VirtualTimer.Pause() - tm.ProfTimer.Pause() - tm.SoftLimitTimer.Pause() - tm.HardLimitTimer.Pause() -} - -// resume is to resume the timers and continue timer signal delivery. -func (tm *TimerManager) resume() { - tm.RealTimer.Resume() - tm.VirtualTimer.Resume() - tm.ProfTimer.Resume() - tm.SoftLimitTimer.Resume() - tm.HardLimitTimer.Resume() -} diff --git a/pkg/sentry/syscalls/linux/sys_rlimit.go b/pkg/sentry/syscalls/linux/sys_rlimit.go index 481e79eaa..d806b58ab 100644 --- a/pkg/sentry/syscalls/linux/sys_rlimit.go +++ b/pkg/sentry/syscalls/linux/sys_rlimit.go @@ -111,7 +111,7 @@ func prlimit64(t *kernel.Task, resource limits.LimitType, newLim *limits.Limit) } if resource == limits.CPU { - t.ThreadGroup().SetCPUTimer(newLim) + t.NotifyRlimitCPUUpdated() } return oldLim, nil } diff --git a/pkg/sentry/syscalls/linux/sys_timer.go b/pkg/sentry/syscalls/linux/sys_timer.go index a12d12d9d..c41074d54 100644 --- a/pkg/sentry/syscalls/linux/sys_timer.go +++ b/pkg/sentry/syscalls/linux/sys_timer.go @@ -21,7 +21,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) @@ -70,34 +69,15 @@ func copyItimerValOut(t *kernel.Task, addr usermem.Addr, itv *linux.ItimerVal) e } } -func findTimer(t *kernel.Task, which int32) (*ktime.Timer, error) { - switch which { - case linux.ITIMER_REAL: - return t.ThreadGroup().Timer().RealTimer, nil - case linux.ITIMER_VIRTUAL: - return t.ThreadGroup().Timer().VirtualTimer, nil - case linux.ITIMER_PROF: - return t.ThreadGroup().Timer().ProfTimer, nil - default: - return nil, syscall.EINVAL - } -} - // Getitimer implements linux syscall getitimer(2). func Getitimer(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { timerID := args[0].Int() val := args[1].Pointer() - timer, err := findTimer(t, timerID) + olditv, err := t.Getitimer(timerID) if err != nil { return 0, nil, err } - value, interval := ktime.SpecFromSetting(timer.Get()) - olditv := linux.ItimerVal{ - Value: linux.DurationToTimeval(value), - Interval: linux.DurationToTimeval(interval), - } - return 0, nil, copyItimerValOut(t, val, &olditv) } @@ -107,29 +87,14 @@ func Setitimer(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys newVal := args[1].Pointer() oldVal := args[2].Pointer() - timer, err := findTimer(t, timerID) + newitv, err := copyItimerValIn(t, newVal) if err != nil { return 0, nil, err } - - itv, err := copyItimerValIn(t, newVal) + olditv, err := t.Setitimer(timerID, newitv) if err != nil { return 0, nil, err } - // Just like linux, we cap the timer value and interval with the max - // number that int64 can represent which is roughly 292 years. - s, err := ktime.SettingFromSpec(itv.Value.ToDuration(), - itv.Interval.ToDuration(), timer.Clock()) - if err != nil { - return 0, nil, err - } - - valueNS, intervalNS := ktime.SpecFromSetting(timer.Swap(s)) - olditv := linux.ItimerVal{ - Value: linux.DurationToTimeval(valueNS), - Interval: linux.DurationToTimeval(intervalNS), - } - return 0, nil, copyItimerValOut(t, oldVal, &olditv) } @@ -137,21 +102,19 @@ func Setitimer(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys func Alarm(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { duration := time.Duration(args[0].Uint()) * time.Second - timer := t.ThreadGroup().Timer().RealTimer - s, err := ktime.SettingFromSpec(duration, 0, timer.Clock()) + olditv, err := t.Setitimer(linux.ITIMER_REAL, linux.ItimerVal{ + Value: linux.DurationToTimeval(duration), + }) if err != nil { return 0, nil, err } - - value, _ := ktime.SpecFromSetting(timer.Swap(s)) - sec := int64(value) / nsecPerSec - nsec := int64(value) % nsecPerSec - // We can't return 0 if we have an alarm pending ... - if (sec == 0 && nsec > 0) || nsec >= nsecPerSec/2 { - sec++ + olddur := olditv.Value.ToDuration() + secs := olddur.Round(time.Second).Nanoseconds() / nsecPerSec + if secs == 0 && olddur != 0 { + // We can't return 0 if an alarm was previously scheduled. + secs = 1 } - - return uintptr(sec), nil, nil + return uintptr(secs), nil, nil } // TimerCreate implements linux syscall timer_create(2). -- cgit v1.2.3 From 8fce67af24945f82378b4c2731cca1788936d074 Mon Sep 17 00:00:00 2001 From: Ian Gudger Date: Fri, 19 Oct 2018 16:34:09 -0700 Subject: Use correct company name in copyright header PiperOrigin-RevId: 217951017 Change-Id: Ie08bf6987f98467d07457bcf35b5f1ff6e43c035 --- kokoro/run_build.sh | 2 +- kokoro/run_tests.sh | 2 +- pkg/abi/abi.go | 2 +- pkg/abi/abi_linux.go | 2 +- pkg/abi/flag.go | 2 +- pkg/abi/linux/aio.go | 2 +- pkg/abi/linux/ashmem.go | 2 +- pkg/abi/linux/binder.go | 2 +- pkg/abi/linux/bpf.go | 2 +- pkg/abi/linux/capability.go | 2 +- pkg/abi/linux/dev.go | 2 +- pkg/abi/linux/elf.go | 2 +- pkg/abi/linux/errors.go | 2 +- pkg/abi/linux/eventfd.go | 2 +- pkg/abi/linux/exec.go | 2 +- pkg/abi/linux/fcntl.go | 2 +- pkg/abi/linux/file.go | 2 +- pkg/abi/linux/fs.go | 2 +- pkg/abi/linux/futex.go | 2 +- pkg/abi/linux/inotify.go | 2 +- pkg/abi/linux/ioctl.go | 2 +- pkg/abi/linux/ip.go | 2 +- pkg/abi/linux/ipc.go | 2 +- pkg/abi/linux/limits.go | 2 +- pkg/abi/linux/linux.go | 2 +- pkg/abi/linux/mm.go | 2 +- pkg/abi/linux/netdevice.go | 2 +- pkg/abi/linux/netlink.go | 2 +- pkg/abi/linux/netlink_route.go | 2 +- pkg/abi/linux/poll.go | 2 +- pkg/abi/linux/prctl.go | 2 +- pkg/abi/linux/ptrace.go | 2 +- pkg/abi/linux/rusage.go | 2 +- pkg/abi/linux/sched.go | 2 +- pkg/abi/linux/seccomp.go | 2 +- pkg/abi/linux/sem.go | 2 +- pkg/abi/linux/shm.go | 2 +- pkg/abi/linux/signal.go | 2 +- pkg/abi/linux/socket.go | 2 +- pkg/abi/linux/time.go | 2 +- pkg/abi/linux/timer.go | 2 +- pkg/abi/linux/tty.go | 2 +- pkg/abi/linux/uio.go | 2 +- pkg/abi/linux/utsname.go | 2 +- pkg/amutex/amutex.go | 2 +- pkg/amutex/amutex_test.go | 2 +- pkg/atomicbitops/atomic_bitops.go | 2 +- pkg/atomicbitops/atomic_bitops_amd64.s | 2 +- pkg/atomicbitops/atomic_bitops_common.go | 2 +- pkg/atomicbitops/atomic_bitops_test.go | 2 +- pkg/binary/binary.go | 2 +- pkg/binary/binary_test.go | 2 +- pkg/bits/bits.go | 2 +- pkg/bits/bits_template.go | 2 +- pkg/bits/uint64_arch_amd64.go | 2 +- pkg/bits/uint64_arch_amd64_asm.s | 2 +- pkg/bits/uint64_arch_generic.go | 2 +- pkg/bits/uint64_test.go | 2 +- pkg/bpf/bpf.go | 2 +- pkg/bpf/decoder.go | 2 +- pkg/bpf/decoder_test.go | 2 +- pkg/bpf/input_bytes.go | 2 +- pkg/bpf/interpreter.go | 2 +- pkg/bpf/interpreter_test.go | 2 +- pkg/bpf/program_builder.go | 2 +- pkg/bpf/program_builder_test.go | 2 +- pkg/compressio/compressio.go | 2 +- pkg/compressio/compressio_test.go | 2 +- pkg/control/client/client.go | 2 +- pkg/control/server/server.go | 2 +- pkg/cpuid/cpu_amd64.s | 2 +- pkg/cpuid/cpuid.go | 2 +- pkg/cpuid/cpuid_parse_test.go | 2 +- pkg/cpuid/cpuid_test.go | 2 +- pkg/dhcp/client.go | 2 +- pkg/dhcp/dhcp.go | 2 +- pkg/dhcp/dhcp_string.go | 2 +- pkg/dhcp/dhcp_test.go | 2 +- pkg/dhcp/server.go | 2 +- pkg/eventchannel/event.go | 2 +- pkg/eventchannel/event.proto | 2 +- pkg/fd/fd.go | 2 +- pkg/fd/fd_test.go | 2 +- pkg/gate/gate.go | 2 +- pkg/gate/gate_test.go | 2 +- pkg/ilist/list.go | 2 +- pkg/ilist/list_test.go | 2 +- pkg/linewriter/linewriter.go | 2 +- pkg/linewriter/linewriter_test.go | 2 +- pkg/log/glog.go | 2 +- pkg/log/glog_unsafe.go | 2 +- pkg/log/json.go | 2 +- pkg/log/json_test.go | 2 +- pkg/log/log.go | 2 +- pkg/log/log_test.go | 2 +- pkg/metric/metric.go | 2 +- pkg/metric/metric.proto | 2 +- pkg/metric/metric_test.go | 2 +- pkg/p9/buffer.go | 2 +- pkg/p9/client.go | 2 +- pkg/p9/client_file.go | 2 +- pkg/p9/client_test.go | 2 +- pkg/p9/file.go | 2 +- pkg/p9/handlers.go | 2 +- pkg/p9/local_server/local_server.go | 2 +- pkg/p9/messages.go | 2 +- pkg/p9/messages_test.go | 2 +- pkg/p9/p9.go | 2 +- pkg/p9/p9_test.go | 2 +- pkg/p9/p9test/client_test.go | 2 +- pkg/p9/p9test/mocks.go | 2 +- pkg/p9/pool.go | 2 +- pkg/p9/pool_test.go | 2 +- pkg/p9/server.go | 2 +- pkg/p9/transport.go | 2 +- pkg/p9/transport_test.go | 2 +- pkg/p9/version.go | 2 +- pkg/p9/version_test.go | 2 +- pkg/rand/rand.go | 2 +- pkg/rand/rand_linux.go | 2 +- pkg/refs/refcounter.go | 2 +- pkg/refs/refcounter_state.go | 2 +- pkg/refs/refcounter_test.go | 2 +- pkg/seccomp/seccomp.go | 2 +- pkg/seccomp/seccomp_rules.go | 2 +- pkg/seccomp/seccomp_test.go | 2 +- pkg/seccomp/seccomp_test_victim.go | 2 +- pkg/seccomp/seccomp_unsafe.go | 2 +- pkg/secio/full_reader.go | 2 +- pkg/secio/secio.go | 2 +- pkg/secio/secio_test.go | 2 +- pkg/segment/range.go | 2 +- pkg/segment/set.go | 2 +- pkg/segment/set_state.go | 2 +- pkg/segment/test/segment_test.go | 2 +- pkg/segment/test/set_functions.go | 2 +- pkg/sentry/arch/aligned.go | 2 +- pkg/sentry/arch/arch.go | 2 +- pkg/sentry/arch/arch_amd64.go | 2 +- pkg/sentry/arch/arch_amd64.s | 2 +- pkg/sentry/arch/arch_state_x86.go | 2 +- pkg/sentry/arch/arch_x86.go | 2 +- pkg/sentry/arch/auxv.go | 2 +- pkg/sentry/arch/registers.proto | 2 +- pkg/sentry/arch/signal_act.go | 2 +- pkg/sentry/arch/signal_amd64.go | 2 +- pkg/sentry/arch/signal_info.go | 2 +- pkg/sentry/arch/signal_stack.go | 2 +- pkg/sentry/arch/stack.go | 2 +- pkg/sentry/arch/syscalls_amd64.go | 2 +- pkg/sentry/context/context.go | 2 +- pkg/sentry/context/contexttest/contexttest.go | 2 +- pkg/sentry/control/control.go | 2 +- pkg/sentry/control/proc.go | 2 +- pkg/sentry/control/proc_test.go | 2 +- pkg/sentry/control/state.go | 2 +- pkg/sentry/device/device.go | 2 +- pkg/sentry/device/device_test.go | 2 +- pkg/sentry/fs/anon/anon.go | 2 +- pkg/sentry/fs/anon/device.go | 2 +- pkg/sentry/fs/ashmem/area.go | 2 +- pkg/sentry/fs/ashmem/device.go | 2 +- pkg/sentry/fs/ashmem/pin_board.go | 2 +- pkg/sentry/fs/ashmem/pin_board_test.go | 2 +- pkg/sentry/fs/attr.go | 2 +- pkg/sentry/fs/binder/binder.go | 2 +- pkg/sentry/fs/context.go | 2 +- pkg/sentry/fs/copy_up.go | 2 +- pkg/sentry/fs/copy_up_test.go | 2 +- pkg/sentry/fs/dentry.go | 2 +- pkg/sentry/fs/dev/dev.go | 2 +- pkg/sentry/fs/dev/device.go | 2 +- pkg/sentry/fs/dev/fs.go | 2 +- pkg/sentry/fs/dev/full.go | 2 +- pkg/sentry/fs/dev/null.go | 2 +- pkg/sentry/fs/dev/random.go | 2 +- pkg/sentry/fs/dirent.go | 2 +- pkg/sentry/fs/dirent_cache.go | 2 +- pkg/sentry/fs/dirent_cache_test.go | 2 +- pkg/sentry/fs/dirent_refs_test.go | 2 +- pkg/sentry/fs/dirent_state.go | 2 +- pkg/sentry/fs/fdpipe/pipe.go | 2 +- pkg/sentry/fs/fdpipe/pipe_opener.go | 2 +- pkg/sentry/fs/fdpipe/pipe_opener_test.go | 2 +- pkg/sentry/fs/fdpipe/pipe_state.go | 2 +- pkg/sentry/fs/fdpipe/pipe_test.go | 2 +- pkg/sentry/fs/file.go | 2 +- pkg/sentry/fs/file_operations.go | 2 +- pkg/sentry/fs/file_overlay.go | 2 +- pkg/sentry/fs/file_overlay_test.go | 2 +- pkg/sentry/fs/file_state.go | 2 +- pkg/sentry/fs/file_test.go | 2 +- pkg/sentry/fs/filesystems.go | 2 +- pkg/sentry/fs/filetest/filetest.go | 2 +- pkg/sentry/fs/flags.go | 2 +- pkg/sentry/fs/fs.go | 2 +- pkg/sentry/fs/fsutil/dirty_set.go | 2 +- pkg/sentry/fs/fsutil/dirty_set_test.go | 2 +- pkg/sentry/fs/fsutil/file.go | 2 +- pkg/sentry/fs/fsutil/file_range_set.go | 2 +- pkg/sentry/fs/fsutil/frame_ref_set.go | 2 +- pkg/sentry/fs/fsutil/fsutil.go | 2 +- pkg/sentry/fs/fsutil/handle.go | 2 +- pkg/sentry/fs/fsutil/handle_test.go | 2 +- pkg/sentry/fs/fsutil/host_file_mapper.go | 2 +- pkg/sentry/fs/fsutil/host_file_mapper_state.go | 2 +- pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go | 2 +- pkg/sentry/fs/fsutil/inode.go | 2 +- pkg/sentry/fs/fsutil/inode_cached.go | 2 +- pkg/sentry/fs/fsutil/inode_cached_test.go | 2 +- pkg/sentry/fs/gofer/attr.go | 2 +- pkg/sentry/fs/gofer/cache_policy.go | 2 +- pkg/sentry/fs/gofer/context_file.go | 2 +- pkg/sentry/fs/gofer/device.go | 2 +- pkg/sentry/fs/gofer/file.go | 2 +- pkg/sentry/fs/gofer/file_state.go | 2 +- pkg/sentry/fs/gofer/fs.go | 2 +- pkg/sentry/fs/gofer/gofer_test.go | 2 +- pkg/sentry/fs/gofer/handles.go | 2 +- pkg/sentry/fs/gofer/inode.go | 2 +- pkg/sentry/fs/gofer/inode_state.go | 2 +- pkg/sentry/fs/gofer/path.go | 2 +- pkg/sentry/fs/gofer/session.go | 2 +- pkg/sentry/fs/gofer/session_state.go | 2 +- pkg/sentry/fs/gofer/socket.go | 2 +- pkg/sentry/fs/gofer/util.go | 2 +- pkg/sentry/fs/host/control.go | 2 +- pkg/sentry/fs/host/descriptor.go | 2 +- pkg/sentry/fs/host/descriptor_state.go | 2 +- pkg/sentry/fs/host/descriptor_test.go | 2 +- pkg/sentry/fs/host/device.go | 2 +- pkg/sentry/fs/host/file.go | 2 +- pkg/sentry/fs/host/fs.go | 2 +- pkg/sentry/fs/host/fs_test.go | 2 +- pkg/sentry/fs/host/inode.go | 2 +- pkg/sentry/fs/host/inode_state.go | 2 +- pkg/sentry/fs/host/inode_test.go | 2 +- pkg/sentry/fs/host/ioctl_unsafe.go | 2 +- pkg/sentry/fs/host/socket.go | 2 +- pkg/sentry/fs/host/socket_iovec.go | 2 +- pkg/sentry/fs/host/socket_state.go | 2 +- pkg/sentry/fs/host/socket_test.go | 2 +- pkg/sentry/fs/host/socket_unsafe.go | 2 +- pkg/sentry/fs/host/tty.go | 2 +- pkg/sentry/fs/host/util.go | 2 +- pkg/sentry/fs/host/util_unsafe.go | 2 +- pkg/sentry/fs/host/wait_test.go | 2 +- pkg/sentry/fs/inode.go | 2 +- pkg/sentry/fs/inode_inotify.go | 2 +- pkg/sentry/fs/inode_operations.go | 2 +- pkg/sentry/fs/inode_overlay.go | 2 +- pkg/sentry/fs/inode_overlay_test.go | 2 +- pkg/sentry/fs/inotify.go | 2 +- pkg/sentry/fs/inotify_event.go | 2 +- pkg/sentry/fs/inotify_watch.go | 2 +- pkg/sentry/fs/lock/lock.go | 2 +- pkg/sentry/fs/lock/lock_range_test.go | 2 +- pkg/sentry/fs/lock/lock_set_functions.go | 2 +- pkg/sentry/fs/lock/lock_test.go | 2 +- pkg/sentry/fs/mock.go | 2 +- pkg/sentry/fs/mount.go | 2 +- pkg/sentry/fs/mount_overlay.go | 2 +- pkg/sentry/fs/mount_state.go | 2 +- pkg/sentry/fs/mount_test.go | 2 +- pkg/sentry/fs/mounts.go | 2 +- pkg/sentry/fs/mounts_test.go | 2 +- pkg/sentry/fs/offset.go | 2 +- pkg/sentry/fs/overlay.go | 2 +- pkg/sentry/fs/path.go | 2 +- pkg/sentry/fs/path_test.go | 2 +- pkg/sentry/fs/proc/cpuinfo.go | 2 +- pkg/sentry/fs/proc/device/device.go | 2 +- pkg/sentry/fs/proc/exec_args.go | 2 +- pkg/sentry/fs/proc/fds.go | 2 +- pkg/sentry/fs/proc/file.go | 2 +- pkg/sentry/fs/proc/filesystems.go | 2 +- pkg/sentry/fs/proc/fs.go | 2 +- pkg/sentry/fs/proc/loadavg.go | 2 +- pkg/sentry/fs/proc/meminfo.go | 2 +- pkg/sentry/fs/proc/mounts.go | 2 +- pkg/sentry/fs/proc/net.go | 2 +- pkg/sentry/fs/proc/net_test.go | 2 +- pkg/sentry/fs/proc/proc.go | 2 +- pkg/sentry/fs/proc/rpcinet_proc.go | 2 +- pkg/sentry/fs/proc/seqfile/seqfile.go | 2 +- pkg/sentry/fs/proc/seqfile/seqfile_test.go | 2 +- pkg/sentry/fs/proc/stat.go | 2 +- pkg/sentry/fs/proc/sys.go | 2 +- pkg/sentry/fs/proc/sys_net.go | 2 +- pkg/sentry/fs/proc/sys_net_test.go | 2 +- pkg/sentry/fs/proc/task.go | 2 +- pkg/sentry/fs/proc/uid_gid_map.go | 2 +- pkg/sentry/fs/proc/uptime.go | 2 +- pkg/sentry/fs/proc/version.go | 2 +- pkg/sentry/fs/ramfs/dir.go | 2 +- pkg/sentry/fs/ramfs/file.go | 2 +- pkg/sentry/fs/ramfs/ramfs.go | 2 +- pkg/sentry/fs/ramfs/socket.go | 2 +- pkg/sentry/fs/ramfs/symlink.go | 2 +- pkg/sentry/fs/ramfs/test/test.go | 2 +- pkg/sentry/fs/ramfs/tree.go | 2 +- pkg/sentry/fs/ramfs/tree_test.go | 2 +- pkg/sentry/fs/restore.go | 2 +- pkg/sentry/fs/save.go | 2 +- pkg/sentry/fs/seek.go | 2 +- pkg/sentry/fs/sync.go | 2 +- pkg/sentry/fs/sys/device.go | 2 +- pkg/sentry/fs/sys/devices.go | 2 +- pkg/sentry/fs/sys/fs.go | 2 +- pkg/sentry/fs/sys/sys.go | 2 +- pkg/sentry/fs/timerfd/timerfd.go | 2 +- pkg/sentry/fs/tmpfs/device.go | 2 +- pkg/sentry/fs/tmpfs/file_regular.go | 2 +- pkg/sentry/fs/tmpfs/file_test.go | 2 +- pkg/sentry/fs/tmpfs/fs.go | 2 +- pkg/sentry/fs/tmpfs/inode_file.go | 2 +- pkg/sentry/fs/tmpfs/tmpfs.go | 2 +- pkg/sentry/fs/tty/dir.go | 2 +- pkg/sentry/fs/tty/fs.go | 2 +- pkg/sentry/fs/tty/inode.go | 2 +- pkg/sentry/fs/tty/line_discipline.go | 2 +- pkg/sentry/fs/tty/master.go | 2 +- pkg/sentry/fs/tty/queue.go | 2 +- pkg/sentry/fs/tty/slave.go | 2 +- pkg/sentry/fs/tty/terminal.go | 2 +- pkg/sentry/fs/tty/tty_test.go | 2 +- pkg/sentry/hostcpu/getcpu_amd64.s | 2 +- pkg/sentry/hostcpu/hostcpu.go | 2 +- pkg/sentry/hostcpu/hostcpu_test.go | 2 +- pkg/sentry/inet/context.go | 2 +- pkg/sentry/inet/inet.go | 2 +- pkg/sentry/inet/test_stack.go | 2 +- pkg/sentry/kernel/abstract_socket_namespace.go | 2 +- pkg/sentry/kernel/auth/auth.go | 2 +- pkg/sentry/kernel/auth/capability_set.go | 2 +- pkg/sentry/kernel/auth/context.go | 2 +- pkg/sentry/kernel/auth/credentials.go | 2 +- pkg/sentry/kernel/auth/id.go | 2 +- pkg/sentry/kernel/auth/id_map.go | 2 +- pkg/sentry/kernel/auth/id_map_functions.go | 2 +- pkg/sentry/kernel/auth/user_namespace.go | 2 +- pkg/sentry/kernel/context.go | 2 +- pkg/sentry/kernel/epoll/epoll.go | 2 +- pkg/sentry/kernel/epoll/epoll_state.go | 2 +- pkg/sentry/kernel/epoll/epoll_test.go | 2 +- pkg/sentry/kernel/eventfd/eventfd.go | 2 +- pkg/sentry/kernel/eventfd/eventfd_test.go | 2 +- pkg/sentry/kernel/fasync/fasync.go | 2 +- pkg/sentry/kernel/fd_map.go | 2 +- pkg/sentry/kernel/fd_map_test.go | 2 +- pkg/sentry/kernel/fs_context.go | 2 +- pkg/sentry/kernel/futex/futex.go | 2 +- pkg/sentry/kernel/futex/futex_test.go | 2 +- pkg/sentry/kernel/ipc_namespace.go | 2 +- pkg/sentry/kernel/kdefs/kdefs.go | 2 +- pkg/sentry/kernel/kernel.go | 2 +- pkg/sentry/kernel/kernel_state.go | 2 +- pkg/sentry/kernel/memevent/memory_events.go | 2 +- pkg/sentry/kernel/memevent/memory_events.proto | 2 +- pkg/sentry/kernel/pending_signals.go | 2 +- pkg/sentry/kernel/pending_signals_state.go | 2 +- pkg/sentry/kernel/pipe/buffers.go | 2 +- pkg/sentry/kernel/pipe/device.go | 2 +- pkg/sentry/kernel/pipe/node.go | 2 +- pkg/sentry/kernel/pipe/node_test.go | 2 +- pkg/sentry/kernel/pipe/pipe.go | 2 +- pkg/sentry/kernel/pipe/pipe_test.go | 2 +- pkg/sentry/kernel/pipe/reader.go | 2 +- pkg/sentry/kernel/pipe/reader_writer.go | 2 +- pkg/sentry/kernel/pipe/writer.go | 2 +- pkg/sentry/kernel/posixtimer.go | 2 +- pkg/sentry/kernel/ptrace.go | 2 +- pkg/sentry/kernel/rseq.go | 2 +- pkg/sentry/kernel/sched/cpuset.go | 2 +- pkg/sentry/kernel/sched/cpuset_test.go | 2 +- pkg/sentry/kernel/sched/sched.go | 2 +- pkg/sentry/kernel/seccomp.go | 2 +- pkg/sentry/kernel/semaphore/semaphore.go | 2 +- pkg/sentry/kernel/semaphore/semaphore_test.go | 2 +- pkg/sentry/kernel/sessions.go | 2 +- pkg/sentry/kernel/shm/device.go | 2 +- pkg/sentry/kernel/shm/shm.go | 2 +- pkg/sentry/kernel/signal.go | 2 +- pkg/sentry/kernel/signal_handlers.go | 2 +- pkg/sentry/kernel/syscalls.go | 2 +- pkg/sentry/kernel/syscalls_state.go | 2 +- pkg/sentry/kernel/syslog.go | 2 +- pkg/sentry/kernel/table_test.go | 2 +- pkg/sentry/kernel/task.go | 2 +- pkg/sentry/kernel/task_acct.go | 2 +- pkg/sentry/kernel/task_block.go | 2 +- pkg/sentry/kernel/task_clone.go | 2 +- pkg/sentry/kernel/task_context.go | 2 +- pkg/sentry/kernel/task_exec.go | 2 +- pkg/sentry/kernel/task_exit.go | 2 +- pkg/sentry/kernel/task_futex.go | 2 +- pkg/sentry/kernel/task_identity.go | 2 +- pkg/sentry/kernel/task_log.go | 2 +- pkg/sentry/kernel/task_net.go | 2 +- pkg/sentry/kernel/task_run.go | 2 +- pkg/sentry/kernel/task_sched.go | 2 +- pkg/sentry/kernel/task_signals.go | 2 +- pkg/sentry/kernel/task_start.go | 2 +- pkg/sentry/kernel/task_stop.go | 2 +- pkg/sentry/kernel/task_syscall.go | 2 +- pkg/sentry/kernel/task_test.go | 2 +- pkg/sentry/kernel/task_usermem.go | 2 +- pkg/sentry/kernel/thread_group.go | 2 +- pkg/sentry/kernel/threads.go | 2 +- pkg/sentry/kernel/time/context.go | 2 +- pkg/sentry/kernel/time/time.go | 2 +- pkg/sentry/kernel/timekeeper.go | 2 +- pkg/sentry/kernel/timekeeper_state.go | 2 +- pkg/sentry/kernel/timekeeper_test.go | 2 +- pkg/sentry/kernel/uts_namespace.go | 2 +- pkg/sentry/kernel/vdso.go | 2 +- pkg/sentry/kernel/version.go | 2 +- pkg/sentry/limits/context.go | 2 +- pkg/sentry/limits/limits.go | 2 +- pkg/sentry/limits/limits_test.go | 2 +- pkg/sentry/limits/linux.go | 2 +- pkg/sentry/loader/elf.go | 2 +- pkg/sentry/loader/interpreter.go | 2 +- pkg/sentry/loader/loader.go | 2 +- pkg/sentry/loader/vdso.go | 2 +- pkg/sentry/loader/vdso_state.go | 2 +- pkg/sentry/memmap/mapping_set.go | 2 +- pkg/sentry/memmap/mapping_set_test.go | 2 +- pkg/sentry/memmap/memmap.go | 2 +- pkg/sentry/memutil/memutil.go | 2 +- pkg/sentry/memutil/memutil_unsafe.go | 2 +- pkg/sentry/mm/address_space.go | 2 +- pkg/sentry/mm/aio_context.go | 2 +- pkg/sentry/mm/aio_context_state.go | 2 +- pkg/sentry/mm/debug.go | 2 +- pkg/sentry/mm/io.go | 2 +- pkg/sentry/mm/lifecycle.go | 2 +- pkg/sentry/mm/metadata.go | 2 +- pkg/sentry/mm/mm.go | 2 +- pkg/sentry/mm/mm_test.go | 2 +- pkg/sentry/mm/pma.go | 2 +- pkg/sentry/mm/proc_pid_maps.go | 2 +- pkg/sentry/mm/save_restore.go | 2 +- pkg/sentry/mm/shm.go | 2 +- pkg/sentry/mm/special_mappable.go | 2 +- pkg/sentry/mm/syscalls.go | 2 +- pkg/sentry/mm/vma.go | 2 +- pkg/sentry/platform/context.go | 2 +- pkg/sentry/platform/filemem/filemem.go | 2 +- pkg/sentry/platform/filemem/filemem_state.go | 2 +- pkg/sentry/platform/filemem/filemem_test.go | 2 +- pkg/sentry/platform/filemem/filemem_unsafe.go | 2 +- pkg/sentry/platform/interrupt/interrupt.go | 2 +- pkg/sentry/platform/interrupt/interrupt_test.go | 2 +- pkg/sentry/platform/kvm/address_space.go | 2 +- pkg/sentry/platform/kvm/allocator.go | 2 +- pkg/sentry/platform/kvm/bluepill.go | 2 +- pkg/sentry/platform/kvm/bluepill_amd64.go | 2 +- pkg/sentry/platform/kvm/bluepill_amd64.s | 2 +- pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go | 2 +- pkg/sentry/platform/kvm/bluepill_fault.go | 2 +- pkg/sentry/platform/kvm/bluepill_unsafe.go | 2 +- pkg/sentry/platform/kvm/context.go | 2 +- pkg/sentry/platform/kvm/host_map.go | 2 +- pkg/sentry/platform/kvm/kvm.go | 2 +- pkg/sentry/platform/kvm/kvm_amd64.go | 2 +- pkg/sentry/platform/kvm/kvm_amd64_unsafe.go | 2 +- pkg/sentry/platform/kvm/kvm_const.go | 2 +- pkg/sentry/platform/kvm/kvm_test.go | 2 +- pkg/sentry/platform/kvm/machine.go | 2 +- pkg/sentry/platform/kvm/machine_amd64.go | 2 +- pkg/sentry/platform/kvm/machine_amd64_unsafe.go | 2 +- pkg/sentry/platform/kvm/machine_unsafe.go | 2 +- pkg/sentry/platform/kvm/physical_map.go | 2 +- pkg/sentry/platform/kvm/testutil/testutil.go | 2 +- pkg/sentry/platform/kvm/testutil/testutil_amd64.go | 2 +- pkg/sentry/platform/kvm/testutil/testutil_amd64.s | 2 +- pkg/sentry/platform/kvm/virtual_map.go | 2 +- pkg/sentry/platform/kvm/virtual_map_test.go | 2 +- pkg/sentry/platform/mmap_min_addr.go | 2 +- pkg/sentry/platform/platform.go | 2 +- pkg/sentry/platform/procid/procid.go | 2 +- pkg/sentry/platform/procid/procid_amd64.s | 2 +- pkg/sentry/platform/procid/procid_net_test.go | 2 +- pkg/sentry/platform/procid/procid_test.go | 2 +- pkg/sentry/platform/ptrace/ptrace.go | 2 +- pkg/sentry/platform/ptrace/ptrace_unsafe.go | 2 +- pkg/sentry/platform/ptrace/stub_amd64.s | 2 +- pkg/sentry/platform/ptrace/stub_unsafe.go | 2 +- pkg/sentry/platform/ptrace/subprocess.go | 2 +- pkg/sentry/platform/ptrace/subprocess_amd64.go | 2 +- pkg/sentry/platform/ptrace/subprocess_linux.go | 2 +- pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go | 2 +- pkg/sentry/platform/ptrace/subprocess_unsafe.go | 2 +- pkg/sentry/platform/ring0/defs.go | 2 +- pkg/sentry/platform/ring0/defs_amd64.go | 2 +- pkg/sentry/platform/ring0/entry_amd64.go | 2 +- pkg/sentry/platform/ring0/entry_amd64.s | 2 +- pkg/sentry/platform/ring0/gen_offsets/main.go | 2 +- pkg/sentry/platform/ring0/kernel.go | 2 +- pkg/sentry/platform/ring0/kernel_amd64.go | 2 +- pkg/sentry/platform/ring0/kernel_unsafe.go | 2 +- pkg/sentry/platform/ring0/lib_amd64.go | 2 +- pkg/sentry/platform/ring0/lib_amd64.s | 2 +- pkg/sentry/platform/ring0/offsets_amd64.go | 2 +- pkg/sentry/platform/ring0/pagetables/allocator.go | 2 +- pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables_test.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables_x86.go | 2 +- pkg/sentry/platform/ring0/pagetables/pcids_x86.go | 2 +- pkg/sentry/platform/ring0/pagetables/walker_amd64.go | 2 +- pkg/sentry/platform/ring0/ring0.go | 2 +- pkg/sentry/platform/ring0/x86.go | 2 +- pkg/sentry/platform/safecopy/atomic_amd64.s | 2 +- pkg/sentry/platform/safecopy/memclr_amd64.s | 2 +- pkg/sentry/platform/safecopy/memcpy_amd64.s | 2 +- pkg/sentry/platform/safecopy/safecopy.go | 2 +- pkg/sentry/platform/safecopy/safecopy_test.go | 2 +- pkg/sentry/platform/safecopy/safecopy_unsafe.go | 2 +- pkg/sentry/platform/safecopy/sighandler_amd64.s | 2 +- pkg/sentry/safemem/block_unsafe.go | 2 +- pkg/sentry/safemem/io.go | 2 +- pkg/sentry/safemem/io_test.go | 2 +- pkg/sentry/safemem/safemem.go | 2 +- pkg/sentry/safemem/seq_test.go | 2 +- pkg/sentry/safemem/seq_unsafe.go | 2 +- pkg/sentry/sighandling/sighandling.go | 2 +- pkg/sentry/sighandling/sighandling_unsafe.go | 2 +- pkg/sentry/socket/control/control.go | 2 +- pkg/sentry/socket/epsocket/device.go | 2 +- pkg/sentry/socket/epsocket/epsocket.go | 2 +- pkg/sentry/socket/epsocket/provider.go | 2 +- pkg/sentry/socket/epsocket/save_restore.go | 2 +- pkg/sentry/socket/epsocket/stack.go | 2 +- pkg/sentry/socket/hostinet/device.go | 2 +- pkg/sentry/socket/hostinet/hostinet.go | 2 +- pkg/sentry/socket/hostinet/save_restore.go | 2 +- pkg/sentry/socket/hostinet/socket.go | 2 +- pkg/sentry/socket/hostinet/socket_unsafe.go | 2 +- pkg/sentry/socket/hostinet/stack.go | 2 +- pkg/sentry/socket/netlink/message.go | 2 +- pkg/sentry/socket/netlink/port/port.go | 2 +- pkg/sentry/socket/netlink/port/port_test.go | 2 +- pkg/sentry/socket/netlink/provider.go | 2 +- pkg/sentry/socket/netlink/route/protocol.go | 2 +- pkg/sentry/socket/netlink/socket.go | 2 +- pkg/sentry/socket/rpcinet/conn/conn.go | 2 +- pkg/sentry/socket/rpcinet/device.go | 2 +- pkg/sentry/socket/rpcinet/notifier/notifier.go | 2 +- pkg/sentry/socket/rpcinet/rpcinet.go | 2 +- pkg/sentry/socket/rpcinet/socket.go | 2 +- pkg/sentry/socket/rpcinet/stack.go | 2 +- pkg/sentry/socket/rpcinet/stack_unsafe.go | 2 +- pkg/sentry/socket/socket.go | 2 +- pkg/sentry/socket/unix/device.go | 2 +- pkg/sentry/socket/unix/io.go | 2 +- pkg/sentry/socket/unix/transport/connectioned.go | 2 +- pkg/sentry/socket/unix/transport/connectioned_state.go | 2 +- pkg/sentry/socket/unix/transport/connectionless.go | 2 +- pkg/sentry/socket/unix/transport/queue.go | 2 +- pkg/sentry/socket/unix/transport/unix.go | 2 +- pkg/sentry/socket/unix/unix.go | 2 +- pkg/sentry/state/state.go | 2 +- pkg/sentry/state/state_metadata.go | 2 +- pkg/sentry/state/state_unsafe.go | 2 +- pkg/sentry/strace/clone.go | 2 +- pkg/sentry/strace/futex.go | 2 +- pkg/sentry/strace/linux64.go | 2 +- pkg/sentry/strace/open.go | 2 +- pkg/sentry/strace/ptrace.go | 2 +- pkg/sentry/strace/socket.go | 2 +- pkg/sentry/strace/strace.go | 2 +- pkg/sentry/strace/strace.proto | 2 +- pkg/sentry/strace/syscalls.go | 2 +- pkg/sentry/syscalls/epoll.go | 2 +- pkg/sentry/syscalls/linux/error.go | 2 +- pkg/sentry/syscalls/linux/flags.go | 2 +- pkg/sentry/syscalls/linux/linux64.go | 2 +- pkg/sentry/syscalls/linux/sigset.go | 2 +- pkg/sentry/syscalls/linux/sys_aio.go | 2 +- pkg/sentry/syscalls/linux/sys_capability.go | 2 +- pkg/sentry/syscalls/linux/sys_epoll.go | 2 +- pkg/sentry/syscalls/linux/sys_eventfd.go | 2 +- pkg/sentry/syscalls/linux/sys_file.go | 2 +- pkg/sentry/syscalls/linux/sys_futex.go | 2 +- pkg/sentry/syscalls/linux/sys_getdents.go | 2 +- pkg/sentry/syscalls/linux/sys_identity.go | 2 +- pkg/sentry/syscalls/linux/sys_inotify.go | 2 +- pkg/sentry/syscalls/linux/sys_lseek.go | 2 +- pkg/sentry/syscalls/linux/sys_mmap.go | 2 +- pkg/sentry/syscalls/linux/sys_mount.go | 2 +- pkg/sentry/syscalls/linux/sys_pipe.go | 2 +- pkg/sentry/syscalls/linux/sys_poll.go | 2 +- pkg/sentry/syscalls/linux/sys_prctl.go | 2 +- pkg/sentry/syscalls/linux/sys_random.go | 2 +- pkg/sentry/syscalls/linux/sys_read.go | 2 +- pkg/sentry/syscalls/linux/sys_rlimit.go | 2 +- pkg/sentry/syscalls/linux/sys_rusage.go | 2 +- pkg/sentry/syscalls/linux/sys_sched.go | 2 +- pkg/sentry/syscalls/linux/sys_seccomp.go | 2 +- pkg/sentry/syscalls/linux/sys_sem.go | 2 +- pkg/sentry/syscalls/linux/sys_shm.go | 2 +- pkg/sentry/syscalls/linux/sys_signal.go | 2 +- pkg/sentry/syscalls/linux/sys_socket.go | 2 +- pkg/sentry/syscalls/linux/sys_stat.go | 2 +- pkg/sentry/syscalls/linux/sys_sync.go | 2 +- pkg/sentry/syscalls/linux/sys_sysinfo.go | 2 +- pkg/sentry/syscalls/linux/sys_syslog.go | 2 +- pkg/sentry/syscalls/linux/sys_thread.go | 2 +- pkg/sentry/syscalls/linux/sys_time.go | 2 +- pkg/sentry/syscalls/linux/sys_timer.go | 2 +- pkg/sentry/syscalls/linux/sys_timerfd.go | 2 +- pkg/sentry/syscalls/linux/sys_tls.go | 2 +- pkg/sentry/syscalls/linux/sys_utsname.go | 2 +- pkg/sentry/syscalls/linux/sys_write.go | 2 +- pkg/sentry/syscalls/linux/timespec.go | 2 +- pkg/sentry/syscalls/polling.go | 2 +- pkg/sentry/syscalls/syscalls.go | 2 +- pkg/sentry/syscalls/unimplemented_syscall.proto | 2 +- pkg/sentry/time/calibrated_clock.go | 2 +- pkg/sentry/time/calibrated_clock_test.go | 2 +- pkg/sentry/time/clock_id.go | 2 +- pkg/sentry/time/clocks.go | 2 +- pkg/sentry/time/muldiv_amd64.s | 2 +- pkg/sentry/time/parameters.go | 2 +- pkg/sentry/time/parameters_test.go | 2 +- pkg/sentry/time/sampler.go | 2 +- pkg/sentry/time/sampler_test.go | 2 +- pkg/sentry/time/sampler_unsafe.go | 2 +- pkg/sentry/time/tsc_amd64.s | 2 +- pkg/sentry/uniqueid/context.go | 2 +- pkg/sentry/usage/cpu.go | 2 +- pkg/sentry/usage/io.go | 2 +- pkg/sentry/usage/memory.go | 2 +- pkg/sentry/usage/memory_unsafe.go | 2 +- pkg/sentry/usage/usage.go | 2 +- pkg/sentry/usermem/access_type.go | 2 +- pkg/sentry/usermem/addr.go | 2 +- pkg/sentry/usermem/addr_range_seq_test.go | 2 +- pkg/sentry/usermem/addr_range_seq_unsafe.go | 2 +- pkg/sentry/usermem/bytes_io.go | 2 +- pkg/sentry/usermem/bytes_io_unsafe.go | 2 +- pkg/sentry/usermem/usermem.go | 2 +- pkg/sentry/usermem/usermem_test.go | 2 +- pkg/sentry/usermem/usermem_x86.go | 2 +- pkg/sentry/watchdog/watchdog.go | 2 +- pkg/sleep/commit_amd64.s | 2 +- pkg/sleep/commit_asm.go | 2 +- pkg/sleep/commit_noasm.go | 2 +- pkg/sleep/empty.s | 2 +- pkg/sleep/sleep_test.go | 2 +- pkg/sleep/sleep_unsafe.go | 2 +- pkg/state/decode.go | 2 +- pkg/state/encode.go | 2 +- pkg/state/encode_unsafe.go | 2 +- pkg/state/map.go | 2 +- pkg/state/object.proto | 2 +- pkg/state/printer.go | 2 +- pkg/state/state.go | 2 +- pkg/state/state_test.go | 2 +- pkg/state/statefile/statefile.go | 2 +- pkg/state/statefile/statefile_test.go | 2 +- pkg/state/stats.go | 2 +- pkg/sync/atomicptr_unsafe.go | 2 +- pkg/sync/atomicptrtest/atomicptr_test.go | 2 +- pkg/sync/memmove_unsafe.go | 2 +- pkg/sync/norace_unsafe.go | 2 +- pkg/sync/race_unsafe.go | 2 +- pkg/sync/seqatomic_unsafe.go | 2 +- pkg/sync/seqatomictest/seqatomic_test.go | 2 +- pkg/sync/seqcount.go | 2 +- pkg/sync/seqcount_test.go | 2 +- pkg/sync/sync.go | 2 +- pkg/syserr/host_linux.go | 2 +- pkg/syserr/netstack.go | 2 +- pkg/syserr/syserr.go | 2 +- pkg/syserror/syserror.go | 2 +- pkg/syserror/syserror_test.go | 2 +- pkg/tcpip/adapters/gonet/gonet.go | 2 +- pkg/tcpip/adapters/gonet/gonet_test.go | 2 +- pkg/tcpip/buffer/prependable.go | 2 +- pkg/tcpip/buffer/view.go | 2 +- pkg/tcpip/buffer/view_test.go | 2 +- pkg/tcpip/checker/checker.go | 2 +- pkg/tcpip/header/arp.go | 2 +- pkg/tcpip/header/checksum.go | 2 +- pkg/tcpip/header/eth.go | 2 +- pkg/tcpip/header/gue.go | 2 +- pkg/tcpip/header/icmpv4.go | 2 +- pkg/tcpip/header/icmpv6.go | 2 +- pkg/tcpip/header/interfaces.go | 2 +- pkg/tcpip/header/ipv4.go | 2 +- pkg/tcpip/header/ipv6.go | 2 +- pkg/tcpip/header/ipv6_fragment.go | 2 +- pkg/tcpip/header/ipversion_test.go | 2 +- pkg/tcpip/header/tcp.go | 2 +- pkg/tcpip/header/tcp_test.go | 2 +- pkg/tcpip/header/udp.go | 2 +- pkg/tcpip/link/channel/channel.go | 2 +- pkg/tcpip/link/fdbased/endpoint.go | 2 +- pkg/tcpip/link/fdbased/endpoint_test.go | 2 +- pkg/tcpip/link/loopback/loopback.go | 2 +- pkg/tcpip/link/rawfile/blockingpoll_amd64.s | 2 +- pkg/tcpip/link/rawfile/blockingpoll_unsafe.go | 2 +- pkg/tcpip/link/rawfile/blockingpoll_unsafe_amd64.go | 2 +- pkg/tcpip/link/rawfile/errors.go | 2 +- pkg/tcpip/link/rawfile/rawfile_unsafe.go | 2 +- pkg/tcpip/link/sharedmem/pipe/pipe.go | 2 +- pkg/tcpip/link/sharedmem/pipe/pipe_test.go | 2 +- pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go | 2 +- pkg/tcpip/link/sharedmem/pipe/rx.go | 2 +- pkg/tcpip/link/sharedmem/pipe/tx.go | 2 +- pkg/tcpip/link/sharedmem/queue/queue_test.go | 2 +- pkg/tcpip/link/sharedmem/queue/rx.go | 2 +- pkg/tcpip/link/sharedmem/queue/tx.go | 2 +- pkg/tcpip/link/sharedmem/rx.go | 2 +- pkg/tcpip/link/sharedmem/sharedmem.go | 2 +- pkg/tcpip/link/sharedmem/sharedmem_test.go | 2 +- pkg/tcpip/link/sharedmem/sharedmem_unsafe.go | 2 +- pkg/tcpip/link/sharedmem/tx.go | 2 +- pkg/tcpip/link/sniffer/pcap.go | 2 +- pkg/tcpip/link/sniffer/sniffer.go | 2 +- pkg/tcpip/link/tun/tun_unsafe.go | 2 +- pkg/tcpip/link/waitable/waitable.go | 2 +- pkg/tcpip/link/waitable/waitable_test.go | 2 +- pkg/tcpip/network/arp/arp.go | 2 +- pkg/tcpip/network/arp/arp_test.go | 2 +- pkg/tcpip/network/fragmentation/frag_heap.go | 2 +- pkg/tcpip/network/fragmentation/frag_heap_test.go | 2 +- pkg/tcpip/network/fragmentation/fragmentation.go | 2 +- pkg/tcpip/network/fragmentation/fragmentation_test.go | 2 +- pkg/tcpip/network/fragmentation/reassembler.go | 2 +- pkg/tcpip/network/fragmentation/reassembler_test.go | 2 +- pkg/tcpip/network/hash/hash.go | 2 +- pkg/tcpip/network/ip_test.go | 2 +- pkg/tcpip/network/ipv4/icmp.go | 2 +- pkg/tcpip/network/ipv4/ipv4.go | 2 +- pkg/tcpip/network/ipv4/ipv4_test.go | 2 +- pkg/tcpip/network/ipv6/icmp.go | 2 +- pkg/tcpip/network/ipv6/icmp_test.go | 2 +- pkg/tcpip/network/ipv6/ipv6.go | 2 +- pkg/tcpip/ports/ports.go | 2 +- pkg/tcpip/ports/ports_test.go | 2 +- pkg/tcpip/sample/tun_tcp_connect/main.go | 2 +- pkg/tcpip/sample/tun_tcp_echo/main.go | 2 +- pkg/tcpip/seqnum/seqnum.go | 2 +- pkg/tcpip/stack/linkaddrcache.go | 2 +- pkg/tcpip/stack/linkaddrcache_test.go | 2 +- pkg/tcpip/stack/nic.go | 2 +- pkg/tcpip/stack/registration.go | 2 +- pkg/tcpip/stack/route.go | 2 +- pkg/tcpip/stack/stack.go | 2 +- pkg/tcpip/stack/stack_global_state.go | 2 +- pkg/tcpip/stack/stack_test.go | 2 +- pkg/tcpip/stack/transport_demuxer.go | 2 +- pkg/tcpip/stack/transport_test.go | 2 +- pkg/tcpip/tcpip.go | 2 +- pkg/tcpip/tcpip_test.go | 2 +- pkg/tcpip/time.s | 2 +- pkg/tcpip/time_unsafe.go | 2 +- pkg/tcpip/transport/ping/endpoint.go | 2 +- pkg/tcpip/transport/ping/endpoint_state.go | 2 +- pkg/tcpip/transport/ping/protocol.go | 2 +- pkg/tcpip/transport/tcp/accept.go | 2 +- pkg/tcpip/transport/tcp/connect.go | 2 +- pkg/tcpip/transport/tcp/cubic.go | 2 +- pkg/tcpip/transport/tcp/dual_stack_test.go | 2 +- pkg/tcpip/transport/tcp/endpoint.go | 2 +- pkg/tcpip/transport/tcp/endpoint_state.go | 2 +- pkg/tcpip/transport/tcp/forwarder.go | 2 +- pkg/tcpip/transport/tcp/protocol.go | 2 +- pkg/tcpip/transport/tcp/rcv.go | 2 +- pkg/tcpip/transport/tcp/reno.go | 2 +- pkg/tcpip/transport/tcp/sack.go | 2 +- pkg/tcpip/transport/tcp/segment.go | 2 +- pkg/tcpip/transport/tcp/segment_heap.go | 2 +- pkg/tcpip/transport/tcp/segment_queue.go | 2 +- pkg/tcpip/transport/tcp/segment_state.go | 2 +- pkg/tcpip/transport/tcp/snd.go | 2 +- pkg/tcpip/transport/tcp/snd_state.go | 2 +- pkg/tcpip/transport/tcp/tcp_sack_test.go | 2 +- pkg/tcpip/transport/tcp/tcp_test.go | 2 +- pkg/tcpip/transport/tcp/tcp_timestamp_test.go | 2 +- pkg/tcpip/transport/tcp/testing/context/context.go | 2 +- pkg/tcpip/transport/tcp/timer.go | 2 +- pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go | 2 +- pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go | 2 +- pkg/tcpip/transport/udp/endpoint.go | 2 +- pkg/tcpip/transport/udp/endpoint_state.go | 2 +- pkg/tcpip/transport/udp/protocol.go | 2 +- pkg/tcpip/transport/udp/udp_test.go | 2 +- pkg/tmutex/tmutex.go | 2 +- pkg/tmutex/tmutex_test.go | 2 +- pkg/unet/unet.go | 2 +- pkg/unet/unet_test.go | 2 +- pkg/unet/unet_unsafe.go | 2 +- pkg/urpc/urpc.go | 2 +- pkg/urpc/urpc_test.go | 2 +- pkg/waiter/fdnotifier/fdnotifier.go | 2 +- pkg/waiter/fdnotifier/poll_unsafe.go | 2 +- pkg/waiter/waiter.go | 2 +- pkg/waiter/waiter_test.go | 2 +- runsc/boot/compat.go | 2 +- runsc/boot/config.go | 2 +- runsc/boot/controller.go | 2 +- runsc/boot/debug.go | 2 +- runsc/boot/events.go | 2 +- runsc/boot/fds.go | 2 +- runsc/boot/filter/config.go | 2 +- runsc/boot/filter/extra_filters.go | 2 +- runsc/boot/filter/extra_filters_msan.go | 2 +- runsc/boot/filter/extra_filters_race.go | 2 +- runsc/boot/filter/filter.go | 2 +- runsc/boot/fs.go | 2 +- runsc/boot/limits.go | 2 +- runsc/boot/loader.go | 2 +- runsc/boot/loader_test.go | 2 +- runsc/boot/network.go | 2 +- runsc/boot/strace.go | 2 +- runsc/cgroup/cgroup.go | 2 +- runsc/cgroup/cgroup_test.go | 2 +- runsc/cmd/boot.go | 2 +- runsc/cmd/capability.go | 2 +- runsc/cmd/capability_test.go | 2 +- runsc/cmd/checkpoint.go | 2 +- runsc/cmd/cmd.go | 2 +- runsc/cmd/create.go | 2 +- runsc/cmd/debug.go | 2 +- runsc/cmd/delete.go | 2 +- runsc/cmd/delete_test.go | 2 +- runsc/cmd/events.go | 2 +- runsc/cmd/exec.go | 2 +- runsc/cmd/exec_test.go | 2 +- runsc/cmd/gofer.go | 2 +- runsc/cmd/kill.go | 2 +- runsc/cmd/list.go | 2 +- runsc/cmd/path.go | 2 +- runsc/cmd/pause.go | 2 +- runsc/cmd/ps.go | 2 +- runsc/cmd/restore.go | 2 +- runsc/cmd/resume.go | 2 +- runsc/cmd/run.go | 2 +- runsc/cmd/spec.go | 2 +- runsc/cmd/start.go | 2 +- runsc/cmd/state.go | 2 +- runsc/cmd/wait.go | 2 +- runsc/console/console.go | 2 +- runsc/container/console_test.go | 2 +- runsc/container/container.go | 2 +- runsc/container/container_test.go | 2 +- runsc/container/fs.go | 2 +- runsc/container/fs_test.go | 2 +- runsc/container/hook.go | 2 +- runsc/container/multi_container_test.go | 2 +- runsc/container/status.go | 2 +- runsc/container/test_app.go | 2 +- runsc/fsgofer/filter/config.go | 2 +- runsc/fsgofer/filter/extra_filters.go | 2 +- runsc/fsgofer/filter/extra_filters_msan.go | 2 +- runsc/fsgofer/filter/extra_filters_race.go | 2 +- runsc/fsgofer/filter/filter.go | 2 +- runsc/fsgofer/fsgofer.go | 2 +- runsc/fsgofer/fsgofer_test.go | 2 +- runsc/fsgofer/fsgofer_unsafe.go | 2 +- runsc/main.go | 2 +- runsc/sandbox/chroot.go | 2 +- runsc/sandbox/network.go | 2 +- runsc/sandbox/sandbox.go | 2 +- runsc/specutils/namespace.go | 2 +- runsc/specutils/specutils.go | 2 +- runsc/specutils/specutils_test.go | 2 +- runsc/test/image/image.go | 2 +- runsc/test/image/image_test.go | 2 +- runsc/test/image/mysql.sql | 2 +- runsc/test/image/ruby.rb | 2 +- runsc/test/image/ruby.sh | 2 +- runsc/test/install.sh | 2 +- runsc/test/integration/exec_test.go | 2 +- runsc/test/integration/integration.go | 2 +- runsc/test/integration/integration_test.go | 2 +- runsc/test/root/cgroup_test.go | 2 +- runsc/test/root/chroot_test.go | 2 +- runsc/test/root/root.go | 2 +- runsc/test/testutil/docker.go | 2 +- runsc/test/testutil/testutil.go | 2 +- runsc/test/testutil/testutil_race.go | 2 +- runsc/tools/dockercfg/dockercfg.go | 2 +- tools/go_generics/generics.go | 2 +- tools/go_generics/generics_tests/all_stmts/input.go | 2 +- tools/go_generics/generics_tests/all_stmts/output/output.go | 2 +- tools/go_generics/generics_tests/all_types/input.go | 2 +- tools/go_generics/generics_tests/all_types/lib/lib.go | 2 +- tools/go_generics/generics_tests/all_types/output/output.go | 2 +- tools/go_generics/generics_tests/consts/input.go | 2 +- tools/go_generics/generics_tests/consts/output/output.go | 2 +- tools/go_generics/generics_tests/imports/input.go | 2 +- tools/go_generics/generics_tests/imports/output/output.go | 2 +- tools/go_generics/generics_tests/remove_typedef/input.go | 2 +- tools/go_generics/generics_tests/remove_typedef/output/output.go | 2 +- tools/go_generics/generics_tests/simple/input.go | 2 +- tools/go_generics/generics_tests/simple/output/output.go | 2 +- tools/go_generics/globals/globals_visitor.go | 2 +- tools/go_generics/globals/scope.go | 2 +- tools/go_generics/go_generics_unittest.sh | 2 +- tools/go_generics/imports.go | 2 +- tools/go_generics/merge.go | 2 +- tools/go_generics/remove.go | 2 +- tools/go_generics/rules_tests/template.go | 2 +- tools/go_generics/rules_tests/template_test.go | 2 +- tools/go_stateify/main.go | 2 +- tools/workspace_status.sh | 2 +- vdso/barrier.h | 2 +- vdso/check_vdso.py | 2 +- vdso/compiler.h | 2 +- vdso/cycle_clock.h | 2 +- vdso/seqlock.h | 2 +- vdso/syscalls.h | 2 +- vdso/vdso.cc | 2 +- vdso/vdso_time.cc | 2 +- vdso/vdso_time.h | 2 +- 923 files changed, 923 insertions(+), 923 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/kokoro/run_build.sh b/kokoro/run_build.sh index f2b719f52..89e24b037 100755 --- a/kokoro/run_build.sh +++ b/kokoro/run_build.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/kokoro/run_tests.sh b/kokoro/run_tests.sh index 3f8841cee..0a0d73d29 100755 --- a/kokoro/run_tests.sh +++ b/kokoro/run_tests.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/pkg/abi/abi.go b/pkg/abi/abi.go index a53c2747b..7770f0405 100644 --- a/pkg/abi/abi.go +++ b/pkg/abi/abi.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/abi_linux.go b/pkg/abi/abi_linux.go index dd5d67b51..9d9f361a4 100644 --- a/pkg/abi/abi_linux.go +++ b/pkg/abi/abi_linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/flag.go b/pkg/abi/flag.go index 0391ccf37..0698e410f 100644 --- a/pkg/abi/flag.go +++ b/pkg/abi/flag.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/aio.go b/pkg/abi/linux/aio.go index 9c39ca2ef..1b7ca714a 100644 --- a/pkg/abi/linux/aio.go +++ b/pkg/abi/linux/aio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ashmem.go b/pkg/abi/linux/ashmem.go index 7fbfd2e68..ced1e44d4 100644 --- a/pkg/abi/linux/ashmem.go +++ b/pkg/abi/linux/ashmem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/binder.go b/pkg/abi/linux/binder.go index b228898f9..522dc6f53 100644 --- a/pkg/abi/linux/binder.go +++ b/pkg/abi/linux/binder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/bpf.go b/pkg/abi/linux/bpf.go index 80e5b1af1..d9cd09948 100644 --- a/pkg/abi/linux/bpf.go +++ b/pkg/abi/linux/bpf.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/capability.go b/pkg/abi/linux/capability.go index b470ce0a5..7d96f013e 100644 --- a/pkg/abi/linux/capability.go +++ b/pkg/abi/linux/capability.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/dev.go b/pkg/abi/linux/dev.go index ea5b16b7b..5b1199aac 100644 --- a/pkg/abi/linux/dev.go +++ b/pkg/abi/linux/dev.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/elf.go b/pkg/abi/linux/elf.go index 76c13b677..928067c04 100644 --- a/pkg/abi/linux/elf.go +++ b/pkg/abi/linux/elf.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/errors.go b/pkg/abi/linux/errors.go index b5ddb2b2f..01e4095b8 100644 --- a/pkg/abi/linux/errors.go +++ b/pkg/abi/linux/errors.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/eventfd.go b/pkg/abi/linux/eventfd.go index bc0fb44d2..5614f5cf1 100644 --- a/pkg/abi/linux/eventfd.go +++ b/pkg/abi/linux/eventfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/exec.go b/pkg/abi/linux/exec.go index 4d81eca54..a07c29243 100644 --- a/pkg/abi/linux/exec.go +++ b/pkg/abi/linux/exec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/fcntl.go b/pkg/abi/linux/fcntl.go index 2a5ad6ed7..c8558933a 100644 --- a/pkg/abi/linux/fcntl.go +++ b/pkg/abi/linux/fcntl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/file.go b/pkg/abi/linux/file.go index 9bf229a57..72e5c6f83 100644 --- a/pkg/abi/linux/file.go +++ b/pkg/abi/linux/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/fs.go b/pkg/abi/linux/fs.go index 32a0812b4..7817bfb52 100644 --- a/pkg/abi/linux/fs.go +++ b/pkg/abi/linux/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/futex.go b/pkg/abi/linux/futex.go index f63f5200c..5dff01fba 100644 --- a/pkg/abi/linux/futex.go +++ b/pkg/abi/linux/futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/inotify.go b/pkg/abi/linux/inotify.go index 072a2d146..79c5d3593 100644 --- a/pkg/abi/linux/inotify.go +++ b/pkg/abi/linux/inotify.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ioctl.go b/pkg/abi/linux/ioctl.go index afd9ee82b..9afc3d1ef 100644 --- a/pkg/abi/linux/ioctl.go +++ b/pkg/abi/linux/ioctl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ip.go b/pkg/abi/linux/ip.go index 6b68999ab..fcec16965 100644 --- a/pkg/abi/linux/ip.go +++ b/pkg/abi/linux/ip.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ipc.go b/pkg/abi/linux/ipc.go index 81e9904dd..10681768b 100644 --- a/pkg/abi/linux/ipc.go +++ b/pkg/abi/linux/ipc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/limits.go b/pkg/abi/linux/limits.go index e1f0932ec..b2e51b9bd 100644 --- a/pkg/abi/linux/limits.go +++ b/pkg/abi/linux/limits.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/linux.go b/pkg/abi/linux/linux.go index de2af80dc..d365f693d 100644 --- a/pkg/abi/linux/linux.go +++ b/pkg/abi/linux/linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/mm.go b/pkg/abi/linux/mm.go index b48e1d18a..3fcdf8235 100644 --- a/pkg/abi/linux/mm.go +++ b/pkg/abi/linux/mm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/netdevice.go b/pkg/abi/linux/netdevice.go index 88654a1b3..e3b6b1e40 100644 --- a/pkg/abi/linux/netdevice.go +++ b/pkg/abi/linux/netdevice.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/netlink.go b/pkg/abi/linux/netlink.go index e823ffa7e..10ceb5bf2 100644 --- a/pkg/abi/linux/netlink.go +++ b/pkg/abi/linux/netlink.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/netlink_route.go b/pkg/abi/linux/netlink_route.go index a5d778748..4200b6506 100644 --- a/pkg/abi/linux/netlink_route.go +++ b/pkg/abi/linux/netlink_route.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/poll.go b/pkg/abi/linux/poll.go index f373cfca1..9f0b15d1c 100644 --- a/pkg/abi/linux/poll.go +++ b/pkg/abi/linux/poll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/prctl.go b/pkg/abi/linux/prctl.go index 074ec03f0..e152c4c27 100644 --- a/pkg/abi/linux/prctl.go +++ b/pkg/abi/linux/prctl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ptrace.go b/pkg/abi/linux/ptrace.go index ba48d4d6d..7db4f5464 100644 --- a/pkg/abi/linux/ptrace.go +++ b/pkg/abi/linux/ptrace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/rusage.go b/pkg/abi/linux/rusage.go index a4a89abda..7fea4b589 100644 --- a/pkg/abi/linux/rusage.go +++ b/pkg/abi/linux/rusage.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/sched.go b/pkg/abi/linux/sched.go index 05fda1604..ef96a3801 100644 --- a/pkg/abi/linux/sched.go +++ b/pkg/abi/linux/sched.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/seccomp.go b/pkg/abi/linux/seccomp.go index a8de9d3d0..9963ceeba 100644 --- a/pkg/abi/linux/seccomp.go +++ b/pkg/abi/linux/seccomp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/sem.go b/pkg/abi/linux/sem.go index 3495f5cd0..d1a0bdb32 100644 --- a/pkg/abi/linux/sem.go +++ b/pkg/abi/linux/sem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/shm.go b/pkg/abi/linux/shm.go index f50b3c2e2..82a80e609 100644 --- a/pkg/abi/linux/shm.go +++ b/pkg/abi/linux/shm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/signal.go b/pkg/abi/linux/signal.go index b2c7230c4..bf9bce6ed 100644 --- a/pkg/abi/linux/signal.go +++ b/pkg/abi/linux/signal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/socket.go b/pkg/abi/linux/socket.go index 19b5fa212..af0761a3b 100644 --- a/pkg/abi/linux/socket.go +++ b/pkg/abi/linux/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/time.go b/pkg/abi/linux/time.go index 4569f4208..bbd21e726 100644 --- a/pkg/abi/linux/time.go +++ b/pkg/abi/linux/time.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/timer.go b/pkg/abi/linux/timer.go index 6c4675c35..a6f420bdb 100644 --- a/pkg/abi/linux/timer.go +++ b/pkg/abi/linux/timer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/tty.go b/pkg/abi/linux/tty.go index f63dc52aa..e6f7c5b2a 100644 --- a/pkg/abi/linux/tty.go +++ b/pkg/abi/linux/tty.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/uio.go b/pkg/abi/linux/uio.go index 93c972774..7e00d9959 100644 --- a/pkg/abi/linux/uio.go +++ b/pkg/abi/linux/uio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/utsname.go b/pkg/abi/linux/utsname.go index 7d33d20de..f80ed7d4a 100644 --- a/pkg/abi/linux/utsname.go +++ b/pkg/abi/linux/utsname.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/amutex/amutex.go b/pkg/amutex/amutex.go index 1cb73359a..26b674435 100644 --- a/pkg/amutex/amutex.go +++ b/pkg/amutex/amutex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/amutex/amutex_test.go b/pkg/amutex/amutex_test.go index 876e47b19..104e0dab1 100644 --- a/pkg/amutex/amutex_test.go +++ b/pkg/amutex/amutex_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/atomicbitops/atomic_bitops.go b/pkg/atomicbitops/atomic_bitops.go index 6635ea0d2..9a57f9599 100644 --- a/pkg/atomicbitops/atomic_bitops.go +++ b/pkg/atomicbitops/atomic_bitops.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/atomicbitops/atomic_bitops_amd64.s b/pkg/atomicbitops/atomic_bitops_amd64.s index 542452bec..b37e3aad3 100644 --- a/pkg/atomicbitops/atomic_bitops_amd64.s +++ b/pkg/atomicbitops/atomic_bitops_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/atomicbitops/atomic_bitops_common.go b/pkg/atomicbitops/atomic_bitops_common.go index 542ff4e83..b03242baa 100644 --- a/pkg/atomicbitops/atomic_bitops_common.go +++ b/pkg/atomicbitops/atomic_bitops_common.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/atomicbitops/atomic_bitops_test.go b/pkg/atomicbitops/atomic_bitops_test.go index ec0c07ee2..ee6207cb3 100644 --- a/pkg/atomicbitops/atomic_bitops_test.go +++ b/pkg/atomicbitops/atomic_bitops_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/binary/binary.go b/pkg/binary/binary.go index 3b18a86ee..02f7e9fb8 100644 --- a/pkg/binary/binary.go +++ b/pkg/binary/binary.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/binary/binary_test.go b/pkg/binary/binary_test.go index 921a0369a..d8d481f32 100644 --- a/pkg/binary/binary_test.go +++ b/pkg/binary/binary_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/bits.go b/pkg/bits/bits.go index 50ca4bff7..eb3c80f49 100644 --- a/pkg/bits/bits.go +++ b/pkg/bits/bits.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/bits_template.go b/pkg/bits/bits_template.go index 0a01f29c2..8c578cca2 100644 --- a/pkg/bits/bits_template.go +++ b/pkg/bits/bits_template.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/uint64_arch_amd64.go b/pkg/bits/uint64_arch_amd64.go index 068597f68..1fef89394 100644 --- a/pkg/bits/uint64_arch_amd64.go +++ b/pkg/bits/uint64_arch_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/uint64_arch_amd64_asm.s b/pkg/bits/uint64_arch_amd64_asm.s index 33885641a..8c7322f0f 100644 --- a/pkg/bits/uint64_arch_amd64_asm.s +++ b/pkg/bits/uint64_arch_amd64_asm.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/uint64_arch_generic.go b/pkg/bits/uint64_arch_generic.go index 862033a4b..cfb47400b 100644 --- a/pkg/bits/uint64_arch_generic.go +++ b/pkg/bits/uint64_arch_generic.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/uint64_test.go b/pkg/bits/uint64_test.go index 906017e1a..d6dbaf602 100644 --- a/pkg/bits/uint64_test.go +++ b/pkg/bits/uint64_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/bpf.go b/pkg/bpf/bpf.go index 757744090..98d44d911 100644 --- a/pkg/bpf/bpf.go +++ b/pkg/bpf/bpf.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/decoder.go b/pkg/bpf/decoder.go index ef41e9edc..ae6b8839a 100644 --- a/pkg/bpf/decoder.go +++ b/pkg/bpf/decoder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/decoder_test.go b/pkg/bpf/decoder_test.go index 18709b944..f093e1e41 100644 --- a/pkg/bpf/decoder_test.go +++ b/pkg/bpf/decoder_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/input_bytes.go b/pkg/bpf/input_bytes.go index 74af038eb..745c0749b 100644 --- a/pkg/bpf/input_bytes.go +++ b/pkg/bpf/input_bytes.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/interpreter.go b/pkg/bpf/interpreter.go index 111ada9d1..86c7add4d 100644 --- a/pkg/bpf/interpreter.go +++ b/pkg/bpf/interpreter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/interpreter_test.go b/pkg/bpf/interpreter_test.go index 9e5e33228..c46a43991 100644 --- a/pkg/bpf/interpreter_test.go +++ b/pkg/bpf/interpreter_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/program_builder.go b/pkg/bpf/program_builder.go index bad56d7ac..b4ce228e1 100644 --- a/pkg/bpf/program_builder.go +++ b/pkg/bpf/program_builder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/program_builder_test.go b/pkg/bpf/program_builder_test.go index 7e4f06584..0e0b79d88 100644 --- a/pkg/bpf/program_builder_test.go +++ b/pkg/bpf/program_builder_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/compressio/compressio.go b/pkg/compressio/compressio.go index 667f17c5c..205536812 100644 --- a/pkg/compressio/compressio.go +++ b/pkg/compressio/compressio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/compressio/compressio_test.go b/pkg/compressio/compressio_test.go index 7cb5f8dc4..1bbabee79 100644 --- a/pkg/compressio/compressio_test.go +++ b/pkg/compressio/compressio_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/control/client/client.go b/pkg/control/client/client.go index f7c2e8776..0d0c9f148 100644 --- a/pkg/control/client/client.go +++ b/pkg/control/client/client.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/control/server/server.go b/pkg/control/server/server.go index d00061ce3..c46b5d70b 100644 --- a/pkg/control/server/server.go +++ b/pkg/control/server/server.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/cpuid/cpu_amd64.s b/pkg/cpuid/cpu_amd64.s index 48a13c6fd..905c1d12e 100644 --- a/pkg/cpuid/cpu_amd64.s +++ b/pkg/cpuid/cpu_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/cpuid/cpuid.go b/pkg/cpuid/cpuid.go index e91e34dc7..5b083a5fb 100644 --- a/pkg/cpuid/cpuid.go +++ b/pkg/cpuid/cpuid.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/cpuid/cpuid_parse_test.go b/pkg/cpuid/cpuid_parse_test.go index c4f52818c..81b06f48c 100644 --- a/pkg/cpuid/cpuid_parse_test.go +++ b/pkg/cpuid/cpuid_parse_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/cpuid/cpuid_test.go b/pkg/cpuid/cpuid_test.go index 02f732f85..0decd8f08 100644 --- a/pkg/cpuid/cpuid_test.go +++ b/pkg/cpuid/cpuid_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/client.go b/pkg/dhcp/client.go index 92c634a14..3330c4998 100644 --- a/pkg/dhcp/client.go +++ b/pkg/dhcp/client.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/dhcp.go b/pkg/dhcp/dhcp.go index ceaba34c3..ad11e178a 100644 --- a/pkg/dhcp/dhcp.go +++ b/pkg/dhcp/dhcp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/dhcp_string.go b/pkg/dhcp/dhcp_string.go index 7cabed29e..8533895bd 100644 --- a/pkg/dhcp/dhcp_string.go +++ b/pkg/dhcp/dhcp_string.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/dhcp_test.go b/pkg/dhcp/dhcp_test.go index d60e3752b..a21dce6bc 100644 --- a/pkg/dhcp/dhcp_test.go +++ b/pkg/dhcp/dhcp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/server.go b/pkg/dhcp/server.go index 26700bdbc..3e06ab4c7 100644 --- a/pkg/dhcp/server.go +++ b/pkg/dhcp/server.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/eventchannel/event.go b/pkg/eventchannel/event.go index bfd28256e..41a7b5ed3 100644 --- a/pkg/eventchannel/event.go +++ b/pkg/eventchannel/event.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/eventchannel/event.proto b/pkg/eventchannel/event.proto index 455f03658..c1679c7e7 100644 --- a/pkg/eventchannel/event.proto +++ b/pkg/eventchannel/event.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/fd/fd.go b/pkg/fd/fd.go index 32d24c41b..f6656ffa1 100644 --- a/pkg/fd/fd.go +++ b/pkg/fd/fd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/fd/fd_test.go b/pkg/fd/fd_test.go index 94b3eb7cc..42bb3ef6c 100644 --- a/pkg/fd/fd_test.go +++ b/pkg/fd/fd_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/gate/gate.go b/pkg/gate/gate.go index 93808c9dd..48122bf5a 100644 --- a/pkg/gate/gate.go +++ b/pkg/gate/gate.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/gate/gate_test.go b/pkg/gate/gate_test.go index 06587339b..95620fa8e 100644 --- a/pkg/gate/gate_test.go +++ b/pkg/gate/gate_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/ilist/list.go b/pkg/ilist/list.go index 4ae02eee9..51c9b6df3 100644 --- a/pkg/ilist/list.go +++ b/pkg/ilist/list.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/ilist/list_test.go b/pkg/ilist/list_test.go index 2c56280f6..4bda570b6 100644 --- a/pkg/ilist/list_test.go +++ b/pkg/ilist/list_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/linewriter/linewriter.go b/pkg/linewriter/linewriter.go index 98f974410..5fbd4e779 100644 --- a/pkg/linewriter/linewriter.go +++ b/pkg/linewriter/linewriter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/linewriter/linewriter_test.go b/pkg/linewriter/linewriter_test.go index ce97cca05..9140ee6af 100644 --- a/pkg/linewriter/linewriter_test.go +++ b/pkg/linewriter/linewriter_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/glog.go b/pkg/log/glog.go index 58b4052e6..fbb58501b 100644 --- a/pkg/log/glog.go +++ b/pkg/log/glog.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/glog_unsafe.go b/pkg/log/glog_unsafe.go index c320190b8..bb06aa7d3 100644 --- a/pkg/log/glog_unsafe.go +++ b/pkg/log/glog_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/json.go b/pkg/log/json.go index 3887f1cd5..96bd13d87 100644 --- a/pkg/log/json.go +++ b/pkg/log/json.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/json_test.go b/pkg/log/json_test.go index 3b167dab0..b8c7a795e 100644 --- a/pkg/log/json_test.go +++ b/pkg/log/json_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/log.go b/pkg/log/log.go index c496e86e4..b8d456aae 100644 --- a/pkg/log/log.go +++ b/pkg/log/log.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/log_test.go b/pkg/log/log_test.go index d93e989dc..a59d457dd 100644 --- a/pkg/log/log_test.go +++ b/pkg/log/log_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/metric/metric.go b/pkg/metric/metric.go index 763cd6bc2..02af75974 100644 --- a/pkg/metric/metric.go +++ b/pkg/metric/metric.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/metric/metric.proto b/pkg/metric/metric.proto index 6108cb7c0..917fda1ac 100644 --- a/pkg/metric/metric.proto +++ b/pkg/metric/metric.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/metric/metric_test.go b/pkg/metric/metric_test.go index 7d156e4a5..40034a589 100644 --- a/pkg/metric/metric_test.go +++ b/pkg/metric/metric_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/buffer.go b/pkg/p9/buffer.go index fc65d2c5f..9575ddf12 100644 --- a/pkg/p9/buffer.go +++ b/pkg/p9/buffer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/client.go b/pkg/p9/client.go index 5fa231bc5..3ebfab82a 100644 --- a/pkg/p9/client.go +++ b/pkg/p9/client.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/client_file.go b/pkg/p9/client_file.go index a46efd27f..066639fda 100644 --- a/pkg/p9/client_file.go +++ b/pkg/p9/client_file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/client_test.go b/pkg/p9/client_test.go index 06302a76a..f7145452d 100644 --- a/pkg/p9/client_test.go +++ b/pkg/p9/client_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/file.go b/pkg/p9/file.go index 9723fa24d..d2e89e373 100644 --- a/pkg/p9/file.go +++ b/pkg/p9/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/handlers.go b/pkg/p9/handlers.go index ea41f97c7..959dff31d 100644 --- a/pkg/p9/handlers.go +++ b/pkg/p9/handlers.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/local_server/local_server.go b/pkg/p9/local_server/local_server.go index cef3701a7..1e6aaa762 100644 --- a/pkg/p9/local_server/local_server.go +++ b/pkg/p9/local_server/local_server.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/messages.go b/pkg/p9/messages.go index b3d76801b..972c37344 100644 --- a/pkg/p9/messages.go +++ b/pkg/p9/messages.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/messages_test.go b/pkg/p9/messages_test.go index f353755f1..dfb41bb76 100644 --- a/pkg/p9/messages_test.go +++ b/pkg/p9/messages_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/p9.go b/pkg/p9/p9.go index c6899c3ce..3b0993ecd 100644 --- a/pkg/p9/p9.go +++ b/pkg/p9/p9.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/p9_test.go b/pkg/p9/p9_test.go index a50ac80a4..02498346c 100644 --- a/pkg/p9/p9_test.go +++ b/pkg/p9/p9_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/p9test/client_test.go b/pkg/p9/p9test/client_test.go index 34ddccd8b..db562b9ba 100644 --- a/pkg/p9/p9test/client_test.go +++ b/pkg/p9/p9test/client_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/p9test/mocks.go b/pkg/p9/p9test/mocks.go index 9d039ac63..9a8c14975 100644 --- a/pkg/p9/p9test/mocks.go +++ b/pkg/p9/p9test/mocks.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/pool.go b/pkg/p9/pool.go index 9a508b898..34ed898e8 100644 --- a/pkg/p9/pool.go +++ b/pkg/p9/pool.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/pool_test.go b/pkg/p9/pool_test.go index 96be2c8bd..71052d8c4 100644 --- a/pkg/p9/pool_test.go +++ b/pkg/p9/pool_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/server.go b/pkg/p9/server.go index 28a273ac6..5c7cb18c8 100644 --- a/pkg/p9/server.go +++ b/pkg/p9/server.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/transport.go b/pkg/p9/transport.go index b5df29961..97396806c 100644 --- a/pkg/p9/transport.go +++ b/pkg/p9/transport.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/transport_test.go b/pkg/p9/transport_test.go index d6d4b6365..3352a5205 100644 --- a/pkg/p9/transport_test.go +++ b/pkg/p9/transport_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/version.go b/pkg/p9/version.go index 8783eaa7e..ceb6fabbf 100644 --- a/pkg/p9/version.go +++ b/pkg/p9/version.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/version_test.go b/pkg/p9/version_test.go index 634ac3ca5..c053614c9 100644 --- a/pkg/p9/version_test.go +++ b/pkg/p9/version_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/rand/rand.go b/pkg/rand/rand.go index e81f0f5db..593a14380 100644 --- a/pkg/rand/rand.go +++ b/pkg/rand/rand.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/rand/rand_linux.go b/pkg/rand/rand_linux.go index a2be66b3b..7ebe8f3b0 100644 --- a/pkg/rand/rand_linux.go +++ b/pkg/rand/rand_linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/refs/refcounter.go b/pkg/refs/refcounter.go index 638a93bab..8f08c74c7 100644 --- a/pkg/refs/refcounter.go +++ b/pkg/refs/refcounter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/refs/refcounter_state.go b/pkg/refs/refcounter_state.go index 093eae785..136f06fbf 100644 --- a/pkg/refs/refcounter_state.go +++ b/pkg/refs/refcounter_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/refs/refcounter_test.go b/pkg/refs/refcounter_test.go index cc11bcd71..abaa87453 100644 --- a/pkg/refs/refcounter_test.go +++ b/pkg/refs/refcounter_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp.go b/pkg/seccomp/seccomp.go index a746dc9b3..1dfbf749e 100644 --- a/pkg/seccomp/seccomp.go +++ b/pkg/seccomp/seccomp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp_rules.go b/pkg/seccomp/seccomp_rules.go index 6b707f195..a9278c64b 100644 --- a/pkg/seccomp/seccomp_rules.go +++ b/pkg/seccomp/seccomp_rules.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp_test.go b/pkg/seccomp/seccomp_test.go index 0188ad4f3..226f30b7b 100644 --- a/pkg/seccomp/seccomp_test.go +++ b/pkg/seccomp/seccomp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp_test_victim.go b/pkg/seccomp/seccomp_test_victim.go index 4f2ae4dac..007038273 100644 --- a/pkg/seccomp/seccomp_test_victim.go +++ b/pkg/seccomp/seccomp_test_victim.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp_unsafe.go b/pkg/seccomp/seccomp_unsafe.go index ae18534bf..dd009221a 100644 --- a/pkg/seccomp/seccomp_unsafe.go +++ b/pkg/seccomp/seccomp_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/secio/full_reader.go b/pkg/secio/full_reader.go index b2dbb8615..90b1772a7 100644 --- a/pkg/secio/full_reader.go +++ b/pkg/secio/full_reader.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/secio/secio.go b/pkg/secio/secio.go index fc625efb8..e5f74a497 100644 --- a/pkg/secio/secio.go +++ b/pkg/secio/secio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/secio/secio_test.go b/pkg/secio/secio_test.go index 64b4cc17d..8304c4f74 100644 --- a/pkg/secio/secio_test.go +++ b/pkg/secio/secio_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/range.go b/pkg/segment/range.go index 34c067265..057bcd7ff 100644 --- a/pkg/segment/range.go +++ b/pkg/segment/range.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/set.go b/pkg/segment/set.go index cffec2a2c..a9a3b8875 100644 --- a/pkg/segment/set.go +++ b/pkg/segment/set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/set_state.go b/pkg/segment/set_state.go index a763d1915..b86e1b75f 100644 --- a/pkg/segment/set_state.go +++ b/pkg/segment/set_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/test/segment_test.go b/pkg/segment/test/segment_test.go index 7ea24b177..0825105db 100644 --- a/pkg/segment/test/segment_test.go +++ b/pkg/segment/test/segment_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/test/set_functions.go b/pkg/segment/test/set_functions.go index 37c196ea1..05ba5fbb9 100644 --- a/pkg/segment/test/set_functions.go +++ b/pkg/segment/test/set_functions.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/aligned.go b/pkg/sentry/arch/aligned.go index 193232e27..c88c034f6 100644 --- a/pkg/sentry/arch/aligned.go +++ b/pkg/sentry/arch/aligned.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go index 21cb84502..575b7ba66 100644 --- a/pkg/sentry/arch/arch.go +++ b/pkg/sentry/arch/arch.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch_amd64.go b/pkg/sentry/arch/arch_amd64.go index 5ba6c19ea..bb80a7bed 100644 --- a/pkg/sentry/arch/arch_amd64.go +++ b/pkg/sentry/arch/arch_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch_amd64.s b/pkg/sentry/arch/arch_amd64.s index 10d621b6d..fa9857df7 100644 --- a/pkg/sentry/arch/arch_amd64.s +++ b/pkg/sentry/arch/arch_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch_state_x86.go b/pkg/sentry/arch/arch_state_x86.go index e9c23a06b..604bd08a6 100644 --- a/pkg/sentry/arch/arch_state_x86.go +++ b/pkg/sentry/arch/arch_state_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch_x86.go b/pkg/sentry/arch/arch_x86.go index b35eec53c..59bf89d99 100644 --- a/pkg/sentry/arch/arch_x86.go +++ b/pkg/sentry/arch/arch_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/auxv.go b/pkg/sentry/arch/auxv.go index 81cfb4a01..5df65a691 100644 --- a/pkg/sentry/arch/auxv.go +++ b/pkg/sentry/arch/auxv.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/registers.proto b/pkg/sentry/arch/registers.proto index 437ff44ca..f4c2f7043 100644 --- a/pkg/sentry/arch/registers.proto +++ b/pkg/sentry/arch/registers.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/signal_act.go b/pkg/sentry/arch/signal_act.go index 36437b965..ad098c746 100644 --- a/pkg/sentry/arch/signal_act.go +++ b/pkg/sentry/arch/signal_act.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index 9ca4c8ed1..f7f054b0b 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/signal_info.go b/pkg/sentry/arch/signal_info.go index ec004ae75..fa0ecbec5 100644 --- a/pkg/sentry/arch/signal_info.go +++ b/pkg/sentry/arch/signal_info.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/signal_stack.go b/pkg/sentry/arch/signal_stack.go index ba43dd1d4..c02ae3b7c 100644 --- a/pkg/sentry/arch/signal_stack.go +++ b/pkg/sentry/arch/signal_stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/stack.go b/pkg/sentry/arch/stack.go index 6c1b9be82..716a3574d 100644 --- a/pkg/sentry/arch/stack.go +++ b/pkg/sentry/arch/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/syscalls_amd64.go b/pkg/sentry/arch/syscalls_amd64.go index 41d8ba0d1..47c31d4b9 100644 --- a/pkg/sentry/arch/syscalls_amd64.go +++ b/pkg/sentry/arch/syscalls_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/context/context.go b/pkg/sentry/context/context.go index 598c5b4ff..12bdcef85 100644 --- a/pkg/sentry/context/context.go +++ b/pkg/sentry/context/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/context/contexttest/contexttest.go b/pkg/sentry/context/contexttest/contexttest.go index b3c6a566b..d2f084ed7 100644 --- a/pkg/sentry/context/contexttest/contexttest.go +++ b/pkg/sentry/context/contexttest/contexttest.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/control.go b/pkg/sentry/control/control.go index a6ee6e649..32d30b6ea 100644 --- a/pkg/sentry/control/control.go +++ b/pkg/sentry/control/control.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go index 0ba730c1e..b6ac2f312 100644 --- a/pkg/sentry/control/proc.go +++ b/pkg/sentry/control/proc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/proc_test.go b/pkg/sentry/control/proc_test.go index 22c826236..5d52cd829 100644 --- a/pkg/sentry/control/proc_test.go +++ b/pkg/sentry/control/proc_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/state.go b/pkg/sentry/control/state.go index cee4db636..0a480c84a 100644 --- a/pkg/sentry/control/state.go +++ b/pkg/sentry/control/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/device/device.go b/pkg/sentry/device/device.go index 21fee8f8a..27e4eb258 100644 --- a/pkg/sentry/device/device.go +++ b/pkg/sentry/device/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/device/device_test.go b/pkg/sentry/device/device_test.go index dfec45046..5d8805c2f 100644 --- a/pkg/sentry/device/device_test.go +++ b/pkg/sentry/device/device_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/anon/anon.go b/pkg/sentry/fs/anon/anon.go index ddc2c0985..743cf511f 100644 --- a/pkg/sentry/fs/anon/anon.go +++ b/pkg/sentry/fs/anon/anon.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/anon/device.go b/pkg/sentry/fs/anon/device.go index 1c666729c..2d1249299 100644 --- a/pkg/sentry/fs/anon/device.go +++ b/pkg/sentry/fs/anon/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index bfd7f2762..5372875ac 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ashmem/device.go b/pkg/sentry/fs/ashmem/device.go index d0986fa11..962da141b 100644 --- a/pkg/sentry/fs/ashmem/device.go +++ b/pkg/sentry/fs/ashmem/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ashmem/pin_board.go b/pkg/sentry/fs/ashmem/pin_board.go index ecba395a0..7c997f533 100644 --- a/pkg/sentry/fs/ashmem/pin_board.go +++ b/pkg/sentry/fs/ashmem/pin_board.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ashmem/pin_board_test.go b/pkg/sentry/fs/ashmem/pin_board_test.go index f4ea5de6d..736e628dc 100644 --- a/pkg/sentry/fs/ashmem/pin_board_test.go +++ b/pkg/sentry/fs/ashmem/pin_board_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/attr.go b/pkg/sentry/fs/attr.go index 091f4ac63..59e060e3c 100644 --- a/pkg/sentry/fs/attr.go +++ b/pkg/sentry/fs/attr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index 502a262dd..42b9e8b26 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/context.go b/pkg/sentry/fs/context.go index da46ad77f..1775d3486 100644 --- a/pkg/sentry/fs/context.go +++ b/pkg/sentry/fs/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/copy_up.go b/pkg/sentry/fs/copy_up.go index 8c949b176..d65dc74bf 100644 --- a/pkg/sentry/fs/copy_up.go +++ b/pkg/sentry/fs/copy_up.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/copy_up_test.go b/pkg/sentry/fs/copy_up_test.go index c3c9d963d..64f030f72 100644 --- a/pkg/sentry/fs/copy_up_test.go +++ b/pkg/sentry/fs/copy_up_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dentry.go b/pkg/sentry/fs/dentry.go index b347468ff..ef6d1a870 100644 --- a/pkg/sentry/fs/dentry.go +++ b/pkg/sentry/fs/dentry.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/dev.go b/pkg/sentry/fs/dev/dev.go index 3f4f2a40a..05a5005ad 100644 --- a/pkg/sentry/fs/dev/dev.go +++ b/pkg/sentry/fs/dev/dev.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/device.go b/pkg/sentry/fs/dev/device.go index 9d935e008..3cecdf6e2 100644 --- a/pkg/sentry/fs/dev/device.go +++ b/pkg/sentry/fs/dev/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/fs.go b/pkg/sentry/fs/dev/fs.go index 2ae49be4e..d96f4f423 100644 --- a/pkg/sentry/fs/dev/fs.go +++ b/pkg/sentry/fs/dev/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index 492b8eb3a..eeda646ab 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 2977c8670..68090f353 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index 47b76218f..33e4913e4 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go index 27fea0019..2c01485a8 100644 --- a/pkg/sentry/fs/dirent.go +++ b/pkg/sentry/fs/dirent.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_cache.go b/pkg/sentry/fs/dirent_cache.go index c680e4828..502b0a09b 100644 --- a/pkg/sentry/fs/dirent_cache.go +++ b/pkg/sentry/fs/dirent_cache.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_cache_test.go b/pkg/sentry/fs/dirent_cache_test.go index 82b7f6bd5..5d0e9d91c 100644 --- a/pkg/sentry/fs/dirent_cache_test.go +++ b/pkg/sentry/fs/dirent_cache_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_refs_test.go b/pkg/sentry/fs/dirent_refs_test.go index f9dcba316..325404e27 100644 --- a/pkg/sentry/fs/dirent_refs_test.go +++ b/pkg/sentry/fs/dirent_refs_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_state.go b/pkg/sentry/fs/dirent_state.go index 04ab197b9..5cf151dab 100644 --- a/pkg/sentry/fs/dirent_state.go +++ b/pkg/sentry/fs/dirent_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index 2e34604e6..bfafff5ec 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe_opener.go b/pkg/sentry/fs/fdpipe/pipe_opener.go index 945cfaf08..92ab6ff0e 100644 --- a/pkg/sentry/fs/fdpipe/pipe_opener.go +++ b/pkg/sentry/fs/fdpipe/pipe_opener.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe_opener_test.go b/pkg/sentry/fs/fdpipe/pipe_opener_test.go index 83f6c1986..69516e048 100644 --- a/pkg/sentry/fs/fdpipe/pipe_opener_test.go +++ b/pkg/sentry/fs/fdpipe/pipe_opener_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe_state.go b/pkg/sentry/fs/fdpipe/pipe_state.go index 99c40d8ed..4395666ad 100644 --- a/pkg/sentry/fs/fdpipe/pipe_state.go +++ b/pkg/sentry/fs/fdpipe/pipe_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe_test.go b/pkg/sentry/fs/fdpipe/pipe_test.go index 6cd314f5b..d3f15be6b 100644 --- a/pkg/sentry/fs/fdpipe/pipe_test.go +++ b/pkg/sentry/fs/fdpipe/pipe_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 36794d378..d6752ed1b 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_operations.go b/pkg/sentry/fs/file_operations.go index d223bb5c7..28e8e233d 100644 --- a/pkg/sentry/fs/file_operations.go +++ b/pkg/sentry/fs/file_operations.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index 41e646ee8..9b958b64b 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_overlay_test.go b/pkg/sentry/fs/file_overlay_test.go index 830458ff9..11e4f7203 100644 --- a/pkg/sentry/fs/file_overlay_test.go +++ b/pkg/sentry/fs/file_overlay_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_state.go b/pkg/sentry/fs/file_state.go index f848d1b79..1c3bae3e8 100644 --- a/pkg/sentry/fs/file_state.go +++ b/pkg/sentry/fs/file_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_test.go b/pkg/sentry/fs/file_test.go index 18aee7101..f3ed9a70b 100644 --- a/pkg/sentry/fs/file_test.go +++ b/pkg/sentry/fs/file_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/filesystems.go b/pkg/sentry/fs/filesystems.go index 5a1e7a270..ba8be85e4 100644 --- a/pkg/sentry/fs/filesystems.go +++ b/pkg/sentry/fs/filesystems.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/filetest/filetest.go b/pkg/sentry/fs/filetest/filetest.go index 1831aa82f..65ca196d9 100644 --- a/pkg/sentry/fs/filetest/filetest.go +++ b/pkg/sentry/fs/filetest/filetest.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/flags.go b/pkg/sentry/fs/flags.go index 1aa271560..bf2a20b33 100644 --- a/pkg/sentry/fs/flags.go +++ b/pkg/sentry/fs/flags.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fs.go b/pkg/sentry/fs/fs.go index 6ec9ff446..b5c72990e 100644 --- a/pkg/sentry/fs/fs.go +++ b/pkg/sentry/fs/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/dirty_set.go b/pkg/sentry/fs/fsutil/dirty_set.go index 8e31e48fd..5add16ac4 100644 --- a/pkg/sentry/fs/fsutil/dirty_set.go +++ b/pkg/sentry/fs/fsutil/dirty_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/dirty_set_test.go b/pkg/sentry/fs/fsutil/dirty_set_test.go index f7693cb19..f5c9d9215 100644 --- a/pkg/sentry/fs/fsutil/dirty_set_test.go +++ b/pkg/sentry/fs/fsutil/dirty_set_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/file.go b/pkg/sentry/fs/fsutil/file.go index d5881613b..46db2e51c 100644 --- a/pkg/sentry/fs/fsutil/file.go +++ b/pkg/sentry/fs/fsutil/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/file_range_set.go b/pkg/sentry/fs/fsutil/file_range_set.go index da6949ccb..dd7ab4b4a 100644 --- a/pkg/sentry/fs/fsutil/file_range_set.go +++ b/pkg/sentry/fs/fsutil/file_range_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/frame_ref_set.go b/pkg/sentry/fs/fsutil/frame_ref_set.go index 14dece315..b6e783614 100644 --- a/pkg/sentry/fs/fsutil/frame_ref_set.go +++ b/pkg/sentry/fs/fsutil/frame_ref_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/fsutil.go b/pkg/sentry/fs/fsutil/fsutil.go index 6fe4ef13d..3d7f3732d 100644 --- a/pkg/sentry/fs/fsutil/fsutil.go +++ b/pkg/sentry/fs/fsutil/fsutil.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/handle.go b/pkg/sentry/fs/fsutil/handle.go index e7efd3c0f..8920b72ee 100644 --- a/pkg/sentry/fs/fsutil/handle.go +++ b/pkg/sentry/fs/fsutil/handle.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/handle_test.go b/pkg/sentry/fs/fsutil/handle_test.go index d94c3eb0d..43e1a3bdf 100644 --- a/pkg/sentry/fs/fsutil/handle_test.go +++ b/pkg/sentry/fs/fsutil/handle_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/host_file_mapper.go b/pkg/sentry/fs/fsutil/host_file_mapper.go index 9c1e2f76f..9599665f0 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/host_file_mapper_state.go b/pkg/sentry/fs/fsutil/host_file_mapper_state.go index 57705decd..bbd15b30b 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper_state.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go b/pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go index 790f3a5a6..86df76822 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index 3acc32752..d4db1c2de 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index 6777c8bf7..b0af44ddd 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/inode_cached_test.go b/pkg/sentry/fs/fsutil/inode_cached_test.go index 996c91849..e388ec3d7 100644 --- a/pkg/sentry/fs/fsutil/inode_cached_test.go +++ b/pkg/sentry/fs/fsutil/inode_cached_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/attr.go b/pkg/sentry/fs/gofer/attr.go index 5e24767f9..98700d014 100644 --- a/pkg/sentry/fs/gofer/attr.go +++ b/pkg/sentry/fs/gofer/attr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/cache_policy.go b/pkg/sentry/fs/gofer/cache_policy.go index 98f43c578..3d380f0e8 100644 --- a/pkg/sentry/fs/gofer/cache_policy.go +++ b/pkg/sentry/fs/gofer/cache_policy.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/context_file.go b/pkg/sentry/fs/gofer/context_file.go index d4b6f6eb7..a0265c2aa 100644 --- a/pkg/sentry/fs/gofer/context_file.go +++ b/pkg/sentry/fs/gofer/context_file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/device.go b/pkg/sentry/fs/gofer/device.go index fac7306d4..52c5acf48 100644 --- a/pkg/sentry/fs/gofer/device.go +++ b/pkg/sentry/fs/gofer/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index c4a210656..6d961813d 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/file_state.go b/pkg/sentry/fs/gofer/file_state.go index 715af8f16..dd4f817bf 100644 --- a/pkg/sentry/fs/gofer/file_state.go +++ b/pkg/sentry/fs/gofer/file_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/fs.go b/pkg/sentry/fs/gofer/fs.go index 3ae93f059..ed30cb1f1 100644 --- a/pkg/sentry/fs/gofer/fs.go +++ b/pkg/sentry/fs/gofer/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/gofer_test.go b/pkg/sentry/fs/gofer/gofer_test.go index c8d7bd773..3190d1e18 100644 --- a/pkg/sentry/fs/gofer/gofer_test.go +++ b/pkg/sentry/fs/gofer/gofer_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/handles.go b/pkg/sentry/fs/gofer/handles.go index a3e52aad6..f32e99ce0 100644 --- a/pkg/sentry/fs/gofer/handles.go +++ b/pkg/sentry/fs/gofer/handles.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go index 7fc8f77b0..5811b8b12 100644 --- a/pkg/sentry/fs/gofer/inode.go +++ b/pkg/sentry/fs/gofer/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/inode_state.go b/pkg/sentry/fs/gofer/inode_state.go index ad11034f9..ad4d3df58 100644 --- a/pkg/sentry/fs/gofer/inode_state.go +++ b/pkg/sentry/fs/gofer/inode_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/path.go b/pkg/sentry/fs/gofer/path.go index 0bf7881da..a324dc990 100644 --- a/pkg/sentry/fs/gofer/path.go +++ b/pkg/sentry/fs/gofer/path.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index 4e2293398..7552216f3 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/session_state.go b/pkg/sentry/fs/gofer/session_state.go index 8e6424492..f657135fc 100644 --- a/pkg/sentry/fs/gofer/session_state.go +++ b/pkg/sentry/fs/gofer/session_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/socket.go b/pkg/sentry/fs/gofer/socket.go index d072da624..76ce58810 100644 --- a/pkg/sentry/fs/gofer/socket.go +++ b/pkg/sentry/fs/gofer/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/util.go b/pkg/sentry/fs/gofer/util.go index d9ed8c81e..1a759370d 100644 --- a/pkg/sentry/fs/gofer/util.go +++ b/pkg/sentry/fs/gofer/util.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/control.go b/pkg/sentry/fs/host/control.go index d2e34a69d..0753640a2 100644 --- a/pkg/sentry/fs/host/control.go +++ b/pkg/sentry/fs/host/control.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/descriptor.go b/pkg/sentry/fs/host/descriptor.go index 148291ba6..7c9d2b299 100644 --- a/pkg/sentry/fs/host/descriptor.go +++ b/pkg/sentry/fs/host/descriptor.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/descriptor_state.go b/pkg/sentry/fs/host/descriptor_state.go index 7fb274451..530c0109f 100644 --- a/pkg/sentry/fs/host/descriptor_state.go +++ b/pkg/sentry/fs/host/descriptor_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/descriptor_test.go b/pkg/sentry/fs/host/descriptor_test.go index f393a8b54..6bc1bd2ae 100644 --- a/pkg/sentry/fs/host/descriptor_test.go +++ b/pkg/sentry/fs/host/descriptor_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/device.go b/pkg/sentry/fs/host/device.go index f2a0b6b15..b5adedf44 100644 --- a/pkg/sentry/fs/host/device.go +++ b/pkg/sentry/fs/host/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index 22a5d9f12..975084c86 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/fs.go b/pkg/sentry/fs/host/fs.go index e46ae433c..fec890964 100644 --- a/pkg/sentry/fs/host/fs.go +++ b/pkg/sentry/fs/host/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/fs_test.go b/pkg/sentry/fs/host/fs_test.go index b08125ca8..e69559aac 100644 --- a/pkg/sentry/fs/host/fs_test.go +++ b/pkg/sentry/fs/host/fs_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index e32497203..08754bd6b 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/inode_state.go b/pkg/sentry/fs/host/inode_state.go index 8bc99d94b..b7c1a9581 100644 --- a/pkg/sentry/fs/host/inode_state.go +++ b/pkg/sentry/fs/host/inode_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/inode_test.go b/pkg/sentry/fs/host/inode_test.go index 0ff87c418..9f1561bd5 100644 --- a/pkg/sentry/fs/host/inode_test.go +++ b/pkg/sentry/fs/host/inode_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/ioctl_unsafe.go b/pkg/sentry/fs/host/ioctl_unsafe.go index bc965a1c2..175dca613 100644 --- a/pkg/sentry/fs/host/ioctl_unsafe.go +++ b/pkg/sentry/fs/host/ioctl_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket.go b/pkg/sentry/fs/host/socket.go index 0eb267c00..af53bf533 100644 --- a/pkg/sentry/fs/host/socket.go +++ b/pkg/sentry/fs/host/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket_iovec.go b/pkg/sentry/fs/host/socket_iovec.go index 1a9587b90..d4ce4a8c1 100644 --- a/pkg/sentry/fs/host/socket_iovec.go +++ b/pkg/sentry/fs/host/socket_iovec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket_state.go b/pkg/sentry/fs/host/socket_state.go index 7fa500bfb..2932c1f16 100644 --- a/pkg/sentry/fs/host/socket_state.go +++ b/pkg/sentry/fs/host/socket_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket_test.go b/pkg/sentry/fs/host/socket_test.go index 483e99dd6..e9a88b124 100644 --- a/pkg/sentry/fs/host/socket_test.go +++ b/pkg/sentry/fs/host/socket_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket_unsafe.go b/pkg/sentry/fs/host/socket_unsafe.go index 5e4c5feed..f35e2492d 100644 --- a/pkg/sentry/fs/host/socket_unsafe.go +++ b/pkg/sentry/fs/host/socket_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/tty.go b/pkg/sentry/fs/host/tty.go index ad1323610..cf3639c46 100644 --- a/pkg/sentry/fs/host/tty.go +++ b/pkg/sentry/fs/host/tty.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/util.go b/pkg/sentry/fs/host/util.go index 74c703eb7..40c450660 100644 --- a/pkg/sentry/fs/host/util.go +++ b/pkg/sentry/fs/host/util.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/util_unsafe.go b/pkg/sentry/fs/host/util_unsafe.go index 2ecb54319..d00da89d6 100644 --- a/pkg/sentry/fs/host/util_unsafe.go +++ b/pkg/sentry/fs/host/util_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/wait_test.go b/pkg/sentry/fs/host/wait_test.go index c5f5c9c0d..9ca8c399f 100644 --- a/pkg/sentry/fs/host/wait_test.go +++ b/pkg/sentry/fs/host/wait_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index 409c81a97..95769ccf8 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode_inotify.go b/pkg/sentry/fs/inode_inotify.go index 683140afe..e213df924 100644 --- a/pkg/sentry/fs/inode_inotify.go +++ b/pkg/sentry/fs/inode_inotify.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode_operations.go b/pkg/sentry/fs/inode_operations.go index 3ee3de10e..77973ce79 100644 --- a/pkg/sentry/fs/inode_operations.go +++ b/pkg/sentry/fs/inode_operations.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode_overlay.go b/pkg/sentry/fs/inode_overlay.go index cf698a4da..78923fb5b 100644 --- a/pkg/sentry/fs/inode_overlay.go +++ b/pkg/sentry/fs/inode_overlay.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode_overlay_test.go b/pkg/sentry/fs/inode_overlay_test.go index 23e5635a4..bba20da14 100644 --- a/pkg/sentry/fs/inode_overlay_test.go +++ b/pkg/sentry/fs/inode_overlay_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 2aabdded8..f251df0d1 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inotify_event.go b/pkg/sentry/fs/inotify_event.go index e9b5e0f56..9e3e9d816 100644 --- a/pkg/sentry/fs/inotify_event.go +++ b/pkg/sentry/fs/inotify_event.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inotify_watch.go b/pkg/sentry/fs/inotify_watch.go index 3e1959e83..b83544c9f 100644 --- a/pkg/sentry/fs/inotify_watch.go +++ b/pkg/sentry/fs/inotify_watch.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/lock/lock.go b/pkg/sentry/fs/lock/lock.go index 439e645db..5ff800d2d 100644 --- a/pkg/sentry/fs/lock/lock.go +++ b/pkg/sentry/fs/lock/lock.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/lock/lock_range_test.go b/pkg/sentry/fs/lock/lock_range_test.go index 06a37c701..b0ab882b9 100644 --- a/pkg/sentry/fs/lock/lock_range_test.go +++ b/pkg/sentry/fs/lock/lock_range_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/lock/lock_set_functions.go b/pkg/sentry/fs/lock/lock_set_functions.go index e16f485be..395592a4b 100644 --- a/pkg/sentry/fs/lock/lock_set_functions.go +++ b/pkg/sentry/fs/lock/lock_set_functions.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/lock/lock_test.go b/pkg/sentry/fs/lock/lock_test.go index c60f5f7a2..67fa4b1dd 100644 --- a/pkg/sentry/fs/lock/lock_test.go +++ b/pkg/sentry/fs/lock/lock_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mock.go b/pkg/sentry/fs/mock.go index 846b6e8bb..6bfcda6bb 100644 --- a/pkg/sentry/fs/mock.go +++ b/pkg/sentry/fs/mock.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index 8345876fc..24e28ddb2 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mount_overlay.go b/pkg/sentry/fs/mount_overlay.go index dbc608c7e..fb91635bc 100644 --- a/pkg/sentry/fs/mount_overlay.go +++ b/pkg/sentry/fs/mount_overlay.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mount_state.go b/pkg/sentry/fs/mount_state.go index f5ed1dd8d..6344d5160 100644 --- a/pkg/sentry/fs/mount_state.go +++ b/pkg/sentry/fs/mount_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mount_test.go b/pkg/sentry/fs/mount_test.go index 968b435ab..a1c9f4f79 100644 --- a/pkg/sentry/fs/mount_test.go +++ b/pkg/sentry/fs/mount_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mounts.go b/pkg/sentry/fs/mounts.go index c0a803b2d..7c5348cce 100644 --- a/pkg/sentry/fs/mounts.go +++ b/pkg/sentry/fs/mounts.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mounts_test.go b/pkg/sentry/fs/mounts_test.go index 8669f3a38..cc7c32c9b 100644 --- a/pkg/sentry/fs/mounts_test.go +++ b/pkg/sentry/fs/mounts_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/offset.go b/pkg/sentry/fs/offset.go index 7cc8398e6..38aee765a 100644 --- a/pkg/sentry/fs/offset.go +++ b/pkg/sentry/fs/offset.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/overlay.go b/pkg/sentry/fs/overlay.go index 5a30af419..036c0f733 100644 --- a/pkg/sentry/fs/overlay.go +++ b/pkg/sentry/fs/overlay.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/path.go b/pkg/sentry/fs/path.go index b74f6ed8c..91a9a8ffd 100644 --- a/pkg/sentry/fs/path.go +++ b/pkg/sentry/fs/path.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/path_test.go b/pkg/sentry/fs/path_test.go index 7ab070855..391b010a7 100644 --- a/pkg/sentry/fs/path_test.go +++ b/pkg/sentry/fs/path_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/cpuinfo.go b/pkg/sentry/fs/proc/cpuinfo.go index 4dfec03a4..f8be06dc3 100644 --- a/pkg/sentry/fs/proc/cpuinfo.go +++ b/pkg/sentry/fs/proc/cpuinfo.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/device/device.go b/pkg/sentry/fs/proc/device/device.go index 6194afe88..04b687bcf 100644 --- a/pkg/sentry/fs/proc/device/device.go +++ b/pkg/sentry/fs/proc/device/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/exec_args.go b/pkg/sentry/fs/proc/exec_args.go index a69cbaa0e..b4896053f 100644 --- a/pkg/sentry/fs/proc/exec_args.go +++ b/pkg/sentry/fs/proc/exec_args.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index dada8f982..5ebb33703 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/file.go b/pkg/sentry/fs/proc/file.go index 4b3448245..f659e590a 100644 --- a/pkg/sentry/fs/proc/file.go +++ b/pkg/sentry/fs/proc/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/filesystems.go b/pkg/sentry/fs/proc/filesystems.go index 49b92fd8a..c050a00be 100644 --- a/pkg/sentry/fs/proc/filesystems.go +++ b/pkg/sentry/fs/proc/filesystems.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/fs.go b/pkg/sentry/fs/proc/fs.go index 061824b8c..63f737ff4 100644 --- a/pkg/sentry/fs/proc/fs.go +++ b/pkg/sentry/fs/proc/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/loadavg.go b/pkg/sentry/fs/proc/loadavg.go index 6fac251d2..78f3a1dc0 100644 --- a/pkg/sentry/fs/proc/loadavg.go +++ b/pkg/sentry/fs/proc/loadavg.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/meminfo.go b/pkg/sentry/fs/proc/meminfo.go index 53dfd59ef..b31258eed 100644 --- a/pkg/sentry/fs/proc/meminfo.go +++ b/pkg/sentry/fs/proc/meminfo.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/mounts.go b/pkg/sentry/fs/proc/mounts.go index 81dcc153a..0b0e87528 100644 --- a/pkg/sentry/fs/proc/mounts.go +++ b/pkg/sentry/fs/proc/mounts.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/net.go b/pkg/sentry/fs/proc/net.go index 8cd6fe9d3..45f2a1211 100644 --- a/pkg/sentry/fs/proc/net.go +++ b/pkg/sentry/fs/proc/net.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/net_test.go b/pkg/sentry/fs/proc/net_test.go index a31a20494..94677cc1d 100644 --- a/pkg/sentry/fs/proc/net_test.go +++ b/pkg/sentry/fs/proc/net_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index 07029a7bb..33030bebf 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/rpcinet_proc.go b/pkg/sentry/fs/proc/rpcinet_proc.go index 50d0271f9..d025069df 100644 --- a/pkg/sentry/fs/proc/rpcinet_proc.go +++ b/pkg/sentry/fs/proc/rpcinet_proc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index 51cae5e37..0499ba65b 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/seqfile/seqfile_test.go b/pkg/sentry/fs/proc/seqfile/seqfile_test.go index d90e3e736..f9a2ca38e 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile_test.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/stat.go b/pkg/sentry/fs/proc/stat.go index bf7650211..f2bbef375 100644 --- a/pkg/sentry/fs/proc/stat.go +++ b/pkg/sentry/fs/proc/stat.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index 384b4ffe1..54562508d 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index beb25be20..801eb6a1e 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/sys_net_test.go b/pkg/sentry/fs/proc/sys_net_test.go index 7ba392346..0ce9d30f1 100644 --- a/pkg/sentry/fs/proc/sys_net_test.go +++ b/pkg/sentry/fs/proc/sys_net_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 748ca4320..404faea0a 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index a7e4cf0a6..f70399686 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/uptime.go b/pkg/sentry/fs/proc/uptime.go index f3a9b81df..80c7ce0b4 100644 --- a/pkg/sentry/fs/proc/uptime.go +++ b/pkg/sentry/fs/proc/uptime.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/version.go b/pkg/sentry/fs/proc/version.go index 00f6a2afd..b6d49d5e9 100644 --- a/pkg/sentry/fs/proc/version.go +++ b/pkg/sentry/fs/proc/version.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index 075e13b01..0a911b155 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/file.go b/pkg/sentry/fs/ramfs/file.go index 0b94d92a1..b7fc98ffc 100644 --- a/pkg/sentry/fs/ramfs/file.go +++ b/pkg/sentry/fs/ramfs/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/ramfs.go b/pkg/sentry/fs/ramfs/ramfs.go index 83cbcab23..d77688a34 100644 --- a/pkg/sentry/fs/ramfs/ramfs.go +++ b/pkg/sentry/fs/ramfs/ramfs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index 9ac00eb18..8c81478c8 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index 1c54d9991..a21fac2c7 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/test/test.go b/pkg/sentry/fs/ramfs/test/test.go index fb669558f..11bff7729 100644 --- a/pkg/sentry/fs/ramfs/test/test.go +++ b/pkg/sentry/fs/ramfs/test/test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/tree.go b/pkg/sentry/fs/ramfs/tree.go index 1fb335f74..29a70f698 100644 --- a/pkg/sentry/fs/ramfs/tree.go +++ b/pkg/sentry/fs/ramfs/tree.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/tree_test.go b/pkg/sentry/fs/ramfs/tree_test.go index 68e2929d5..d5567d9e1 100644 --- a/pkg/sentry/fs/ramfs/tree_test.go +++ b/pkg/sentry/fs/ramfs/tree_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/restore.go b/pkg/sentry/fs/restore.go index b4ac85a27..da2df7e1d 100644 --- a/pkg/sentry/fs/restore.go +++ b/pkg/sentry/fs/restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/save.go b/pkg/sentry/fs/save.go index bf2a85143..90988d385 100644 --- a/pkg/sentry/fs/save.go +++ b/pkg/sentry/fs/save.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/seek.go b/pkg/sentry/fs/seek.go index 1268726c2..72f3fb632 100644 --- a/pkg/sentry/fs/seek.go +++ b/pkg/sentry/fs/seek.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sync.go b/pkg/sentry/fs/sync.go index 9738a8f22..6dcc2fe8d 100644 --- a/pkg/sentry/fs/sync.go +++ b/pkg/sentry/fs/sync.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sys/device.go b/pkg/sentry/fs/sys/device.go index 54e414d1b..38ecd0c18 100644 --- a/pkg/sentry/fs/sys/device.go +++ b/pkg/sentry/fs/sys/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sys/devices.go b/pkg/sentry/fs/sys/devices.go index 2cf3a6f98..e64aa0edc 100644 --- a/pkg/sentry/fs/sys/devices.go +++ b/pkg/sentry/fs/sys/devices.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sys/fs.go b/pkg/sentry/fs/sys/fs.go index 625525540..5ce33f87f 100644 --- a/pkg/sentry/fs/sys/fs.go +++ b/pkg/sentry/fs/sys/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sys/sys.go b/pkg/sentry/fs/sys/sys.go index 7b9697668..7cc1942c7 100644 --- a/pkg/sentry/fs/sys/sys.go +++ b/pkg/sentry/fs/sys/sys.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/timerfd/timerfd.go b/pkg/sentry/fs/timerfd/timerfd.go index 767db95a0..7423e816c 100644 --- a/pkg/sentry/fs/timerfd/timerfd.go +++ b/pkg/sentry/fs/timerfd/timerfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/device.go b/pkg/sentry/fs/tmpfs/device.go index e588b3440..aade93c26 100644 --- a/pkg/sentry/fs/tmpfs/device.go +++ b/pkg/sentry/fs/tmpfs/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/file_regular.go b/pkg/sentry/fs/tmpfs/file_regular.go index 342688f81..1f9d69909 100644 --- a/pkg/sentry/fs/tmpfs/file_regular.go +++ b/pkg/sentry/fs/tmpfs/file_regular.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/file_test.go b/pkg/sentry/fs/tmpfs/file_test.go index f064eb1ac..b5830d3df 100644 --- a/pkg/sentry/fs/tmpfs/file_test.go +++ b/pkg/sentry/fs/tmpfs/file_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go index ca620e65e..7c91e248b 100644 --- a/pkg/sentry/fs/tmpfs/fs.go +++ b/pkg/sentry/fs/tmpfs/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 1e4fe47d2..42a7d7b9c 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 38be6db46..91b782540 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index 7c0c0b0c1..e32b05c1d 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/fs.go b/pkg/sentry/fs/tty/fs.go index d9f8f02f3..0c412eb21 100644 --- a/pkg/sentry/fs/tty/fs.go +++ b/pkg/sentry/fs/tty/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/inode.go b/pkg/sentry/fs/tty/inode.go index c0fa2b407..d5d1caafc 100644 --- a/pkg/sentry/fs/tty/inode.go +++ b/pkg/sentry/fs/tty/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/line_discipline.go b/pkg/sentry/fs/tty/line_discipline.go index 31804571e..484366f85 100644 --- a/pkg/sentry/fs/tty/line_discipline.go +++ b/pkg/sentry/fs/tty/line_discipline.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index ae7540eff..dad0cad79 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/queue.go b/pkg/sentry/fs/tty/queue.go index 01dc8d1ac..a09ca0119 100644 --- a/pkg/sentry/fs/tty/queue.go +++ b/pkg/sentry/fs/tty/queue.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index 4a0d4fdb9..9de3168bf 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/terminal.go b/pkg/sentry/fs/tty/terminal.go index 3cb135124..79f9d76d7 100644 --- a/pkg/sentry/fs/tty/terminal.go +++ b/pkg/sentry/fs/tty/terminal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/tty_test.go b/pkg/sentry/fs/tty/tty_test.go index 32e1b1556..ad535838f 100644 --- a/pkg/sentry/fs/tty/tty_test.go +++ b/pkg/sentry/fs/tty/tty_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/hostcpu/getcpu_amd64.s b/pkg/sentry/hostcpu/getcpu_amd64.s index 7f6247d81..409db1450 100644 --- a/pkg/sentry/hostcpu/getcpu_amd64.s +++ b/pkg/sentry/hostcpu/getcpu_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/hostcpu/hostcpu.go b/pkg/sentry/hostcpu/hostcpu.go index fa46499ad..3adc847bb 100644 --- a/pkg/sentry/hostcpu/hostcpu.go +++ b/pkg/sentry/hostcpu/hostcpu.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/hostcpu/hostcpu_test.go b/pkg/sentry/hostcpu/hostcpu_test.go index a82e1a271..38de0e1f6 100644 --- a/pkg/sentry/hostcpu/hostcpu_test.go +++ b/pkg/sentry/hostcpu/hostcpu_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/inet/context.go b/pkg/sentry/inet/context.go index 370381f41..d05e96f15 100644 --- a/pkg/sentry/inet/context.go +++ b/pkg/sentry/inet/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/inet/inet.go b/pkg/sentry/inet/inet.go index 30ca4e0c0..8206377cc 100644 --- a/pkg/sentry/inet/inet.go +++ b/pkg/sentry/inet/inet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/inet/test_stack.go b/pkg/sentry/inet/test_stack.go index bc10926ee..05c1a1792 100644 --- a/pkg/sentry/inet/test_stack.go +++ b/pkg/sentry/inet/test_stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/abstract_socket_namespace.go b/pkg/sentry/kernel/abstract_socket_namespace.go index 45088c988..1ea2cee36 100644 --- a/pkg/sentry/kernel/abstract_socket_namespace.go +++ b/pkg/sentry/kernel/abstract_socket_namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/auth.go b/pkg/sentry/kernel/auth/auth.go index c49a6b852..19f15fd36 100644 --- a/pkg/sentry/kernel/auth/auth.go +++ b/pkg/sentry/kernel/auth/auth.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/capability_set.go b/pkg/sentry/kernel/auth/capability_set.go index 5b8164c49..88d6243aa 100644 --- a/pkg/sentry/kernel/auth/capability_set.go +++ b/pkg/sentry/kernel/auth/capability_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/context.go b/pkg/sentry/kernel/auth/context.go index 914589b28..f7e945599 100644 --- a/pkg/sentry/kernel/auth/context.go +++ b/pkg/sentry/kernel/auth/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go index f18f7dac9..de33f1953 100644 --- a/pkg/sentry/kernel/auth/credentials.go +++ b/pkg/sentry/kernel/auth/credentials.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/id.go b/pkg/sentry/kernel/auth/id.go index 37522b018..e5bed44d7 100644 --- a/pkg/sentry/kernel/auth/id.go +++ b/pkg/sentry/kernel/auth/id.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/id_map.go b/pkg/sentry/kernel/auth/id_map.go index bd0090e0f..43f439825 100644 --- a/pkg/sentry/kernel/auth/id_map.go +++ b/pkg/sentry/kernel/auth/id_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/id_map_functions.go b/pkg/sentry/kernel/auth/id_map_functions.go index 889291d96..8f1a189ec 100644 --- a/pkg/sentry/kernel/auth/id_map_functions.go +++ b/pkg/sentry/kernel/auth/id_map_functions.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/user_namespace.go b/pkg/sentry/kernel/auth/user_namespace.go index d359f3f31..5bb9c44c0 100644 --- a/pkg/sentry/kernel/auth/user_namespace.go +++ b/pkg/sentry/kernel/auth/user_namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/context.go b/pkg/sentry/kernel/context.go index 261ca6f7a..b629521eb 100644 --- a/pkg/sentry/kernel/context.go +++ b/pkg/sentry/kernel/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index a8eb114c0..9c13ecfcc 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/epoll/epoll_state.go b/pkg/sentry/kernel/epoll/epoll_state.go index dabb32f49..7f3e2004a 100644 --- a/pkg/sentry/kernel/epoll/epoll_state.go +++ b/pkg/sentry/kernel/epoll/epoll_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/epoll/epoll_test.go b/pkg/sentry/kernel/epoll/epoll_test.go index bc869fc13..d89c1b745 100644 --- a/pkg/sentry/kernel/epoll/epoll_test.go +++ b/pkg/sentry/kernel/epoll/epoll_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index a4ada0e78..26dc59a85 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/eventfd/eventfd_test.go b/pkg/sentry/kernel/eventfd/eventfd_test.go index 71326b62f..14e8996d9 100644 --- a/pkg/sentry/kernel/eventfd/eventfd_test.go +++ b/pkg/sentry/kernel/eventfd/eventfd_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/fasync/fasync.go b/pkg/sentry/kernel/fasync/fasync.go index f77339cae..aa4aac109 100644 --- a/pkg/sentry/kernel/fasync/fasync.go +++ b/pkg/sentry/kernel/fasync/fasync.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/fd_map.go b/pkg/sentry/kernel/fd_map.go index cad0b0a20..715f4714d 100644 --- a/pkg/sentry/kernel/fd_map.go +++ b/pkg/sentry/kernel/fd_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/fd_map_test.go b/pkg/sentry/kernel/fd_map_test.go index 95123aef3..b49996137 100644 --- a/pkg/sentry/kernel/fd_map_test.go +++ b/pkg/sentry/kernel/fd_map_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/fs_context.go b/pkg/sentry/kernel/fs_context.go index f3f05e8f5..3cf0db280 100644 --- a/pkg/sentry/kernel/fs_context.go +++ b/pkg/sentry/kernel/fs_context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index 54b1982a0..ea69d433b 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/futex/futex_test.go b/pkg/sentry/kernel/futex/futex_test.go index 726c26990..ea506a29b 100644 --- a/pkg/sentry/kernel/futex/futex_test.go +++ b/pkg/sentry/kernel/futex/futex_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/ipc_namespace.go b/pkg/sentry/kernel/ipc_namespace.go index 5eef49f59..9ceb9bd92 100644 --- a/pkg/sentry/kernel/ipc_namespace.go +++ b/pkg/sentry/kernel/ipc_namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/kdefs/kdefs.go b/pkg/sentry/kernel/kdefs/kdefs.go index bbb476544..8eafe810b 100644 --- a/pkg/sentry/kernel/kdefs/kdefs.go +++ b/pkg/sentry/kernel/kdefs/kdefs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 5d6856f3c..bad558d48 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/kernel_state.go b/pkg/sentry/kernel/kernel_state.go index bb2d5102d..a0a69b498 100644 --- a/pkg/sentry/kernel/kernel_state.go +++ b/pkg/sentry/kernel/kernel_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/memevent/memory_events.go b/pkg/sentry/kernel/memevent/memory_events.go index f7a183a1d..f05ef1b64 100644 --- a/pkg/sentry/kernel/memevent/memory_events.go +++ b/pkg/sentry/kernel/memevent/memory_events.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/memevent/memory_events.proto b/pkg/sentry/kernel/memevent/memory_events.proto index abc565054..43b8deb76 100644 --- a/pkg/sentry/kernel/memevent/memory_events.proto +++ b/pkg/sentry/kernel/memevent/memory_events.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go index bb5db0309..373e11772 100644 --- a/pkg/sentry/kernel/pending_signals.go +++ b/pkg/sentry/kernel/pending_signals.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pending_signals_state.go b/pkg/sentry/kernel/pending_signals_state.go index 6d90ed033..72be6702f 100644 --- a/pkg/sentry/kernel/pending_signals_state.go +++ b/pkg/sentry/kernel/pending_signals_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/buffers.go b/pkg/sentry/kernel/pipe/buffers.go index a82e45c3f..fa8045910 100644 --- a/pkg/sentry/kernel/pipe/buffers.go +++ b/pkg/sentry/kernel/pipe/buffers.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/device.go b/pkg/sentry/kernel/pipe/device.go index 8d383577a..eec5c5de8 100644 --- a/pkg/sentry/kernel/pipe/device.go +++ b/pkg/sentry/kernel/pipe/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index 23d692da1..4b0e00b85 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/node_test.go b/pkg/sentry/kernel/pipe/node_test.go index cc1ebf4f6..eda551594 100644 --- a/pkg/sentry/kernel/pipe/node_test.go +++ b/pkg/sentry/kernel/pipe/node_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index ced2559a7..126054826 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/pipe_test.go b/pkg/sentry/kernel/pipe/pipe_test.go index 49ef8c8ac..3b9895927 100644 --- a/pkg/sentry/kernel/pipe/pipe_test.go +++ b/pkg/sentry/kernel/pipe/pipe_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/reader.go b/pkg/sentry/kernel/pipe/reader.go index 1fa5e9a32..f27379969 100644 --- a/pkg/sentry/kernel/pipe/reader.go +++ b/pkg/sentry/kernel/pipe/reader.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index 82607367b..63efc5bbe 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/writer.go b/pkg/sentry/kernel/pipe/writer.go index d93324b53..6fea9769c 100644 --- a/pkg/sentry/kernel/pipe/writer.go +++ b/pkg/sentry/kernel/pipe/writer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/posixtimer.go b/pkg/sentry/kernel/posixtimer.go index 0ab958529..40b5acca3 100644 --- a/pkg/sentry/kernel/posixtimer.go +++ b/pkg/sentry/kernel/posixtimer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index 9fe28f435..20bac2b70 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/rseq.go b/pkg/sentry/kernel/rseq.go index 1f3de58e3..46b03c700 100644 --- a/pkg/sentry/kernel/rseq.go +++ b/pkg/sentry/kernel/rseq.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/sched/cpuset.go b/pkg/sentry/kernel/sched/cpuset.go index 0a97603f0..69aee9127 100644 --- a/pkg/sentry/kernel/sched/cpuset.go +++ b/pkg/sentry/kernel/sched/cpuset.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/sched/cpuset_test.go b/pkg/sentry/kernel/sched/cpuset_test.go index 8a6e12958..a036ed513 100644 --- a/pkg/sentry/kernel/sched/cpuset_test.go +++ b/pkg/sentry/kernel/sched/cpuset_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/sched/sched.go b/pkg/sentry/kernel/sched/sched.go index f1de1da60..e59909baf 100644 --- a/pkg/sentry/kernel/sched/sched.go +++ b/pkg/sentry/kernel/sched/sched.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/seccomp.go b/pkg/sentry/kernel/seccomp.go index d77c05e2f..37dd3e4c9 100644 --- a/pkg/sentry/kernel/seccomp.go +++ b/pkg/sentry/kernel/seccomp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index aa07946cf..232a276dc 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/semaphore/semaphore_test.go b/pkg/sentry/kernel/semaphore/semaphore_test.go index f9eb382e9..5f886bf31 100644 --- a/pkg/sentry/kernel/semaphore/semaphore_test.go +++ b/pkg/sentry/kernel/semaphore/semaphore_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index a9b4e7647..78a5b4063 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/shm/device.go b/pkg/sentry/kernel/shm/device.go index b0dacdbe0..bbc653ed8 100644 --- a/pkg/sentry/kernel/shm/device.go +++ b/pkg/sentry/kernel/shm/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 77973951e..8d0d14e45 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/signal.go b/pkg/sentry/kernel/signal.go index e3a2a777a..b066df132 100644 --- a/pkg/sentry/kernel/signal.go +++ b/pkg/sentry/kernel/signal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/signal_handlers.go b/pkg/sentry/kernel/signal_handlers.go index 3649f5e4d..3f1ac9898 100644 --- a/pkg/sentry/kernel/signal_handlers.go +++ b/pkg/sentry/kernel/signal_handlers.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go index 4c7811b6c..19b711e9c 100644 --- a/pkg/sentry/kernel/syscalls.go +++ b/pkg/sentry/kernel/syscalls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/syscalls_state.go b/pkg/sentry/kernel/syscalls_state.go index 826809a70..981455d46 100644 --- a/pkg/sentry/kernel/syscalls_state.go +++ b/pkg/sentry/kernel/syscalls_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/syslog.go b/pkg/sentry/kernel/syslog.go index 6531bd5d2..2aecf3eea 100644 --- a/pkg/sentry/kernel/syslog.go +++ b/pkg/sentry/kernel/syslog.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/table_test.go b/pkg/sentry/kernel/table_test.go index 71ca75555..3b29d3c6a 100644 --- a/pkg/sentry/kernel/table_test.go +++ b/pkg/sentry/kernel/table_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 4f0b7fe3f..e22ec768d 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_acct.go b/pkg/sentry/kernel/task_acct.go index d2052921e..24230af89 100644 --- a/pkg/sentry/kernel/task_acct.go +++ b/pkg/sentry/kernel/task_acct.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_block.go b/pkg/sentry/kernel/task_block.go index 6dc7b938e..e5027e551 100644 --- a/pkg/sentry/kernel/task_block.go +++ b/pkg/sentry/kernel/task_block.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index de3aef40d..755fe0370 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index d2df7e9d1..45b8d2b04 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index 1b760aba4..a9b74da8e 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index 65969ca9b..44fbb487c 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_futex.go b/pkg/sentry/kernel/task_futex.go index 62ebbcb0d..5a11ca3df 100644 --- a/pkg/sentry/kernel/task_futex.go +++ b/pkg/sentry/kernel/task_futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_identity.go b/pkg/sentry/kernel/task_identity.go index b0921b2eb..8f90ed786 100644 --- a/pkg/sentry/kernel/task_identity.go +++ b/pkg/sentry/kernel/task_identity.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_log.go b/pkg/sentry/kernel/task_log.go index 1769da210..f4c881c2d 100644 --- a/pkg/sentry/kernel/task_log.go +++ b/pkg/sentry/kernel/task_log.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_net.go b/pkg/sentry/kernel/task_net.go index 4df2e53d3..fc7cefc1f 100644 --- a/pkg/sentry/kernel/task_net.go +++ b/pkg/sentry/kernel/task_net.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 49ac933b7..596b9aa16 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_sched.go b/pkg/sentry/kernel/task_sched.go index 19dcc963a..3b3cdc24a 100644 --- a/pkg/sentry/kernel/task_sched.go +++ b/pkg/sentry/kernel/task_sched.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index e2925a708..fe24f7542 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_start.go b/pkg/sentry/kernel/task_start.go index 6c8d7d316..c82a32c78 100644 --- a/pkg/sentry/kernel/task_start.go +++ b/pkg/sentry/kernel/task_start.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_stop.go b/pkg/sentry/kernel/task_stop.go index feaf6cae4..36846484c 100644 --- a/pkg/sentry/kernel/task_stop.go +++ b/pkg/sentry/kernel/task_stop.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index f0373c375..0318adb35 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_test.go b/pkg/sentry/kernel/task_test.go index 82ef858a1..3f37f505d 100644 --- a/pkg/sentry/kernel/task_test.go +++ b/pkg/sentry/kernel/task_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_usermem.go b/pkg/sentry/kernel/task_usermem.go index 2b4954869..c8e973bd5 100644 --- a/pkg/sentry/kernel/task_usermem.go +++ b/pkg/sentry/kernel/task_usermem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index dfff7b52d..d7652f57c 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index 4e3d19e97..bdb907905 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/time/context.go b/pkg/sentry/kernel/time/context.go index ac4dc01d8..3675ea20d 100644 --- a/pkg/sentry/kernel/time/context.go +++ b/pkg/sentry/kernel/time/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go index 52e0dfba1..ca0f4ba2e 100644 --- a/pkg/sentry/kernel/time/time.go +++ b/pkg/sentry/kernel/time/time.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go index 2167f3efe..6bff80f13 100644 --- a/pkg/sentry/kernel/timekeeper.go +++ b/pkg/sentry/kernel/timekeeper.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/timekeeper_state.go b/pkg/sentry/kernel/timekeeper_state.go index 2e7fed4d8..f3a3ed543 100644 --- a/pkg/sentry/kernel/timekeeper_state.go +++ b/pkg/sentry/kernel/timekeeper_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/timekeeper_test.go b/pkg/sentry/kernel/timekeeper_test.go index 34a5cec27..71674c21c 100644 --- a/pkg/sentry/kernel/timekeeper_test.go +++ b/pkg/sentry/kernel/timekeeper_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/uts_namespace.go b/pkg/sentry/kernel/uts_namespace.go index 7e0fe0d21..ed5f0c031 100644 --- a/pkg/sentry/kernel/uts_namespace.go +++ b/pkg/sentry/kernel/uts_namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/vdso.go b/pkg/sentry/kernel/vdso.go index 971e8bc59..0ec858a4a 100644 --- a/pkg/sentry/kernel/vdso.go +++ b/pkg/sentry/kernel/vdso.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/version.go b/pkg/sentry/kernel/version.go index a9e84673f..72bb0f93c 100644 --- a/pkg/sentry/kernel/version.go +++ b/pkg/sentry/kernel/version.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/limits/context.go b/pkg/sentry/limits/context.go index 75e97bf92..bf413eb7d 100644 --- a/pkg/sentry/limits/context.go +++ b/pkg/sentry/limits/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/limits/limits.go b/pkg/sentry/limits/limits.go index 02c8b60e3..ba0b7d4fd 100644 --- a/pkg/sentry/limits/limits.go +++ b/pkg/sentry/limits/limits.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/limits/limits_test.go b/pkg/sentry/limits/limits_test.go index dd6f80750..d41f62554 100644 --- a/pkg/sentry/limits/limits_test.go +++ b/pkg/sentry/limits/limits_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/limits/linux.go b/pkg/sentry/limits/linux.go index 8e6a24341..511db6733 100644 --- a/pkg/sentry/limits/linux.go +++ b/pkg/sentry/limits/linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/elf.go b/pkg/sentry/loader/elf.go index 849be5a3d..9b1e81dc9 100644 --- a/pkg/sentry/loader/elf.go +++ b/pkg/sentry/loader/elf.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/interpreter.go b/pkg/sentry/loader/interpreter.go index 54534952b..06a3c7156 100644 --- a/pkg/sentry/loader/interpreter.go +++ b/pkg/sentry/loader/interpreter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/loader.go b/pkg/sentry/loader/loader.go index 62b39e52b..d1417c4f1 100644 --- a/pkg/sentry/loader/loader.go +++ b/pkg/sentry/loader/loader.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index a06e27ac9..437cc5da1 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/vdso_state.go b/pkg/sentry/loader/vdso_state.go index dc71e1c2d..b327f0e1e 100644 --- a/pkg/sentry/loader/vdso_state.go +++ b/pkg/sentry/loader/vdso_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memmap/mapping_set.go b/pkg/sentry/memmap/mapping_set.go index c9483905d..33cf16f91 100644 --- a/pkg/sentry/memmap/mapping_set.go +++ b/pkg/sentry/memmap/mapping_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memmap/mapping_set_test.go b/pkg/sentry/memmap/mapping_set_test.go index 10668d404..49ee34548 100644 --- a/pkg/sentry/memmap/mapping_set_test.go +++ b/pkg/sentry/memmap/mapping_set_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memmap/memmap.go b/pkg/sentry/memmap/memmap.go index cdc5f2b27..05349a77f 100644 --- a/pkg/sentry/memmap/memmap.go +++ b/pkg/sentry/memmap/memmap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memutil/memutil.go b/pkg/sentry/memutil/memutil.go index 4f245cf3c..286d50ca4 100644 --- a/pkg/sentry/memutil/memutil.go +++ b/pkg/sentry/memutil/memutil.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memutil/memutil_unsafe.go b/pkg/sentry/memutil/memutil_unsafe.go index 32c27eb2f..8d9fc64fb 100644 --- a/pkg/sentry/memutil/memutil_unsafe.go +++ b/pkg/sentry/memutil/memutil_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/address_space.go b/pkg/sentry/mm/address_space.go index 27554f163..7488f7c4a 100644 --- a/pkg/sentry/mm/address_space.go +++ b/pkg/sentry/mm/address_space.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index b42156d45..87942af0e 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/aio_context_state.go b/pkg/sentry/mm/aio_context_state.go index 1a5e56f8e..192a6f744 100644 --- a/pkg/sentry/mm/aio_context_state.go +++ b/pkg/sentry/mm/aio_context_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/debug.go b/pkg/sentry/mm/debug.go index 56d0490f0..d341b9c07 100644 --- a/pkg/sentry/mm/debug.go +++ b/pkg/sentry/mm/debug.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/io.go b/pkg/sentry/mm/io.go index 6741db594..6600ddd78 100644 --- a/pkg/sentry/mm/io.go +++ b/pkg/sentry/mm/io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/lifecycle.go b/pkg/sentry/mm/lifecycle.go index a4b5cb443..b248b76e7 100644 --- a/pkg/sentry/mm/lifecycle.go +++ b/pkg/sentry/mm/lifecycle.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/metadata.go b/pkg/sentry/mm/metadata.go index 32d5e2ff6..5ef1ba0b1 100644 --- a/pkg/sentry/mm/metadata.go +++ b/pkg/sentry/mm/metadata.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index 3299ae164..aab697f9e 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/mm_test.go b/pkg/sentry/mm/mm_test.go index b47aa7263..f2db43196 100644 --- a/pkg/sentry/mm/mm_test.go +++ b/pkg/sentry/mm/mm_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/pma.go b/pkg/sentry/mm/pma.go index 9febb25ac..5690fe6b4 100644 --- a/pkg/sentry/mm/pma.go +++ b/pkg/sentry/mm/pma.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/proc_pid_maps.go b/pkg/sentry/mm/proc_pid_maps.go index 5840b257c..0bf1cdb51 100644 --- a/pkg/sentry/mm/proc_pid_maps.go +++ b/pkg/sentry/mm/proc_pid_maps.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/save_restore.go b/pkg/sentry/mm/save_restore.go index 36fed8f1c..6e7080a84 100644 --- a/pkg/sentry/mm/save_restore.go +++ b/pkg/sentry/mm/save_restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/shm.go b/pkg/sentry/mm/shm.go index bab137a5a..3bc48c7e7 100644 --- a/pkg/sentry/mm/shm.go +++ b/pkg/sentry/mm/shm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index 5d7bd33bd..e511472f4 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index b0622b0c3..a721cc456 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/vma.go b/pkg/sentry/mm/vma.go index b81e861f1..dafdbd0e4 100644 --- a/pkg/sentry/mm/vma.go +++ b/pkg/sentry/mm/vma.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/context.go b/pkg/sentry/platform/context.go index 0d200a5e2..cca21a23e 100644 --- a/pkg/sentry/platform/context.go +++ b/pkg/sentry/platform/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/filemem/filemem.go b/pkg/sentry/platform/filemem/filemem.go index f278c8d63..97da31e70 100644 --- a/pkg/sentry/platform/filemem/filemem.go +++ b/pkg/sentry/platform/filemem/filemem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/filemem/filemem_state.go b/pkg/sentry/platform/filemem/filemem_state.go index e28e021c9..964e2aaaa 100644 --- a/pkg/sentry/platform/filemem/filemem_state.go +++ b/pkg/sentry/platform/filemem/filemem_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/filemem/filemem_test.go b/pkg/sentry/platform/filemem/filemem_test.go index 4b165dc48..9becec25f 100644 --- a/pkg/sentry/platform/filemem/filemem_test.go +++ b/pkg/sentry/platform/filemem/filemem_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/filemem/filemem_unsafe.go b/pkg/sentry/platform/filemem/filemem_unsafe.go index a23b9825a..776aed74d 100644 --- a/pkg/sentry/platform/filemem/filemem_unsafe.go +++ b/pkg/sentry/platform/filemem/filemem_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/interrupt/interrupt.go b/pkg/sentry/platform/interrupt/interrupt.go index ca4f42087..9c83f41eb 100644 --- a/pkg/sentry/platform/interrupt/interrupt.go +++ b/pkg/sentry/platform/interrupt/interrupt.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/interrupt/interrupt_test.go b/pkg/sentry/platform/interrupt/interrupt_test.go index 7c49eeea6..fb3284395 100644 --- a/pkg/sentry/platform/interrupt/interrupt_test.go +++ b/pkg/sentry/platform/interrupt/interrupt_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/address_space.go b/pkg/sentry/platform/kvm/address_space.go index c4293c517..72e897a9a 100644 --- a/pkg/sentry/platform/kvm/address_space.go +++ b/pkg/sentry/platform/kvm/address_space.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/allocator.go b/pkg/sentry/platform/kvm/allocator.go index f5cebd5b3..b25cad155 100644 --- a/pkg/sentry/platform/kvm/allocator.go +++ b/pkg/sentry/platform/kvm/allocator.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill.go b/pkg/sentry/platform/kvm/bluepill.go index ecc33d7dd..9f1c9510b 100644 --- a/pkg/sentry/platform/kvm/bluepill.go +++ b/pkg/sentry/platform/kvm/bluepill.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_amd64.go b/pkg/sentry/platform/kvm/bluepill_amd64.go index b364e3ef7..f013d1dc9 100644 --- a/pkg/sentry/platform/kvm/bluepill_amd64.go +++ b/pkg/sentry/platform/kvm/bluepill_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_amd64.s b/pkg/sentry/platform/kvm/bluepill_amd64.s index 0881bd5f5..ec017f6c2 100644 --- a/pkg/sentry/platform/kvm/bluepill_amd64.s +++ b/pkg/sentry/platform/kvm/bluepill_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go b/pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go index 61ca61dcb..cd00a47f2 100644 --- a/pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go +++ b/pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_fault.go b/pkg/sentry/platform/kvm/bluepill_fault.go index 8650cd78f..e79a30ef2 100644 --- a/pkg/sentry/platform/kvm/bluepill_fault.go +++ b/pkg/sentry/platform/kvm/bluepill_fault.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_unsafe.go b/pkg/sentry/platform/kvm/bluepill_unsafe.go index 216d4b4b6..747a95997 100644 --- a/pkg/sentry/platform/kvm/bluepill_unsafe.go +++ b/pkg/sentry/platform/kvm/bluepill_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/context.go b/pkg/sentry/platform/kvm/context.go index aac84febf..be902be88 100644 --- a/pkg/sentry/platform/kvm/context.go +++ b/pkg/sentry/platform/kvm/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/host_map.go b/pkg/sentry/platform/kvm/host_map.go index fc16ad2de..ee6a1a42d 100644 --- a/pkg/sentry/platform/kvm/host_map.go +++ b/pkg/sentry/platform/kvm/host_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm.go b/pkg/sentry/platform/kvm/kvm.go index 0c4dff308..d4f50024d 100644 --- a/pkg/sentry/platform/kvm/kvm.go +++ b/pkg/sentry/platform/kvm/kvm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm_amd64.go b/pkg/sentry/platform/kvm/kvm_amd64.go index 3d56ed895..70d0ac63b 100644 --- a/pkg/sentry/platform/kvm/kvm_amd64.go +++ b/pkg/sentry/platform/kvm/kvm_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go b/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go index 476e783a0..c0a0af92d 100644 --- a/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go +++ b/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm_const.go b/pkg/sentry/platform/kvm/kvm_const.go index ca44c31b3..8c53c6f06 100644 --- a/pkg/sentry/platform/kvm/kvm_const.go +++ b/pkg/sentry/platform/kvm/kvm_const.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm_test.go b/pkg/sentry/platform/kvm/kvm_test.go index 52448839f..45eeb96ff 100644 --- a/pkg/sentry/platform/kvm/kvm_test.go +++ b/pkg/sentry/platform/kvm/kvm_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/machine.go b/pkg/sentry/platform/kvm/machine.go index 9f60b6b31..fc7ad258f 100644 --- a/pkg/sentry/platform/kvm/machine.go +++ b/pkg/sentry/platform/kvm/machine.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/machine_amd64.go b/pkg/sentry/platform/kvm/machine_amd64.go index bcd29a947..e0aec42b8 100644 --- a/pkg/sentry/platform/kvm/machine_amd64.go +++ b/pkg/sentry/platform/kvm/machine_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/machine_amd64_unsafe.go b/pkg/sentry/platform/kvm/machine_amd64_unsafe.go index 8b9041f13..50e513f3b 100644 --- a/pkg/sentry/platform/kvm/machine_amd64_unsafe.go +++ b/pkg/sentry/platform/kvm/machine_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/machine_unsafe.go b/pkg/sentry/platform/kvm/machine_unsafe.go index 86323c891..4f5b01321 100644 --- a/pkg/sentry/platform/kvm/machine_unsafe.go +++ b/pkg/sentry/platform/kvm/machine_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/physical_map.go b/pkg/sentry/platform/kvm/physical_map.go index 81a98656d..b908cae6a 100644 --- a/pkg/sentry/platform/kvm/physical_map.go +++ b/pkg/sentry/platform/kvm/physical_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/testutil/testutil.go b/pkg/sentry/platform/kvm/testutil/testutil.go index 8a614e25d..0d496561d 100644 --- a/pkg/sentry/platform/kvm/testutil/testutil.go +++ b/pkg/sentry/platform/kvm/testutil/testutil.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/testutil/testutil_amd64.go b/pkg/sentry/platform/kvm/testutil/testutil_amd64.go index 39286a0af..fcba33813 100644 --- a/pkg/sentry/platform/kvm/testutil/testutil_amd64.go +++ b/pkg/sentry/platform/kvm/testutil/testutil_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/testutil/testutil_amd64.s b/pkg/sentry/platform/kvm/testutil/testutil_amd64.s index 3b5ad8817..f1da41a44 100644 --- a/pkg/sentry/platform/kvm/testutil/testutil_amd64.s +++ b/pkg/sentry/platform/kvm/testutil/testutil_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/virtual_map.go b/pkg/sentry/platform/kvm/virtual_map.go index 0d3fbe043..0343e9267 100644 --- a/pkg/sentry/platform/kvm/virtual_map.go +++ b/pkg/sentry/platform/kvm/virtual_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/virtual_map_test.go b/pkg/sentry/platform/kvm/virtual_map_test.go index 7875bd3e9..935e0eb93 100644 --- a/pkg/sentry/platform/kvm/virtual_map_test.go +++ b/pkg/sentry/platform/kvm/virtual_map_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/mmap_min_addr.go b/pkg/sentry/platform/mmap_min_addr.go index 6398e5e01..1bcc1f8e9 100644 --- a/pkg/sentry/platform/mmap_min_addr.go +++ b/pkg/sentry/platform/mmap_min_addr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/platform.go b/pkg/sentry/platform/platform.go index 8a1620d93..f16588e6e 100644 --- a/pkg/sentry/platform/platform.go +++ b/pkg/sentry/platform/platform.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid.go b/pkg/sentry/platform/procid/procid.go index 5f861908f..3f49ab093 100644 --- a/pkg/sentry/platform/procid/procid.go +++ b/pkg/sentry/platform/procid/procid.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid_amd64.s b/pkg/sentry/platform/procid/procid_amd64.s index 5b1ba1f24..fd88ce82e 100644 --- a/pkg/sentry/platform/procid/procid_amd64.s +++ b/pkg/sentry/platform/procid/procid_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid_net_test.go b/pkg/sentry/platform/procid/procid_net_test.go index 2d1605a08..e8dcc479d 100644 --- a/pkg/sentry/platform/procid/procid_net_test.go +++ b/pkg/sentry/platform/procid/procid_net_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid_test.go b/pkg/sentry/platform/procid/procid_test.go index 5e44da36f..7a57c7cdc 100644 --- a/pkg/sentry/platform/procid/procid_test.go +++ b/pkg/sentry/platform/procid/procid_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/ptrace.go b/pkg/sentry/platform/ptrace/ptrace.go index 4f20716f7..00d92b092 100644 --- a/pkg/sentry/platform/ptrace/ptrace.go +++ b/pkg/sentry/platform/ptrace/ptrace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/ptrace_unsafe.go b/pkg/sentry/platform/ptrace/ptrace_unsafe.go index 46a8bda8e..7a3cb8f49 100644 --- a/pkg/sentry/platform/ptrace/ptrace_unsafe.go +++ b/pkg/sentry/platform/ptrace/ptrace_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/stub_amd64.s b/pkg/sentry/platform/ptrace/stub_amd64.s index 9bf87b6f6..63f98e40d 100644 --- a/pkg/sentry/platform/ptrace/stub_amd64.s +++ b/pkg/sentry/platform/ptrace/stub_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/stub_unsafe.go b/pkg/sentry/platform/ptrace/stub_unsafe.go index c868a2d68..48c16c4a1 100644 --- a/pkg/sentry/platform/ptrace/stub_unsafe.go +++ b/pkg/sentry/platform/ptrace/stub_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess.go b/pkg/sentry/platform/ptrace/subprocess.go index 6d5ad6b71..6a9da5db8 100644 --- a/pkg/sentry/platform/ptrace/subprocess.go +++ b/pkg/sentry/platform/ptrace/subprocess.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess_amd64.go b/pkg/sentry/platform/ptrace/subprocess_amd64.go index c38dc1ff8..d23a1133e 100644 --- a/pkg/sentry/platform/ptrace/subprocess_amd64.go +++ b/pkg/sentry/platform/ptrace/subprocess_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess_linux.go b/pkg/sentry/platform/ptrace/subprocess_linux.go index 53adadadd..7523487e7 100644 --- a/pkg/sentry/platform/ptrace/subprocess_linux.go +++ b/pkg/sentry/platform/ptrace/subprocess_linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go b/pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go index 697431472..0c9263060 100644 --- a/pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go +++ b/pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess_unsafe.go b/pkg/sentry/platform/ptrace/subprocess_unsafe.go index fe41641d3..ca6c4ac97 100644 --- a/pkg/sentry/platform/ptrace/subprocess_unsafe.go +++ b/pkg/sentry/platform/ptrace/subprocess_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/defs.go b/pkg/sentry/platform/ring0/defs.go index f09d045eb..18137e55d 100644 --- a/pkg/sentry/platform/ring0/defs.go +++ b/pkg/sentry/platform/ring0/defs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/defs_amd64.go b/pkg/sentry/platform/ring0/defs_amd64.go index 84819f132..67242b92b 100644 --- a/pkg/sentry/platform/ring0/defs_amd64.go +++ b/pkg/sentry/platform/ring0/defs_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/entry_amd64.go b/pkg/sentry/platform/ring0/entry_amd64.go index a3e992e0d..4a9affe64 100644 --- a/pkg/sentry/platform/ring0/entry_amd64.go +++ b/pkg/sentry/platform/ring0/entry_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/entry_amd64.s b/pkg/sentry/platform/ring0/entry_amd64.s index 08c15ad65..d48fbd2d1 100644 --- a/pkg/sentry/platform/ring0/entry_amd64.s +++ b/pkg/sentry/platform/ring0/entry_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/gen_offsets/main.go b/pkg/sentry/platform/ring0/gen_offsets/main.go index ffa7eaf77..11c49855f 100644 --- a/pkg/sentry/platform/ring0/gen_offsets/main.go +++ b/pkg/sentry/platform/ring0/gen_offsets/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/kernel.go b/pkg/sentry/platform/ring0/kernel.go index 62e67005e..e70eafde2 100644 --- a/pkg/sentry/platform/ring0/kernel.go +++ b/pkg/sentry/platform/ring0/kernel.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/kernel_amd64.go b/pkg/sentry/platform/ring0/kernel_amd64.go index 0d2b0f7dc..ab562bca7 100644 --- a/pkg/sentry/platform/ring0/kernel_amd64.go +++ b/pkg/sentry/platform/ring0/kernel_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/kernel_unsafe.go b/pkg/sentry/platform/ring0/kernel_unsafe.go index cfb3ad853..faf4240e5 100644 --- a/pkg/sentry/platform/ring0/kernel_unsafe.go +++ b/pkg/sentry/platform/ring0/kernel_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/lib_amd64.go b/pkg/sentry/platform/ring0/lib_amd64.go index 989e3e383..2b95a0141 100644 --- a/pkg/sentry/platform/ring0/lib_amd64.go +++ b/pkg/sentry/platform/ring0/lib_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/lib_amd64.s b/pkg/sentry/platform/ring0/lib_amd64.s index 6f143ea5a..98a130525 100644 --- a/pkg/sentry/platform/ring0/lib_amd64.s +++ b/pkg/sentry/platform/ring0/lib_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/offsets_amd64.go b/pkg/sentry/platform/ring0/offsets_amd64.go index ca5fd456b..753d31ef8 100644 --- a/pkg/sentry/platform/ring0/offsets_amd64.go +++ b/pkg/sentry/platform/ring0/offsets_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/allocator.go b/pkg/sentry/platform/ring0/pagetables/allocator.go index 049fd0247..ee6e90a11 100644 --- a/pkg/sentry/platform/ring0/pagetables/allocator.go +++ b/pkg/sentry/platform/ring0/pagetables/allocator.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go b/pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go index aca778913..f48647b3a 100644 --- a/pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go +++ b/pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables.go b/pkg/sentry/platform/ring0/pagetables/pagetables.go index ff5787f89..c7207ec18 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go index 878463018..746f614e5 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go index a7f2ad9a4..2f82c4353 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_test.go b/pkg/sentry/platform/ring0/pagetables/pagetables_test.go index dca3f69ef..3e5dc7dc7 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_test.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_x86.go b/pkg/sentry/platform/ring0/pagetables/pagetables_x86.go index ca49d20f8..6bd8c3584 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_x86.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pcids_x86.go b/pkg/sentry/platform/ring0/pagetables/pcids_x86.go index fa068e35e..0d9a51aa5 100644 --- a/pkg/sentry/platform/ring0/pagetables/pcids_x86.go +++ b/pkg/sentry/platform/ring0/pagetables/pcids_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/walker_amd64.go b/pkg/sentry/platform/ring0/pagetables/walker_amd64.go index afa4d473a..c4c71d23e 100644 --- a/pkg/sentry/platform/ring0/pagetables/walker_amd64.go +++ b/pkg/sentry/platform/ring0/pagetables/walker_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/ring0.go b/pkg/sentry/platform/ring0/ring0.go index 4991031c5..10c51e88d 100644 --- a/pkg/sentry/platform/ring0/ring0.go +++ b/pkg/sentry/platform/ring0/ring0.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/x86.go b/pkg/sentry/platform/ring0/x86.go index f489fcecb..7c88010d8 100644 --- a/pkg/sentry/platform/ring0/x86.go +++ b/pkg/sentry/platform/ring0/x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/atomic_amd64.s b/pkg/sentry/platform/safecopy/atomic_amd64.s index 69947dec3..873ffa046 100644 --- a/pkg/sentry/platform/safecopy/atomic_amd64.s +++ b/pkg/sentry/platform/safecopy/atomic_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/memclr_amd64.s b/pkg/sentry/platform/safecopy/memclr_amd64.s index 7d1019f60..488b6e666 100644 --- a/pkg/sentry/platform/safecopy/memclr_amd64.s +++ b/pkg/sentry/platform/safecopy/memclr_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/memcpy_amd64.s b/pkg/sentry/platform/safecopy/memcpy_amd64.s index 96ef2eefc..0bf26fd7b 100644 --- a/pkg/sentry/platform/safecopy/memcpy_amd64.s +++ b/pkg/sentry/platform/safecopy/memcpy_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/safecopy.go b/pkg/sentry/platform/safecopy/safecopy.go index 90a2aad7b..c60f73103 100644 --- a/pkg/sentry/platform/safecopy/safecopy.go +++ b/pkg/sentry/platform/safecopy/safecopy.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/safecopy_test.go b/pkg/sentry/platform/safecopy/safecopy_test.go index 67df36121..1a682d28a 100644 --- a/pkg/sentry/platform/safecopy/safecopy_test.go +++ b/pkg/sentry/platform/safecopy/safecopy_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/safecopy_unsafe.go b/pkg/sentry/platform/safecopy/safecopy_unsafe.go index 72f243f8d..df1c35b66 100644 --- a/pkg/sentry/platform/safecopy/safecopy_unsafe.go +++ b/pkg/sentry/platform/safecopy/safecopy_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/sighandler_amd64.s b/pkg/sentry/platform/safecopy/sighandler_amd64.s index a65cb0c26..06614f1b4 100644 --- a/pkg/sentry/platform/safecopy/sighandler_amd64.s +++ b/pkg/sentry/platform/safecopy/sighandler_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/block_unsafe.go b/pkg/sentry/safemem/block_unsafe.go index 0b58f6497..e91ff66ae 100644 --- a/pkg/sentry/safemem/block_unsafe.go +++ b/pkg/sentry/safemem/block_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/io.go b/pkg/sentry/safemem/io.go index fd917648b..6cb52439f 100644 --- a/pkg/sentry/safemem/io.go +++ b/pkg/sentry/safemem/io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/io_test.go b/pkg/sentry/safemem/io_test.go index edac4c1d7..2eda8c3bb 100644 --- a/pkg/sentry/safemem/io_test.go +++ b/pkg/sentry/safemem/io_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/safemem.go b/pkg/sentry/safemem/safemem.go index 2f8002004..090932d3e 100644 --- a/pkg/sentry/safemem/safemem.go +++ b/pkg/sentry/safemem/safemem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/seq_test.go b/pkg/sentry/safemem/seq_test.go index 3e83b3851..fddcaf714 100644 --- a/pkg/sentry/safemem/seq_test.go +++ b/pkg/sentry/safemem/seq_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/seq_unsafe.go b/pkg/sentry/safemem/seq_unsafe.go index e0d29a0b3..83a6b7183 100644 --- a/pkg/sentry/safemem/seq_unsafe.go +++ b/pkg/sentry/safemem/seq_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/sighandling/sighandling.go b/pkg/sentry/sighandling/sighandling.go index 29bcf55ab..6b5d5f993 100644 --- a/pkg/sentry/sighandling/sighandling.go +++ b/pkg/sentry/sighandling/sighandling.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/sighandling/sighandling_unsafe.go b/pkg/sentry/sighandling/sighandling_unsafe.go index a455b919f..5913d47a8 100644 --- a/pkg/sentry/sighandling/sighandling_unsafe.go +++ b/pkg/sentry/sighandling/sighandling_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/control/control.go b/pkg/sentry/socket/control/control.go index db97e95f2..d44f5e88a 100644 --- a/pkg/sentry/socket/control/control.go +++ b/pkg/sentry/socket/control/control.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/device.go b/pkg/sentry/socket/epsocket/device.go index 17f2c9559..3cc138eb0 100644 --- a/pkg/sentry/socket/epsocket/device.go +++ b/pkg/sentry/socket/epsocket/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index 47c575e7b..e90ef4835 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/provider.go b/pkg/sentry/socket/epsocket/provider.go index dbc232d26..686554437 100644 --- a/pkg/sentry/socket/epsocket/provider.go +++ b/pkg/sentry/socket/epsocket/provider.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/save_restore.go b/pkg/sentry/socket/epsocket/save_restore.go index 2613f90de..34d9a7cf0 100644 --- a/pkg/sentry/socket/epsocket/save_restore.go +++ b/pkg/sentry/socket/epsocket/save_restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/stack.go b/pkg/sentry/socket/epsocket/stack.go index e4ed52fc8..c0081c819 100644 --- a/pkg/sentry/socket/epsocket/stack.go +++ b/pkg/sentry/socket/epsocket/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/device.go b/pkg/sentry/socket/hostinet/device.go index a9a673316..c5133f3bb 100644 --- a/pkg/sentry/socket/hostinet/device.go +++ b/pkg/sentry/socket/hostinet/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/hostinet.go b/pkg/sentry/socket/hostinet/hostinet.go index 67c6c8066..7858892ab 100644 --- a/pkg/sentry/socket/hostinet/hostinet.go +++ b/pkg/sentry/socket/hostinet/hostinet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/save_restore.go b/pkg/sentry/socket/hostinet/save_restore.go index 0821a794a..3827f082a 100644 --- a/pkg/sentry/socket/hostinet/save_restore.go +++ b/pkg/sentry/socket/hostinet/save_restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index e82624b44..e4e950fbb 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/socket_unsafe.go b/pkg/sentry/socket/hostinet/socket_unsafe.go index f8bb75636..59c8910ca 100644 --- a/pkg/sentry/socket/hostinet/socket_unsafe.go +++ b/pkg/sentry/socket/hostinet/socket_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/stack.go b/pkg/sentry/socket/hostinet/stack.go index f64809d39..4ce73c1f1 100644 --- a/pkg/sentry/socket/hostinet/stack.go +++ b/pkg/sentry/socket/hostinet/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/message.go b/pkg/sentry/socket/netlink/message.go index b902d7ec9..a95172cba 100644 --- a/pkg/sentry/socket/netlink/message.go +++ b/pkg/sentry/socket/netlink/message.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/port/port.go b/pkg/sentry/socket/netlink/port/port.go index 1c5d4c3a5..20b9a6e37 100644 --- a/pkg/sentry/socket/netlink/port/port.go +++ b/pkg/sentry/socket/netlink/port/port.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/port/port_test.go b/pkg/sentry/socket/netlink/port/port_test.go index 34565e2f9..49b3b48ab 100644 --- a/pkg/sentry/socket/netlink/port/port_test.go +++ b/pkg/sentry/socket/netlink/port/port_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/provider.go b/pkg/sentry/socket/netlink/provider.go index 5d0a04a07..06786bd50 100644 --- a/pkg/sentry/socket/netlink/provider.go +++ b/pkg/sentry/socket/netlink/provider.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/route/protocol.go b/pkg/sentry/socket/netlink/route/protocol.go index 70322b9ed..7e70b09b2 100644 --- a/pkg/sentry/socket/netlink/route/protocol.go +++ b/pkg/sentry/socket/netlink/route/protocol.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index 0c03997f2..4d4130a4c 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/conn/conn.go b/pkg/sentry/socket/rpcinet/conn/conn.go index f4c8489b1..9c749b888 100644 --- a/pkg/sentry/socket/rpcinet/conn/conn.go +++ b/pkg/sentry/socket/rpcinet/conn/conn.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/device.go b/pkg/sentry/socket/rpcinet/device.go index f7b63436e..d2b9f9222 100644 --- a/pkg/sentry/socket/rpcinet/device.go +++ b/pkg/sentry/socket/rpcinet/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/notifier/notifier.go b/pkg/sentry/socket/rpcinet/notifier/notifier.go index f88a908ed..73c255c33 100644 --- a/pkg/sentry/socket/rpcinet/notifier/notifier.go +++ b/pkg/sentry/socket/rpcinet/notifier/notifier.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/rpcinet.go b/pkg/sentry/socket/rpcinet/rpcinet.go index 10b0dedc2..6c98e6acb 100644 --- a/pkg/sentry/socket/rpcinet/rpcinet.go +++ b/pkg/sentry/socket/rpcinet/rpcinet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index c7e761d54..44fa5c620 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/stack.go b/pkg/sentry/socket/rpcinet/stack.go index bcb89fb34..cb8344ec6 100644 --- a/pkg/sentry/socket/rpcinet/stack.go +++ b/pkg/sentry/socket/rpcinet/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/stack_unsafe.go b/pkg/sentry/socket/rpcinet/stack_unsafe.go index 9a896c623..d04fb2069 100644 --- a/pkg/sentry/socket/rpcinet/stack_unsafe.go +++ b/pkg/sentry/socket/rpcinet/stack_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index 31f8d42d7..a235c5249 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/device.go b/pkg/sentry/socket/unix/device.go index e8bcc7a9f..41820dbb3 100644 --- a/pkg/sentry/socket/unix/device.go +++ b/pkg/sentry/socket/unix/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/io.go b/pkg/sentry/socket/unix/io.go index 06333e14b..7d6434696 100644 --- a/pkg/sentry/socket/unix/io.go +++ b/pkg/sentry/socket/unix/io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/connectioned.go b/pkg/sentry/socket/unix/transport/connectioned.go index 566e3d57b..4c913effc 100644 --- a/pkg/sentry/socket/unix/transport/connectioned.go +++ b/pkg/sentry/socket/unix/transport/connectioned.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/connectioned_state.go b/pkg/sentry/socket/unix/transport/connectioned_state.go index 7e6c73dcc..608a6a97a 100644 --- a/pkg/sentry/socket/unix/transport/connectioned_state.go +++ b/pkg/sentry/socket/unix/transport/connectioned_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/connectionless.go b/pkg/sentry/socket/unix/transport/connectionless.go index 86cd05199..cd4633106 100644 --- a/pkg/sentry/socket/unix/transport/connectionless.go +++ b/pkg/sentry/socket/unix/transport/connectionless.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/queue.go b/pkg/sentry/socket/unix/transport/queue.go index c4d7d863c..5b4dfab68 100644 --- a/pkg/sentry/socket/unix/transport/queue.go +++ b/pkg/sentry/socket/unix/transport/queue.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/unix.go b/pkg/sentry/socket/unix/transport/unix.go index 2934101a2..157133b65 100644 --- a/pkg/sentry/socket/unix/transport/unix.go +++ b/pkg/sentry/socket/unix/transport/unix.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 668363864..3543dd81f 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/state/state.go b/pkg/sentry/state/state.go index 43e88a713..70b33f190 100644 --- a/pkg/sentry/state/state.go +++ b/pkg/sentry/state/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/state/state_metadata.go b/pkg/sentry/state/state_metadata.go index afa21672a..7f047b808 100644 --- a/pkg/sentry/state/state_metadata.go +++ b/pkg/sentry/state/state_metadata.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/state/state_unsafe.go b/pkg/sentry/state/state_unsafe.go index 3ff7d24c8..f02e12b2a 100644 --- a/pkg/sentry/state/state_unsafe.go +++ b/pkg/sentry/state/state_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/clone.go b/pkg/sentry/strace/clone.go index b82ca1ad1..e18ce84dc 100644 --- a/pkg/sentry/strace/clone.go +++ b/pkg/sentry/strace/clone.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/futex.go b/pkg/sentry/strace/futex.go index 3da108cb7..ceb3dc21d 100644 --- a/pkg/sentry/strace/futex.go +++ b/pkg/sentry/strace/futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/linux64.go b/pkg/sentry/strace/linux64.go index 1df148e7d..99714f12c 100644 --- a/pkg/sentry/strace/linux64.go +++ b/pkg/sentry/strace/linux64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/open.go b/pkg/sentry/strace/open.go index 839d5eda7..5a72a940c 100644 --- a/pkg/sentry/strace/open.go +++ b/pkg/sentry/strace/open.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/ptrace.go b/pkg/sentry/strace/ptrace.go index fcdb7e9f4..c572aafb4 100644 --- a/pkg/sentry/strace/ptrace.go +++ b/pkg/sentry/strace/ptrace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/socket.go b/pkg/sentry/strace/socket.go index 26831edd6..375418dc1 100644 --- a/pkg/sentry/strace/socket.go +++ b/pkg/sentry/strace/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/strace.go b/pkg/sentry/strace/strace.go index f7bfa3a1f..4286f0df7 100644 --- a/pkg/sentry/strace/strace.go +++ b/pkg/sentry/strace/strace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/strace.proto b/pkg/sentry/strace/strace.proto index 914e8c7b0..f1fc539d6 100644 --- a/pkg/sentry/strace/strace.proto +++ b/pkg/sentry/strace/strace.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/syscalls.go b/pkg/sentry/strace/syscalls.go index 8be4fa318..9eeb18a03 100644 --- a/pkg/sentry/strace/syscalls.go +++ b/pkg/sentry/strace/syscalls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/epoll.go b/pkg/sentry/syscalls/epoll.go index 01dd6fa71..b90d191b7 100644 --- a/pkg/sentry/syscalls/epoll.go +++ b/pkg/sentry/syscalls/epoll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/error.go b/pkg/sentry/syscalls/linux/error.go index 013b385bc..9fd002955 100644 --- a/pkg/sentry/syscalls/linux/error.go +++ b/pkg/sentry/syscalls/linux/error.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/flags.go b/pkg/sentry/syscalls/linux/flags.go index f01483cd3..d1e0833fc 100644 --- a/pkg/sentry/syscalls/linux/flags.go +++ b/pkg/sentry/syscalls/linux/flags.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index 4465549ad..75e87f5ec 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sigset.go b/pkg/sentry/syscalls/linux/sigset.go index bfb541634..a033b7c70 100644 --- a/pkg/sentry/syscalls/linux/sigset.go +++ b/pkg/sentry/syscalls/linux/sigset.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_aio.go b/pkg/sentry/syscalls/linux/sys_aio.go index 54e4afa9e..355071131 100644 --- a/pkg/sentry/syscalls/linux/sys_aio.go +++ b/pkg/sentry/syscalls/linux/sys_aio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_capability.go b/pkg/sentry/syscalls/linux/sys_capability.go index 89c81ac90..cf972dc28 100644 --- a/pkg/sentry/syscalls/linux/sys_capability.go +++ b/pkg/sentry/syscalls/linux/sys_capability.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_epoll.go b/pkg/sentry/syscalls/linux/sys_epoll.go index e69dfc77a..62272efcd 100644 --- a/pkg/sentry/syscalls/linux/sys_epoll.go +++ b/pkg/sentry/syscalls/linux/sys_epoll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_eventfd.go b/pkg/sentry/syscalls/linux/sys_eventfd.go index 60fe5a133..903172890 100644 --- a/pkg/sentry/syscalls/linux/sys_eventfd.go +++ b/pkg/sentry/syscalls/linux/sys_eventfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 64704bb88..a70f35be0 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index d35dcecbe..cf04428bc 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_getdents.go b/pkg/sentry/syscalls/linux/sys_getdents.go index 29c0d7a39..4b441b31b 100644 --- a/pkg/sentry/syscalls/linux/sys_getdents.go +++ b/pkg/sentry/syscalls/linux/sys_getdents.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_identity.go b/pkg/sentry/syscalls/linux/sys_identity.go index 4fd0ed794..8d594aa83 100644 --- a/pkg/sentry/syscalls/linux/sys_identity.go +++ b/pkg/sentry/syscalls/linux/sys_identity.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_inotify.go b/pkg/sentry/syscalls/linux/sys_inotify.go index 725204dff..26a505782 100644 --- a/pkg/sentry/syscalls/linux/sys_inotify.go +++ b/pkg/sentry/syscalls/linux/sys_inotify.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_lseek.go b/pkg/sentry/syscalls/linux/sys_lseek.go index 97b51ba7c..ad3bfd761 100644 --- a/pkg/sentry/syscalls/linux/sys_lseek.go +++ b/pkg/sentry/syscalls/linux/sys_lseek.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_mmap.go b/pkg/sentry/syscalls/linux/sys_mmap.go index 1a98328dc..f8d9c43fd 100644 --- a/pkg/sentry/syscalls/linux/sys_mmap.go +++ b/pkg/sentry/syscalls/linux/sys_mmap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_mount.go b/pkg/sentry/syscalls/linux/sys_mount.go index 57cedccc1..bf0df7302 100644 --- a/pkg/sentry/syscalls/linux/sys_mount.go +++ b/pkg/sentry/syscalls/linux/sys_mount.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_pipe.go b/pkg/sentry/syscalls/linux/sys_pipe.go index 2b544f145..3652c429e 100644 --- a/pkg/sentry/syscalls/linux/sys_pipe.go +++ b/pkg/sentry/syscalls/linux/sys_pipe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_poll.go b/pkg/sentry/syscalls/linux/sys_poll.go index b9bdefadb..bf0958435 100644 --- a/pkg/sentry/syscalls/linux/sys_poll.go +++ b/pkg/sentry/syscalls/linux/sys_poll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_prctl.go b/pkg/sentry/syscalls/linux/sys_prctl.go index a1242acd3..c7b39ede8 100644 --- a/pkg/sentry/syscalls/linux/sys_prctl.go +++ b/pkg/sentry/syscalls/linux/sys_prctl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_random.go b/pkg/sentry/syscalls/linux/sys_random.go index be31e6b17..452dff058 100644 --- a/pkg/sentry/syscalls/linux/sys_random.go +++ b/pkg/sentry/syscalls/linux/sys_random.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_read.go b/pkg/sentry/syscalls/linux/sys_read.go index 0be2d195a..b2e5a5449 100644 --- a/pkg/sentry/syscalls/linux/sys_read.go +++ b/pkg/sentry/syscalls/linux/sys_read.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_rlimit.go b/pkg/sentry/syscalls/linux/sys_rlimit.go index d806b58ab..2f16e1791 100644 --- a/pkg/sentry/syscalls/linux/sys_rlimit.go +++ b/pkg/sentry/syscalls/linux/sys_rlimit.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_rusage.go b/pkg/sentry/syscalls/linux/sys_rusage.go index 82e42b589..ab07c77f9 100644 --- a/pkg/sentry/syscalls/linux/sys_rusage.go +++ b/pkg/sentry/syscalls/linux/sys_rusage.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_sched.go b/pkg/sentry/syscalls/linux/sys_sched.go index ff9e46077..e679a6694 100644 --- a/pkg/sentry/syscalls/linux/sys_sched.go +++ b/pkg/sentry/syscalls/linux/sys_sched.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_seccomp.go b/pkg/sentry/syscalls/linux/sys_seccomp.go index 4323a4df4..969acaa36 100644 --- a/pkg/sentry/syscalls/linux/sys_seccomp.go +++ b/pkg/sentry/syscalls/linux/sys_seccomp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_sem.go b/pkg/sentry/syscalls/linux/sys_sem.go index a8983705b..4ed52c4a7 100644 --- a/pkg/sentry/syscalls/linux/sys_sem.go +++ b/pkg/sentry/syscalls/linux/sys_sem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_shm.go b/pkg/sentry/syscalls/linux/sys_shm.go index 48ff1d5f0..b13d48b98 100644 --- a/pkg/sentry/syscalls/linux/sys_shm.go +++ b/pkg/sentry/syscalls/linux/sys_shm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_signal.go b/pkg/sentry/syscalls/linux/sys_signal.go index ecdec5d3a..a539354c5 100644 --- a/pkg/sentry/syscalls/linux/sys_signal.go +++ b/pkg/sentry/syscalls/linux/sys_signal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go index 5fa5ddce6..0a7551742 100644 --- a/pkg/sentry/syscalls/linux/sys_socket.go +++ b/pkg/sentry/syscalls/linux/sys_socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_stat.go b/pkg/sentry/syscalls/linux/sys_stat.go index 619a14d7c..9c433c45d 100644 --- a/pkg/sentry/syscalls/linux/sys_stat.go +++ b/pkg/sentry/syscalls/linux/sys_stat.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_sync.go b/pkg/sentry/syscalls/linux/sys_sync.go index 902d210db..826c6869d 100644 --- a/pkg/sentry/syscalls/linux/sys_sync.go +++ b/pkg/sentry/syscalls/linux/sys_sync.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_sysinfo.go b/pkg/sentry/syscalls/linux/sys_sysinfo.go index 6560bac57..5eeb3ba58 100644 --- a/pkg/sentry/syscalls/linux/sys_sysinfo.go +++ b/pkg/sentry/syscalls/linux/sys_sysinfo.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_syslog.go b/pkg/sentry/syscalls/linux/sys_syslog.go index 792040c81..7193b7aed 100644 --- a/pkg/sentry/syscalls/linux/sys_syslog.go +++ b/pkg/sentry/syscalls/linux/sys_syslog.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index 550f63a43..820ca680e 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_time.go b/pkg/sentry/syscalls/linux/sys_time.go index 8e6683444..063fbb106 100644 --- a/pkg/sentry/syscalls/linux/sys_time.go +++ b/pkg/sentry/syscalls/linux/sys_time.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_timer.go b/pkg/sentry/syscalls/linux/sys_timer.go index c41074d54..6baf4599b 100644 --- a/pkg/sentry/syscalls/linux/sys_timer.go +++ b/pkg/sentry/syscalls/linux/sys_timer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_timerfd.go b/pkg/sentry/syscalls/linux/sys_timerfd.go index 92c6a3d60..f70d13682 100644 --- a/pkg/sentry/syscalls/linux/sys_timerfd.go +++ b/pkg/sentry/syscalls/linux/sys_timerfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_tls.go b/pkg/sentry/syscalls/linux/sys_tls.go index b95d62320..27ddb3808 100644 --- a/pkg/sentry/syscalls/linux/sys_tls.go +++ b/pkg/sentry/syscalls/linux/sys_tls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_utsname.go b/pkg/sentry/syscalls/linux/sys_utsname.go index 899116374..689f2f838 100644 --- a/pkg/sentry/syscalls/linux/sys_utsname.go +++ b/pkg/sentry/syscalls/linux/sys_utsname.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_write.go b/pkg/sentry/syscalls/linux/sys_write.go index caa7b01ea..08e263112 100644 --- a/pkg/sentry/syscalls/linux/sys_write.go +++ b/pkg/sentry/syscalls/linux/sys_write.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/timespec.go b/pkg/sentry/syscalls/linux/timespec.go index e865c6fc0..752ec326d 100644 --- a/pkg/sentry/syscalls/linux/timespec.go +++ b/pkg/sentry/syscalls/linux/timespec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/polling.go b/pkg/sentry/syscalls/polling.go index fd90184ef..2b33d6c19 100644 --- a/pkg/sentry/syscalls/polling.go +++ b/pkg/sentry/syscalls/polling.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/syscalls.go b/pkg/sentry/syscalls/syscalls.go index 1176f858d..bae32d727 100644 --- a/pkg/sentry/syscalls/syscalls.go +++ b/pkg/sentry/syscalls/syscalls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/unimplemented_syscall.proto b/pkg/sentry/syscalls/unimplemented_syscall.proto index d6febf5b1..41579b016 100644 --- a/pkg/sentry/syscalls/unimplemented_syscall.proto +++ b/pkg/sentry/syscalls/unimplemented_syscall.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/calibrated_clock.go b/pkg/sentry/time/calibrated_clock.go index cbb95e2d7..c8cf4eca4 100644 --- a/pkg/sentry/time/calibrated_clock.go +++ b/pkg/sentry/time/calibrated_clock.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/calibrated_clock_test.go b/pkg/sentry/time/calibrated_clock_test.go index 8b6dd5592..a9237630e 100644 --- a/pkg/sentry/time/calibrated_clock_test.go +++ b/pkg/sentry/time/calibrated_clock_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/clock_id.go b/pkg/sentry/time/clock_id.go index 500102e58..1317a5dad 100644 --- a/pkg/sentry/time/clock_id.go +++ b/pkg/sentry/time/clock_id.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/clocks.go b/pkg/sentry/time/clocks.go index 9925b407d..e26386520 100644 --- a/pkg/sentry/time/clocks.go +++ b/pkg/sentry/time/clocks.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/muldiv_amd64.s b/pkg/sentry/time/muldiv_amd64.s index 291940b1d..bfcb8c724 100644 --- a/pkg/sentry/time/muldiv_amd64.s +++ b/pkg/sentry/time/muldiv_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/parameters.go b/pkg/sentry/time/parameters.go index 594b4874b..f3ad58454 100644 --- a/pkg/sentry/time/parameters.go +++ b/pkg/sentry/time/parameters.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/parameters_test.go b/pkg/sentry/time/parameters_test.go index 7394fc5ee..4a0c4e880 100644 --- a/pkg/sentry/time/parameters_test.go +++ b/pkg/sentry/time/parameters_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/sampler.go b/pkg/sentry/time/sampler.go index cf581b5fa..445690d49 100644 --- a/pkg/sentry/time/sampler.go +++ b/pkg/sentry/time/sampler.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/sampler_test.go b/pkg/sentry/time/sampler_test.go index caf7e5c53..ec0e442b6 100644 --- a/pkg/sentry/time/sampler_test.go +++ b/pkg/sentry/time/sampler_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/sampler_unsafe.go b/pkg/sentry/time/sampler_unsafe.go index 7ea19d387..0f8eb4fc8 100644 --- a/pkg/sentry/time/sampler_unsafe.go +++ b/pkg/sentry/time/sampler_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/tsc_amd64.s b/pkg/sentry/time/tsc_amd64.s index 4cc604392..e53d477f7 100644 --- a/pkg/sentry/time/tsc_amd64.s +++ b/pkg/sentry/time/tsc_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/uniqueid/context.go b/pkg/sentry/uniqueid/context.go index e48fabc2d..399d98c29 100644 --- a/pkg/sentry/uniqueid/context.go +++ b/pkg/sentry/uniqueid/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/cpu.go b/pkg/sentry/usage/cpu.go index ed7b04b9e..cbd7cfe19 100644 --- a/pkg/sentry/usage/cpu.go +++ b/pkg/sentry/usage/cpu.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/io.go b/pkg/sentry/usage/io.go index 49faa507d..8e27a0a88 100644 --- a/pkg/sentry/usage/io.go +++ b/pkg/sentry/usage/io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/memory.go b/pkg/sentry/usage/memory.go index 92a478d85..7e065cb76 100644 --- a/pkg/sentry/usage/memory.go +++ b/pkg/sentry/usage/memory.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/memory_unsafe.go b/pkg/sentry/usage/memory_unsafe.go index f990a7750..a3ae668a5 100644 --- a/pkg/sentry/usage/memory_unsafe.go +++ b/pkg/sentry/usage/memory_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/usage.go b/pkg/sentry/usage/usage.go index 3b3118659..ab327f8e2 100644 --- a/pkg/sentry/usage/usage.go +++ b/pkg/sentry/usage/usage.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/access_type.go b/pkg/sentry/usermem/access_type.go index 75346d854..c71d05afe 100644 --- a/pkg/sentry/usermem/access_type.go +++ b/pkg/sentry/usermem/access_type.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/addr.go b/pkg/sentry/usermem/addr.go index fc94bee80..2a75aa60c 100644 --- a/pkg/sentry/usermem/addr.go +++ b/pkg/sentry/usermem/addr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/addr_range_seq_test.go b/pkg/sentry/usermem/addr_range_seq_test.go index cf9d785ed..bd6a1ec8a 100644 --- a/pkg/sentry/usermem/addr_range_seq_test.go +++ b/pkg/sentry/usermem/addr_range_seq_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/addr_range_seq_unsafe.go b/pkg/sentry/usermem/addr_range_seq_unsafe.go index 13b2998b3..f5fd446fa 100644 --- a/pkg/sentry/usermem/addr_range_seq_unsafe.go +++ b/pkg/sentry/usermem/addr_range_seq_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/bytes_io.go b/pkg/sentry/usermem/bytes_io.go index 01a746404..274f568d0 100644 --- a/pkg/sentry/usermem/bytes_io.go +++ b/pkg/sentry/usermem/bytes_io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/bytes_io_unsafe.go b/pkg/sentry/usermem/bytes_io_unsafe.go index efd71fcbc..8bdf3a508 100644 --- a/pkg/sentry/usermem/bytes_io_unsafe.go +++ b/pkg/sentry/usermem/bytes_io_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/usermem.go b/pkg/sentry/usermem/usermem.go index 5d8a1c558..1d6c0b4d6 100644 --- a/pkg/sentry/usermem/usermem.go +++ b/pkg/sentry/usermem/usermem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/usermem_test.go b/pkg/sentry/usermem/usermem_test.go index 563560da8..1991a9641 100644 --- a/pkg/sentry/usermem/usermem_test.go +++ b/pkg/sentry/usermem/usermem_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/usermem_x86.go b/pkg/sentry/usermem/usermem_x86.go index 2484b0d82..9ec90f9ff 100644 --- a/pkg/sentry/usermem/usermem_x86.go +++ b/pkg/sentry/usermem/usermem_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/watchdog/watchdog.go b/pkg/sentry/watchdog/watchdog.go index 5b620693d..75b11237f 100644 --- a/pkg/sentry/watchdog/watchdog.go +++ b/pkg/sentry/watchdog/watchdog.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/commit_amd64.s b/pkg/sleep/commit_amd64.s index d525e5b79..d08df7f37 100644 --- a/pkg/sleep/commit_amd64.s +++ b/pkg/sleep/commit_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/commit_asm.go b/pkg/sleep/commit_asm.go index 39a55df7e..90eef4cbc 100644 --- a/pkg/sleep/commit_asm.go +++ b/pkg/sleep/commit_asm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/commit_noasm.go b/pkg/sleep/commit_noasm.go index 584866cd8..967d22e24 100644 --- a/pkg/sleep/commit_noasm.go +++ b/pkg/sleep/commit_noasm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/empty.s b/pkg/sleep/empty.s index 8aca31bee..85d52cd9c 100644 --- a/pkg/sleep/empty.s +++ b/pkg/sleep/empty.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/sleep_test.go b/pkg/sleep/sleep_test.go index bc1738371..8feb9ffc2 100644 --- a/pkg/sleep/sleep_test.go +++ b/pkg/sleep/sleep_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/sleep_unsafe.go b/pkg/sleep/sleep_unsafe.go index b12cce681..45fb6f0ea 100644 --- a/pkg/sleep/sleep_unsafe.go +++ b/pkg/sleep/sleep_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/decode.go b/pkg/state/decode.go index 3ef59610b..54b5ad8b8 100644 --- a/pkg/state/decode.go +++ b/pkg/state/decode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/encode.go b/pkg/state/encode.go index fd052db12..577aaf051 100644 --- a/pkg/state/encode.go +++ b/pkg/state/encode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/encode_unsafe.go b/pkg/state/encode_unsafe.go index d96ba56d4..be94742a8 100644 --- a/pkg/state/encode_unsafe.go +++ b/pkg/state/encode_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/map.go b/pkg/state/map.go index c3d165501..0035d7250 100644 --- a/pkg/state/map.go +++ b/pkg/state/map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/object.proto b/pkg/state/object.proto index c78efed2a..d3b46ea97 100644 --- a/pkg/state/object.proto +++ b/pkg/state/object.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/printer.go b/pkg/state/printer.go index 2c8ce60a5..aee4b69fb 100644 --- a/pkg/state/printer.go +++ b/pkg/state/printer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/state.go b/pkg/state/state.go index 23a0b5922..4b141777e 100644 --- a/pkg/state/state.go +++ b/pkg/state/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/state_test.go b/pkg/state/state_test.go index 38ad9da9c..22bcad9e1 100644 --- a/pkg/state/state_test.go +++ b/pkg/state/state_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/statefile/statefile.go b/pkg/state/statefile/statefile.go index 9c86c1934..99158fd02 100644 --- a/pkg/state/statefile/statefile.go +++ b/pkg/state/statefile/statefile.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/statefile/statefile_test.go b/pkg/state/statefile/statefile_test.go index fa3fb9f2c..b4f400e01 100644 --- a/pkg/state/statefile/statefile_test.go +++ b/pkg/state/statefile/statefile_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/stats.go b/pkg/state/stats.go index ddcc49f78..17ca258fc 100644 --- a/pkg/state/stats.go +++ b/pkg/state/stats.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/atomicptr_unsafe.go b/pkg/sync/atomicptr_unsafe.go index f12e9cb67..d943b7ff4 100644 --- a/pkg/sync/atomicptr_unsafe.go +++ b/pkg/sync/atomicptr_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/atomicptrtest/atomicptr_test.go b/pkg/sync/atomicptrtest/atomicptr_test.go index b458382b1..3262785ce 100644 --- a/pkg/sync/atomicptrtest/atomicptr_test.go +++ b/pkg/sync/atomicptrtest/atomicptr_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/memmove_unsafe.go b/pkg/sync/memmove_unsafe.go index 0c992d5a4..cd7a02dca 100644 --- a/pkg/sync/memmove_unsafe.go +++ b/pkg/sync/memmove_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/norace_unsafe.go b/pkg/sync/norace_unsafe.go index 968665078..1593b9e5d 100644 --- a/pkg/sync/norace_unsafe.go +++ b/pkg/sync/norace_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/race_unsafe.go b/pkg/sync/race_unsafe.go index d143a21c7..473eaddc6 100644 --- a/pkg/sync/race_unsafe.go +++ b/pkg/sync/race_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/seqatomic_unsafe.go b/pkg/sync/seqatomic_unsafe.go index a18e1229a..bea31adc5 100644 --- a/pkg/sync/seqatomic_unsafe.go +++ b/pkg/sync/seqatomic_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/seqatomictest/seqatomic_test.go b/pkg/sync/seqatomictest/seqatomic_test.go index b785d2344..f5e1fbfff 100644 --- a/pkg/sync/seqatomictest/seqatomic_test.go +++ b/pkg/sync/seqatomictest/seqatomic_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/seqcount.go b/pkg/sync/seqcount.go index 8e3304d69..732e856a4 100644 --- a/pkg/sync/seqcount.go +++ b/pkg/sync/seqcount.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/seqcount_test.go b/pkg/sync/seqcount_test.go index fa4abed1d..b14a8878e 100644 --- a/pkg/sync/seqcount_test.go +++ b/pkg/sync/seqcount_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sync/sync.go b/pkg/sync/sync.go index 36d4c4dee..22c5348d7 100644 --- a/pkg/sync/sync.go +++ b/pkg/sync/sync.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserr/host_linux.go b/pkg/syserr/host_linux.go index 22009a799..74bbe9f5b 100644 --- a/pkg/syserr/host_linux.go +++ b/pkg/syserr/host_linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserr/netstack.go b/pkg/syserr/netstack.go index b9786b48f..20e756edb 100644 --- a/pkg/syserr/netstack.go +++ b/pkg/syserr/netstack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserr/syserr.go b/pkg/syserr/syserr.go index dba6cb7de..6a66e23a2 100644 --- a/pkg/syserr/syserr.go +++ b/pkg/syserr/syserr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserror/syserror.go b/pkg/syserror/syserror.go index 5bc74e65e..4228707f4 100644 --- a/pkg/syserror/syserror.go +++ b/pkg/syserror/syserror.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserror/syserror_test.go b/pkg/syserror/syserror_test.go index fb7d8d5ee..0f0da5781 100644 --- a/pkg/syserror/syserror_test.go +++ b/pkg/syserror/syserror_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/adapters/gonet/gonet.go b/pkg/tcpip/adapters/gonet/gonet.go index b64dce720..81428770b 100644 --- a/pkg/tcpip/adapters/gonet/gonet.go +++ b/pkg/tcpip/adapters/gonet/gonet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/adapters/gonet/gonet_test.go b/pkg/tcpip/adapters/gonet/gonet_test.go index 79b7c77ee..05a730a05 100644 --- a/pkg/tcpip/adapters/gonet/gonet_test.go +++ b/pkg/tcpip/adapters/gonet/gonet_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/buffer/prependable.go b/pkg/tcpip/buffer/prependable.go index c5dd2819f..d3a9a0f88 100644 --- a/pkg/tcpip/buffer/prependable.go +++ b/pkg/tcpip/buffer/prependable.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/buffer/view.go b/pkg/tcpip/buffer/view.go index cea4e3657..24479ea40 100644 --- a/pkg/tcpip/buffer/view.go +++ b/pkg/tcpip/buffer/view.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/buffer/view_test.go b/pkg/tcpip/buffer/view_test.go index 02c264593..74a0a96fc 100644 --- a/pkg/tcpip/buffer/view_test.go +++ b/pkg/tcpip/buffer/view_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/checker/checker.go b/pkg/tcpip/checker/checker.go index 206531f20..5dfb3ca1d 100644 --- a/pkg/tcpip/checker/checker.go +++ b/pkg/tcpip/checker/checker.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/arp.go b/pkg/tcpip/header/arp.go index ae373f112..22b259ccb 100644 --- a/pkg/tcpip/header/arp.go +++ b/pkg/tcpip/header/arp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/checksum.go b/pkg/tcpip/header/checksum.go index e67c50f50..12f208fde 100644 --- a/pkg/tcpip/header/checksum.go +++ b/pkg/tcpip/header/checksum.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/eth.go b/pkg/tcpip/header/eth.go index 99c29b750..77365bc41 100644 --- a/pkg/tcpip/header/eth.go +++ b/pkg/tcpip/header/eth.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/gue.go b/pkg/tcpip/header/gue.go index aac4593c5..2ad13955a 100644 --- a/pkg/tcpip/header/gue.go +++ b/pkg/tcpip/header/gue.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/icmpv4.go b/pkg/tcpip/header/icmpv4.go index af1e94b7f..3ac89cdae 100644 --- a/pkg/tcpip/header/icmpv4.go +++ b/pkg/tcpip/header/icmpv4.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/icmpv6.go b/pkg/tcpip/header/icmpv6.go index 7d35caff7..e317975e8 100644 --- a/pkg/tcpip/header/icmpv6.go +++ b/pkg/tcpip/header/icmpv6.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/interfaces.go b/pkg/tcpip/header/interfaces.go index 042006983..ac327d8a5 100644 --- a/pkg/tcpip/header/interfaces.go +++ b/pkg/tcpip/header/interfaces.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/ipv4.go b/pkg/tcpip/header/ipv4.go index 29570cc34..1b882d3d8 100644 --- a/pkg/tcpip/header/ipv4.go +++ b/pkg/tcpip/header/ipv4.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/ipv6.go b/pkg/tcpip/header/ipv6.go index 66c778fe1..d985b745d 100644 --- a/pkg/tcpip/header/ipv6.go +++ b/pkg/tcpip/header/ipv6.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/ipv6_fragment.go b/pkg/tcpip/header/ipv6_fragment.go index 44b28b326..e36d5177b 100644 --- a/pkg/tcpip/header/ipv6_fragment.go +++ b/pkg/tcpip/header/ipv6_fragment.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/ipversion_test.go b/pkg/tcpip/header/ipversion_test.go index 3ae9b7e4a..8301ba5cf 100644 --- a/pkg/tcpip/header/ipversion_test.go +++ b/pkg/tcpip/header/ipversion_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/tcp.go b/pkg/tcpip/header/tcp.go index 6689a6dc5..567a21167 100644 --- a/pkg/tcpip/header/tcp.go +++ b/pkg/tcpip/header/tcp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/tcp_test.go b/pkg/tcpip/header/tcp_test.go index 7854d3523..7cd98df3b 100644 --- a/pkg/tcpip/header/tcp_test.go +++ b/pkg/tcpip/header/tcp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/udp.go b/pkg/tcpip/header/udp.go index cf2602e50..31c8ef456 100644 --- a/pkg/tcpip/header/udp.go +++ b/pkg/tcpip/header/udp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go index 113cbbf5e..da34032cc 100644 --- a/pkg/tcpip/link/channel/channel.go +++ b/pkg/tcpip/link/channel/channel.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go index ee99ada07..24af428dd 100644 --- a/pkg/tcpip/link/fdbased/endpoint.go +++ b/pkg/tcpip/link/fdbased/endpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/fdbased/endpoint_test.go b/pkg/tcpip/link/fdbased/endpoint_test.go index 52e532ebb..19b007a9e 100644 --- a/pkg/tcpip/link/fdbased/endpoint_test.go +++ b/pkg/tcpip/link/fdbased/endpoint_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/loopback/loopback.go b/pkg/tcpip/link/loopback/loopback.go index fc3f80c01..e6585be66 100644 --- a/pkg/tcpip/link/loopback/loopback.go +++ b/pkg/tcpip/link/loopback/loopback.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/blockingpoll_amd64.s b/pkg/tcpip/link/rawfile/blockingpoll_amd64.s index fc5231831..63b8c4451 100644 --- a/pkg/tcpip/link/rawfile/blockingpoll_amd64.s +++ b/pkg/tcpip/link/rawfile/blockingpoll_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/blockingpoll_unsafe.go b/pkg/tcpip/link/rawfile/blockingpoll_unsafe.go index a0a9d4acd..6a3e956ad 100644 --- a/pkg/tcpip/link/rawfile/blockingpoll_unsafe.go +++ b/pkg/tcpip/link/rawfile/blockingpoll_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/blockingpoll_unsafe_amd64.go b/pkg/tcpip/link/rawfile/blockingpoll_unsafe_amd64.go index 1f143c0db..89a8a9954 100644 --- a/pkg/tcpip/link/rawfile/blockingpoll_unsafe_amd64.go +++ b/pkg/tcpip/link/rawfile/blockingpoll_unsafe_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/errors.go b/pkg/tcpip/link/rawfile/errors.go index de7593d9c..f42ff98db 100644 --- a/pkg/tcpip/link/rawfile/errors.go +++ b/pkg/tcpip/link/rawfile/errors.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/rawfile_unsafe.go b/pkg/tcpip/link/rawfile/rawfile_unsafe.go index cea3cd6a1..be4a4fa9c 100644 --- a/pkg/tcpip/link/rawfile/rawfile_unsafe.go +++ b/pkg/tcpip/link/rawfile/rawfile_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/pipe.go b/pkg/tcpip/link/sharedmem/pipe/pipe.go index 1a0edbaba..e014324cc 100644 --- a/pkg/tcpip/link/sharedmem/pipe/pipe.go +++ b/pkg/tcpip/link/sharedmem/pipe/pipe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/pipe_test.go b/pkg/tcpip/link/sharedmem/pipe/pipe_test.go index db0737c98..30742ccb1 100644 --- a/pkg/tcpip/link/sharedmem/pipe/pipe_test.go +++ b/pkg/tcpip/link/sharedmem/pipe/pipe_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go b/pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go index 480dc4a23..f491d74a2 100644 --- a/pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go +++ b/pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/rx.go b/pkg/tcpip/link/sharedmem/pipe/rx.go index ff778cecd..8d641c76f 100644 --- a/pkg/tcpip/link/sharedmem/pipe/rx.go +++ b/pkg/tcpip/link/sharedmem/pipe/rx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/tx.go b/pkg/tcpip/link/sharedmem/pipe/tx.go index 717f5a4b1..e75175d98 100644 --- a/pkg/tcpip/link/sharedmem/pipe/tx.go +++ b/pkg/tcpip/link/sharedmem/pipe/tx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/queue/queue_test.go b/pkg/tcpip/link/sharedmem/queue/queue_test.go index 3d5909cef..391165bc3 100644 --- a/pkg/tcpip/link/sharedmem/queue/queue_test.go +++ b/pkg/tcpip/link/sharedmem/queue/queue_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/queue/rx.go b/pkg/tcpip/link/sharedmem/queue/rx.go index c40d62c33..d3a5da08a 100644 --- a/pkg/tcpip/link/sharedmem/queue/rx.go +++ b/pkg/tcpip/link/sharedmem/queue/rx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/queue/tx.go b/pkg/tcpip/link/sharedmem/queue/tx.go index 39b595e56..845108db1 100644 --- a/pkg/tcpip/link/sharedmem/queue/tx.go +++ b/pkg/tcpip/link/sharedmem/queue/tx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/rx.go b/pkg/tcpip/link/sharedmem/rx.go index b8e39eca1..3eeab769e 100644 --- a/pkg/tcpip/link/sharedmem/rx.go +++ b/pkg/tcpip/link/sharedmem/rx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/sharedmem.go b/pkg/tcpip/link/sharedmem/sharedmem.go index ce6e86767..27d7eb3b9 100644 --- a/pkg/tcpip/link/sharedmem/sharedmem.go +++ b/pkg/tcpip/link/sharedmem/sharedmem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/sharedmem_test.go b/pkg/tcpip/link/sharedmem/sharedmem_test.go index ad987d382..4b8061b13 100644 --- a/pkg/tcpip/link/sharedmem/sharedmem_test.go +++ b/pkg/tcpip/link/sharedmem/sharedmem_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/sharedmem_unsafe.go b/pkg/tcpip/link/sharedmem/sharedmem_unsafe.go index f0be2dc73..b91adbaf7 100644 --- a/pkg/tcpip/link/sharedmem/sharedmem_unsafe.go +++ b/pkg/tcpip/link/sharedmem/sharedmem_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/tx.go b/pkg/tcpip/link/sharedmem/tx.go index 42a21cb43..37da34831 100644 --- a/pkg/tcpip/link/sharedmem/tx.go +++ b/pkg/tcpip/link/sharedmem/tx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sniffer/pcap.go b/pkg/tcpip/link/sniffer/pcap.go index 04f3d494e..3d0d8d852 100644 --- a/pkg/tcpip/link/sniffer/pcap.go +++ b/pkg/tcpip/link/sniffer/pcap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sniffer/sniffer.go b/pkg/tcpip/link/sniffer/sniffer.go index a30e57a32..1bd174bc3 100644 --- a/pkg/tcpip/link/sniffer/sniffer.go +++ b/pkg/tcpip/link/sniffer/sniffer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/tun/tun_unsafe.go b/pkg/tcpip/link/tun/tun_unsafe.go index 1dec41982..e4c589dda 100644 --- a/pkg/tcpip/link/tun/tun_unsafe.go +++ b/pkg/tcpip/link/tun/tun_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/waitable/waitable.go b/pkg/tcpip/link/waitable/waitable.go index ef8c88561..9ffb7b7e9 100644 --- a/pkg/tcpip/link/waitable/waitable.go +++ b/pkg/tcpip/link/waitable/waitable.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/waitable/waitable_test.go b/pkg/tcpip/link/waitable/waitable_test.go index 0a15c40de..5ebe09664 100644 --- a/pkg/tcpip/link/waitable/waitable_test.go +++ b/pkg/tcpip/link/waitable/waitable_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/arp/arp.go b/pkg/tcpip/network/arp/arp.go index 9d0881e11..2e0024925 100644 --- a/pkg/tcpip/network/arp/arp.go +++ b/pkg/tcpip/network/arp/arp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/arp/arp_test.go b/pkg/tcpip/network/arp/arp_test.go index 50628e4a2..5894f9114 100644 --- a/pkg/tcpip/network/arp/arp_test.go +++ b/pkg/tcpip/network/arp/arp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/frag_heap.go b/pkg/tcpip/network/fragmentation/frag_heap.go index 6c7faafe4..55615c8e6 100644 --- a/pkg/tcpip/network/fragmentation/frag_heap.go +++ b/pkg/tcpip/network/fragmentation/frag_heap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/frag_heap_test.go b/pkg/tcpip/network/fragmentation/frag_heap_test.go index a15540634..1b1b72e88 100644 --- a/pkg/tcpip/network/fragmentation/frag_heap_test.go +++ b/pkg/tcpip/network/fragmentation/frag_heap_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/fragmentation.go b/pkg/tcpip/network/fragmentation/fragmentation.go index 885e3cca2..a5dda0398 100644 --- a/pkg/tcpip/network/fragmentation/fragmentation.go +++ b/pkg/tcpip/network/fragmentation/fragmentation.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/fragmentation_test.go b/pkg/tcpip/network/fragmentation/fragmentation_test.go index fc62a15dd..5bf3463a9 100644 --- a/pkg/tcpip/network/fragmentation/fragmentation_test.go +++ b/pkg/tcpip/network/fragmentation/fragmentation_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/reassembler.go b/pkg/tcpip/network/fragmentation/reassembler.go index b57fe82ec..c9ad2bef6 100644 --- a/pkg/tcpip/network/fragmentation/reassembler.go +++ b/pkg/tcpip/network/fragmentation/reassembler.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/reassembler_test.go b/pkg/tcpip/network/fragmentation/reassembler_test.go index 4c137828f..a2bc9707a 100644 --- a/pkg/tcpip/network/fragmentation/reassembler_test.go +++ b/pkg/tcpip/network/fragmentation/reassembler_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/hash/hash.go b/pkg/tcpip/network/hash/hash.go index eddf7ca4d..07960ddf0 100644 --- a/pkg/tcpip/network/hash/hash.go +++ b/pkg/tcpip/network/hash/hash.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ip_test.go b/pkg/tcpip/network/ip_test.go index e3c7af1f9..5c1e88e56 100644 --- a/pkg/tcpip/network/ip_test.go +++ b/pkg/tcpip/network/ip_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv4/icmp.go b/pkg/tcpip/network/ipv4/icmp.go index ee8172ac8..f82dc098f 100644 --- a/pkg/tcpip/network/ipv4/icmp.go +++ b/pkg/tcpip/network/ipv4/icmp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go index d4eeeb5d9..d7801ec19 100644 --- a/pkg/tcpip/network/ipv4/ipv4.go +++ b/pkg/tcpip/network/ipv4/ipv4.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv4/ipv4_test.go b/pkg/tcpip/network/ipv4/ipv4_test.go index 2b7067a50..190d548eb 100644 --- a/pkg/tcpip/network/ipv4/ipv4_test.go +++ b/pkg/tcpip/network/ipv4/ipv4_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go index 81aba0923..14107443b 100644 --- a/pkg/tcpip/network/ipv6/icmp.go +++ b/pkg/tcpip/network/ipv6/icmp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv6/icmp_test.go b/pkg/tcpip/network/ipv6/icmp_test.go index fabbdc8c7..12c818b48 100644 --- a/pkg/tcpip/network/ipv6/icmp_test.go +++ b/pkg/tcpip/network/ipv6/icmp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go index 25bd998e5..4d0b6ee9c 100644 --- a/pkg/tcpip/network/ipv6/ipv6.go +++ b/pkg/tcpip/network/ipv6/ipv6.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/ports/ports.go b/pkg/tcpip/ports/ports.go index 4e24efddb..41ef32921 100644 --- a/pkg/tcpip/ports/ports.go +++ b/pkg/tcpip/ports/ports.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/ports/ports_test.go b/pkg/tcpip/ports/ports_test.go index 4ab6a1fa2..72577dfcb 100644 --- a/pkg/tcpip/ports/ports_test.go +++ b/pkg/tcpip/ports/ports_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/sample/tun_tcp_connect/main.go b/pkg/tcpip/sample/tun_tcp_connect/main.go index c4707736e..67e8f0b9e 100644 --- a/pkg/tcpip/sample/tun_tcp_connect/main.go +++ b/pkg/tcpip/sample/tun_tcp_connect/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/sample/tun_tcp_echo/main.go b/pkg/tcpip/sample/tun_tcp_echo/main.go index 910d1257f..ab40e9e0b 100644 --- a/pkg/tcpip/sample/tun_tcp_echo/main.go +++ b/pkg/tcpip/sample/tun_tcp_echo/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/seqnum/seqnum.go b/pkg/tcpip/seqnum/seqnum.go index e507d02f7..f2b988839 100644 --- a/pkg/tcpip/seqnum/seqnum.go +++ b/pkg/tcpip/seqnum/seqnum.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/linkaddrcache.go b/pkg/tcpip/stack/linkaddrcache.go index 3a147a75f..cb7b7116b 100644 --- a/pkg/tcpip/stack/linkaddrcache.go +++ b/pkg/tcpip/stack/linkaddrcache.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/linkaddrcache_test.go b/pkg/tcpip/stack/linkaddrcache_test.go index e46267f12..651fa17ac 100644 --- a/pkg/tcpip/stack/linkaddrcache_test.go +++ b/pkg/tcpip/stack/linkaddrcache_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go index dba95369c..3da99ac67 100644 --- a/pkg/tcpip/stack/nic.go +++ b/pkg/tcpip/stack/nic.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/registration.go b/pkg/tcpip/stack/registration.go index 0acec2984..b6266eb55 100644 --- a/pkg/tcpip/stack/registration.go +++ b/pkg/tcpip/stack/registration.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/route.go b/pkg/tcpip/stack/route.go index 6c6400c33..2b4185014 100644 --- a/pkg/tcpip/stack/route.go +++ b/pkg/tcpip/stack/route.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index d1ec6a660..d4da980a9 100644 --- a/pkg/tcpip/stack/stack.go +++ b/pkg/tcpip/stack/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/stack_global_state.go b/pkg/tcpip/stack/stack_global_state.go index b6c095efb..f2c6c9a8d 100644 --- a/pkg/tcpip/stack/stack_global_state.go +++ b/pkg/tcpip/stack/stack_global_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/stack_test.go b/pkg/tcpip/stack/stack_test.go index a0b3399a8..74bf2c99e 100644 --- a/pkg/tcpip/stack/stack_test.go +++ b/pkg/tcpip/stack/stack_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/transport_demuxer.go b/pkg/tcpip/stack/transport_demuxer.go index a7470d606..c8522ad9e 100644 --- a/pkg/tcpip/stack/transport_demuxer.go +++ b/pkg/tcpip/stack/transport_demuxer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/transport_test.go b/pkg/tcpip/stack/transport_test.go index 98cc3b120..f09760180 100644 --- a/pkg/tcpip/stack/transport_test.go +++ b/pkg/tcpip/stack/transport_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index bf11c2175..413aee6c6 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/tcpip_test.go b/pkg/tcpip/tcpip_test.go index d283f71c7..361e359d4 100644 --- a/pkg/tcpip/tcpip_test.go +++ b/pkg/tcpip/tcpip_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/time.s b/pkg/tcpip/time.s index 8aca31bee..85d52cd9c 100644 --- a/pkg/tcpip/time.s +++ b/pkg/tcpip/time.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/time_unsafe.go b/pkg/tcpip/time_unsafe.go index 2102e9633..231151bf3 100644 --- a/pkg/tcpip/time_unsafe.go +++ b/pkg/tcpip/time_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/ping/endpoint.go b/pkg/tcpip/transport/ping/endpoint.go index 055daa918..b3f54cfe0 100644 --- a/pkg/tcpip/transport/ping/endpoint.go +++ b/pkg/tcpip/transport/ping/endpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/ping/endpoint_state.go b/pkg/tcpip/transport/ping/endpoint_state.go index a16087304..80721d227 100644 --- a/pkg/tcpip/transport/ping/endpoint_state.go +++ b/pkg/tcpip/transport/ping/endpoint_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/ping/protocol.go b/pkg/tcpip/transport/ping/protocol.go index 549b1b2d3..1d504773b 100644 --- a/pkg/tcpip/transport/ping/protocol.go +++ b/pkg/tcpip/transport/ping/protocol.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/accept.go b/pkg/tcpip/transport/tcp/accept.go index c22ed5ea7..5a88d25d0 100644 --- a/pkg/tcpip/transport/tcp/accept.go +++ b/pkg/tcpip/transport/tcp/accept.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go index 27dbcace2..800d2409e 100644 --- a/pkg/tcpip/transport/tcp/connect.go +++ b/pkg/tcpip/transport/tcp/connect.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/cubic.go b/pkg/tcpip/transport/tcp/cubic.go index 8cea416d2..003525d86 100644 --- a/pkg/tcpip/transport/tcp/cubic.go +++ b/pkg/tcpip/transport/tcp/cubic.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/dual_stack_test.go b/pkg/tcpip/transport/tcp/dual_stack_test.go index c88e98977..d3120c1d8 100644 --- a/pkg/tcpip/transport/tcp/dual_stack_test.go +++ b/pkg/tcpip/transport/tcp/dual_stack_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index 707d6be96..673a65c31 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go index bed7ec6a6..e32c73aae 100644 --- a/pkg/tcpip/transport/tcp/endpoint_state.go +++ b/pkg/tcpip/transport/tcp/endpoint_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/forwarder.go b/pkg/tcpip/transport/tcp/forwarder.go index c80f3c7d6..2f90839e9 100644 --- a/pkg/tcpip/transport/tcp/forwarder.go +++ b/pkg/tcpip/transport/tcp/forwarder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/protocol.go b/pkg/tcpip/transport/tcp/protocol.go index abdc825cd..753e1419e 100644 --- a/pkg/tcpip/transport/tcp/protocol.go +++ b/pkg/tcpip/transport/tcp/protocol.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/rcv.go b/pkg/tcpip/transport/tcp/rcv.go index 92ef9c6f7..05ff9e0d7 100644 --- a/pkg/tcpip/transport/tcp/rcv.go +++ b/pkg/tcpip/transport/tcp/rcv.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/reno.go b/pkg/tcpip/transport/tcp/reno.go index feb593234..e4f8b7d5a 100644 --- a/pkg/tcpip/transport/tcp/reno.go +++ b/pkg/tcpip/transport/tcp/reno.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/sack.go b/pkg/tcpip/transport/tcp/sack.go index 05bac08cb..24e48fe7b 100644 --- a/pkg/tcpip/transport/tcp/sack.go +++ b/pkg/tcpip/transport/tcp/sack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index 51a3d6aba..fc87a05fd 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/segment_heap.go b/pkg/tcpip/transport/tcp/segment_heap.go index e3a3405ef..98422fadf 100644 --- a/pkg/tcpip/transport/tcp/segment_heap.go +++ b/pkg/tcpip/transport/tcp/segment_heap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/segment_queue.go b/pkg/tcpip/transport/tcp/segment_queue.go index 6a2d7bc0b..0c637d7ad 100644 --- a/pkg/tcpip/transport/tcp/segment_queue.go +++ b/pkg/tcpip/transport/tcp/segment_queue.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/segment_state.go b/pkg/tcpip/transport/tcp/segment_state.go index 22f0bbf18..46b6d85a6 100644 --- a/pkg/tcpip/transport/tcp/segment_state.go +++ b/pkg/tcpip/transport/tcp/segment_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index 0bd421ff4..eefe93d48 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/snd_state.go b/pkg/tcpip/transport/tcp/snd_state.go index d536839af..86bbd643f 100644 --- a/pkg/tcpip/transport/tcp/snd_state.go +++ b/pkg/tcpip/transport/tcp/snd_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/tcp_sack_test.go b/pkg/tcpip/transport/tcp/tcp_sack_test.go index a61d0ca64..06b0702c5 100644 --- a/pkg/tcpip/transport/tcp/tcp_sack_test.go +++ b/pkg/tcpip/transport/tcp/tcp_sack_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go index 48852ea47..04e046257 100644 --- a/pkg/tcpip/transport/tcp/tcp_test.go +++ b/pkg/tcpip/transport/tcp/tcp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/tcp_timestamp_test.go b/pkg/tcpip/transport/tcp/tcp_timestamp_test.go index ca16fc8fa..b08df0fec 100644 --- a/pkg/tcpip/transport/tcp/tcp_timestamp_test.go +++ b/pkg/tcpip/transport/tcp/tcp_timestamp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/testing/context/context.go b/pkg/tcpip/transport/tcp/testing/context/context.go index 5b25534f4..0695e8150 100644 --- a/pkg/tcpip/transport/tcp/testing/context/context.go +++ b/pkg/tcpip/transport/tcp/testing/context/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/timer.go b/pkg/tcpip/transport/tcp/timer.go index 938c0bcef..38240d2d5 100644 --- a/pkg/tcpip/transport/tcp/timer.go +++ b/pkg/tcpip/transport/tcp/timer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go b/pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go index 5f8f1a64d..f7b2900de 100644 --- a/pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go +++ b/pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go b/pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go index 514722ab7..aaeae9b18 100644 --- a/pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go +++ b/pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 840e95302..d777a80d0 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/endpoint_state.go b/pkg/tcpip/transport/udp/endpoint_state.go index 70a37c7f2..db1e281ad 100644 --- a/pkg/tcpip/transport/udp/endpoint_state.go +++ b/pkg/tcpip/transport/udp/endpoint_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/protocol.go b/pkg/tcpip/transport/udp/protocol.go index 1334fec8a..b3fbed6e4 100644 --- a/pkg/tcpip/transport/udp/protocol.go +++ b/pkg/tcpip/transport/udp/protocol.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/udp_test.go b/pkg/tcpip/transport/udp/udp_test.go index c3f592bd4..58a346cd9 100644 --- a/pkg/tcpip/transport/udp/udp_test.go +++ b/pkg/tcpip/transport/udp/udp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tmutex/tmutex.go b/pkg/tmutex/tmutex.go index bd5c681dd..df61d89f5 100644 --- a/pkg/tmutex/tmutex.go +++ b/pkg/tmutex/tmutex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tmutex/tmutex_test.go b/pkg/tmutex/tmutex_test.go index a9dc9972f..a4537cb3b 100644 --- a/pkg/tmutex/tmutex_test.go +++ b/pkg/tmutex/tmutex_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/unet/unet.go b/pkg/unet/unet.go index f4800e0d9..deeea078d 100644 --- a/pkg/unet/unet.go +++ b/pkg/unet/unet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/unet/unet_test.go b/pkg/unet/unet_test.go index 6c546825f..ecc670925 100644 --- a/pkg/unet/unet_test.go +++ b/pkg/unet/unet_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/unet/unet_unsafe.go b/pkg/unet/unet_unsafe.go index fa15cf744..1d69de542 100644 --- a/pkg/unet/unet_unsafe.go +++ b/pkg/unet/unet_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/urpc/urpc.go b/pkg/urpc/urpc.go index 1ec06dd4c..753366be2 100644 --- a/pkg/urpc/urpc.go +++ b/pkg/urpc/urpc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/urpc/urpc_test.go b/pkg/urpc/urpc_test.go index d9cfc512e..f1b9a85ca 100644 --- a/pkg/urpc/urpc_test.go +++ b/pkg/urpc/urpc_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/waiter/fdnotifier/fdnotifier.go b/pkg/waiter/fdnotifier/fdnotifier.go index 8bb93e39b..624b1a0c5 100644 --- a/pkg/waiter/fdnotifier/fdnotifier.go +++ b/pkg/waiter/fdnotifier/fdnotifier.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/waiter/fdnotifier/poll_unsafe.go b/pkg/waiter/fdnotifier/poll_unsafe.go index 26bca2b53..8459d4c74 100644 --- a/pkg/waiter/fdnotifier/poll_unsafe.go +++ b/pkg/waiter/fdnotifier/poll_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/waiter/waiter.go b/pkg/waiter/waiter.go index 832b6a5a9..93390b299 100644 --- a/pkg/waiter/waiter.go +++ b/pkg/waiter/waiter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/waiter/waiter_test.go b/pkg/waiter/waiter_test.go index c45f22889..60853f9c1 100644 --- a/pkg/waiter/waiter_test.go +++ b/pkg/waiter/waiter_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/compat.go b/runsc/boot/compat.go index 3250cdcdc..6766953b3 100644 --- a/runsc/boot/compat.go +++ b/runsc/boot/compat.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/config.go b/runsc/boot/config.go index 51d20d06d..9ebbde424 100644 --- a/runsc/boot/config.go +++ b/runsc/boot/config.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index bee82f344..6dd7fadd9 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/debug.go b/runsc/boot/debug.go index 971962c91..d224d08b7 100644 --- a/runsc/boot/debug.go +++ b/runsc/boot/debug.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/events.go b/runsc/boot/events.go index 595846b10..f954b8c0b 100644 --- a/runsc/boot/events.go +++ b/runsc/boot/events.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/fds.go b/runsc/boot/fds.go index 9416e3a5c..a3d21d963 100644 --- a/runsc/boot/fds.go +++ b/runsc/boot/fds.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/config.go b/runsc/boot/filter/config.go index 92a73db9a..378396b9b 100644 --- a/runsc/boot/filter/config.go +++ b/runsc/boot/filter/config.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/extra_filters.go b/runsc/boot/filter/extra_filters.go index 82cf00dfb..67f3101fe 100644 --- a/runsc/boot/filter/extra_filters.go +++ b/runsc/boot/filter/extra_filters.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/extra_filters_msan.go b/runsc/boot/filter/extra_filters_msan.go index 76f3f6865..fb95283ab 100644 --- a/runsc/boot/filter/extra_filters_msan.go +++ b/runsc/boot/filter/extra_filters_msan.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/extra_filters_race.go b/runsc/boot/filter/extra_filters_race.go index ebd56c553..02a122c95 100644 --- a/runsc/boot/filter/extra_filters_race.go +++ b/runsc/boot/filter/extra_filters_race.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/filter.go b/runsc/boot/filter/filter.go index b656883ad..dc7294b1d 100644 --- a/runsc/boot/filter/filter.go +++ b/runsc/boot/filter/filter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/fs.go b/runsc/boot/fs.go index ea825e571..e52c89fe4 100644 --- a/runsc/boot/fs.go +++ b/runsc/boot/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/limits.go b/runsc/boot/limits.go index 510497eba..8ecda6d0e 100644 --- a/runsc/boot/limits.go +++ b/runsc/boot/limits.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index c79b95bde..fa3de0133 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/loader_test.go b/runsc/boot/loader_test.go index 41ff3681b..c342ee005 100644 --- a/runsc/boot/loader_test.go +++ b/runsc/boot/loader_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/network.go b/runsc/boot/network.go index 6a2678ac9..89f186139 100644 --- a/runsc/boot/network.go +++ b/runsc/boot/network.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/strace.go b/runsc/boot/strace.go index 1e898672b..028bcc1f4 100644 --- a/runsc/boot/strace.go +++ b/runsc/boot/strace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cgroup/cgroup.go b/runsc/cgroup/cgroup.go index 7a75a189a..d6058a8a2 100644 --- a/runsc/cgroup/cgroup.go +++ b/runsc/cgroup/cgroup.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cgroup/cgroup_test.go b/runsc/cgroup/cgroup_test.go index cde915329..4a4713d4f 100644 --- a/runsc/cgroup/cgroup_test.go +++ b/runsc/cgroup/cgroup_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/boot.go b/runsc/cmd/boot.go index 023b63dc0..7c14857ba 100644 --- a/runsc/cmd/boot.go +++ b/runsc/cmd/boot.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/capability.go b/runsc/cmd/capability.go index 0b18c5481..e5da021e5 100644 --- a/runsc/cmd/capability.go +++ b/runsc/cmd/capability.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/capability_test.go b/runsc/cmd/capability_test.go index 3329b308d..dd278b32d 100644 --- a/runsc/cmd/capability_test.go +++ b/runsc/cmd/capability_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/checkpoint.go b/runsc/cmd/checkpoint.go index 023ab2455..d49d0169b 100644 --- a/runsc/cmd/checkpoint.go +++ b/runsc/cmd/checkpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/cmd.go b/runsc/cmd/cmd.go index 2937ae1c4..a1c3491a3 100644 --- a/runsc/cmd/cmd.go +++ b/runsc/cmd/cmd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/create.go b/runsc/cmd/create.go index 275a96f57..b84185b43 100644 --- a/runsc/cmd/create.go +++ b/runsc/cmd/create.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/debug.go b/runsc/cmd/debug.go index cb7d81057..288cbe435 100644 --- a/runsc/cmd/debug.go +++ b/runsc/cmd/debug.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/delete.go b/runsc/cmd/delete.go index 92b609c3c..ea1ca1278 100644 --- a/runsc/cmd/delete.go +++ b/runsc/cmd/delete.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/delete_test.go b/runsc/cmd/delete_test.go index f6d164394..4a5b4774a 100644 --- a/runsc/cmd/delete_test.go +++ b/runsc/cmd/delete_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/events.go b/runsc/cmd/events.go index df65ea31d..df03415ec 100644 --- a/runsc/cmd/events.go +++ b/runsc/cmd/events.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/exec.go b/runsc/cmd/exec.go index 336edf3f6..9a395e6f1 100644 --- a/runsc/cmd/exec.go +++ b/runsc/cmd/exec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/exec_test.go b/runsc/cmd/exec_test.go index 623461e78..686c5e150 100644 --- a/runsc/cmd/exec_test.go +++ b/runsc/cmd/exec_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/gofer.go b/runsc/cmd/gofer.go index fd4eee546..3842fdf64 100644 --- a/runsc/cmd/gofer.go +++ b/runsc/cmd/gofer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/kill.go b/runsc/cmd/kill.go index 7a98d10a2..1f1086250 100644 --- a/runsc/cmd/kill.go +++ b/runsc/cmd/kill.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/list.go b/runsc/cmd/list.go index 4d4a5cb0b..fd59b73e6 100644 --- a/runsc/cmd/list.go +++ b/runsc/cmd/list.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/path.go b/runsc/cmd/path.go index c207b80da..baba937a8 100644 --- a/runsc/cmd/path.go +++ b/runsc/cmd/path.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/pause.go b/runsc/cmd/pause.go index ac393b48e..5ff6f059c 100644 --- a/runsc/cmd/pause.go +++ b/runsc/cmd/pause.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/ps.go b/runsc/cmd/ps.go index 5d219bfdc..fd76cf975 100644 --- a/runsc/cmd/ps.go +++ b/runsc/cmd/ps.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/restore.go b/runsc/cmd/restore.go index 6dc044672..cc99b3503 100644 --- a/runsc/cmd/restore.go +++ b/runsc/cmd/restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/resume.go b/runsc/cmd/resume.go index a12adf1a3..274b5d084 100644 --- a/runsc/cmd/resume.go +++ b/runsc/cmd/resume.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/run.go b/runsc/cmd/run.go index 9a87cf240..b6a12f5d6 100644 --- a/runsc/cmd/run.go +++ b/runsc/cmd/run.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/spec.go b/runsc/cmd/spec.go index 6281fc49d..57ee37c86 100644 --- a/runsc/cmd/spec.go +++ b/runsc/cmd/spec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/start.go b/runsc/cmd/start.go index 97ea91fff..48bd4c401 100644 --- a/runsc/cmd/start.go +++ b/runsc/cmd/start.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/state.go b/runsc/cmd/state.go index 265014e1b..f8ce8c3d8 100644 --- a/runsc/cmd/state.go +++ b/runsc/cmd/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/wait.go b/runsc/cmd/wait.go index 956349140..121c54554 100644 --- a/runsc/cmd/wait.go +++ b/runsc/cmd/wait.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/console/console.go b/runsc/console/console.go index 3df184742..9f4f9214d 100644 --- a/runsc/console/console.go +++ b/runsc/console/console.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/console_test.go b/runsc/container/console_test.go index 8f019b54a..0b0dfb4cb 100644 --- a/runsc/container/console_test.go +++ b/runsc/container/console_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/container.go b/runsc/container/container.go index f76bad1aa..cb4c9b5c1 100644 --- a/runsc/container/container.go +++ b/runsc/container/container.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/container_test.go b/runsc/container/container_test.go index 662591b3b..243528d35 100644 --- a/runsc/container/container_test.go +++ b/runsc/container/container_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/fs.go b/runsc/container/fs.go index 2ed42fd93..41022686b 100644 --- a/runsc/container/fs.go +++ b/runsc/container/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/fs_test.go b/runsc/container/fs_test.go index 84bde18fb..87cdb078e 100644 --- a/runsc/container/fs_test.go +++ b/runsc/container/fs_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/hook.go b/runsc/container/hook.go index 3d93ca0be..6b9e5550a 100644 --- a/runsc/container/hook.go +++ b/runsc/container/hook.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/multi_container_test.go b/runsc/container/multi_container_test.go index 1781a4602..4548eb106 100644 --- a/runsc/container/multi_container_test.go +++ b/runsc/container/multi_container_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/status.go b/runsc/container/status.go index bf177e78a..234ffb0dd 100644 --- a/runsc/container/status.go +++ b/runsc/container/status.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/test_app.go b/runsc/container/test_app.go index cc3b087e1..b5071ada6 100644 --- a/runsc/container/test_app.go +++ b/runsc/container/test_app.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/config.go b/runsc/fsgofer/filter/config.go index 35698f21f..75a087848 100644 --- a/runsc/fsgofer/filter/config.go +++ b/runsc/fsgofer/filter/config.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/extra_filters.go b/runsc/fsgofer/filter/extra_filters.go index 82cf00dfb..67f3101fe 100644 --- a/runsc/fsgofer/filter/extra_filters.go +++ b/runsc/fsgofer/filter/extra_filters.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/extra_filters_msan.go b/runsc/fsgofer/filter/extra_filters_msan.go index 169a79ed8..7e142b790 100644 --- a/runsc/fsgofer/filter/extra_filters_msan.go +++ b/runsc/fsgofer/filter/extra_filters_msan.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/extra_filters_race.go b/runsc/fsgofer/filter/extra_filters_race.go index 9e6512d8c..3cd29472a 100644 --- a/runsc/fsgofer/filter/extra_filters_race.go +++ b/runsc/fsgofer/filter/extra_filters_race.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/filter.go b/runsc/fsgofer/filter/filter.go index 6f341f688..f50b6bc87 100644 --- a/runsc/fsgofer/filter/filter.go +++ b/runsc/fsgofer/filter/filter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/fsgofer.go b/runsc/fsgofer/fsgofer.go index 9c4864cf1..e03bb7752 100644 --- a/runsc/fsgofer/fsgofer.go +++ b/runsc/fsgofer/fsgofer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/fsgofer_test.go b/runsc/fsgofer/fsgofer_test.go index a500a2976..48860f952 100644 --- a/runsc/fsgofer/fsgofer_test.go +++ b/runsc/fsgofer/fsgofer_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/fsgofer_unsafe.go b/runsc/fsgofer/fsgofer_unsafe.go index e676809ac..99bc25ec1 100644 --- a/runsc/fsgofer/fsgofer_unsafe.go +++ b/runsc/fsgofer/fsgofer_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/main.go b/runsc/main.go index 62b1f01b3..4a92db7c0 100644 --- a/runsc/main.go +++ b/runsc/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/sandbox/chroot.go b/runsc/sandbox/chroot.go index 35b19a0b1..354049871 100644 --- a/runsc/sandbox/chroot.go +++ b/runsc/sandbox/chroot.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/sandbox/network.go b/runsc/sandbox/network.go index 86a52c6ae..52fe8fc0f 100644 --- a/runsc/sandbox/network.go +++ b/runsc/sandbox/network.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 923a52f7f..0fe85cfe1 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/specutils/namespace.go b/runsc/specutils/namespace.go index 00293d45b..73fab13e1 100644 --- a/runsc/specutils/namespace.go +++ b/runsc/specutils/namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/specutils/specutils.go b/runsc/specutils/specutils.go index b29802fde..ab14ed1fc 100644 --- a/runsc/specutils/specutils.go +++ b/runsc/specutils/specutils.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/specutils/specutils_test.go b/runsc/specutils/specutils_test.go index 64e2172c8..b61f1ca62 100644 --- a/runsc/specutils/specutils_test.go +++ b/runsc/specutils/specutils_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/image/image.go b/runsc/test/image/image.go index 069d08013..bcb6f876f 100644 --- a/runsc/test/image/image.go +++ b/runsc/test/image/image.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/image/image_test.go b/runsc/test/image/image_test.go index d89d80a86..763152b47 100644 --- a/runsc/test/image/image_test.go +++ b/runsc/test/image/image_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/image/mysql.sql b/runsc/test/image/mysql.sql index dd5bfaa4e..c1271e719 100644 --- a/runsc/test/image/mysql.sql +++ b/runsc/test/image/mysql.sql @@ -1,4 +1,4 @@ -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/runsc/test/image/ruby.rb b/runsc/test/image/ruby.rb index ae5de3419..25d1ac129 100644 --- a/runsc/test/image/ruby.rb +++ b/runsc/test/image/ruby.rb @@ -1,4 +1,4 @@ -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/runsc/test/image/ruby.sh b/runsc/test/image/ruby.sh index 54be2c931..d3a9b5656 100644 --- a/runsc/test/image/ruby.sh +++ b/runsc/test/image/ruby.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/runsc/test/install.sh b/runsc/test/install.sh index c239588d4..32e1e884e 100755 --- a/runsc/test/install.sh +++ b/runsc/test/install.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/runsc/test/integration/exec_test.go b/runsc/test/integration/exec_test.go index 3cac674d0..fac8337f4 100644 --- a/runsc/test/integration/exec_test.go +++ b/runsc/test/integration/exec_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/integration/integration.go b/runsc/test/integration/integration.go index 49c3c893a..e15321c87 100644 --- a/runsc/test/integration/integration.go +++ b/runsc/test/integration/integration.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/integration/integration_test.go b/runsc/test/integration/integration_test.go index 536bb17e0..526b3a7a1 100644 --- a/runsc/test/integration/integration_test.go +++ b/runsc/test/integration/integration_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/cgroup_test.go b/runsc/test/root/cgroup_test.go index 5cb4b794f..fdb94ff64 100644 --- a/runsc/test/root/cgroup_test.go +++ b/runsc/test/root/cgroup_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/chroot_test.go b/runsc/test/root/chroot_test.go index 8831e6a78..0ffaaf87b 100644 --- a/runsc/test/root/chroot_test.go +++ b/runsc/test/root/chroot_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/root.go b/runsc/test/root/root.go index 790f62c29..586ea0fe3 100644 --- a/runsc/test/root/root.go +++ b/runsc/test/root/root.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/testutil/docker.go b/runsc/test/testutil/docker.go index 7d6a72e5f..3f74e0770 100644 --- a/runsc/test/testutil/docker.go +++ b/runsc/test/testutil/docker.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/testutil/testutil.go b/runsc/test/testutil/testutil.go index 4d7ac3bc9..1b5a02c0f 100644 --- a/runsc/test/testutil/testutil.go +++ b/runsc/test/testutil/testutil.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/testutil/testutil_race.go b/runsc/test/testutil/testutil_race.go index 59cfdaa7b..9267af150 100644 --- a/runsc/test/testutil/testutil_race.go +++ b/runsc/test/testutil/testutil_race.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/tools/dockercfg/dockercfg.go b/runsc/tools/dockercfg/dockercfg.go index 0bd6cad93..110a581ff 100644 --- a/runsc/tools/dockercfg/dockercfg.go +++ b/runsc/tools/dockercfg/dockercfg.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics.go b/tools/go_generics/generics.go index cc61a7537..eaf5c4970 100644 --- a/tools/go_generics/generics.go +++ b/tools/go_generics/generics.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_stmts/input.go b/tools/go_generics/generics_tests/all_stmts/input.go index 870af3b6c..19184a3fe 100644 --- a/tools/go_generics/generics_tests/all_stmts/input.go +++ b/tools/go_generics/generics_tests/all_stmts/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_stmts/output/output.go b/tools/go_generics/generics_tests/all_stmts/output/output.go index e4e670bf1..51582346c 100644 --- a/tools/go_generics/generics_tests/all_stmts/output/output.go +++ b/tools/go_generics/generics_tests/all_stmts/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_types/input.go b/tools/go_generics/generics_tests/all_types/input.go index 3a8643e3d..ed6e97c29 100644 --- a/tools/go_generics/generics_tests/all_types/input.go +++ b/tools/go_generics/generics_tests/all_types/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_types/lib/lib.go b/tools/go_generics/generics_tests/all_types/lib/lib.go index d3911d12d..7e73e678e 100644 --- a/tools/go_generics/generics_tests/all_types/lib/lib.go +++ b/tools/go_generics/generics_tests/all_types/lib/lib.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_types/output/output.go b/tools/go_generics/generics_tests/all_types/output/output.go index b89840936..ec09a6be4 100644 --- a/tools/go_generics/generics_tests/all_types/output/output.go +++ b/tools/go_generics/generics_tests/all_types/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/consts/input.go b/tools/go_generics/generics_tests/consts/input.go index dabf76e1e..394bcc262 100644 --- a/tools/go_generics/generics_tests/consts/input.go +++ b/tools/go_generics/generics_tests/consts/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/consts/output/output.go b/tools/go_generics/generics_tests/consts/output/output.go index 72865607e..91a07fdc2 100644 --- a/tools/go_generics/generics_tests/consts/output/output.go +++ b/tools/go_generics/generics_tests/consts/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/imports/input.go b/tools/go_generics/generics_tests/imports/input.go index 66b43fee5..22e6641a6 100644 --- a/tools/go_generics/generics_tests/imports/input.go +++ b/tools/go_generics/generics_tests/imports/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/imports/output/output.go b/tools/go_generics/generics_tests/imports/output/output.go index 5f20d43ce..2555c0004 100644 --- a/tools/go_generics/generics_tests/imports/output/output.go +++ b/tools/go_generics/generics_tests/imports/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/remove_typedef/input.go b/tools/go_generics/generics_tests/remove_typedef/input.go index c02307d32..d9c9b8530 100644 --- a/tools/go_generics/generics_tests/remove_typedef/input.go +++ b/tools/go_generics/generics_tests/remove_typedef/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/remove_typedef/output/output.go b/tools/go_generics/generics_tests/remove_typedef/output/output.go index d20a89abd..f111a9426 100644 --- a/tools/go_generics/generics_tests/remove_typedef/output/output.go +++ b/tools/go_generics/generics_tests/remove_typedef/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/simple/input.go b/tools/go_generics/generics_tests/simple/input.go index 670161d6e..711687cf5 100644 --- a/tools/go_generics/generics_tests/simple/input.go +++ b/tools/go_generics/generics_tests/simple/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/simple/output/output.go b/tools/go_generics/generics_tests/simple/output/output.go index 75b5467cd..139c9bf9d 100644 --- a/tools/go_generics/generics_tests/simple/output/output.go +++ b/tools/go_generics/generics_tests/simple/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/globals/globals_visitor.go b/tools/go_generics/globals/globals_visitor.go index fc0de4381..daaa17b1d 100644 --- a/tools/go_generics/globals/globals_visitor.go +++ b/tools/go_generics/globals/globals_visitor.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/globals/scope.go b/tools/go_generics/globals/scope.go index 18743bdee..b75a91689 100644 --- a/tools/go_generics/globals/scope.go +++ b/tools/go_generics/globals/scope.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/go_generics_unittest.sh b/tools/go_generics/go_generics_unittest.sh index 699e1f631..e7553a071 100755 --- a/tools/go_generics/go_generics_unittest.sh +++ b/tools/go_generics/go_generics_unittest.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/tools/go_generics/imports.go b/tools/go_generics/imports.go index 97267098b..57f7c3dce 100644 --- a/tools/go_generics/imports.go +++ b/tools/go_generics/imports.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/merge.go b/tools/go_generics/merge.go index ebe7cf4e4..2f83facf8 100644 --- a/tools/go_generics/merge.go +++ b/tools/go_generics/merge.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/remove.go b/tools/go_generics/remove.go index 2a66de762..139d03955 100644 --- a/tools/go_generics/remove.go +++ b/tools/go_generics/remove.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/rules_tests/template.go b/tools/go_generics/rules_tests/template.go index 73c024f0e..f3f31ae8e 100644 --- a/tools/go_generics/rules_tests/template.go +++ b/tools/go_generics/rules_tests/template.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/rules_tests/template_test.go b/tools/go_generics/rules_tests/template_test.go index 76c4cdb64..3a38c8629 100644 --- a/tools/go_generics/rules_tests/template_test.go +++ b/tools/go_generics/rules_tests/template_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_stateify/main.go b/tools/go_stateify/main.go index 5646b879a..9e2c8e106 100644 --- a/tools/go_stateify/main.go +++ b/tools/go_stateify/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/workspace_status.sh b/tools/workspace_status.sh index d89db1f99..7d44dad37 100755 --- a/tools/workspace_status.sh +++ b/tools/workspace_status.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/vdso/barrier.h b/vdso/barrier.h index db8185b2e..7866af414 100644 --- a/vdso/barrier.h +++ b/vdso/barrier.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/check_vdso.py b/vdso/check_vdso.py index 9a3142ab8..6f7d7e7ec 100644 --- a/vdso/check_vdso.py +++ b/vdso/check_vdso.py @@ -1,4 +1,4 @@ -# Copyright 2018 Google Inc. +# Copyright 2018 Google LLC # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/vdso/compiler.h b/vdso/compiler.h index a661516c3..d65f148fb 100644 --- a/vdso/compiler.h +++ b/vdso/compiler.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/cycle_clock.h b/vdso/cycle_clock.h index 93c5f2c0d..dfb5b427d 100644 --- a/vdso/cycle_clock.h +++ b/vdso/cycle_clock.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/seqlock.h b/vdso/seqlock.h index b527bdbca..ab2f3fda3 100644 --- a/vdso/seqlock.h +++ b/vdso/seqlock.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/syscalls.h b/vdso/syscalls.h index fd79c4642..0be8a7f9b 100644 --- a/vdso/syscalls.h +++ b/vdso/syscalls.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/vdso.cc b/vdso/vdso.cc index db3bdef01..f30dc26a2 100644 --- a/vdso/vdso.cc +++ b/vdso/vdso.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/vdso_time.cc b/vdso/vdso_time.cc index 5d5c8de65..a59771bff 100644 --- a/vdso/vdso_time.cc +++ b/vdso/vdso_time.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/vdso_time.h b/vdso/vdso_time.h index 71d6e2f64..464dadff2 100644 --- a/vdso/vdso_time.h +++ b/vdso/vdso_time.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google Inc. +// Copyright 2018 Google LLC // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. -- cgit v1.2.3 From b2068cf5a5d43f3898cf389ab2d6151cf61908ac Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Sat, 20 Oct 2018 11:12:26 -0700 Subject: Add more unimplemented syscall events Added events for *ctl syscalls that may have multiple different commands. For runsc, each syscall event is only logged once. For *ctl syscalls, use the cmd as identifier, not only the syscall number. PiperOrigin-RevId: 218015941 Change-Id: Ie3c19131ae36124861e9b492a7dbe1765d9e5e59 --- pkg/abi/linux/ioctl.go | 75 +++++++++++++------ pkg/abi/linux/prctl.go | 99 +++++++++++++++++++++++-- pkg/sentry/fs/host/BUILD | 1 + pkg/sentry/fs/host/tty.go | 30 ++++++++ pkg/sentry/fs/tty/BUILD | 1 + pkg/sentry/fs/tty/master.go | 45 ++++++++++- pkg/sentry/fs/tty/slave.go | 3 +- pkg/sentry/kernel/BUILD | 2 + pkg/sentry/kernel/kernel.go | 17 +++++ pkg/sentry/kernel/pipe/reader_writer.go | 3 +- pkg/sentry/kernel/task.go | 3 + pkg/sentry/socket/epsocket/BUILD | 1 + pkg/sentry/socket/epsocket/epsocket.go | 3 + pkg/sentry/socket/rpcinet/BUILD | 1 + pkg/sentry/socket/rpcinet/socket.go | 5 ++ pkg/sentry/syscalls/BUILD | 18 ----- pkg/sentry/syscalls/linux/linux64.go | 2 +- pkg/sentry/syscalls/linux/sys_prctl.go | 39 +++++++++- pkg/sentry/syscalls/linux/sys_shm.go | 1 + pkg/sentry/syscalls/linux/sys_tls.go | 3 + pkg/sentry/syscalls/syscalls.go | 15 +--- pkg/sentry/syscalls/unimplemented_syscall.proto | 27 ------- pkg/sentry/unimpl/BUILD | 30 ++++++++ pkg/sentry/unimpl/events.go | 45 +++++++++++ pkg/sentry/unimpl/unimplemented_syscall.proto | 27 +++++++ runsc/boot/BUILD | 9 ++- runsc/boot/compat.go | 72 ++++++++++++++++-- runsc/boot/compat_amd64.go | 54 ++++++++++++++ runsc/boot/compat_test.go | 66 +++++++++++++++++ 29 files changed, 596 insertions(+), 101 deletions(-) delete mode 100644 pkg/sentry/syscalls/unimplemented_syscall.proto create mode 100644 pkg/sentry/unimpl/BUILD create mode 100644 pkg/sentry/unimpl/events.go create mode 100644 pkg/sentry/unimpl/unimplemented_syscall.proto create mode 100644 runsc/boot/compat_amd64.go create mode 100644 runsc/boot/compat_test.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/ioctl.go b/pkg/abi/linux/ioctl.go index 9afc3d1ef..191b26e4d 100644 --- a/pkg/abi/linux/ioctl.go +++ b/pkg/abi/linux/ioctl.go @@ -18,28 +18,59 @@ package linux // // These are ordered by request number (low byte). const ( - TCGETS = 0x00005401 - TCSETS = 0x00005402 - TCSETSW = 0x00005403 - TCSETSF = 0x00005404 - TIOCSCTTY = 0x0000540e - TIOCGPGRP = 0x0000540f - TIOCSPGRP = 0x00005410 - TIOCOUTQ = 0x00005411 - TIOCGWINSZ = 0x00005413 - TIOCSWINSZ = 0x00005414 - TIOCINQ = 0x0000541b - FIONREAD = TIOCINQ - FIONBIO = 0x00005421 - TIOCGPTN = 0x80045430 - TIOCSPTLCK = 0x40045431 - FIONCLEX = 0x00005450 - FIOCLEX = 0x00005451 - FIOASYNC = 0x00005452 - FIOSETOWN = 0x00008901 - SIOCSPGRP = 0x00008902 - FIOGETOWN = 0x00008903 - SIOCGPGRP = 0x00008904 + TCGETS = 0x00005401 + TCSETS = 0x00005402 + TCSETSW = 0x00005403 + TCSETSF = 0x00005404 + TCSBRK = 0x00005409 + TIOCEXCL = 0x0000540c + TIOCNXCL = 0x0000540d + TIOCSCTTY = 0x0000540e + TIOCGPGRP = 0x0000540f + TIOCSPGRP = 0x00005410 + TIOCOUTQ = 0x00005411 + TIOCSTI = 0x00005412 + TIOCGWINSZ = 0x00005413 + TIOCSWINSZ = 0x00005414 + TIOCMGET = 0x00005415 + TIOCMBIS = 0x00005416 + TIOCMBIC = 0x00005417 + TIOCMSET = 0x00005418 + TIOCINQ = 0x0000541b + FIONREAD = TIOCINQ + FIONBIO = 0x00005421 + TIOCSETD = 0x00005423 + TIOCNOTTY = 0x00005422 + TIOCGETD = 0x00005424 + TCSBRKP = 0x00005425 + TIOCSBRK = 0x00005427 + TIOCCBRK = 0x00005428 + TIOCGSID = 0x00005429 + TIOCGPTN = 0x80045430 + TIOCSPTLCK = 0x40045431 + TIOCGDEV = 0x80045432 + TIOCVHANGUP = 0x00005437 + TCFLSH = 0x0000540b + TIOCCONS = 0x0000541d + TIOCSSERIAL = 0x0000541f + TIOCGEXCL = 0x80045440 + TIOCGPTPEER = 0x80045441 + TIOCGICOUNT = 0x0000545d + FIONCLEX = 0x00005450 + FIOCLEX = 0x00005451 + FIOASYNC = 0x00005452 + FIOSETOWN = 0x00008901 + SIOCSPGRP = 0x00008902 + FIOGETOWN = 0x00008903 + SIOCGPGRP = 0x00008904 +) + +// ioctl(2) requests provided by uapi/linux/sockios.h +const ( + SIOCGIFMEM = 0x891f + SIOCGIFPFLAGS = 0x8935 + SIOCGMIIPHY = 0x8947 + SIOCGMIIREG = 0x8948 ) // ioctl(2) requests provided by uapi/linux/android/binder.h diff --git a/pkg/abi/linux/prctl.go b/pkg/abi/linux/prctl.go index e152c4c27..db3206f36 100644 --- a/pkg/abi/linux/prctl.go +++ b/pkg/abi/linux/prctl.go @@ -22,26 +22,102 @@ const ( // PR_GET_PDEATHSIG will get the process' death signal. PR_GET_PDEATHSIG = 2 + // PR_GET_DUMPABLE will get the process's dumpable flag. + PR_GET_DUMPABLE = 3 + + // PR_SET_DUMPABLE will set the process's dumpable flag. + PR_SET_DUMPABLE = 4 + // PR_GET_KEEPCAPS will get the value of the keep capabilities flag. PR_GET_KEEPCAPS = 7 // PR_SET_KEEPCAPS will set the value of the keep capabilities flag. PR_SET_KEEPCAPS = 8 + // PR_GET_TIMING will get the process's timing method. + PR_GET_TIMING = 13 + + // PR_SET_TIMING will set the process's timing method. + PR_SET_TIMING = 14 + // PR_SET_NAME will set the process' name. PR_SET_NAME = 15 // PR_GET_NAME will get the process' name. PR_GET_NAME = 16 + // PR_GET_SECCOMP will get a process' seccomp mode. + PR_GET_SECCOMP = 21 + + // PR_SET_SECCOMP will set a process' seccomp mode. + PR_SET_SECCOMP = 22 + + // PR_CAPBSET_READ will get the capability bounding set. + PR_CAPBSET_READ = 23 + + // PR_CAPBSET_DROP will set the capability bounding set. + PR_CAPBSET_DROP = 24 + + // PR_GET_TSC will get the the value of the flag determining whether the + // timestamp counter can be read. + PR_GET_TSC = 25 + + // PR_SET_TSC will set the the value of the flag determining whether the + // timestamp counter can be read. + PR_SET_TSC = 26 + + // PR_SET_TIMERSLACK set the process's time slack. + PR_SET_TIMERSLACK = 29 + + // PR_GET_TIMERSLACK get the process's time slack. + PR_GET_TIMERSLACK = 30 + + // PR_TASK_PERF_EVENTS_DISABLE disable all performance counters attached to + // the calling process. + PR_TASK_PERF_EVENTS_DISABLE = 31 + + // PR_TASK_PERF_EVENTS_ENABLE enable all performance counters attached to + // the calling process. + PR_TASK_PERF_EVENTS_ENABLE = 32 + + // PR_MCE_KILL set the machine check memory corruption kill policy for the + // calling thread. + PR_MCE_KILL = 33 + + // PR_MCE_KILL_GET get the machine check memory corruption kill policy for the + // calling thread. + PR_MCE_KILL_GET = 34 + // PR_SET_MM will modify certain kernel memory map descriptor fields of the // calling process. See prctl(2) for more information. PR_SET_MM = 35 + PR_SET_MM_START_CODE = 1 + PR_SET_MM_END_CODE = 2 + PR_SET_MM_START_DATA = 3 + PR_SET_MM_END_DATA = 4 + PR_SET_MM_START_STACK = 5 + PR_SET_MM_START_BRK = 6 + PR_SET_MM_BRK = 7 + PR_SET_MM_ARG_START = 8 + PR_SET_MM_ARG_END = 9 + PR_SET_MM_ENV_START = 10 + PR_SET_MM_ENV_END = 11 + PR_SET_MM_AUXV = 12 // PR_SET_MM_EXE_FILE will supersede the /proc/pid/exe symbolic link with a // new one pointing to a new executable file identified by the file descriptor // provided in arg3 argument. See prctl(2) for more information. PR_SET_MM_EXE_FILE = 13 + PR_SET_MM_MAP = 14 + PR_SET_MM_MAP_SIZE = 15 + + // PR_SET_CHILD_SUBREAPER set the "child subreaper" attribute of the calling + // process. + PR_SET_CHILD_SUBREAPER = 36 + + // PR_GET_CHILD_SUBREAPER get the "child subreaper" attribute of the calling + // process. + PR_GET_CHILD_SUBREAPER = 37 // PR_SET_NO_NEW_PRIVS will set the calling thread's no_new_privs bit. PR_SET_NO_NEW_PRIVS = 38 @@ -49,17 +125,24 @@ const ( // PR_GET_NO_NEW_PRIVS will get the calling thread's no_new_privs bit. PR_GET_NO_NEW_PRIVS = 39 - // PR_SET_SECCOMP will set a process' seccomp mode. - PR_SET_SECCOMP = 22 + // PR_GET_TID_ADDRESS retrieve the clear_child_tid address. + PR_GET_TID_ADDRESS = 40 - // PR_GET_SECCOMP will get a process' seccomp mode. - PR_GET_SECCOMP = 21 + // PR_SET_THP_DISABLE set the state of the "THP disable" flag for the calling + // thread. + PR_SET_THP_DISABLE = 41 - // PR_CAPBSET_READ will get the capability bounding set. - PR_CAPBSET_READ = 23 + // PR_GET_THP_DISABLE get the state of the "THP disable" flag for the calling + // thread. + PR_GET_THP_DISABLE = 42 - // PR_CAPBSET_DROP will set the capability bounding set. - PR_CAPBSET_DROP = 24 + // PR_MPX_ENABLE_MANAGEMENT enable kernel management of Memory Protection + // eXtensions (MPX) bounds tables. + PR_MPX_ENABLE_MANAGEMENT = 43 + + // PR_MPX_DISABLE_MANAGEMENTdisable kernel management of Memory Protection + // eXtensions (MPX) bounds tables. + PR_MPX_DISABLE_MANAGEMENT = 44 ) // From diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 4f264a024..d1eb9bd64 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -43,6 +43,7 @@ go_library( "//pkg/sentry/socket/control", "//pkg/sentry/socket/unix", "//pkg/sentry/socket/unix/transport", + "//pkg/sentry/unimpl", "//pkg/sentry/uniqueid", "//pkg/sentry/usermem", "//pkg/syserr", diff --git a/pkg/sentry/fs/host/tty.go b/pkg/sentry/fs/host/tty.go index cf3639c46..f0bcdc908 100644 --- a/pkg/sentry/fs/host/tty.go +++ b/pkg/sentry/fs/host/tty.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -179,6 +180,35 @@ func (t *TTYFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch. err := ioctlSetWinsize(fd, &winsize) return 0, err + // Unimplemented commands. + case linux.TIOCSETD, + linux.TIOCSBRK, + linux.TIOCCBRK, + linux.TCSBRK, + linux.TCSBRKP, + linux.TIOCSTI, + linux.TIOCCONS, + linux.FIONBIO, + linux.TIOCEXCL, + linux.TIOCNXCL, + linux.TIOCGEXCL, + linux.TIOCNOTTY, + linux.TIOCSCTTY, + linux.TIOCGSID, + linux.TIOCGETD, + linux.TIOCVHANGUP, + linux.TIOCGDEV, + linux.TIOCMGET, + linux.TIOCMSET, + linux.TIOCMBIC, + linux.TIOCMBIS, + linux.TIOCGICOUNT, + linux.TCFLSH, + linux.TIOCSSERIAL, + linux.TIOCGPTPEER: + + unimpl.EmitUnimplementedEvent(ctx) + fallthrough default: return 0, syserror.ENOTTY } diff --git a/pkg/sentry/fs/tty/BUILD b/pkg/sentry/fs/tty/BUILD index d4dd20e30..2b45069a6 100644 --- a/pkg/sentry/fs/tty/BUILD +++ b/pkg/sentry/fs/tty/BUILD @@ -27,6 +27,7 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/socket/unix/transport", + "//pkg/sentry/unimpl", "//pkg/sentry/usermem", "//pkg/syserror", "//pkg/waiter", diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index dad0cad79..00bec4c2c 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -20,6 +20,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/waiter" @@ -149,7 +150,7 @@ func (mf *masterFileOperations) Write(ctx context.Context, _ *fs.File, src userm // Ioctl implements fs.FileOperations.Ioctl. func (mf *masterFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { - switch args[1].Uint() { + switch cmd := args[1].Uint(); cmd { case linux.FIONREAD: // linux.FIONREAD == linux.TIOCINQ // Get the number of bytes in the output queue read buffer. return 0, mf.t.ld.outputQueueReadSize(ctx, io, args) @@ -177,6 +178,48 @@ func (mf *masterFileOperations) Ioctl(ctx context.Context, io usermem.IO, args a case linux.TIOCSWINSZ: return 0, mf.t.ld.setWindowSize(ctx, io, args) default: + maybeEmitUnimplementedEvent(ctx, cmd) return 0, syserror.ENOTTY } } + +// maybeEmitUnimplementedEvent emits unimplemented event if cmd is valid. +func maybeEmitUnimplementedEvent(ctx context.Context, cmd uint32) { + switch cmd { + case linux.TCGETS, + linux.TCSETS, + linux.TCSETSW, + linux.TCSETSF, + linux.TIOCGPGRP, + linux.TIOCSPGRP, + linux.TIOCGWINSZ, + linux.TIOCSWINSZ, + linux.TIOCSETD, + linux.TIOCSBRK, + linux.TIOCCBRK, + linux.TCSBRK, + linux.TCSBRKP, + linux.TIOCSTI, + linux.TIOCCONS, + linux.FIONBIO, + linux.TIOCEXCL, + linux.TIOCNXCL, + linux.TIOCGEXCL, + linux.TIOCNOTTY, + linux.TIOCSCTTY, + linux.TIOCGSID, + linux.TIOCGETD, + linux.TIOCVHANGUP, + linux.TIOCGDEV, + linux.TIOCMGET, + linux.TIOCMSET, + linux.TIOCMBIC, + linux.TIOCMBIS, + linux.TIOCGICOUNT, + linux.TCFLSH, + linux.TIOCSSERIAL, + linux.TIOCGPTPEER: + + unimpl.EmitUnimplementedEvent(ctx) + } +} diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index 9de3168bf..a696fbb51 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -134,7 +134,7 @@ func (sf *slaveFileOperations) Write(ctx context.Context, _ *fs.File, src userme // Ioctl implements fs.FileOperations.Ioctl. func (sf *slaveFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { - switch args[1].Uint() { + switch cmd := args[1].Uint(); cmd { case linux.FIONREAD: // linux.FIONREAD == linux.TIOCINQ // Get the number of bytes in the input queue read buffer. return 0, sf.si.t.ld.inputQueueReadSize(ctx, io, args) @@ -161,6 +161,7 @@ func (sf *slaveFileOperations) Ioctl(ctx context.Context, io usermem.IO, args ar // control. return 0, nil default: + maybeEmitUnimplementedEvent(ctx, cmd) return 0, syserror.ENOTTY } } diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index e2fb61ba6..389824b25 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -157,6 +157,8 @@ go_library( "//pkg/sentry/socket/netlink/port", "//pkg/sentry/socket/unix/transport", "//pkg/sentry/time", + "//pkg/sentry/unimpl", + "//pkg/sentry/unimpl:unimplemented_syscall_go_proto", "//pkg/sentry/uniqueid", "//pkg/sentry/usage", "//pkg/sentry/usermem", diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index bad558d48..17425e656 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -40,6 +40,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/cpuid" + "gvisor.googlesource.com/gvisor/pkg/eventchannel" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" @@ -58,6 +59,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/port" sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl" + uspb "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl/unimplemented_syscall_go_proto" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/state" "gvisor.googlesource.com/gvisor/pkg/tcpip" @@ -595,6 +598,8 @@ func (ctx *createProcessContext) Value(key interface{}) interface{} { return ctx.k case uniqueid.CtxInotifyCookie: return ctx.k.GenerateInotifyCookie() + case unimpl.CtxEvents: + return ctx.k default: return nil } @@ -1033,6 +1038,16 @@ func (k *Kernel) SupervisorContext() context.Context { } } +// EmitUnimplementedEvent emits an UnimplementedSyscall event via the event +// channel. +func (k *Kernel) EmitUnimplementedEvent(ctx context.Context) { + t := TaskFromContext(ctx) + eventchannel.Emit(&uspb.UnimplementedSyscall{ + Tid: int32(t.ThreadID()), + Registers: t.Arch().StateData().Proto(), + }) +} + type supervisorContext struct { context.NoopSleeper log.Logger @@ -1073,6 +1088,8 @@ func (ctx supervisorContext) Value(key interface{}) interface{} { return ctx.k case uniqueid.CtxInotifyCookie: return ctx.k.GenerateInotifyCookie() + case unimpl.CtxEvents: + return ctx.k default: return nil } diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index 63efc5bbe..36be1efc3 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -19,6 +19,7 @@ import ( "math" "syscall" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" @@ -77,7 +78,7 @@ func (rw *ReaderWriter) Readiness(mask waiter.EventMask) waiter.EventMask { func (rw *ReaderWriter) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { // Switch on ioctl request. switch int(args[1].Int()) { - case syscall.TIOCINQ: + case linux.FIONREAD: v := rw.queuedSize() if v > math.MaxInt32 { panic(fmt.Sprintf("Impossibly large pipe queued size: %d", v)) diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index e22ec768d..73ba8bee9 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -30,6 +30,7 @@ import ( ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -594,6 +595,8 @@ func (t *Task) Value(key interface{}) interface{} { return t.k case uniqueid.CtxInotifyCookie: return t.k.GenerateInotifyCookie() + case unimpl.CtxEvents: + return t.k default: return nil } diff --git a/pkg/sentry/socket/epsocket/BUILD b/pkg/sentry/socket/epsocket/BUILD index dbabc931c..da4aaf510 100644 --- a/pkg/sentry/socket/epsocket/BUILD +++ b/pkg/sentry/socket/epsocket/BUILD @@ -32,6 +32,7 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/socket", "//pkg/sentry/socket/unix/transport", + "//pkg/sentry/unimpl", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/syserror", diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index e90ef4835..39a0b9941 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -45,6 +45,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" + "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" @@ -1184,6 +1185,8 @@ func Ioctl(ctx context.Context, ep commonEndpoint, io usermem.IO, args arch.Sysc }) return 0, err + case linux.SIOCGIFMEM, linux.SIOCGIFPFLAGS, linux.SIOCGMIIPHY, linux.SIOCGMIIREG: + unimpl.EmitUnimplementedEvent(ctx) } return 0, syserror.ENOTTY diff --git a/pkg/sentry/socket/rpcinet/BUILD b/pkg/sentry/socket/rpcinet/BUILD index 3ea433360..38fa54283 100644 --- a/pkg/sentry/socket/rpcinet/BUILD +++ b/pkg/sentry/socket/rpcinet/BUILD @@ -32,6 +32,7 @@ go_library( "//pkg/sentry/socket/rpcinet/conn", "//pkg/sentry/socket/rpcinet/notifier", "//pkg/sentry/socket/unix/transport", + "//pkg/sentry/unimpl", "//pkg/sentry/usermem", "//pkg/syserr", "//pkg/syserror", diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index 44fa5c620..788d853c9 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -32,6 +32,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/socket/rpcinet/notifier" pb "gvisor.googlesource.com/gvisor/pkg/sentry/socket/rpcinet/syscall_rpc_go_proto" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" + "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" @@ -555,6 +556,10 @@ func (s *socketOperations) Ioctl(ctx context.Context, io usermem.IO, args arch.S }) return 0, err + + case linux.SIOCGIFMEM, linux.SIOCGIFPFLAGS, linux.SIOCGMIIPHY, linux.SIOCGMIIREG: + unimpl.EmitUnimplementedEvent(ctx) + default: return 0, syserror.ENOTTY } diff --git a/pkg/sentry/syscalls/BUILD b/pkg/sentry/syscalls/BUILD index 22a757095..2a9f0915e 100644 --- a/pkg/sentry/syscalls/BUILD +++ b/pkg/sentry/syscalls/BUILD @@ -1,7 +1,6 @@ package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_stateify:defs.bzl", "go_library") -load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") go_library( name = "syscalls", @@ -13,9 +12,7 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls", visibility = ["//:sandbox"], deps = [ - ":unimplemented_syscall_go_proto", "//pkg/abi/linux", - "//pkg/eventchannel", "//pkg/sentry/arch", "//pkg/sentry/fs", "//pkg/sentry/kernel", @@ -26,18 +23,3 @@ go_library( "//pkg/waiter", ], ) - -proto_library( - name = "unimplemented_syscall_proto", - srcs = ["unimplemented_syscall.proto"], - visibility = ["//visibility:public"], - deps = ["//pkg/sentry/arch:registers_proto"], -) - -go_proto_library( - name = "unimplemented_syscall_go_proto", - importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls/unimplemented_syscall_go_proto", - proto = ":unimplemented_syscall_proto", - visibility = ["//visibility:public"], - deps = ["//pkg/sentry/arch:registers_go_proto"], -) diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index 75e87f5ec..11bf81f88 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -369,7 +369,7 @@ var AMD64 = &kernel.SyscallTable{ 0xffffffffff600800: 309, // vsyscall getcpu(2) }, Missing: func(t *kernel.Task, sysno uintptr, args arch.SyscallArguments) (uintptr, error) { - syscalls.UnimplementedEvent(t) + t.Kernel().EmitUnimplementedEvent(t) return 0, syserror.ENOSYS }, } diff --git a/pkg/sentry/syscalls/linux/sys_prctl.go b/pkg/sentry/syscalls/linux/sys_prctl.go index c7b39ede8..91e852049 100644 --- a/pkg/sentry/syscalls/linux/sys_prctl.go +++ b/pkg/sentry/syscalls/linux/sys_prctl.go @@ -104,6 +104,22 @@ func Prctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall // Set the underlying executable. t.MemoryManager().SetExecutable(file.Dirent) + + case linux.PR_SET_MM_AUXV, + linux.PR_SET_MM_START_CODE, + linux.PR_SET_MM_END_CODE, + linux.PR_SET_MM_START_DATA, + linux.PR_SET_MM_END_DATA, + linux.PR_SET_MM_START_STACK, + linux.PR_SET_MM_START_BRK, + linux.PR_SET_MM_BRK, + linux.PR_SET_MM_ARG_START, + linux.PR_SET_MM_ARG_END, + linux.PR_SET_MM_ENV_START, + linux.PR_SET_MM_ENV_END: + + t.Kernel().EmitUnimplementedEvent(t) + fallthrough default: return 0, nil, syscall.EINVAL } @@ -151,8 +167,29 @@ func Prctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall } return 0, nil, t.DropBoundingCapability(cp) + case linux.PR_GET_DUMPABLE, + linux.PR_SET_DUMPABLE, + linux.PR_GET_TIMING, + linux.PR_SET_TIMING, + linux.PR_GET_TSC, + linux.PR_SET_TSC, + linux.PR_TASK_PERF_EVENTS_DISABLE, + linux.PR_TASK_PERF_EVENTS_ENABLE, + linux.PR_GET_TIMERSLACK, + linux.PR_SET_TIMERSLACK, + linux.PR_MCE_KILL, + linux.PR_MCE_KILL_GET, + linux.PR_GET_TID_ADDRESS, + linux.PR_SET_CHILD_SUBREAPER, + linux.PR_GET_CHILD_SUBREAPER, + linux.PR_GET_THP_DISABLE, + linux.PR_SET_THP_DISABLE, + linux.PR_MPX_ENABLE_MANAGEMENT, + linux.PR_MPX_DISABLE_MANAGEMENT: + + t.Kernel().EmitUnimplementedEvent(t) + fallthrough default: - t.Warningf("Unsupported prctl %d", option) return 0, nil, syscall.EINVAL } diff --git a/pkg/sentry/syscalls/linux/sys_shm.go b/pkg/sentry/syscalls/linux/sys_shm.go index b13d48b98..5f887523a 100644 --- a/pkg/sentry/syscalls/linux/sys_shm.go +++ b/pkg/sentry/syscalls/linux/sys_shm.go @@ -147,6 +147,7 @@ func Shmctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal // We currently do not support memmory locking anywhere. // mlock(2)/munlock(2) are currently stubbed out as no-ops so do the // same here. + t.Kernel().EmitUnimplementedEvent(t) return 0, nil, nil default: diff --git a/pkg/sentry/syscalls/linux/sys_tls.go b/pkg/sentry/syscalls/linux/sys_tls.go index 27ddb3808..40e84825b 100644 --- a/pkg/sentry/syscalls/linux/sys_tls.go +++ b/pkg/sentry/syscalls/linux/sys_tls.go @@ -45,6 +45,9 @@ func ArchPrctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys regs.Fs = 0 regs.Fs_base = fsbase + case linux.ARCH_GET_GS, linux.ARCH_SET_GS: + t.Kernel().EmitUnimplementedEvent(t) + fallthrough default: return 0, nil, syscall.EINVAL } diff --git a/pkg/sentry/syscalls/syscalls.go b/pkg/sentry/syscalls/syscalls.go index bae32d727..425ce900c 100644 --- a/pkg/sentry/syscalls/syscalls.go +++ b/pkg/sentry/syscalls/syscalls.go @@ -26,10 +26,8 @@ package syscalls import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/eventchannel" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - uspb "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls/unimplemented_syscall_go_proto" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -44,7 +42,7 @@ func Error(err error) kernel.SyscallFn { // syscall event via the event channel and returns the passed error. func ErrorWithEvent(err error) kernel.SyscallFn { return func(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { - UnimplementedEvent(t) + t.Kernel().EmitUnimplementedEvent(t) return 0, nil, err } } @@ -57,16 +55,7 @@ func CapError(c linux.Capability) kernel.SyscallFn { if !t.HasCapability(c) { return 0, nil, syserror.EPERM } - UnimplementedEvent(t) + t.Kernel().EmitUnimplementedEvent(t) return 0, nil, syserror.ENOSYS } } - -// UnimplementedEvent emits an UnimplementedSyscall event via the event -// channel. -func UnimplementedEvent(t *kernel.Task) { - eventchannel.Emit(&uspb.UnimplementedSyscall{ - Tid: int32(t.ThreadID()), - Registers: t.Arch().StateData().Proto(), - }) -} diff --git a/pkg/sentry/syscalls/unimplemented_syscall.proto b/pkg/sentry/syscalls/unimplemented_syscall.proto deleted file mode 100644 index 41579b016..000000000 --- a/pkg/sentry/syscalls/unimplemented_syscall.proto +++ /dev/null @@ -1,27 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -syntax = "proto3"; - -package gvisor; - -import "pkg/sentry/arch/registers.proto"; - -message UnimplementedSyscall { - // Task ID. - int32 tid = 1; - - // Registers at the time of the call. - Registers registers = 2; -} diff --git a/pkg/sentry/unimpl/BUILD b/pkg/sentry/unimpl/BUILD new file mode 100644 index 000000000..63da5e81f --- /dev/null +++ b/pkg/sentry/unimpl/BUILD @@ -0,0 +1,30 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("//tools/go_stateify:defs.bzl", "go_library") +load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") + +proto_library( + name = "unimplemented_syscall_proto", + srcs = ["unimplemented_syscall.proto"], + visibility = ["//visibility:public"], + deps = ["//pkg/sentry/arch:registers_proto"], +) + +go_proto_library( + name = "unimplemented_syscall_go_proto", + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl/unimplemented_syscall_go_proto", + proto = ":unimplemented_syscall_proto", + visibility = ["//visibility:public"], + deps = ["//pkg/sentry/arch:registers_go_proto"], +) + +go_library( + name = "unimpl", + srcs = ["events.go"], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl", + visibility = ["//:sandbox"], + deps = [ + "//pkg/log", + "//pkg/sentry/context", + ], +) diff --git a/pkg/sentry/unimpl/events.go b/pkg/sentry/unimpl/events.go new file mode 100644 index 000000000..f78f8c981 --- /dev/null +++ b/pkg/sentry/unimpl/events.go @@ -0,0 +1,45 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package unimpl contains interface to emit events about unimplemented +// features. +package unimpl + +import ( + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" +) + +// contextID is the events package's type for context.Context.Value keys. +type contextID int + +const ( + // CtxEvents is a Context.Value key for a Events. + CtxEvents contextID = iota +) + +// Events interface defines method to emit unsupported events. +type Events interface { + EmitUnimplementedEvent(context.Context) +} + +// EmitUnimplementedEvent emits unsupported syscall event to the context. +func EmitUnimplementedEvent(ctx context.Context) { + e := ctx.Value(CtxEvents) + if e == nil { + log.Warningf("Context.Value(CtxEvents) not present, unimplemented syscall event not reported.") + return + } + e.(Events).EmitUnimplementedEvent(ctx) +} diff --git a/pkg/sentry/unimpl/unimplemented_syscall.proto b/pkg/sentry/unimpl/unimplemented_syscall.proto new file mode 100644 index 000000000..41579b016 --- /dev/null +++ b/pkg/sentry/unimpl/unimplemented_syscall.proto @@ -0,0 +1,27 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +syntax = "proto3"; + +package gvisor; + +import "pkg/sentry/arch/registers.proto"; + +message UnimplementedSyscall { + // Task ID. + int32 tid = 1; + + // Registers at the time of the call. + Registers registers = 2; +} diff --git a/runsc/boot/BUILD b/runsc/boot/BUILD index f8f848ebf..04cc0e854 100644 --- a/runsc/boot/BUILD +++ b/runsc/boot/BUILD @@ -6,6 +6,7 @@ go_library( name = "boot", srcs = [ "compat.go", + "compat_amd64.go", "config.go", "controller.go", "debug.go", @@ -59,9 +60,9 @@ go_library( "//pkg/sentry/socket/unix", "//pkg/sentry/state", "//pkg/sentry/strace", - "//pkg/sentry/syscalls:unimplemented_syscall_go_proto", "//pkg/sentry/syscalls/linux", "//pkg/sentry/time", + "//pkg/sentry/unimpl:unimplemented_syscall_go_proto", "//pkg/sentry/usage", "//pkg/sentry/watchdog", "//pkg/syserror", @@ -87,12 +88,16 @@ go_library( go_test( name = "boot_test", size = "small", - srcs = ["loader_test.go"], + srcs = [ + "compat_test.go", + "loader_test.go", + ], embed = [":boot"], deps = [ "//pkg/control/server", "//pkg/log", "//pkg/p9", + "//pkg/sentry/arch:registers_go_proto", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/unet", diff --git a/runsc/boot/compat.go b/runsc/boot/compat.go index 6766953b3..d18c2f802 100644 --- a/runsc/boot/compat.go +++ b/runsc/boot/compat.go @@ -17,6 +17,8 @@ package boot import ( "fmt" "os" + "sync" + "syscall" "github.com/golang/protobuf/proto" "gvisor.googlesource.com/gvisor/pkg/abi" @@ -25,7 +27,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/arch" rpb "gvisor.googlesource.com/gvisor/pkg/sentry/arch/registers_go_proto" "gvisor.googlesource.com/gvisor/pkg/sentry/strace" - spb "gvisor.googlesource.com/gvisor/pkg/sentry/syscalls/unimplemented_syscall_go_proto" + spb "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl/unimplemented_syscall_go_proto" ) func initCompatLogs(fd int) error { @@ -40,15 +42,27 @@ func initCompatLogs(fd int) error { type compatEmitter struct { sink *log.BasicLogger nameMap strace.SyscallMap + + // mu protects the fields below. + mu sync.Mutex + + // trackers map syscall number to the respective tracker instance. + // Protected by 'mu'. + trackers map[uint64]syscallTracker } func newCompatEmitter(logFD int) (*compatEmitter, error) { - // Always logs to default logger. nameMap, ok := strace.Lookup(abi.Linux, arch.AMD64) if !ok { return nil, fmt.Errorf("amd64 Linux syscall table not found") } - c := &compatEmitter{sink: log.Log(), nameMap: nameMap} + + c := &compatEmitter{ + // Always logs to default logger. + sink: log.Log(), + nameMap: nameMap, + trackers: make(map[uint64]syscallTracker), + } if logFD > 0 { f := os.NewFile(uintptr(logFD), "user log file") @@ -61,10 +75,33 @@ func newCompatEmitter(logFD int) (*compatEmitter, error) { // Emit implements eventchannel.Emitter. func (c *compatEmitter) Emit(msg proto.Message) (hangup bool, err error) { // Only interested in UnimplementedSyscall, skip the rest. - if us, ok := msg.(*spb.UnimplementedSyscall); ok { - regs := us.Registers.GetArch().(*rpb.Registers_Amd64).Amd64 - sysnr := regs.OrigRax + us, ok := msg.(*spb.UnimplementedSyscall) + if !ok { + return false, nil + } + regs := us.Registers.GetArch().(*rpb.Registers_Amd64).Amd64 + + c.mu.Lock() + defer c.mu.Unlock() + + sysnr := regs.OrigRax + tr := c.trackers[sysnr] + if tr == nil { + switch sysnr { + case syscall.SYS_PRCTL, syscall.SYS_ARCH_PRCTL: + tr = newCmdTracker(0) + + case syscall.SYS_IOCTL, syscall.SYS_EPOLL_CTL, syscall.SYS_SHMCTL: + tr = newCmdTracker(1) + + default: + tr = &onceTracker{} + } + c.trackers[sysnr] = tr + } + if tr.shouldReport(regs) { c.sink.Infof("Unsupported syscall: %s, regs: %+v", c.nameMap.Name(uintptr(sysnr)), regs) + tr.onReported(regs) } return false, nil } @@ -74,3 +111,26 @@ func (c *compatEmitter) Close() error { c.sink = nil return nil } + +// syscallTracker interface allows filters to apply differently depending on +// the syscall and arguments. +type syscallTracker interface { + // shouldReport returns true is the syscall should be reported. + shouldReport(regs *rpb.AMD64Registers) bool + + // onReported marks the syscall as reported. + onReported(regs *rpb.AMD64Registers) +} + +// onceTracker reports only a single time, used for most syscalls. +type onceTracker struct { + reported bool +} + +func (o *onceTracker) shouldReport(_ *rpb.AMD64Registers) bool { + return !o.reported +} + +func (o *onceTracker) onReported(_ *rpb.AMD64Registers) { + o.reported = true +} diff --git a/runsc/boot/compat_amd64.go b/runsc/boot/compat_amd64.go new file mode 100644 index 000000000..2bb769a49 --- /dev/null +++ b/runsc/boot/compat_amd64.go @@ -0,0 +1,54 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package boot + +import ( + "fmt" + + rpb "gvisor.googlesource.com/gvisor/pkg/sentry/arch/registers_go_proto" +) + +// cmdTracker reports only a single time for each different command argument in +// the syscall. It's used for generic syscalls like ioctl to report once per +// 'cmd' +type cmdTracker struct { + // argIdx is the syscall argument index where the command is located. + argIdx int + cmds map[uint32]struct{} +} + +func newCmdTracker(argIdx int) *cmdTracker { + return &cmdTracker{argIdx: argIdx, cmds: make(map[uint32]struct{})} +} + +// cmd returns the command based on the syscall argument index. +func (c *cmdTracker) cmd(regs *rpb.AMD64Registers) uint32 { + switch c.argIdx { + case 0: + return uint32(regs.Rdi) + case 1: + return uint32(regs.Rsi) + } + panic(fmt.Sprintf("unsupported syscall argument index %d", c.argIdx)) +} + +func (c *cmdTracker) shouldReport(regs *rpb.AMD64Registers) bool { + _, ok := c.cmds[c.cmd(regs)] + return !ok +} + +func (c *cmdTracker) onReported(regs *rpb.AMD64Registers) { + c.cmds[c.cmd(regs)] = struct{}{} +} diff --git a/runsc/boot/compat_test.go b/runsc/boot/compat_test.go new file mode 100644 index 000000000..30b94798a --- /dev/null +++ b/runsc/boot/compat_test.go @@ -0,0 +1,66 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package boot + +import ( + "testing" + + rpb "gvisor.googlesource.com/gvisor/pkg/sentry/arch/registers_go_proto" +) + +func TestOnceTracker(t *testing.T) { + o := onceTracker{} + if !o.shouldReport(nil) { + t.Error("first call to checkAndMark, got: false, want: true") + } + o.onReported(nil) + for i := 0; i < 2; i++ { + if o.shouldReport(nil) { + t.Error("after first call to checkAndMark, got: true, want: false") + } + } +} + +func TestCmdTracker(t *testing.T) { + for _, tc := range []struct { + name string + idx int + rdi1 uint64 + rdi2 uint64 + rsi1 uint64 + rsi2 uint64 + want bool + }{ + {name: "same rdi", idx: 0, rdi1: 123, rdi2: 123, want: false}, + {name: "same rsi", idx: 1, rsi1: 123, rsi2: 123, want: false}, + {name: "diff rdi", idx: 0, rdi1: 123, rdi2: 321, want: true}, + {name: "diff rsi", idx: 1, rsi1: 123, rsi2: 321, want: true}, + {name: "cmd is uint32", idx: 0, rsi1: 0xdead00000123, rsi2: 0xbeef00000123, want: false}, + } { + t.Run(tc.name, func(t *testing.T) { + c := newCmdTracker(tc.idx) + regs := &rpb.AMD64Registers{Rdi: tc.rdi1, Rsi: tc.rsi1} + if !c.shouldReport(regs) { + t.Error("first call to checkAndMark, got: false, want: true") + } + c.onReported(regs) + + regs.Rdi, regs.Rsi = tc.rdi2, tc.rsi2 + if got := c.shouldReport(regs); tc.want != got { + t.Errorf("after first call to checkAndMark, got: %t, want: %t", got, tc.want) + } + }) + } +} -- cgit v1.2.3 From 75cd70ecc9abfd5daaefea04da5070a0e0d620dd Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Tue, 23 Oct 2018 00:19:11 -0700 Subject: Track paths and provide a rename hook. This change also adds extensive testing to the p9 package via mocks. The sanity checks and type checks are moved from the gofer into the core package, where they can be more easily validated. PiperOrigin-RevId: 218296768 Change-Id: I4fc3c326e7bf1e0e140a454cbacbcc6fd617ab55 --- WORKSPACE | 20 +- pkg/amutex/BUILD | 4 +- pkg/atomicbitops/BUILD | 4 +- pkg/binary/BUILD | 4 +- pkg/bits/BUILD | 3 +- pkg/compressio/BUILD | 4 +- pkg/control/client/BUILD | 4 +- pkg/control/server/BUILD | 4 +- pkg/dhcp/BUILD | 4 +- pkg/eventchannel/BUILD | 4 +- pkg/fd/BUILD | 4 +- pkg/gate/BUILD | 4 +- pkg/ilist/BUILD | 4 +- pkg/linewriter/BUILD | 4 +- pkg/log/BUILD | 4 +- pkg/metric/BUILD | 4 +- pkg/p9/BUILD | 2 + pkg/p9/buffer_test.go | 31 + pkg/p9/client.go | 6 + pkg/p9/client_file.go | 4 +- pkg/p9/file.go | 151 +- pkg/p9/handlers.go | 697 ++++++-- pkg/p9/local_server/BUILD | 4 +- pkg/p9/local_server/local_server.go | 5 + pkg/p9/messages_test.go | 37 + pkg/p9/p9.go | 24 + pkg/p9/p9test/BUILD | 76 +- pkg/p9/p9test/client_test.go | 2263 ++++++++++++++++++++++--- pkg/p9/p9test/mocks.go | 489 ------ pkg/p9/p9test/p9test.go | 329 ++++ pkg/p9/path_tree.go | 109 ++ pkg/p9/server.go | 228 ++- pkg/p9/transport.go | 10 +- pkg/rand/BUILD | 4 +- pkg/seccomp/BUILD | 4 +- pkg/secio/BUILD | 4 +- pkg/sentry/arch/BUILD | 3 +- pkg/sentry/context/BUILD | 4 +- pkg/sentry/control/BUILD | 4 +- pkg/sentry/device/BUILD | 4 +- pkg/sentry/fs/anon/BUILD | 4 +- pkg/sentry/fs/gofer/BUILD | 4 - pkg/sentry/fs/gofer/context_file.go | 7 - pkg/sentry/fs/gofer/gofer_test.go | 894 ++-------- pkg/sentry/fs/gofer/session.go | 9 +- pkg/sentry/fs/gofer/session_state.go | 4 +- pkg/sentry/fs/proc/device/BUILD | 4 +- pkg/sentry/hostcpu/BUILD | 4 +- pkg/sentry/kernel/kdefs/BUILD | 4 +- pkg/sentry/kernel/memevent/BUILD | 4 +- pkg/sentry/kernel/sched/BUILD | 4 +- pkg/sentry/loader/BUILD | 3 +- pkg/sentry/memutil/BUILD | 4 +- pkg/sentry/platform/interrupt/BUILD | 4 +- pkg/sentry/platform/kvm/BUILD | 3 +- pkg/sentry/platform/kvm/testutil/BUILD | 4 +- pkg/sentry/platform/procid/BUILD | 4 +- pkg/sentry/platform/ptrace/BUILD | 4 +- pkg/sentry/platform/ring0/BUILD | 3 +- pkg/sentry/platform/ring0/gen_offsets/BUILD | 3 +- pkg/sentry/platform/ring0/pagetables/BUILD | 3 +- pkg/sentry/platform/safecopy/BUILD | 4 +- pkg/sentry/safemem/BUILD | 4 +- pkg/sentry/sighandling/BUILD | 4 +- pkg/sentry/socket/rpcinet/BUILD | 4 +- pkg/sentry/socket/rpcinet/conn/BUILD | 4 +- pkg/sentry/socket/rpcinet/notifier/BUILD | 4 +- pkg/sentry/state/BUILD | 4 +- pkg/sentry/strace/BUILD | 4 +- pkg/sentry/syscalls/BUILD | 4 +- pkg/sentry/time/BUILD | 3 +- pkg/sentry/unimpl/BUILD | 4 +- pkg/sentry/uniqueid/BUILD | 4 +- pkg/sentry/watchdog/BUILD | 4 +- pkg/sleep/BUILD | 4 +- pkg/state/BUILD | 5 +- pkg/state/statefile/BUILD | 4 +- pkg/sync/atomicptrtest/BUILD | 3 +- pkg/sync/seqatomictest/BUILD | 3 +- pkg/syserr/BUILD | 4 +- pkg/syserror/BUILD | 4 +- pkg/tcpip/adapters/gonet/BUILD | 4 +- pkg/tcpip/checker/BUILD | 4 +- pkg/tcpip/link/channel/BUILD | 4 +- pkg/tcpip/link/fdbased/BUILD | 4 +- pkg/tcpip/link/loopback/BUILD | 4 +- pkg/tcpip/link/rawfile/BUILD | 4 +- pkg/tcpip/link/sharedmem/BUILD | 4 +- pkg/tcpip/link/sharedmem/pipe/BUILD | 4 +- pkg/tcpip/link/sharedmem/queue/BUILD | 4 +- pkg/tcpip/link/sniffer/BUILD | 4 +- pkg/tcpip/link/tun/BUILD | 4 +- pkg/tcpip/link/waitable/BUILD | 4 +- pkg/tcpip/network/BUILD | 4 +- pkg/tcpip/network/arp/BUILD | 4 +- pkg/tcpip/network/hash/BUILD | 4 +- pkg/tcpip/network/ipv4/BUILD | 4 +- pkg/tcpip/network/ipv6/BUILD | 4 +- pkg/tcpip/ports/BUILD | 4 +- pkg/tcpip/sample/tun_tcp_connect/BUILD | 4 +- pkg/tcpip/sample/tun_tcp_echo/BUILD | 4 +- pkg/tcpip/transport/tcp/testing/context/BUILD | 4 +- pkg/tcpip/transport/tcpconntrack/BUILD | 4 +- pkg/tmutex/BUILD | 4 +- pkg/unet/BUILD | 4 +- pkg/urpc/BUILD | 4 +- pkg/waiter/fdnotifier/BUILD | 4 +- runsc/boot/BUILD | 4 +- runsc/boot/filter/BUILD | 4 +- runsc/cgroup/BUILD | 4 +- runsc/cmd/BUILD | 4 +- runsc/console/BUILD | 4 +- runsc/container/BUILD | 4 +- runsc/fsgofer/BUILD | 4 +- runsc/fsgofer/filter/BUILD | 4 +- runsc/fsgofer/fsgofer.go | 98 +- runsc/fsgofer/fsgofer_test.go | 78 +- runsc/sandbox/BUILD | 4 +- runsc/specutils/BUILD | 4 +- runsc/test/image/BUILD | 4 +- runsc/test/integration/BUILD | 4 +- runsc/test/root/BUILD | 4 +- runsc/test/testutil/BUILD | 4 +- runsc/tools/dockercfg/BUILD | 4 +- tools/go_generics/BUILD | 4 +- tools/go_generics/globals/BUILD | 4 +- tools/go_generics/rules_tests/BUILD | 3 +- tools/go_stateify/BUILD | 4 +- 128 files changed, 3834 insertions(+), 2147 deletions(-) create mode 100644 pkg/p9/buffer_test.go delete mode 100644 pkg/p9/p9test/mocks.go create mode 100644 pkg/p9/p9test/p9test.go create mode 100644 pkg/p9/path_tree.go (limited to 'pkg/sentry/kernel') diff --git a/WORKSPACE b/WORKSPACE index 48e0d3436..841a23e06 100644 --- a/WORKSPACE +++ b/WORKSPACE @@ -15,7 +15,7 @@ go_register_toolchains(go_version="1.11.1") load("@bazel_gazelle//:deps.bzl", "gazelle_dependencies", "go_repository") gazelle_dependencies() -# Add dependencies on external repositories. +# External repositories, in sorted order. go_repository( name = "com_github_cenkalti_backoff", importpath = "github.com/cenkalti/backoff", @@ -28,6 +28,12 @@ go_repository( commit = "886344bea0798d02ff3fae16a922be5f6b26cee0" ) +go_repository( + name = "com_github_golang_mock", + importpath = "github.com/golang/mock", + commit = "600781dde9cca80734169b9e969d9054ccc57937", +) + go_repository( name = "com_github_google_go-cmp", importpath = "github.com/google/go-cmp", @@ -58,6 +64,12 @@ go_repository( commit = "b2d941ef6a780da2d9982c1fb28d77ad97f54fc7", ) +go_repository( + name = "com_github_syndtr_gocapability", + importpath = "github.com/syndtr/gocapability", + commit = "d98352740cb2c55f81556b63d4a1ec64c5a319c2", +) + go_repository( name = "com_github_vishvananda_netlink", importpath = "github.com/vishvananda/netlink", @@ -81,9 +93,3 @@ go_repository( importpath = "golang.org/x/sys", commit = "0dd5e194bbf5eb84a39666eb4c98a4d007e4203a", ) - -go_repository( - name = "com_github_syndtr_gocapability", - importpath = "github.com/syndtr/gocapability", - commit = "d98352740cb2c55f81556b63d4a1ec64c5a319c2", -) diff --git a/pkg/amutex/BUILD b/pkg/amutex/BUILD index 84e6b79a5..815ee3a69 100644 --- a/pkg/amutex/BUILD +++ b/pkg/amutex/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "amutex", srcs = ["amutex.go"], diff --git a/pkg/atomicbitops/BUILD b/pkg/atomicbitops/BUILD index a8dd17825..235188531 100644 --- a/pkg/atomicbitops/BUILD +++ b/pkg/atomicbitops/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "atomicbitops", srcs = [ diff --git a/pkg/binary/BUILD b/pkg/binary/BUILD index 586d05634..571151f72 100644 --- a/pkg/binary/BUILD +++ b/pkg/binary/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "binary", srcs = ["binary.go"], diff --git a/pkg/bits/BUILD b/pkg/bits/BUILD index 8c943b615..46794bdb8 100644 --- a/pkg/bits/BUILD +++ b/pkg/bits/BUILD @@ -1,6 +1,7 @@ +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") + package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") go_library( diff --git a/pkg/compressio/BUILD b/pkg/compressio/BUILD index d70f982c1..72952d735 100644 --- a/pkg/compressio/BUILD +++ b/pkg/compressio/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "compressio", srcs = ["compressio.go"], diff --git a/pkg/control/client/BUILD b/pkg/control/client/BUILD index d58cd1b71..32853875d 100644 --- a/pkg/control/client/BUILD +++ b/pkg/control/client/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "client", srcs = [ diff --git a/pkg/control/server/BUILD b/pkg/control/server/BUILD index c3f74a532..ba2b1be9f 100644 --- a/pkg/control/server/BUILD +++ b/pkg/control/server/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "server", srcs = ["server.go"], diff --git a/pkg/dhcp/BUILD b/pkg/dhcp/BUILD index 711a72c99..c97dfc14b 100644 --- a/pkg/dhcp/BUILD +++ b/pkg/dhcp/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "dhcp", srcs = [ diff --git a/pkg/eventchannel/BUILD b/pkg/eventchannel/BUILD index 9d531ce12..18348ef54 100644 --- a/pkg/eventchannel/BUILD +++ b/pkg/eventchannel/BUILD @@ -1,8 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "eventchannel", srcs = [ diff --git a/pkg/fd/BUILD b/pkg/fd/BUILD index 435b6fa34..06cfd445e 100644 --- a/pkg/fd/BUILD +++ b/pkg/fd/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "fd", srcs = ["fd.go"], diff --git a/pkg/gate/BUILD b/pkg/gate/BUILD index 872eff531..9a87a3a31 100644 --- a/pkg/gate/BUILD +++ b/pkg/gate/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "gate", srcs = [ diff --git a/pkg/ilist/BUILD b/pkg/ilist/BUILD index 1bd71b800..a67aa2cff 100644 --- a/pkg/ilist/BUILD +++ b/pkg/ilist/BUILD @@ -1,8 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "ilist", srcs = [ diff --git a/pkg/linewriter/BUILD b/pkg/linewriter/BUILD index 6c3795432..3f28ba867 100644 --- a/pkg/linewriter/BUILD +++ b/pkg/linewriter/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "linewriter", srcs = ["linewriter.go"], diff --git a/pkg/log/BUILD b/pkg/log/BUILD index fc9281079..bf85b4494 100644 --- a/pkg/log/BUILD +++ b/pkg/log/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "log", srcs = [ diff --git a/pkg/metric/BUILD b/pkg/metric/BUILD index c0cd40c7b..d96e5563b 100644 --- a/pkg/metric/BUILD +++ b/pkg/metric/BUILD @@ -1,8 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "metric", srcs = ["metric.go"], diff --git a/pkg/p9/BUILD b/pkg/p9/BUILD index 1cf5c6458..2c224e65b 100644 --- a/pkg/p9/BUILD +++ b/pkg/p9/BUILD @@ -15,6 +15,7 @@ go_library( "handlers.go", "messages.go", "p9.go", + "path_tree.go", "pool.go", "server.go", "transport.go", @@ -32,6 +33,7 @@ go_test( name = "p9_test", size = "small", srcs = [ + "buffer_test.go", "client_test.go", "messages_test.go", "p9_test.go", diff --git a/pkg/p9/buffer_test.go b/pkg/p9/buffer_test.go new file mode 100644 index 000000000..97eceefa7 --- /dev/null +++ b/pkg/p9/buffer_test.go @@ -0,0 +1,31 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package p9 + +import ( + "testing" +) + +func TestBufferOverrun(t *testing.T) { + buf := &buffer{ + // This header indicates that a large string should follow, but + // it is only two bytes. Reading a string should cause an + // overrun. + data: []byte{0x0, 0x16}, + } + if s := buf.ReadString(); s != "" { + t.Errorf("overrun read got %s, want empty", s) + } +} diff --git a/pkg/p9/client.go b/pkg/p9/client.go index 3ebfab82a..67887874a 100644 --- a/pkg/p9/client.go +++ b/pkg/p9/client.go @@ -116,6 +116,7 @@ func NewClient(socket *unet.Socket, messageSize uint32, version string) (*Client msize: largestFixedSize, } } + // Compute a payload size and round to 512 (normal block size) // if it's larger than a single block. payloadSize := messageSize - largestFixedSize @@ -299,3 +300,8 @@ func (c *Client) sendRecv(t message, r message) error { func (c *Client) Version() uint32 { return c.version } + +// Close closes the underlying socket. +func (c *Client) Close() error { + return c.socket.Close() +} diff --git a/pkg/p9/client_file.go b/pkg/p9/client_file.go index 066639fda..992d1daf7 100644 --- a/pkg/p9/client_file.go +++ b/pkg/p9/client_file.go @@ -172,6 +172,9 @@ func (c *clientFile) SetAttr(valid SetAttrMask, attr SetAttr) error { } // Remove implements File.Remove. +// +// N.B. This method is no longer part of the file interface and should be +// considered deprecated. func (c *clientFile) Remove() error { // Avoid double close. if !atomic.CompareAndSwapUint32(&c.closed, 0, 1) { @@ -181,7 +184,6 @@ func (c *clientFile) Remove() error { // Send the remove message. if err := c.client.sendRecv(&Tremove{FID: c.fid}, &Rremove{}); err != nil { - log.Warningf("Tremove failed, losing FID %v: %v", c.fid, err) return err } diff --git a/pkg/p9/file.go b/pkg/p9/file.go index d2e89e373..55ceb52e1 100644 --- a/pkg/p9/file.go +++ b/pkg/p9/file.go @@ -31,35 +31,63 @@ type Attacher interface { // File is a set of operations corresponding to a single node. // -// Functions below MUST return syscall.Errno values. -// TODO: Enforce that with the type. +// Note that on the server side, the server logic places constraints on +// concurrent operations to make things easier. This may reduce the need for +// complex, error-prone locking and logic in the backend. These are documented +// for each method. // -// These must be implemented in all circumstances. +// There are three different types of guarantees provided: +// +// none: There is no concurrency guarantee. The method may be invoked +// concurrently with any other method on any other file. +// +// read: The method is guaranteed to be exclusive of any write or global +// operation that is mutating the state of the directory tree starting at this +// node. For example, this means creating new files, symlinks, directories or +// renaming a directory entry (or renaming in to this target), but the method +// may be called concurrently with other read methods. +// +// write: The method is guaranteed to be exclusive of any read, write or global +// operation that is mutating the state of the directory tree starting at this +// node, as described in read above. There may however, be other write +// operations executing concurrently on other components in the directory tree. +// +// global: The method is guaranteed to be exclusive of any read, write or +// global operation. type File interface { // Walk walks to the path components given in names. // // Walk returns QIDs in the same order that the names were passed in. // // An empty list of arguments should return a copy of the current file. + // + // On the server, Walk has a read concurrency guarantee. Walk(names []string) ([]QID, File, error) + // WalkGetAttr walks to the next file and returns its maximal set of + // attributes. + // + // Server-side p9.Files may return syscall.ENOSYS to indicate that Walk + // and GetAttr should be used separately to satisfy this request. + // + // On the server, WalkGetAttr has a read concurrency guarantee. + WalkGetAttr([]string) ([]QID, File, AttrMask, Attr, error) + // StatFS returns information about the file system associated with // this file. + // + // On the server, StatFS has no concurrency guarantee. StatFS() (FSStat, error) // GetAttr returns attributes of this node. + // + // On the server, GetAttr has a read concurrency guarantee. GetAttr(req AttrMask) (QID, AttrMask, Attr, error) // SetAttr sets attributes on this node. - SetAttr(valid SetAttrMask, attr SetAttr) error - - // Remove removes the file. // - // This is deprecated in favor of UnlinkAt below. - Remove() error - - // Rename renames the file. - Rename(directory File, name string) error + // On the server, SetAttr has a write concurrency guarantee. + SetAttr(valid SetAttrMask, attr SetAttr) error // Close is called when all references are dropped on the server side, // and Close should be called by the client to drop all references. @@ -67,65 +95,93 @@ type File interface { // For server-side implementations of Close, the error is ignored. // // Close must be called even when Open has not been called. + // + // On the server, Close has no concurrency guarantee. Close() error - // Open is called prior to using read/write. + // Open must be called prior to using Read, Write or Readdir. Once Open + // is called, some operations, such as Walk, will no longer work. // - // The *fd.FD may be nil. If an *fd.FD is provided, ownership now - // belongs to the caller and the FD must be non-blocking. + // On the client, Open should be called only once. The fd return is + // optional, and may be nil. // - // If Open returns a non-nil *fd.FD, it should do so for all possible - // OpenFlags. If Open returns a nil *fd.FD, it should similarly return - // a nil *fd.FD for all possible OpenFlags. + // On the server, Open has a read concurrency guarantee. If an *fd.FD + // is provided, ownership now belongs to the caller. Open is guaranteed + // to be called only once. // - // This can be assumed to be one-shot only. + // N.B. The server must resolve any lazy paths when open is called. + // After this point, read and write may be called on files with no + // deletion check, so resolving in the data path is not viable. Open(mode OpenFlags) (*fd.FD, QID, uint32, error) - // Read reads from this file. + // Read reads from this file. Open must be called first. // // This may return io.EOF in addition to syscall.Errno values. // - // Preconditions: Open has been called and returned success. + // On the server, ReadAt has a read concurrency guarantee. See Open for + // additional requirements regarding lazy path resolution. ReadAt(p []byte, offset uint64) (int, error) - // Write writes to this file. + // Write writes to this file. Open must be called first. // // This may return io.EOF in addition to syscall.Errno values. // - // Preconditions: Open has been called and returned success. + // On the server, WriteAt has a read concurrency guarantee. See Open + // for additional requirements regarding lazy path resolution. WriteAt(p []byte, offset uint64) (int, error) - // FSync syncs this node. + // FSync syncs this node. Open must be called first. // - // Preconditions: Open has been called and returned success. + // On the server, FSync has a read concurrency guarantee. FSync() error // Create creates a new regular file and opens it according to the - // flags given. + // flags given. This file is already Open. + // + // N.B. On the client, the returned file is a reference to the current + // file, which now represents the created file. This is not the case on + // the server. These semantics are very subtle and can easily lead to + // bugs, but are a consequence of the 9P create operation. // // See p9.File.Open for a description of *fd.FD. + // + // On the server, Create has a write concurrency guarantee. Create(name string, flags OpenFlags, permissions FileMode, uid UID, gid GID) (*fd.FD, File, QID, uint32, error) // Mkdir creates a subdirectory. + // + // On the server, Mkdir has a write concurrency guarantee. Mkdir(name string, permissions FileMode, uid UID, gid GID) (QID, error) // Symlink makes a new symbolic link. - Symlink(oldname string, newname string, uid UID, gid GID) (QID, error) + // + // On the server, Symlink has a write concurrency guarantee. + Symlink(oldName string, newName string, uid UID, gid GID) (QID, error) // Link makes a new hard link. - Link(target File, newname string) error + // + // On the server, Link has a write concurrency guarantee. + Link(target File, newName string) error // Mknod makes a new device node. + // + // On the server, Mknod has a write concurrency guarantee. Mknod(name string, permissions FileMode, major uint32, minor uint32, uid UID, gid GID) (QID, error) + // Rename renames the file. + // + // Rename will never be called on the server, and RenameAt will always + // be used instead. + Rename(newDir File, newName string) error + // RenameAt renames a given file to a new name in a potentially new // directory. // - // oldname must be a name relative to this file, which must be a - // directory. newname is a name relative to newdir. + // oldName must be a name relative to this file, which must be a + // directory. newName is a name relative to newDir. // - // This is deprecated in favor of Rename. - RenameAt(oldname string, newdir File, newname string) error + // On the server, RenameAt has a global concurrency guarantee. + RenameAt(oldName string, newDir File, newName string) error // UnlinkAt the given named file. // @@ -133,16 +189,20 @@ type File interface { // // Flags are implementation-specific (e.g. O_DIRECTORY), but are // generally Linux unlinkat(2) flags. + // + // On the server, UnlinkAt has a write concurrency guarantee. UnlinkAt(name string, flags uint32) error // Readdir reads directory entries. // // This may return io.EOF in addition to syscall.Errno values. // - // Preconditions: Open has been called and returned success. + // On the server, Readdir has a read concurrency guarantee. Readdir(offset uint64, count uint32) ([]Dirent, error) // Readlink reads the link target. + // + // On the server, Readlink has a read concurrency guarantee. Readlink() (string, error) // Flush is called prior to Close. @@ -150,16 +210,11 @@ type File interface { // Whereas Close drops all references to the file, Flush cleans up the // file state. Behavior is implementation-specific. // - // Flush is not related to flush(9p). Flush is an extension to 9P2000.L, + // Flush is not related to flush(9p). Flush is an extension to 9P2000.L, // see version.go. - Flush() error - - // WalkGetAttr walks to the next file and returns its maximal set of - // attributes. // - // Server-side p9.Files may return syscall.ENOSYS to indicate that Walk - // and GetAttr should be used separately to satisfy this request. - WalkGetAttr([]string) ([]QID, File, AttrMask, Attr, error) + // On the server, Flush has a read concurrency guarantee. + Flush() error // Connect establishes a new host-socket backed connection with a // socket. A File does not need to be opened before it can be connected @@ -170,8 +225,22 @@ type File interface { // // The returned FD must be non-blocking. // - // flags indicates the requested type of socket. + // Flags indicates the requested type of socket. + // + // On the server, Connect has a read concurrency guarantee. Connect(flags ConnectFlags) (*fd.FD, error) + + // Renamed is called when this node is renamed. + // + // This may not fail. The file will hold a reference to its parent + // within the p9 package, and is therefore safe to use for the lifetime + // of this File (until Close is called). + // + // This method should not be called by clients, who should use the + // relevant Rename methods. (Although the method will be a no-op.) + // + // On the server, Renamed has a global concurrency guarantee. + Renamed(newDir File, newName string) } // DefaultWalkGetAttr implements File.WalkGetAttr to return ENOSYS for server-side Files. diff --git a/pkg/p9/handlers.go b/pkg/p9/handlers.go index 959dff31d..0d7a6138f 100644 --- a/pkg/p9/handlers.go +++ b/pkg/p9/handlers.go @@ -15,6 +15,7 @@ package p9 import ( + "fmt" "io" "os" "path" @@ -22,22 +23,43 @@ import ( "sync/atomic" "syscall" + "gvisor.googlesource.com/gvisor/pkg/fd" "gvisor.googlesource.com/gvisor/pkg/log" ) -// newErr returns a new error message from an error. -func newErr(err error) *Rlerror { +const maximumNameLength = 255 + +// ExtractErrno extracts a syscall.Errno from a error, best effort. +func ExtractErrno(err error) syscall.Errno { + switch err { + case os.ErrNotExist: + return syscall.ENOENT + case os.ErrExist: + return syscall.EEXIST + case os.ErrPermission: + return syscall.EACCES + case os.ErrInvalid: + return syscall.EINVAL + } + + // Attempt to unwrap. switch e := err.(type) { case syscall.Errno: - return &Rlerror{Error: uint32(e)} + return e case *os.PathError: - return newErr(e.Err) + return ExtractErrno(e.Err) case *os.SyscallError: - return newErr(e.Err) - default: - log.Warningf("unknown error: %v", err) - return &Rlerror{Error: uint32(syscall.EIO)} + return ExtractErrno(e.Err) } + + // Default case. + log.Warningf("unknown error: %v", err) + return syscall.EIO +} + +// newErr returns a new error message from an error. +func newErr(err error) *Rlerror { + return &Rlerror{Error: uint32(ExtractErrno(err))} } // handler is implemented for server-handled messages. @@ -85,13 +107,15 @@ func (t *Tflush) handle(cs *connState) message { return &Rflush{} } -// isSafeName returns true iff the name does not contain directory characters. -// -// We permit walks only on safe names and store the sequence of paths used for -// any given walk in each FID. (This is immutable.) We use this to mark -// relevant FIDs as moved when a successful rename occurs. -func isSafeName(name string) bool { - return name != "" && !strings.Contains(name, "/") && name != "." && name != ".." +// checkSafeName validates the name and returns nil or returns an error. +func checkSafeName(name string) error { + if name == "" || strings.Contains(name, "/") || name == "." || name == ".." { + return syscall.EINVAL + } + if len(name) > maximumNameLength { + return syscall.ENAMETOOLONG + } + return nil } // handle implements handler.handle. @@ -110,22 +134,54 @@ func (t *Tremove) handle(cs *connState) message { } defer ref.DecRef() + // Frustratingly, because we can't be guaranteed that a rename is not + // occurring simultaneously with this removal, we need to acquire the + // global rename lock for this kind of remove operation to ensure that + // ref.parent does not change out from underneath us. + // + // This is why Tremove is a bad idea, and clients should generally use + // Tunlinkat. All p9 clients will use Tunlinkat. + err := ref.safelyGlobal(func() error { + // Is this a root? Can't remove that. + if ref.isRoot() { + return syscall.EINVAL + } + + // N.B. this remove operation is permitted, even if the file is open. + // See also rename below for reasoning. + + // Is this file already deleted? + if ref.isDeleted() { + return syscall.EINVAL + } + + // Retrieve the file's proper name. + name := ref.parent.pathNode.nameFor(ref) + + // Attempt the removal. + if err := ref.parent.file.UnlinkAt(name, 0); err != nil { + return err + } + + // Mark all relevant fids as deleted. We don't need to lock any + // individual nodes because we already hold the global lock. + ref.parent.markChildDeleted(name) + return nil + }) + // "The remove request asks the file server both to remove the file // represented by fid and to clunk the fid, even if the remove fails." // // "It is correct to consider remove to be a clunk with the side effect // of removing the file if permissions allow." // https://swtch.com/plan9port/man/man9/remove.html - err := ref.file.Remove() - - // Clunk the FID regardless of Remove error. if !cs.DeleteFID(t.FID) { return newErr(syscall.EBADF) } - if err != nil { return newErr(err) } + return &Rremove{} } @@ -168,9 +224,12 @@ func (t *Tattach) handle(cs *connState) message { // Build a transient reference. root := &fidRef{ + server: cs.server, + parent: nil, file: sf, refs: 1, - walkable: attr.Mode.IsDir(), + mode: attr.Mode.FileType(), + pathNode: &cs.server.pathTree, } defer root.DecRef() @@ -183,20 +242,24 @@ func (t *Tattach) handle(cs *connState) message { // We want the same traversal checks to apply on attach, so always // attach at the root and use the regular walk paths. names := strings.Split(t.Auth.AttachName, "/") - _, target, _, attr, err := doWalk(cs, root, names) + _, newRef, _, attr, err := doWalk(cs, root, names) if err != nil { return newErr(err) } + defer newRef.DecRef() // Insert the FID. - cs.InsertFID(t.FID, &fidRef{ - file: target, - walkable: attr.Mode.IsDir(), - }) - + cs.InsertFID(t.FID, newRef) return &Rattach{} } +// CanOpen returns whether this file open can be opened, read and written to. +// +// This includes everything except symlinks and sockets. +func CanOpen(mode FileMode) bool { + return mode.IsRegular() || mode.IsDir() || mode.IsNamedPipe() || mode.IsBlockDevice() || mode.IsCharacterDevice() +} + // handle implements handler.handle. func (t *Tlopen) handle(cs *connState) message { // Lookup the FID. @@ -210,13 +273,35 @@ func (t *Tlopen) handle(cs *connState) message { defer ref.openedMu.Unlock() // Has it been opened already? - if ref.opened { + if ref.opened || !CanOpen(ref.mode) { return newErr(syscall.EINVAL) } - // Do the open. - osFile, qid, ioUnit, err := ref.file.Open(t.Flags) - if err != nil { + // Are flags valid? + if t.Flags&^OpenFlagsModeMask != 0 { + return newErr(syscall.EINVAL) + } + + // Is this an attempt to open a directory as writable? Don't accept. + if ref.mode.IsDir() && t.Flags != ReadOnly { + return newErr(syscall.EINVAL) + } + + var ( + qid QID + ioUnit uint32 + osFile *fd.FD + ) + if err := ref.safelyRead(func() (err error) { + // Has it been deleted already? + if ref.isDeleted() { + return syscall.EINVAL + } + + // Do the open. + osFile, qid, ioUnit, err = ref.file.Open(t.Flags) + return err + }); err != nil { return newErr(err) } @@ -229,8 +314,8 @@ func (t *Tlopen) handle(cs *connState) message { func (t *Tlcreate) do(cs *connState, uid UID) (*Rlcreate, error) { // Don't allow complex names. - if !isSafeName(t.Name) { - return nil, syscall.EINVAL + if err := checkSafeName(t.Name); err != nil { + return nil, err } // Lookup the FID. @@ -240,20 +325,48 @@ func (t *Tlcreate) do(cs *connState, uid UID) (*Rlcreate, error) { } defer ref.DecRef() - // Do the create. - osFile, nsf, qid, ioUnit, err := ref.file.Create(t.Name, t.OpenFlags, t.Permissions, uid, t.GID) - if err != nil { + var ( + osFile *fd.FD + nsf File + qid QID + ioUnit uint32 + newRef *fidRef + ) + if err := ref.safelyWrite(func() (err error) { + // Don't allow creation from non-directories or deleted directories. + if ref.isDeleted() || !ref.mode.IsDir() { + return syscall.EINVAL + } + + // Not allowed on open directories. + if _, opened := ref.OpenFlags(); opened { + return syscall.EINVAL + } + + // Do the create. + osFile, nsf, qid, ioUnit, err = ref.file.Create(t.Name, t.OpenFlags, t.Permissions, uid, t.GID) + if err != nil { + return err + } + + newRef = &fidRef{ + server: cs.server, + parent: ref, + file: nsf, + opened: true, + openFlags: t.OpenFlags, + mode: ModeRegular, + pathNode: ref.pathNode.pathNodeFor(t.Name), + } + ref.pathNode.addChild(newRef, t.Name) + ref.IncRef() // Acquire parent reference. + return nil + }); err != nil { return nil, err } // Replace the FID reference. - // - // The new file will be opened already. - cs.InsertFID(t.FID, &fidRef{ - file: nsf, - opened: true, - openFlags: t.OpenFlags, - }) + cs.InsertFID(t.FID, newRef) return &Rlcreate{Rlopen: Rlopen{QID: qid, IoUnit: ioUnit, File: osFile}}, nil } @@ -278,8 +391,8 @@ func (t *Tsymlink) handle(cs *connState) message { func (t *Tsymlink) do(cs *connState, uid UID) (*Rsymlink, error) { // Don't allow complex names. - if !isSafeName(t.Name) { - return nil, syscall.EINVAL + if err := checkSafeName(t.Name); err != nil { + return nil, err } // Lookup the FID. @@ -289,9 +402,22 @@ func (t *Tsymlink) do(cs *connState, uid UID) (*Rsymlink, error) { } defer ref.DecRef() - // Do the symlink. - qid, err := ref.file.Symlink(t.Target, t.Name, uid, t.GID) - if err != nil { + var qid QID + if err := ref.safelyWrite(func() (err error) { + // Don't allow symlinks from non-directories or deleted directories. + if ref.isDeleted() || !ref.mode.IsDir() { + return syscall.EINVAL + } + + // Not allowed on open directories. + if _, opened := ref.OpenFlags(); opened { + return syscall.EINVAL + } + + // Do the symlink. + qid, err = ref.file.Symlink(t.Target, t.Name, uid, t.GID) + return err + }); err != nil { return nil, err } @@ -301,8 +427,8 @@ func (t *Tsymlink) do(cs *connState, uid UID) (*Rsymlink, error) { // handle implements handler.handle. func (t *Tlink) handle(cs *connState) message { // Don't allow complex names. - if !isSafeName(t.Name) { - return newErr(syscall.EINVAL) + if err := checkSafeName(t.Name); err != nil { + return newErr(err) } // Lookup the FID. @@ -319,8 +445,20 @@ func (t *Tlink) handle(cs *connState) message { } defer refTarget.DecRef() - // Do the link. - if err := ref.file.Link(refTarget.file, t.Name); err != nil { + if err := ref.safelyWrite(func() (err error) { + // Don't allow create links from non-directories or deleted directories. + if ref.isDeleted() || !ref.mode.IsDir() { + return syscall.EINVAL + } + + // Not allowed on open directories. + if _, opened := ref.OpenFlags(); opened { + return syscall.EINVAL + } + + // Do the link. + return ref.file.Link(refTarget.file, t.Name) + }); err != nil { return newErr(err) } @@ -330,8 +468,11 @@ func (t *Tlink) handle(cs *connState) message { // handle implements handler.handle. func (t *Trenameat) handle(cs *connState) message { // Don't allow complex names. - if !isSafeName(t.OldName) || !isSafeName(t.NewName) { - return newErr(syscall.EINVAL) + if err := checkSafeName(t.OldName); err != nil { + return newErr(err) + } + if err := checkSafeName(t.NewName); err != nil { + return newErr(err) } // Lookup the FID. @@ -348,8 +489,32 @@ func (t *Trenameat) handle(cs *connState) message { } defer refTarget.DecRef() - // Do the rename. - if err := ref.file.RenameAt(t.OldName, refTarget.file, t.NewName); err != nil { + // Perform the rename holding the global lock. + if err := ref.safelyGlobal(func() (err error) { + // Don't allow renaming across deleted directories. + if ref.isDeleted() || !ref.mode.IsDir() || refTarget.isDeleted() || !refTarget.mode.IsDir() { + return syscall.EINVAL + } + + // Not allowed on open directories. + if _, opened := ref.OpenFlags(); opened { + return syscall.EINVAL + } + + // Is this the same file? If yes, short-circuit and return success. + if ref.pathNode == refTarget.pathNode && t.OldName == t.NewName { + return nil + } + + // Attempt the actual rename. + if err := ref.file.RenameAt(t.OldName, refTarget.file, t.NewName); err != nil { + return err + } + + // Update the path tree. + ref.renameChildTo(t.OldName, refTarget, t.NewName) + return nil + }); err != nil { return newErr(err) } @@ -359,8 +524,8 @@ func (t *Trenameat) handle(cs *connState) message { // handle implements handler.handle. func (t *Tunlinkat) handle(cs *connState) message { // Don't allow complex names. - if !isSafeName(t.Name) { - return newErr(syscall.EINVAL) + if err := checkSafeName(t.Name); err != nil { + return newErr(err) } // Lookup the FID. @@ -370,8 +535,40 @@ func (t *Tunlinkat) handle(cs *connState) message { } defer ref.DecRef() - // Do the unlink. - if err := ref.file.UnlinkAt(t.Name, t.Flags); err != nil { + if err := ref.safelyWrite(func() (err error) { + // Don't allow deletion from non-directories or deleted directories. + if ref.isDeleted() || !ref.mode.IsDir() { + return syscall.EINVAL + } + + // Not allowed on open directories. + if _, opened := ref.OpenFlags(); opened { + return syscall.EINVAL + } + + // Before we do the unlink itself, we need to ensure that there + // are no operations in flight on associated path node. The + // child's path node lock must be held to ensure that the + // unlink at marking the child deleted below is atomic with + // respect to any other read or write operations. + // + // This is one case where we have a lock ordering issue, but + // since we always acquire deeper in the hierarchy, we know + // that we are free of lock cycles. + childPathNode := ref.pathNode.pathNodeFor(t.Name) + childPathNode.mu.Lock() + defer childPathNode.mu.Unlock() + + // Do the unlink. + err = ref.file.UnlinkAt(t.Name, t.Flags) + if err != nil { + return err + } + + // Mark the path as deleted. + ref.markChildDeleted(t.Name) + return nil + }); err != nil { return newErr(err) } @@ -381,8 +578,8 @@ func (t *Tunlinkat) handle(cs *connState) message { // handle implements handler.handle. func (t *Trename) handle(cs *connState) message { // Don't allow complex names. - if !isSafeName(t.Name) { - return newErr(syscall.EINVAL) + if err := checkSafeName(t.Name); err != nil { + return newErr(err) } // Lookup the FID. @@ -399,8 +596,43 @@ func (t *Trename) handle(cs *connState) message { } defer refTarget.DecRef() - // Call the rename method. - if err := ref.file.Rename(refTarget.file, t.Name); err != nil { + if err := ref.safelyGlobal(func() (err error) { + // Don't allow a root rename. + if ref.isRoot() { + return syscall.EINVAL + } + + // Don't allow renaming deleting entries, or target non-directories. + if ref.isDeleted() || refTarget.isDeleted() || !refTarget.mode.IsDir() { + return syscall.EINVAL + } + + // If the parent is deleted, but we not, something is seriously wrong. + // It's fail to die at this point with an assertion failure. + if ref.parent.isDeleted() { + panic(fmt.Sprintf("parent %+v deleted, child %+v is not", ref.parent, ref)) + } + + // N.B. The rename operation is allowed to proceed on open files. It + // does impact the state of its parent, but this is merely a sanity + // check in any case, and the operation is safe. There may be other + // files corresponding to the same path that are renamed anyways. + + // Check for the exact same file and short-circuit. + oldName := ref.parent.pathNode.nameFor(ref) + if ref.parent.pathNode == refTarget.pathNode && oldName == t.Name { + return nil + } + + // Call the rename method on the parent. + if err := ref.parent.file.RenameAt(oldName, refTarget.file, t.Name); err != nil { + return err + } + + // Update the path tree. + ref.parent.renameChildTo(oldName, refTarget, t.Name) + return nil + }); err != nil { return newErr(err) } @@ -416,9 +648,19 @@ func (t *Treadlink) handle(cs *connState) message { } defer ref.DecRef() - // Do the read. - target, err := ref.file.Readlink() - if err != nil { + var target string + if err := ref.safelyRead(func() (err error) { + // Don't allow readlink on deleted files. There is no need to + // check if this file is opened because symlinks cannot be + // opened. + if ref.isDeleted() || !ref.mode.IsSymlink() { + return syscall.EINVAL + } + + // Do the read. + target, err = ref.file.Readlink() + return err + }); err != nil { return newErr(err) } @@ -434,26 +676,30 @@ func (t *Tread) handle(cs *connState) message { } defer ref.DecRef() - // Has it been opened already? - openFlags, opened := ref.OpenFlags() - if !opened { - return newErr(syscall.EINVAL) - } - - // Can it be read? Check permissions. - if openFlags&OpenFlagsModeMask == WriteOnly { - return newErr(syscall.EPERM) - } - // Constrain the size of the read buffer. if int(t.Count) > int(maximumLength) { return newErr(syscall.ENOBUFS) } - // Do the read. - data := make([]byte, t.Count) - n, err := ref.file.ReadAt(data, t.Offset) - if err != nil && err != io.EOF { + var ( + data = make([]byte, t.Count) + n int + ) + if err := ref.safelyRead(func() (err error) { + // Has it been opened already? + openFlags, opened := ref.OpenFlags() + if !opened { + return syscall.EINVAL + } + + // Can it be read? Check permissions. + if openFlags&OpenFlagsModeMask == WriteOnly { + return syscall.EPERM + } + + n, err = ref.file.ReadAt(data, t.Offset) + return err + }); err != nil && err != io.EOF { return newErr(err) } @@ -469,20 +715,22 @@ func (t *Twrite) handle(cs *connState) message { } defer ref.DecRef() - // Has it been opened already? - openFlags, opened := ref.OpenFlags() - if !opened { - return newErr(syscall.EINVAL) - } + var n int + if err := ref.safelyRead(func() (err error) { + // Has it been opened already? + openFlags, opened := ref.OpenFlags() + if !opened { + return syscall.EINVAL + } - // Can it be write? Check permissions. - if openFlags&OpenFlagsModeMask == ReadOnly { - return newErr(syscall.EPERM) - } + // Can it be write? Check permissions. + if openFlags&OpenFlagsModeMask == ReadOnly { + return syscall.EPERM + } - // Do the write. - n, err := ref.file.WriteAt(t.Data, t.Offset) - if err != nil { + n, err = ref.file.WriteAt(t.Data, t.Offset) + return err + }); err != nil { return newErr(err) } @@ -500,8 +748,8 @@ func (t *Tmknod) handle(cs *connState) message { func (t *Tmknod) do(cs *connState, uid UID) (*Rmknod, error) { // Don't allow complex names. - if !isSafeName(t.Name) { - return nil, syscall.EINVAL + if err := checkSafeName(t.Name); err != nil { + return nil, err } // Lookup the FID. @@ -511,9 +759,22 @@ func (t *Tmknod) do(cs *connState, uid UID) (*Rmknod, error) { } defer ref.DecRef() - // Do the mknod. - qid, err := ref.file.Mknod(t.Name, t.Permissions, t.Major, t.Minor, uid, t.GID) - if err != nil { + var qid QID + if err := ref.safelyWrite(func() (err error) { + // Don't allow mknod on deleted files. + if ref.isDeleted() || !ref.mode.IsDir() { + return syscall.EINVAL + } + + // Not allowed on open directories. + if _, opened := ref.OpenFlags(); opened { + return syscall.EINVAL + } + + // Do the mknod. + qid, err = ref.file.Mknod(t.Name, t.Permissions, t.Major, t.Minor, uid, t.GID) + return err + }); err != nil { return nil, err } @@ -531,8 +792,8 @@ func (t *Tmkdir) handle(cs *connState) message { func (t *Tmkdir) do(cs *connState, uid UID) (*Rmkdir, error) { // Don't allow complex names. - if !isSafeName(t.Name) { - return nil, syscall.EINVAL + if err := checkSafeName(t.Name); err != nil { + return nil, err } // Lookup the FID. @@ -542,9 +803,22 @@ func (t *Tmkdir) do(cs *connState, uid UID) (*Rmkdir, error) { } defer ref.DecRef() - // Do the mkdir. - qid, err := ref.file.Mkdir(t.Name, t.Permissions, uid, t.GID) - if err != nil { + var qid QID + if err := ref.safelyWrite(func() (err error) { + // Don't allow mkdir on deleted files. + if ref.isDeleted() || !ref.mode.IsDir() { + return syscall.EINVAL + } + + // Not allowed on open directories. + if _, opened := ref.OpenFlags(); opened { + return syscall.EINVAL + } + + // Do the mkdir. + qid, err = ref.file.Mkdir(t.Name, t.Permissions, uid, t.GID) + return err + }); err != nil { return nil, err } @@ -560,9 +834,20 @@ func (t *Tgetattr) handle(cs *connState) message { } defer ref.DecRef() - // Get attributes. - qid, valid, attr, err := ref.file.GetAttr(t.AttrMask) - if err != nil { + // We allow getattr on deleted files. Depending on the backing + // implementation, it's possible that races exist that might allow + // fetching attributes of other files. But we need to generally allow + // refreshing attributes and this is a minor leak, if at all. + + var ( + qid QID + valid AttrMask + attr Attr + ) + if err := ref.safelyRead(func() (err error) { + qid, valid, attr, err = ref.file.GetAttr(t.AttrMask) + return err + }); err != nil { return newErr(err) } @@ -578,8 +863,18 @@ func (t *Tsetattr) handle(cs *connState) message { } defer ref.DecRef() - // Set attributes. - if err := ref.file.SetAttr(t.Valid, t.SetAttr); err != nil { + if err := ref.safelyWrite(func() error { + // We don't allow setattr on files that have been deleted. + // This might be technically incorrect, as it's possible that + // there were multiple links and you can still change the + // corresponding inode information. + if ref.isDeleted() { + return syscall.EINVAL + } + + // Set the attributes. + return ref.file.SetAttr(t.Valid, t.SetAttr) + }); err != nil { return newErr(err) } @@ -621,14 +916,25 @@ func (t *Treaddir) handle(cs *connState) message { } defer ref.DecRef() - // Has it been opened already? - if _, opened := ref.OpenFlags(); !opened { - return newErr(syscall.EINVAL) - } + var entries []Dirent + if err := ref.safelyRead(func() (err error) { + // Don't allow reading deleted directories. + if ref.isDeleted() || !ref.mode.IsDir() { + return syscall.EINVAL + } + + // Has it been opened already? + if _, opened := ref.OpenFlags(); !opened { + return syscall.EINVAL + } - // Read the entries. - entries, err := ref.file.Readdir(t.Offset, t.Count) - if err != nil && err != io.EOF { + // Read the entries. + entries, err = ref.file.Readdir(t.Offset, t.Count) + if err != nil && err != io.EOF { + return err + } + return nil + }); err != nil { return newErr(err) } @@ -644,13 +950,15 @@ func (t *Tfsync) handle(cs *connState) message { } defer ref.DecRef() - // Has it been opened already? - if _, opened := ref.OpenFlags(); !opened { - return newErr(syscall.EINVAL) - } + if err := ref.safelyRead(func() (err error) { + // Has it been opened already? + if _, opened := ref.OpenFlags(); !opened { + return syscall.EINVAL + } - err := ref.file.FSync() - if err != nil { + // Perform the sync. + return ref.file.FSync() + }); err != nil { return newErr(err) } @@ -671,6 +979,11 @@ func (t *Tstatfs) handle(cs *connState) message { return newErr(err) } + // Constrain the name length. + if st.NameLength > maximumNameLength { + st.NameLength = maximumNameLength + } + return &Rstatfs{st} } @@ -682,7 +995,7 @@ func (t *Tflushf) handle(cs *connState) message { } defer ref.DecRef() - if err := ref.file.Flush(); err != nil { + if err := ref.safelyRead(ref.file.Flush); err != nil { return newErr(err) } @@ -726,12 +1039,14 @@ func walkOne(qids []QID, from File, names []string) ([]QID, File, AttrMask, Attr // doWalk walks from a given fidRef. // -// This enforces that all intermediate nodes are walkable (directories). -func doWalk(cs *connState, ref *fidRef, names []string) (qids []QID, sf File, valid AttrMask, attr Attr, err error) { +// This enforces that all intermediate nodes are walkable (directories). The +// fidRef returned (newRef) has a reference associated with it that is now +// owned by the caller and must be handled appropriately. +func doWalk(cs *connState, ref *fidRef, names []string) (qids []QID, newRef *fidRef, valid AttrMask, attr Attr, err error) { // Check the names. for _, name := range names { - if !isSafeName(name) { - err = syscall.EINVAL + err = checkSafeName(name) + if err != nil { return } } @@ -745,44 +1060,88 @@ func doWalk(cs *connState, ref *fidRef, names []string) (qids []QID, sf File, va // Is this an empty list? Handle specially. We don't actually need to // validate anything since this is always permitted. if len(names) == 0 { - return walkOne(nil, ref.file, nil) - } - - // Is it walkable? - if !ref.walkable { - err = syscall.EINVAL - return + var sf File // Temporary. + if err := ref.maybeParent().safelyRead(func() (err error) { + // Clone the single element. + qids, sf, valid, attr, err = walkOne(nil, ref.file, nil) + if err != nil { + return err + } + + newRef = &fidRef{ + server: cs.server, + parent: ref.parent, + file: sf, + mode: ref.mode, + pathNode: ref.pathNode, + + // For the clone case, the cloned fid must + // preserve the deleted property of the + // original FID. + deleted: ref.deleted, + } + if !ref.isRoot() { + if !newRef.isDeleted() { + // Add only if a non-root node; the same node. + ref.parent.pathNode.addChild(newRef, ref.parent.pathNode.nameFor(ref)) + } + ref.parent.IncRef() // Acquire parent reference. + } + // doWalk returns a reference. + newRef.IncRef() + return nil + }); err != nil { + return nil, nil, AttrMask{}, Attr{}, err + } + return qids, newRef, valid, attr, nil } - from := ref.file // Start at the passed ref. - // Do the walk, one element at a time. + walkRef := ref + walkRef.IncRef() for i := 0; i < len(names); i++ { - qids, sf, valid, attr, err = walkOne(qids, from, names[i:i+1]) - - // Close the intermediate file. Note that we don't close the - // first file because in that case we are walking from the - // existing reference. - if i > 0 { - from.Close() - } - from = sf // Use the new file. - - // Was there an error walking? - if err != nil { - return nil, nil, AttrMask{}, Attr{}, err - } - // We won't allow beyond past symlinks; stop here if this isn't // a proper directory and we have additional paths to walk. - if !valid.Mode || (!attr.Mode.IsDir() && i < len(names)-1) { - from.Close() // Not using the file object. + if !walkRef.mode.IsDir() { + walkRef.DecRef() // Drop walk reference; no lock required. return nil, nil, AttrMask{}, Attr{}, syscall.EINVAL } + + var sf File // Temporary. + if err := walkRef.safelyRead(func() (err error) { + qids, sf, valid, attr, err = walkOne(qids, walkRef.file, names[i:i+1]) + if err != nil { + return err + } + + // Note that we don't need to acquire a lock on any of + // these individual instances. That's because they are + // not actually addressable via a FID. They are + // anonymous. They exist in the tree for tracking + // purposes. + newRef := &fidRef{ + server: cs.server, + parent: walkRef, + file: sf, + mode: attr.Mode.FileType(), + pathNode: walkRef.pathNode.pathNodeFor(names[i]), + } + walkRef.pathNode.addChild(newRef, names[i]) + // We allow our walk reference to become the new parent + // reference here and so we don't IncRef. Instead, just + // set walkRef to the newRef above and acquire a new + // walk reference. + walkRef = newRef + walkRef.IncRef() + return nil + }); err != nil { + walkRef.DecRef() // Drop the old walkRef. + return nil, nil, AttrMask{}, Attr{}, err + } } // Success. - return qids, sf, valid, attr, nil + return qids, walkRef, valid, attr, nil } // handle implements handler.handle. @@ -795,17 +1154,14 @@ func (t *Twalk) handle(cs *connState) message { defer ref.DecRef() // Do the walk. - qids, sf, _, attr, err := doWalk(cs, ref, t.Names) + qids, newRef, _, _, err := doWalk(cs, ref, t.Names) if err != nil { return newErr(err) } + defer newRef.DecRef() // Install the new FID. - cs.InsertFID(t.NewFID, &fidRef{ - file: sf, - walkable: attr.Mode.IsDir(), - }) - + cs.InsertFID(t.NewFID, newRef) return &Rwalk{QIDs: qids} } @@ -819,17 +1175,14 @@ func (t *Twalkgetattr) handle(cs *connState) message { defer ref.DecRef() // Do the walk. - qids, sf, valid, attr, err := doWalk(cs, ref, t.Names) + qids, newRef, valid, attr, err := doWalk(cs, ref, t.Names) if err != nil { return newErr(err) } + defer newRef.DecRef() // Install the new FID. - cs.InsertFID(t.NewFID, &fidRef{ - file: sf, - walkable: attr.Mode.IsDir(), - }) - + cs.InsertFID(t.NewFID, newRef) return &Rwalkgetattr{QIDs: qids, Valid: valid, Attr: attr} } @@ -878,9 +1231,17 @@ func (t *Tlconnect) handle(cs *connState) message { } defer ref.DecRef() - // Do the connect. - osFile, err := ref.file.Connect(t.Flags) - if err != nil { + var osFile *fd.FD + if err := ref.safelyRead(func() (err error) { + // Don't allow connecting to deleted files. + if ref.isDeleted() || !ref.mode.IsSocket() { + return syscall.EINVAL + } + + // Do the connect. + osFile, err = ref.file.Connect(t.Flags) + return err + }); err != nil { return newErr(err) } diff --git a/pkg/p9/local_server/BUILD b/pkg/p9/local_server/BUILD index 8229e6308..b17ebb79d 100644 --- a/pkg/p9/local_server/BUILD +++ b/pkg/p9/local_server/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_binary") +package(licenses = ["notice"]) # Apache 2.0 + go_binary( name = "local_server", srcs = ["local_server.go"], diff --git a/pkg/p9/local_server/local_server.go b/pkg/p9/local_server/local_server.go index 1e6aaa762..69b90c6cd 100644 --- a/pkg/p9/local_server/local_server.go +++ b/pkg/p9/local_server/local_server.go @@ -318,6 +318,11 @@ func (l *local) Connect(p9.ConnectFlags) (*fd.FD, error) { return nil, syscall.ECONNREFUSED } +// Renamed implements p9.File.Renamed. +func (l *local) Renamed(parent p9.File, newName string) { + l.path = path.Join(parent.(*local).path, newName) +} + func main() { log.SetLevel(log.Debug) diff --git a/pkg/p9/messages_test.go b/pkg/p9/messages_test.go index dfb41bb76..c0d65d82c 100644 --- a/pkg/p9/messages_test.go +++ b/pkg/p9/messages_test.go @@ -15,6 +15,7 @@ package p9 import ( + "fmt" "reflect" "testing" ) @@ -186,6 +187,13 @@ func TestEncodeDecode(t *testing.T) { &Rxattrwalk{ Size: 1, }, + &Txattrcreate{ + FID: 1, + Name: "a", + AttrSize: 2, + Flags: 3, + }, + &Rxattrcreate{}, &Treaddir{ Directory: 1, Offset: 2, @@ -389,3 +397,32 @@ func TestEncodeDecode(t *testing.T) { } } } + +func TestMessageStrings(t *testing.T) { + for typ, fn := range messageRegistry { + name := fmt.Sprintf("%+v", typ) + t.Run(name, func(t *testing.T) { + defer func() { // Ensure no panic. + if r := recover(); r != nil { + t.Errorf("printing %s failed: %v", name, r) + } + }() + m := fn() + _ = fmt.Sprintf("%v", m) + err := ErrInvalidMsgType{typ} + _ = err.Error() + }) + } +} + +func TestRegisterDuplicate(t *testing.T) { + defer func() { + if r := recover(); r == nil { + // We expect a panic. + t.FailNow() + } + }() + + // Register a duplicate. + register(&Rlerror{}) +} diff --git a/pkg/p9/p9.go b/pkg/p9/p9.go index 3b0993ecd..be644e7bf 100644 --- a/pkg/p9/p9.go +++ b/pkg/p9/p9.go @@ -984,6 +984,30 @@ func (s *SetAttr) Encode(b *buffer) { b.Write64(s.MTimeNanoSeconds) } +// Apply applies this to the given Attr. +func (a *Attr) Apply(mask SetAttrMask, attr SetAttr) { + if mask.Permissions { + a.Mode = a.Mode&^PermissionsMask | (attr.Permissions & PermissionsMask) + } + if mask.UID { + a.UID = attr.UID + } + if mask.GID { + a.GID = attr.GID + } + if mask.Size { + a.Size = attr.Size + } + if mask.ATime { + a.ATimeSeconds = attr.ATimeSeconds + a.ATimeNanoSeconds = attr.ATimeNanoSeconds + } + if mask.MTime { + a.MTimeSeconds = attr.MTimeSeconds + a.MTimeNanoSeconds = attr.MTimeNanoSeconds + } +} + // Dirent is used for readdir. type Dirent struct { // QID is the entry QID. diff --git a/pkg/p9/p9test/BUILD b/pkg/p9/p9test/BUILD index d6f428e11..7c4b875ce 100644 --- a/pkg/p9/p9test/BUILD +++ b/pkg/p9/p9test/BUILD @@ -1,16 +1,60 @@ +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +load("@io_bazel_rules_go//go:def.bzl", "go_binary") + package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +alias( + name = "mockgen", + actual = "@com_github_golang_mock//mockgen:mockgen", +) -go_test( - name = "p9test_test", - size = "small", - srcs = ["client_test.go"], - embed = [":p9test"], +MOCK_SRC_PACKAGE = "gvisor.googlesource.com/gvisor/pkg/p9" + +# mockgen_reflect is a source file that contains mock generation code that +# imports the p9 package and generates a specification via reflection. The +# usual generation path must be split into two distinct parts because the full +# source tree is not available to all build targets. Only declared depencies +# are available (and even then, not the Go source files). +genrule( + name = "mockgen_reflect", + testonly = 1, + outs = ["mockgen_reflect.go"], + cmd = ( + "$(location :mockgen) " + + "-package p9test " + + "-prog_only " + MOCK_SRC_PACKAGE + " " + + "Attacher,File > $@" + ), + tools = [":mockgen"], +) + +# mockgen_exec is the binary that includes the above reflection generator. +# Running this binary will emit an encoded version of the p9 Attacher and File +# structures. This is consumed by the mocks genrule, below. +go_binary( + name = "mockgen_exec", + testonly = 1, + srcs = ["mockgen_reflect.go"], deps = [ - "//pkg/fd", "//pkg/p9", - "//pkg/unet", + "@com_github_golang_mock//mockgen/model:go_default_library", + ], +) + +# mocks consumes the encoded output above, and generates the full source for a +# set of mocks. These are included directly in the p9test library. +genrule( + name = "mocks", + testonly = 1, + outs = ["mocks.go"], + cmd = ( + "$(location :mockgen) " + + "-package p9test " + + "-exec_only $(location :mockgen_exec) " + MOCK_SRC_PACKAGE + " File > $@" + ), + tools = [ + ":mockgen", + ":mockgen_exec", ], ) @@ -18,11 +62,27 @@ go_library( name = "p9test", srcs = [ "mocks.go", + "p9test.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/p9/p9test", visibility = ["//:sandbox"], + deps = [ + "//pkg/fd", + "//pkg/log", + "//pkg/p9", + "//pkg/unet", + "@com_github_golang_mock//gomock:go_default_library", + ], +) + +go_test( + name = "client_test", + size = "small", + srcs = ["client_test.go"], + embed = [":p9test"], deps = [ "//pkg/fd", "//pkg/p9", + "@com_github_golang_mock//gomock:go_default_library", ], ) diff --git a/pkg/p9/p9test/client_test.go b/pkg/p9/p9test/client_test.go index db562b9ba..242d81b95 100644 --- a/pkg/p9/p9test/client_test.go +++ b/pkg/p9/p9test/client_test.go @@ -15,360 +15,2059 @@ package p9test import ( - "io/ioutil" + "bytes" + "fmt" + "io" + "math/rand" "os" "reflect" + "strings" + "sync" "syscall" "testing" + "time" + "github.com/golang/mock/gomock" "gvisor.googlesource.com/gvisor/pkg/fd" "gvisor.googlesource.com/gvisor/pkg/p9" - "gvisor.googlesource.com/gvisor/pkg/unet" ) -func TestDonateFD(t *testing.T) { - // Temporary file. - osFile, err := ioutil.TempFile("", "p9") +func TestPanic(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + // Create a new root. + d := h.NewDirectory(nil)(nil) + defer d.Close() // Needed manually. + h.Attacher.EXPECT().Attach().Return(d, nil).Do(func() { + // Panic here, and ensure that we get back EFAULT. + panic("handler") + }) + + // Attach to the client. + if _, err := c.Attach("/"); err != syscall.EFAULT { + t.Fatalf("got attach err %v, want EFAULT", err) + } +} + +func TestAttachNoLeak(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + // Create a new root. + d := h.NewDirectory(nil)(nil) + h.Attacher.EXPECT().Attach().Return(d, nil).Times(1) + + // Attach to the client. + f, err := c.Attach("/") + if err != nil { + t.Fatalf("got attach err %v, want nil", err) + } + + // Don't close the file. This should be closed automatically when the + // client disconnects. The mock asserts that everything is closed + // exactly once. This statement just removes the unused variable error. + _ = f +} + +func TestBadAttach(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + // Return an error on attach. + h.Attacher.EXPECT().Attach().Return(nil, syscall.EINVAL).Times(1) + + // Attach to the client. + if _, err := c.Attach("/"); err != syscall.EINVAL { + t.Fatalf("got attach err %v, want syscall.EINVAL", err) + } +} + +func TestWalkAttach(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + // Create a new root. + d := h.NewDirectory(map[string]Generator{ + "a": h.NewDirectory(map[string]Generator{ + "b": h.NewFile(), + }), + })(nil) + h.Attacher.EXPECT().Attach().Return(d, nil).Times(1) + + // Attach to the client as a non-root, and ensure that the walk above + // occurs as expected. We should get back b, and all references should + // be dropped when the file is closed. + f, err := c.Attach("/a/b") + if err != nil { + t.Fatalf("got attach err %v, want nil", err) + } + defer f.Close() + + // Check that's a regular file. + if _, _, attr, err := f.GetAttr(p9.AttrMaskAll()); err != nil { + t.Errorf("got err %v, want nil", err) + } else if !attr.Mode.IsRegular() { + t.Errorf("got mode %v, want regular file", err) + } +} + +// newTypeMap returns a new type map dictionary. +func newTypeMap(h *Harness) map[string]Generator { + return map[string]Generator{ + "directory": h.NewDirectory(map[string]Generator{}), + "file": h.NewFile(), + "symlink": h.NewSymlink(), + "block-device": h.NewBlockDevice(), + "character-device": h.NewCharacterDevice(), + "named-pipe": h.NewNamedPipe(), + "socket": h.NewSocket(), + } +} + +// newRoot returns a new root filesystem. +// +// This is set up in a deterministic way for testing most operations. +// +// The represented file system looks like: +// - file +// - symlink +// - directory +// ... +// + one +// - file +// - symlink +// - directory +// ... +// + two +// - file +// - symlink +// - directory +// ... +// + three +// - file +// - symlink +// - directory +// ... +func newRoot(h *Harness, c *p9.Client) (*Mock, p9.File) { + root := newTypeMap(h) + one := newTypeMap(h) + two := newTypeMap(h) + three := newTypeMap(h) + one["two"] = h.NewDirectory(two) // Will be nested in one. + root["one"] = h.NewDirectory(one) // Top level. + root["three"] = h.NewDirectory(three) // Alternate top-level. + + // Create a new root. + rootBackend := h.NewDirectory(root)(nil) + h.Attacher.EXPECT().Attach().Return(rootBackend, nil) + + // Attach to the client. + r, err := c.Attach("/") + if err != nil { + h.t.Fatalf("got attach err %v, want nil", err) + } + + return rootBackend, r +} + +func allInvalidNames(from string) []string { + return []string{ + from + "/other", + from + "/..", + from + "/.", + from + "/", + "other/" + from, + "/" + from, + "./" + from, + "../" + from, + ".", + "..", + "/", + "", + } +} + +func TestWalkInvalid(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + // Run relevant tests. + for name := range newTypeMap(h) { + // These are all the various ways that one might attempt to + // construct compound paths. They should all be rejected, as + // any compound that contains a / is not allowed, as well as + // the singular paths of '.' and '..'. + if _, _, err := root.Walk([]string{".", name}); err != syscall.EINVAL { + t.Errorf("Walk through . %s wanted EINVAL, got %v", name, err) + } + if _, _, err := root.Walk([]string{"..", name}); err != syscall.EINVAL { + t.Errorf("Walk through . %s wanted EINVAL, got %v", name, err) + } + if _, _, err := root.Walk([]string{name, "."}); err != syscall.EINVAL { + t.Errorf("Walk through %s . wanted EINVAL, got %v", name, err) + } + if _, _, err := root.Walk([]string{name, ".."}); err != syscall.EINVAL { + t.Errorf("Walk through %s .. wanted EINVAL, got %v", name, err) + } + for _, invalidName := range allInvalidNames(name) { + if _, _, err := root.Walk([]string{invalidName}); err != syscall.EINVAL { + t.Errorf("Walk through %s wanted EINVAL, got %v", invalidName, err) + } + } + wantErr := syscall.EINVAL + if name == "directory" { + // We can attempt a walk through a directory. However, + // we should never see a file named "other", so we + // expect this to return ENOENT. + wantErr = syscall.ENOENT + } + if _, _, err := root.Walk([]string{name, "other"}); err != wantErr { + t.Errorf("Walk through %s/other wanted %v, got %v", name, wantErr, err) + } + + // Do a successful walk. + _, f, err := root.Walk([]string{name}) + if err != nil { + t.Errorf("Walk to %s wanted nil, got %v", name, err) + } + defer f.Close() + local := h.Pop(f) + + // Check that the file matches. + _, localMask, localAttr, localErr := local.GetAttr(p9.AttrMaskAll()) + if _, mask, attr, err := f.GetAttr(p9.AttrMaskAll()); mask != localMask || attr != localAttr || err != localErr { + t.Errorf("GetAttr got (%v, %v, %v), wanted (%v, %v, %v)", + mask, attr, err, localMask, localAttr, localErr) + } + + // Ensure we can't walk backwards. + if _, _, err := f.Walk([]string{"."}); err != syscall.EINVAL { + t.Errorf("Walk through %s/. wanted EINVAL, got %v", name, err) + } + if _, _, err := f.Walk([]string{".."}); err != syscall.EINVAL { + t.Errorf("Walk through %s/.. wanted EINVAL, got %v", name, err) + } + } +} + +// fileGenerator is a function to generate files via walk or create. +// +// Examples are: +// - walkHelper +// - walkAndOpenHelper +// - createHelper +type fileGenerator func(*Harness, string, p9.File) (*Mock, *Mock, p9.File) + +// walkHelper walks to the given file. +// +// The backends of the parent and walked file are returned, as well as the +// walked client file. +func walkHelper(h *Harness, name string, dir p9.File) (parentBackend *Mock, walkedBackend *Mock, walked p9.File) { + _, parent, err := dir.Walk(nil) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer parent.Close() + parentBackend = h.Pop(parent) + + _, walked, err = parent.Walk([]string{name}) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + walkedBackend = h.Pop(walked) + + return parentBackend, walkedBackend, walked +} + +// walkAndOpenHelper additionally opens the walked file, if possible. +func walkAndOpenHelper(h *Harness, name string, dir p9.File) (*Mock, *Mock, p9.File) { + parentBackend, walkedBackend, walked := walkHelper(h, name, dir) + if p9.CanOpen(walkedBackend.Attr.Mode) { + // Open for all file types that we can. We stick to a read-only + // open here because directories may not be opened otherwise. + walkedBackend.EXPECT().Open(p9.ReadOnly).Times(1) + if _, _, _, err := walked.Open(p9.ReadOnly); err != nil { + h.t.Errorf("got open err %v, want nil", err) + } + } else { + // ... or assert an error for others. + if _, _, _, err := walked.Open(p9.ReadOnly); err != syscall.EINVAL { + h.t.Errorf("got open err %v, want EINVAL", err) + } + } + return parentBackend, walkedBackend, walked +} + +// createHelper creates the given file and returns the parent directory, +// created file and client file, which must be closed when done. +func createHelper(h *Harness, name string, dir p9.File) (*Mock, *Mock, p9.File) { + // Clone the directory first, since Create replaces the existing file. + // We change the type after calling create. + _, dirThenFile, err := dir.Walk(nil) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + + // Create a new server-side file. On the server-side, the a new file is + // returned from a create call. The client will reuse the same file, + // but we still expect the normal chain of closes. This complicates + // things a bit because the "parent" will always chain to the cloned + // dir above. + dirBackend := h.Pop(dirThenFile) // New backend directory. + newFile := h.NewFile()(dirBackend) // New file with backend parent. + dirBackend.EXPECT().Create(name, gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any()).Return(nil, newFile, newFile.QID, uint32(0), nil) + + // Create via the client. + _, dirThenFile, _, _, err = dirThenFile.Create(name, p9.ReadOnly, 0, 0, 0) if err != nil { - t.Fatalf("could not create temporary file: %v", err) + h.t.Fatalf("got create err %v, want nil", err) + } + + // Ensure subsequent walks succeed. + dirBackend.AddChild(name, h.NewFile()) + return dirBackend, newFile, dirThenFile +} + +// deprecatedRemover allows us to access the deprecated Remove operation within +// the p9.File client object. +type deprecatedRemover interface { + Remove() error +} + +// checkDeleted asserts that relevant methods fail for an unlinked file. +// +// This function will close the file at the end. +func checkDeleted(h *Harness, file p9.File) { + defer file.Close() // See doc. + + if _, _, _, err := file.Open(p9.ReadOnly); err != syscall.EINVAL { + h.t.Errorf("open while deleted, got %v, want EINVAL", err) + } + if _, _, _, _, err := file.Create("created", p9.ReadOnly, 0, 0, 0); err != syscall.EINVAL { + h.t.Errorf("create while deleted, got %v, want EINVAL", err) + } + if _, err := file.Symlink("old", "new", 0, 0); err != syscall.EINVAL { + h.t.Errorf("symlink while deleted, got %v, want EINVAL", err) + } + // N.B. This link is technically invalid, but if a call to link is + // actually made in the backend then the mock will panic. + if err := file.Link(file, "new"); err != syscall.EINVAL { + h.t.Errorf("link while deleted, got %v, want EINVAL", err) + } + if err := file.RenameAt("src", file, "dst"); err != syscall.EINVAL { + h.t.Errorf("renameAt while deleted, got %v, want EINVAL", err) + } + if err := file.UnlinkAt("file", 0); err != syscall.EINVAL { + h.t.Errorf("unlinkAt while deleted, got %v, want EINVAL", err) + } + if err := file.Rename(file, "dst"); err != syscall.EINVAL { + h.t.Errorf("rename while deleted, got %v, want EINVAL", err) + } + if _, err := file.Readlink(); err != syscall.EINVAL { + h.t.Errorf("readlink while deleted, got %v, want EINVAL", err) + } + if _, err := file.Mkdir("dir", p9.ModeDirectory, 0, 0); err != syscall.EINVAL { + h.t.Errorf("mkdir while deleted, got %v, want EINVAL", err) + } + if _, err := file.Mknod("dir", p9.ModeDirectory, 0, 0, 0, 0); err != syscall.EINVAL { + h.t.Errorf("mknod while deleted, got %v, want EINVAL", err) + } + if _, err := file.Readdir(0, 1); err != syscall.EINVAL { + h.t.Errorf("readdir while deleted, got %v, want EINVAL", err) + } + if _, err := file.Connect(p9.ConnectFlags(0)); err != syscall.EINVAL { + h.t.Errorf("connect while deleted, got %v, want EINVAL", err) + } + + // The remove method is technically deprecated, but we want to ensure + // that it still checks for deleted appropriately. We must first clone + // the file because remove is equivalent to close. + _, newFile, err := file.Walk(nil) + if err == syscall.EBUSY { + // We can't walk from here because this reference is open + // aleady. Okay, we will also have unopened cases through + // TestUnlink, just skip the remove operation for now. + return + } else if err != nil { + h.t.Fatalf("clone failed, got %v, want nil", err) + } + if err := newFile.(deprecatedRemover).Remove(); err != syscall.EINVAL { + h.t.Errorf("remove while deleted, got %v, want EINVAL", err) + } +} + +// deleter is a function to remove a file. +type deleter func(parent p9.File, name string) error + +// unlinkAt is a deleter. +func unlinkAt(parent p9.File, name string) error { + // Call unlink. Note that a filesystem may normally impose additional + // constaints on unlinkat success, such as ensuring that a directory is + // empty, requiring AT_REMOVEDIR in flags to remove a directory, etc. + // None of that is required internally (entire trees can be marked + // deleted when this operation succeeds), so the mock will succeed. + return parent.UnlinkAt(name, 0) +} + +// remove is a deleter. +func remove(parent p9.File, name string) error { + // See notes above re: remove. + _, newFile, err := parent.Walk([]string{name}) + if err != nil { + // Should not be expected. + return err + } + + // Do the actual remove. + if err := newFile.(deprecatedRemover).Remove(); err != nil { + return err + } + + // Ensure that the remove closed the file. + if err := newFile.(deprecatedRemover).Remove(); err != syscall.EBADF { + return syscall.EBADF // Propagate this code. + } + + return nil +} + +// unlinkHelper unlinks the noted path, and ensures that all relevant +// operations on that path, acquired from multiple paths, start failing. +func unlinkHelper(h *Harness, root p9.File, targetNames []string, targetGen fileGenerator, deleteFn deleter) { + // name is the file to be unlinked. + name := targetNames[len(targetNames)-1] + + // Walk to the directory containing the target. + _, parent, err := root.Walk(targetNames[:len(targetNames)-1]) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer parent.Close() + parentBackend := h.Pop(parent) + + // Walk to or generate the target file. + _, _, target := targetGen(h, name, parent) + defer checkDeleted(h, target) + + // Walk to a second reference. + _, second, err := parent.Walk([]string{name}) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer checkDeleted(h, second) + + // Walk to a third reference, from the start. + _, third, err := root.Walk(targetNames) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer checkDeleted(h, third) + + // This will be translated in the backend to an unlinkat. + parentBackend.EXPECT().UnlinkAt(name, uint32(0)).Return(nil) + + // Actually perform the deletion. + if err := deleteFn(parent, name); err != nil { + h.t.Fatalf("got delete err %v, want nil", err) + } +} + +func unlinkTest(t *testing.T, targetNames []string, targetGen fileGenerator) { + t.Run(fmt.Sprintf("unlinkAt(%s)", strings.Join(targetNames, "/")), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + unlinkHelper(h, root, targetNames, targetGen, unlinkAt) + }) + t.Run(fmt.Sprintf("remove(%s)", strings.Join(targetNames, "/")), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + unlinkHelper(h, root, targetNames, targetGen, remove) + }) +} + +func TestUnlink(t *testing.T) { + // Unlink all files. + for name := range newTypeMap(nil) { + unlinkTest(t, []string{name}, walkHelper) + unlinkTest(t, []string{name}, walkAndOpenHelper) + unlinkTest(t, []string{"one", name}, walkHelper) + unlinkTest(t, []string{"one", name}, walkAndOpenHelper) + unlinkTest(t, []string{"one", "two", name}, walkHelper) + unlinkTest(t, []string{"one", "two", name}, walkAndOpenHelper) + } + + // Unlink a directory. + unlinkTest(t, []string{"one"}, walkHelper) + unlinkTest(t, []string{"one"}, walkAndOpenHelper) + unlinkTest(t, []string{"one", "two"}, walkHelper) + unlinkTest(t, []string{"one", "two"}, walkAndOpenHelper) + + // Unlink created files. + unlinkTest(t, []string{"created"}, createHelper) + unlinkTest(t, []string{"one", "created"}, createHelper) + unlinkTest(t, []string{"one", "two", "created"}, createHelper) +} + +func TestUnlinkAtInvalid(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + for name := range newTypeMap(nil) { + for _, invalidName := range allInvalidNames(name) { + if err := root.UnlinkAt(invalidName, 0); err != syscall.EINVAL { + t.Errorf("got %v for name %q, want EINVAL", err, invalidName) + } + } + } +} + +// expectRenamed asserts an ordered sequence of rename calls, based on all the +// elements in elements being the source, and the first element therein +// changing to dstName, parented at dstParent. +func expectRenamed(file *Mock, elements []string, dstParent *Mock, dstName string) *gomock.Call { + if len(elements) > 0 { + // Recurse to the parent, if necessary. + call := expectRenamed(file.parent, elements[:len(elements)-1], dstParent, dstName) + + // Recursive case: this element is unchanged, but should have + // it's hook called after the parent. + return file.EXPECT().Renamed(file.parent, elements[len(elements)-1]).Do(func(p p9.File, _ string) { + file.parent = p.(*Mock) + }).After(call) + } + + // Base case: this is the changed element. + return file.EXPECT().Renamed(dstParent, dstName).Do(func(p p9.File, name string) { + file.parent = p.(*Mock) + }) +} + +// renamer is a rename function. +type renamer func(h *Harness, srcParent, dstParent p9.File, origName, newName string, selfRename bool) error + +// renameAt is a renamer. +func renameAt(_ *Harness, srcParent, dstParent p9.File, srcName, dstName string, selfRename bool) error { + return srcParent.RenameAt(srcName, dstParent, dstName) +} + +// rename is a renamer. +func rename(h *Harness, srcParent, dstParent p9.File, srcName, dstName string, selfRename bool) error { + _, f, err := srcParent.Walk([]string{srcName}) + if err != nil { + return err + } + defer f.Close() + if !selfRename { + backend := h.Pop(f) + backend.EXPECT().Renamed(gomock.Any(), dstName).Do(func(p p9.File, name string) { + backend.parent = p.(*Mock) // Required for close ordering. + }) + } + return f.Rename(dstParent, dstName) +} + +// renameHelper executes a rename, and asserts that all relevant elements +// receive expected notifications. If overwriting a file, this includes +// ensuring that the target has been appropriately marked as unlinked. +func renameHelper(h *Harness, root p9.File, srcNames []string, dstNames []string, target fileGenerator, renameFn renamer) { + // Walk to the directory containing the target. + srcQID, targetParent, err := root.Walk(srcNames[:len(srcNames)-1]) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer targetParent.Close() + targetParentBackend := h.Pop(targetParent) + + // Walk to or generate the target file. + _, targetBackend, src := target(h, srcNames[len(srcNames)-1], targetParent) + defer src.Close() + + // Walk to a second reference. + _, second, err := targetParent.Walk([]string{srcNames[len(srcNames)-1]}) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer second.Close() + secondBackend := h.Pop(second) + + // Walk to a third reference, from the start. + _, third, err := root.Walk(srcNames) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer third.Close() + thirdBackend := h.Pop(third) + + // Find the common suffix to identify the rename parent. + var ( + renameDestPath []string + renameSrcPath []string + selfRename bool + ) + for i := 1; i <= len(srcNames) && i <= len(dstNames); i++ { + if srcNames[len(srcNames)-i] != dstNames[len(dstNames)-i] { + // Take the full prefix of dstNames up until this + // point, including the first mismatched name. The + // first mismatch must be the renamed entry. + renameDestPath = dstNames[:len(dstNames)-i+1] + renameSrcPath = srcNames[:len(srcNames)-i+1] + + // Does the renameDestPath fully contain the + // renameSrcPath here? If yes, then this is a mismatch. + // We can't rename the src to some subpath of itself. + if len(renameDestPath) > len(renameSrcPath) && + reflect.DeepEqual(renameDestPath[:len(renameSrcPath)], renameSrcPath) { + renameDestPath = nil + renameSrcPath = nil + continue + } + break + } + } + if len(renameSrcPath) == 0 || len(renameDestPath) == 0 { + // This must be a rename to self, or a tricky look-alike. This + // happens iff we fail to find a suitable divergence in the two + // paths. It's a true self move if the path length is the same. + renameDestPath = dstNames + renameSrcPath = srcNames + selfRename = len(srcNames) == len(dstNames) + } + + // Walk to the source parent. + _, srcParent, err := root.Walk(renameSrcPath[:len(renameSrcPath)-1]) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer srcParent.Close() + srcParentBackend := h.Pop(srcParent) + + // Walk to the destination parent. + _, dstParent, err := root.Walk(renameDestPath[:len(renameDestPath)-1]) + if err != nil { + h.t.Fatalf("got walk err %v, want nil", err) + } + defer dstParent.Close() + dstParentBackend := h.Pop(dstParent) + + // expectedErr is the result of the rename operation. + var expectedErr error + + // Walk to the target file, if one exists. + dstQID, dst, err := root.Walk(renameDestPath) + if err == nil { + if !selfRename && srcQID[0].Type == dstQID[0].Type { + // If there is a destination file, and is it of the + // same type as the source file, then we expect the + // rename to succeed. We expect the destination file to + // be deleted, so we run a deletion test on it in this + // case. + defer checkDeleted(h, dst) + } else { + if !selfRename { + // If the type is different than the + // destination, then we expect the rename to + // fail. We expect ensure that this is + // returned. + expectedErr = syscall.EINVAL + } else { + // This is the file being renamed to itself. + // This is technically allowed and a no-op, but + // all the triggers will fire. + } + dst.Close() + } + } + dstName := renameDestPath[len(renameDestPath)-1] // Renamed element. + srcName := renameSrcPath[len(renameSrcPath)-1] // Renamed element. + if expectedErr == nil && !selfRename { + // Expect all to be renamed appropriately. Note that if this is + // a final file being renamed, then we expect the file to be + // called with the new parent. If not, then we expect the + // rename hook to be called, but the parent will remain + // unchanged. + elements := srcNames[len(renameSrcPath):] + expectRenamed(targetBackend, elements, dstParentBackend, dstName) + expectRenamed(secondBackend, elements, dstParentBackend, dstName) + expectRenamed(thirdBackend, elements, dstParentBackend, dstName) + + // The target parent has also been opened, and may be moved + // directly or indirectly. + if len(elements) > 1 { + expectRenamed(targetParentBackend, elements[:len(elements)-1], dstParentBackend, dstName) + } + } + + // Expect the rename if it's not the same file. Note that like unlink, + // renames are always translated to the at variant in the backend. + if !selfRename { + srcParentBackend.EXPECT().RenameAt(srcName, dstParentBackend, dstName).Return(expectedErr) + } + + // Perform the actual rename; everything has been lined up. + if err := renameFn(h, srcParent, dstParent, srcName, dstName, selfRename); err != expectedErr { + h.t.Fatalf("got rename err %v, want %v", err, expectedErr) + } +} + +func renameTest(t *testing.T, srcNames []string, dstNames []string, target fileGenerator) { + t.Run(fmt.Sprintf("renameAt(%s->%s)", strings.Join(srcNames, "/"), strings.Join(dstNames, "/")), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + renameHelper(h, root, srcNames, dstNames, target, renameAt) + }) + t.Run(fmt.Sprintf("rename(%s->%s)", strings.Join(srcNames, "/"), strings.Join(dstNames, "/")), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + renameHelper(h, root, srcNames, dstNames, target, rename) + }) +} + +func TestRename(t *testing.T) { + // In-directory rename, simple case. + for name := range newTypeMap(nil) { + // Within the root. + renameTest(t, []string{name}, []string{"renamed"}, walkHelper) + renameTest(t, []string{name}, []string{"renamed"}, walkAndOpenHelper) + + // Within a subdirectory. + renameTest(t, []string{"one", name}, []string{"one", "renamed"}, walkHelper) + renameTest(t, []string{"one", name}, []string{"one", "renamed"}, walkAndOpenHelper) + } + + // ... with created files. + renameTest(t, []string{"created"}, []string{"renamed"}, createHelper) + renameTest(t, []string{"one", "created"}, []string{"one", "renamed"}, createHelper) + + // Across directories. + for name := range newTypeMap(nil) { + // Down one level. + renameTest(t, []string{"one", name}, []string{"one", "two", "renamed"}, walkHelper) + renameTest(t, []string{"one", name}, []string{"one", "two", "renamed"}, walkAndOpenHelper) + + // Up one level. + renameTest(t, []string{"one", "two", name}, []string{"one", "renamed"}, walkHelper) + renameTest(t, []string{"one", "two", name}, []string{"one", "renamed"}, walkAndOpenHelper) + + // Across at the same level. + renameTest(t, []string{"one", name}, []string{"three", "renamed"}, walkHelper) + renameTest(t, []string{"one", name}, []string{"three", "renamed"}, walkAndOpenHelper) + } + + // ... with created files. + renameTest(t, []string{"one", "created"}, []string{"one", "two", "renamed"}, createHelper) + renameTest(t, []string{"one", "two", "created"}, []string{"one", "renamed"}, createHelper) + renameTest(t, []string{"one", "created"}, []string{"three", "renamed"}, createHelper) + + // Renaming parents. + for name := range newTypeMap(nil) { + // Rename a parent. + renameTest(t, []string{"one", name}, []string{"renamed", name}, walkHelper) + renameTest(t, []string{"one", name}, []string{"renamed", name}, walkAndOpenHelper) + + // Rename a super parent. + renameTest(t, []string{"one", "two", name}, []string{"renamed", name}, walkHelper) + renameTest(t, []string{"one", "two", name}, []string{"renamed", name}, walkAndOpenHelper) + } + + // ... with created files. + renameTest(t, []string{"one", "created"}, []string{"renamed", "created"}, createHelper) + renameTest(t, []string{"one", "two", "created"}, []string{"renamed", "created"}, createHelper) + + // Over existing files, including itself. + for name := range newTypeMap(nil) { + for other := range newTypeMap(nil) { + // Overwrite the noted file (may be itself). + renameTest(t, []string{"one", name}, []string{"one", other}, walkHelper) + renameTest(t, []string{"one", name}, []string{"one", other}, walkAndOpenHelper) + + // Overwrite other files in another directory. + renameTest(t, []string{"one", name}, []string{"one", "two", other}, walkHelper) + renameTest(t, []string{"one", name}, []string{"one", "two", other}, walkAndOpenHelper) + } + + // Overwrite by moving the parent. + renameTest(t, []string{"three", name}, []string{"one", name}, walkHelper) + renameTest(t, []string{"three", name}, []string{"one", name}, walkAndOpenHelper) + + // Create over the types. + renameTest(t, []string{"one", "created"}, []string{"one", name}, createHelper) + renameTest(t, []string{"one", "created"}, []string{"one", "two", name}, createHelper) + renameTest(t, []string{"three", "created"}, []string{"one", name}, createHelper) + } +} + +func TestRenameInvalid(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + for name := range newTypeMap(nil) { + for _, invalidName := range allInvalidNames(name) { + if err := root.Rename(root, invalidName); err != syscall.EINVAL { + t.Errorf("got %v for name %q, want EINVAL", err, invalidName) + } + } + } +} + +func TestRenameAtInvalid(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + for name := range newTypeMap(nil) { + for _, invalidName := range allInvalidNames(name) { + if err := root.RenameAt(invalidName, root, "okay"); err != syscall.EINVAL { + t.Errorf("got %v for name %q, want EINVAL", err, invalidName) + } + if err := root.RenameAt("okay", root, invalidName); err != syscall.EINVAL { + t.Errorf("got %v for name %q, want EINVAL", err, invalidName) + } + } + } +} + +func TestReadlink(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + // Walk to the file normally. + _, f, err := root.Walk([]string{name}) + if err != nil { + t.Fatalf("walk failed: got %v, wanted nil", err) + } + defer f.Close() + backend := h.Pop(f) + + const symlinkTarget = "symlink-target" + + if backend.Attr.Mode.IsSymlink() { + // This should only go through on symlinks. + backend.EXPECT().Readlink().Return(symlinkTarget, nil) + } + + // Attempt a Readlink operation. + target, err := f.Readlink() + if err != nil && err != syscall.EINVAL { + t.Errorf("readlink got %v, wanted EINVAL", err) + } else if err == nil && target != symlinkTarget { + t.Errorf("readlink got %v, wanted %v", target, symlinkTarget) + } + }) + } +} + +// fdTest is a wrapper around operations that may send file descriptors. This +// asserts that the file descriptors are working as intended. +func fdTest(t *testing.T, sendFn func(*fd.FD) *fd.FD) { + // Create a pipe that we can read from. + r, w, err := os.Pipe() + if err != nil { + t.Fatalf("unable to create pipe: %v", err) + } + defer r.Close() + defer w.Close() + + // Attempt to send the write end. + wFD, err := fd.NewFromFile(w) + if err != nil { + t.Fatalf("unable to convert file: %v", err) + } + defer wFD.Close() // This is a copy. + + // Send wFD and receive newFD. + newFD := sendFn(wFD) + defer newFD.Close() + + // Attempt to write. + const message = "hello" + if _, err := newFD.Write([]byte(message)); err != nil { + t.Fatalf("write got %v, wanted nil", err) + } + + // Should see the message on our end. + buffer := []byte(message) + if _, err := io.ReadFull(r, buffer); err != nil { + t.Fatalf("read got %v, wanted nil", err) + } + if string(buffer) != message { + t.Errorf("got message %v, wanted %v", string(buffer), message) + } +} + +func TestConnect(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + // Walk to the file normally. + _, backend, f := walkHelper(h, name, root) + defer f.Close() + + // Catch all the non-socket cases. + if !backend.Attr.Mode.IsSocket() { + // This has been set up to fail if Connect is called. + if _, err := f.Connect(p9.ConnectFlags(0)); err != syscall.EINVAL { + t.Errorf("connect got %v, wanted EINVAL", err) + } + return + } + + // Ensure the fd exchange works. + fdTest(t, func(send *fd.FD) *fd.FD { + backend.EXPECT().Connect(p9.ConnectFlags(0)).Return(send, nil) + recv, err := backend.Connect(p9.ConnectFlags(0)) + if err != nil { + t.Fatalf("connect got %v, wanted nil", err) + } + return recv + }) + }) + } +} + +func TestReaddir(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + // Walk to the file normally. + _, backend, f := walkHelper(h, name, root) + defer f.Close() + + // Catch all the non-directory cases. + if !backend.Attr.Mode.IsDir() { + // This has also been set up to fail if Readdir is called. + if _, err := f.Readdir(0, 1); err != syscall.EINVAL { + t.Errorf("readdir got %v, wanted EINVAL", err) + } + return + } + + // Ensure that readdir works for directories. + if _, err := f.Readdir(0, 1); err != syscall.EINVAL { + t.Errorf("readdir got %v, wanted EINVAL", err) + } + if _, _, _, err := f.Open(p9.ReadWrite); err != syscall.EINVAL { + t.Errorf("readdir got %v, wanted EINVAL", err) + } + if _, _, _, err := f.Open(p9.WriteOnly); err != syscall.EINVAL { + t.Errorf("readdir got %v, wanted EINVAL", err) + } + backend.EXPECT().Open(p9.ReadOnly).Times(1) + if _, _, _, err := f.Open(p9.ReadOnly); err != nil { + t.Errorf("readdir got %v, wanted nil", err) + } + backend.EXPECT().Readdir(uint64(0), uint32(1)).Times(1) + if _, err := f.Readdir(0, 1); err != nil { + t.Errorf("readdir got %v, wanted nil", err) + } + }) + } +} + +func TestOpen(t *testing.T) { + type openTest struct { + name string + mode p9.OpenFlags + err error + match func(p9.FileMode) bool + } + + cases := []openTest{ + { + name: "invalid", + mode: ^p9.OpenFlagsModeMask, + err: syscall.EINVAL, + match: func(p9.FileMode) bool { return true }, + }, + { + name: "not-openable-read-only", + mode: p9.ReadOnly, + err: syscall.EINVAL, + match: func(mode p9.FileMode) bool { return !p9.CanOpen(mode) }, + }, + { + name: "not-openable-write-only", + mode: p9.WriteOnly, + err: syscall.EINVAL, + match: func(mode p9.FileMode) bool { return !p9.CanOpen(mode) }, + }, + { + name: "not-openable-read-write", + mode: p9.ReadWrite, + err: syscall.EINVAL, + match: func(mode p9.FileMode) bool { return !p9.CanOpen(mode) }, + }, + { + name: "directory-read-only", + mode: p9.ReadOnly, + err: nil, + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + }, + { + name: "directory-read-write", + mode: p9.ReadWrite, + err: syscall.EINVAL, + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + }, + { + name: "directory-write-only", + mode: p9.WriteOnly, + err: syscall.EINVAL, + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + }, + { + name: "read-only", + mode: p9.ReadOnly, + err: nil, + match: func(mode p9.FileMode) bool { return p9.CanOpen(mode) }, + }, + { + name: "write-only", + mode: p9.WriteOnly, + err: nil, + match: func(mode p9.FileMode) bool { return p9.CanOpen(mode) && !mode.IsDir() }, + }, + { + name: "read-write", + mode: p9.ReadWrite, + err: nil, + match: func(mode p9.FileMode) bool { return p9.CanOpen(mode) && !mode.IsDir() }, + }, + } + + // Open(mode OpenFlags) (*fd.FD, QID, uint32, error) + // - only works on Regular, NamedPipe, BLockDevice, CharacterDevice + // - returning a file works as expected + for name := range newTypeMap(nil) { + for _, tc := range cases { + t.Run(fmt.Sprintf("%s-%s", tc.name, name), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + // Walk to the file normally. + _, backend, f := walkHelper(h, name, root) + defer f.Close() + + // Does this match the case? + if !tc.match(backend.Attr.Mode) { + t.SkipNow() + } + + // Ensure open-required operations fail. + if _, err := f.ReadAt([]byte("hello"), 0); err != syscall.EINVAL { + t.Errorf("readAt got %v, wanted EINVAL", err) + } + if _, err := f.WriteAt(make([]byte, 6), 0); err != syscall.EINVAL { + t.Errorf("writeAt got %v, wanted EINVAL", err) + } + if err := f.FSync(); err != syscall.EINVAL { + t.Errorf("fsync got %v, wanted EINVAL", err) + } + if _, err := f.Readdir(0, 1); err != syscall.EINVAL { + t.Errorf("readdir got %v, wanted EINVAL", err) + } + + // Attempt the given open. + if tc.err != nil { + // We expect an error, just test and return. + if _, _, _, err := f.Open(tc.mode); err != tc.err { + t.Fatalf("open with mode %v got %v, want %v", tc.mode, err, tc.err) + } + return + } + + // Run an FD test, since we expect success. + fdTest(t, func(send *fd.FD) *fd.FD { + backend.EXPECT().Open(tc.mode).Return(send, p9.QID{}, uint32(0), nil).Times(1) + recv, _, _, err := f.Open(tc.mode) + if err != tc.err { + t.Fatalf("open with mode %v got %v, want %v", tc.mode, err, tc.err) + } + return recv + }) + + // If the open was successful, attempt another one. + if _, _, _, err := f.Open(tc.mode); err != syscall.EINVAL { + t.Errorf("second open with mode %v got %v, want EINVAL", tc.mode, err) + } + + // Ensure that all illegal operations fail. + if _, _, err := f.Walk(nil); err != syscall.EINVAL && err != syscall.EBUSY { + t.Errorf("walk got %v, wanted EINVAL or EBUSY", err) + } + if _, _, _, _, err := f.WalkGetAttr(nil); err != syscall.EINVAL && err != syscall.EBUSY { + t.Errorf("walkgetattr got %v, wanted EINVAL or EBUSY", err) + } + }) + } + } +} + +func TestClose(t *testing.T) { + type closeTest struct { + name string + closeFn func(backend *Mock, f p9.File) + } + + cases := []closeTest{ + { + name: "close", + closeFn: func(_ *Mock, f p9.File) { + f.Close() + }, + }, + { + name: "remove", + closeFn: func(backend *Mock, f p9.File) { + // Allow the rename call in the parent, automatically translated. + backend.parent.EXPECT().UnlinkAt(gomock.Any(), gomock.Any()).Times(1) + f.(deprecatedRemover).Remove() + }, + }, } - os.Remove(osFile.Name()) - hfi, err := osFile.Stat() - if err != nil { - osFile.Close() - t.Fatalf("stat failed: %v", err) + for name := range newTypeMap(nil) { + for _, tc := range cases { + t.Run(fmt.Sprintf("%s(%s)", tc.name, name), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + // Walk to the file normally. + _, backend, f := walkHelper(h, name, root) + + // Close via the prescribed method. + tc.closeFn(backend, f) + + // Everything should fail with EBADF. + if _, _, err := f.Walk(nil); err != syscall.EBADF { + t.Errorf("walk got %v, wanted EBADF", err) + } + if _, err := f.StatFS(); err != syscall.EBADF { + t.Errorf("statfs got %v, wanted EBADF", err) + } + if _, _, _, err := f.GetAttr(p9.AttrMaskAll()); err != syscall.EBADF { + t.Errorf("getattr got %v, wanted EBADF", err) + } + if err := f.SetAttr(p9.SetAttrMask{}, p9.SetAttr{}); err != syscall.EBADF { + t.Errorf("setattrk got %v, wanted EBADF", err) + } + if err := f.Rename(root, "new-name"); err != syscall.EBADF { + t.Errorf("rename got %v, wanted EBADF", err) + } + if err := f.Close(); err != syscall.EBADF { + t.Errorf("close got %v, wanted EBADF", err) + } + if _, _, _, err := f.Open(p9.ReadOnly); err != syscall.EBADF { + t.Errorf("open got %v, wanted EBADF", err) + } + if _, err := f.ReadAt([]byte("hello"), 0); err != syscall.EBADF { + t.Errorf("readAt got %v, wanted EBADF", err) + } + if _, err := f.WriteAt(make([]byte, 6), 0); err != syscall.EBADF { + t.Errorf("writeAt got %v, wanted EBADF", err) + } + if err := f.FSync(); err != syscall.EBADF { + t.Errorf("fsync got %v, wanted EBADF", err) + } + if _, _, _, _, err := f.Create("new-file", p9.ReadWrite, 0, 0, 0); err != syscall.EBADF { + t.Errorf("create got %v, wanted EBADF", err) + } + if _, err := f.Mkdir("new-directory", 0, 0, 0); err != syscall.EBADF { + t.Errorf("mkdir got %v, wanted EBADF", err) + } + if _, err := f.Symlink("old-name", "new-name", 0, 0); err != syscall.EBADF { + t.Errorf("symlink got %v, wanted EBADF", err) + } + if err := f.Link(root, "new-name"); err != syscall.EBADF { + t.Errorf("link got %v, wanted EBADF", err) + } + if _, err := f.Mknod("new-block-device", 0, 0, 0, 0, 0); err != syscall.EBADF { + t.Errorf("mknod got %v, wanted EBADF", err) + } + if err := f.RenameAt("old-name", root, "new-name"); err != syscall.EBADF { + t.Errorf("renameAt got %v, wanted EBADF", err) + } + if err := f.UnlinkAt("name", 0); err != syscall.EBADF { + t.Errorf("unlinkAt got %v, wanted EBADF", err) + } + if _, err := f.Readdir(0, 1); err != syscall.EBADF { + t.Errorf("readdir got %v, wanted EBADF", err) + } + if _, err := f.Readlink(); err != syscall.EBADF { + t.Errorf("readlink got %v, wanted EBADF", err) + } + if err := f.Flush(); err != syscall.EBADF { + t.Errorf("flush got %v, wanted EBADF", err) + } + if _, _, _, _, err := f.WalkGetAttr(nil); err != syscall.EBADF { + t.Errorf("walkgetattr got %v, wanted EBADF", err) + } + if _, err := f.Connect(p9.ConnectFlags(0)); err != syscall.EBADF { + t.Errorf("connect got %v, wanted EBADF", err) + } + }) + } + } +} + +// onlyWorksOnOpenThings is a helper test method for operations that should +// only work on files that have been explicitly opened. +func onlyWorksOnOpenThings(h *Harness, t *testing.T, name string, root p9.File, mode p9.OpenFlags, expectedErr error, fn func(backend *Mock, f p9.File, shouldSucceed bool) error) { + // Walk to the file normally. + _, backend, f := walkHelper(h, name, root) + defer f.Close() + + // Does it work before opening? + if err := fn(backend, f, false); err != syscall.EINVAL { + t.Errorf("operation got %v, wanted EINVAL", err) } - osFileStat := hfi.Sys().(*syscall.Stat_t) - f, err := fd.NewFromFile(osFile) - // osFile should always be closed. - osFile.Close() - if err != nil { - t.Fatalf("unable to create file: %v", err) + // Is this openable? + if !p9.CanOpen(backend.Attr.Mode) { + return // Nothing to do. + } + + // If this is a directory, we can't handle writing. + if backend.Attr.Mode.IsDir() && (mode == p9.ReadWrite || mode == p9.WriteOnly) { + return // Skip. + } + + // Open the file. + backend.EXPECT().Open(mode) + if _, _, _, err := f.Open(mode); err != nil { + t.Fatalf("open got %v, wanted nil", err) + } + + // Attempt the operation. + if err := fn(backend, f, expectedErr == nil); err != expectedErr { + t.Fatalf("operation got %v, wanted %v", err, expectedErr) + } +} + +func TestRead(t *testing.T) { + type readTest struct { + name string + mode p9.OpenFlags + err error + } + + cases := []readTest{ + { + name: "read-only", + mode: p9.ReadOnly, + err: nil, + }, + { + name: "read-write", + mode: p9.ReadWrite, + err: nil, + }, + { + name: "write-only", + mode: p9.WriteOnly, + err: syscall.EPERM, + }, + } + + for name := range newTypeMap(nil) { + for _, tc := range cases { + t.Run(fmt.Sprintf("%s-%s", tc.name, name), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + const message = "hello" + + onlyWorksOnOpenThings(h, t, name, root, tc.mode, tc.err, func(backend *Mock, f p9.File, shouldSucceed bool) error { + if !shouldSucceed { + _, err := f.ReadAt([]byte(message), 0) + return err + } + + // Prepare for the call to readAt in the backend. + backend.EXPECT().ReadAt(gomock.Any(), uint64(0)).Do(func(p []byte, offset uint64) { + copy(p, message) + }).Return(len(message), nil) + + // Make the client call. + p := make([]byte, 2*len(message)) // Double size. + n, err := f.ReadAt(p, 0) + + // Sanity check result. + if err != nil { + return err + } + if n != len(message) { + t.Fatalf("message length incorrect, got %d, want %d", n, len(message)) + } + if !bytes.Equal(p[:n], []byte(message)) { + t.Fatalf("message incorrect, got %v, want %v", p, []byte(message)) + } + return nil // Success. + }) + }) + } + } +} + +func TestWrite(t *testing.T) { + type writeTest struct { + name string + mode p9.OpenFlags + err error } - // Craft attacher to attach to the mocked file which will return our - // temporary file. - fileMock := &FileMock{ - OpenMock: OpenMock{File: f}, - GetAttrMock: GetAttrMock{ - // The mode must be valid always. - Valid: p9.AttrMask{Mode: true}, + cases := []writeTest{ + { + name: "read-only", + mode: p9.ReadOnly, + err: syscall.EPERM, + }, + { + name: "read-write", + mode: p9.ReadWrite, + err: nil, + }, + { + name: "write-only", + mode: p9.WriteOnly, + err: nil, }, } - attacher := &AttachMock{ - File: fileMock, + + for name := range newTypeMap(nil) { + for _, tc := range cases { + t.Run(fmt.Sprintf("%s-%s", tc.name, name), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + const message = "hello" + + onlyWorksOnOpenThings(h, t, name, root, tc.mode, tc.err, func(backend *Mock, f p9.File, shouldSucceed bool) error { + if !shouldSucceed { + _, err := f.WriteAt([]byte(message), 0) + return err + } + + // Prepare for the call to readAt in the backend. + var output []byte // Saved by Do below. + backend.EXPECT().WriteAt(gomock.Any(), uint64(0)).Do(func(p []byte, offset uint64) { + output = p + }).Return(len(message), nil) + + // Make the client call. + n, err := f.WriteAt([]byte(message), 0) + + // Sanity check result. + if err != nil { + return err + } + if n != len(message) { + t.Fatalf("message length incorrect, got %d, want %d", n, len(message)) + } + if !bytes.Equal(output, []byte(message)) { + t.Fatalf("message incorrect, got %v, want %v", output, []byte(message)) + } + return nil // Success. + }) + }) + } } +} - // Make socket pair. - serverSocket, clientSocket, err := unet.SocketPair(false) - if err != nil { - t.Fatalf("socketpair got err %v wanted nil", err) +func TestFSync(t *testing.T) { + for name := range newTypeMap(nil) { + for _, mode := range []p9.OpenFlags{p9.ReadOnly, p9.WriteOnly, p9.ReadWrite} { + t.Run(fmt.Sprintf("%s-%s", mode, name), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + onlyWorksOnOpenThings(h, t, name, root, mode, nil, func(backend *Mock, f p9.File, shouldSucceed bool) error { + if shouldSucceed { + backend.EXPECT().FSync().Times(1) + } + return f.FSync() + }) + }) + } } - defer clientSocket.Close() - server := p9.NewServer(attacher) - go server.Handle(serverSocket) - client, err := p9.NewClient(clientSocket, 1024*1024 /* 1M message size */, p9.HighestVersionString()) - if err != nil { - t.Fatalf("new client got %v, expected nil", err) +} + +func TestFlush(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + _, backend, f := walkHelper(h, name, root) + defer f.Close() + + backend.EXPECT().Flush() + f.Flush() + }) } +} - // Attach to the mocked file. - cFile, err := client.Attach("") - if err != nil { - t.Fatalf("attach failed: %v", err) +// onlyWorksOnDirectories is a helper test method for operations that should +// only work on unopened directories, such as create, mkdir and symlink. +func onlyWorksOnDirectories(h *Harness, t *testing.T, name string, root p9.File, fn func(backend *Mock, f p9.File, shouldSucceed bool) error) { + // Walk to the file normally. + _, backend, f := walkHelper(h, name, root) + defer f.Close() + + // Only directories support mknod. + if !backend.Attr.Mode.IsDir() { + if err := fn(backend, f, false); err != syscall.EINVAL { + t.Errorf("operation got %v, wanted EINVAL", err) + } + return // Nothing else to do. } - // Try to open the mocked file. - clientHostFile, _, _, err := cFile.Open(0) - if err != nil { - t.Fatalf("open failed: %v", err) + // Should succeed. + if err := fn(backend, f, true); err != nil { + t.Fatalf("operation got %v, wanted nil", err) } - var clientStat syscall.Stat_t - if err := syscall.Fstat(clientHostFile.FD(), &clientStat); err != nil { - t.Fatalf("stat failed: %v", err) + + // Open the directory. + backend.EXPECT().Open(p9.ReadOnly).Times(1) + if _, _, _, err := f.Open(p9.ReadOnly); err != nil { + t.Fatalf("open got %v, wanted nil", err) } - // Compare inode nums to make sure it's the same file. - if clientStat.Ino != osFileStat.Ino { - t.Errorf("fd donation failed") + // Should not work again. + if err := fn(backend, f, false); err != syscall.EINVAL { + t.Fatalf("operation got %v, wanted EINVAL", err) } } -// TestClient is a megatest. -// -// This allows us to probe various edge cases, while changing the state of the -// underlying server in expected ways. The test slowly builds server state and -// is documented inline. -// -// We wind up with the following, after probing edge cases: -// -// FID 1: ServerFile (sf). -// FID 2: Directory (d). -// FID 3: File (f). -// FID 4: Symlink (s). -// -// Although you should use the FID method on the individual files. -func TestClient(t *testing.T) { +func TestCreate(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + onlyWorksOnDirectories(h, t, name, root, func(backend *Mock, f p9.File, shouldSucceed bool) error { + if !shouldSucceed { + _, _, _, _, err := f.Create("new-file", p9.ReadWrite, 0, 1, 2) + return err + } + + // If the create is going to succeed, then we + // need to create a new backend file, and we + // clone to ensure that we don't close the + // original. + _, newF, err := f.Walk(nil) + if err != nil { + t.Fatalf("clone got %v, wanted nil", err) + } + defer newF.Close() + newBackend := h.Pop(newF) + + // Run a regular FD test to validate that path. + fdTest(t, func(send *fd.FD) *fd.FD { + // Return the send FD on success. + newFile := h.NewFile()(backend) // New file with the parent backend. + newBackend.EXPECT().Create("new-file", p9.ReadWrite, p9.FileMode(0), p9.UID(1), p9.GID(2)).Return(send, newFile, p9.QID{}, uint32(0), nil) + + // Receive the fd back. + recv, _, _, _, err := newF.Create("new-file", p9.ReadWrite, 0, 1, 2) + if err != nil { + t.Fatalf("create got %v, wanted nil", err) + } + return recv + }) + + // The above will fail via normal test flow, so + // we can assume that it passed. + return nil + }) + }) + } +} + +func TestCreateInvalid(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + for name := range newTypeMap(nil) { + for _, invalidName := range allInvalidNames(name) { + if _, _, _, _, err := root.Create(invalidName, p9.ReadWrite, 0, 0, 0); err != syscall.EINVAL { + t.Errorf("got %v for name %q, want EINVAL", err, invalidName) + } + } + } +} + +func TestMkdir(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + onlyWorksOnDirectories(h, t, name, root, func(backend *Mock, f p9.File, shouldSucceed bool) error { + if shouldSucceed { + backend.EXPECT().Mkdir("new-directory", p9.FileMode(0), p9.UID(1), p9.GID(2)) + } + _, err := f.Mkdir("new-directory", 0, 1, 2) + return err + }) + }) + } +} + +func TestMkdirInvalid(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + for name := range newTypeMap(nil) { + for _, invalidName := range allInvalidNames(name) { + if _, err := root.Mkdir(invalidName, 0, 0, 0); err != syscall.EINVAL { + t.Errorf("got %v for name %q, want EINVAL", err, invalidName) + } + } + } +} + +func TestSymlink(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + onlyWorksOnDirectories(h, t, name, root, func(backend *Mock, f p9.File, shouldSucceed bool) error { + if shouldSucceed { + backend.EXPECT().Symlink("old-name", "new-name", p9.UID(1), p9.GID(2)) + } + _, err := f.Symlink("old-name", "new-name", 1, 2) + return err + }) + }) + } +} + +func TestSyminkInvalid(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + for name := range newTypeMap(nil) { + for _, invalidName := range allInvalidNames(name) { + // We need only test for invalid names in the new name, + // the target can be an arbitrary string and we don't + // need to sanity check it. + if _, err := root.Symlink("old-name", invalidName, 0, 0); err != syscall.EINVAL { + t.Errorf("got %v for name %q, want EINVAL", err, invalidName) + } + } + } +} + +func TestLink(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + onlyWorksOnDirectories(h, t, name, root, func(backend *Mock, f p9.File, shouldSucceed bool) error { + if shouldSucceed { + backend.EXPECT().Link(gomock.Any(), "new-link") + } + return f.Link(f, "new-link") + }) + }) + } +} + +func TestLinkInvalid(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + for name := range newTypeMap(nil) { + for _, invalidName := range allInvalidNames(name) { + if err := root.Link(root, invalidName); err != syscall.EINVAL { + t.Errorf("got %v for name %q, want EINVAL", err, invalidName) + } + } + } +} + +func TestMknod(t *testing.T) { + for name := range newTypeMap(nil) { + t.Run(name, func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + onlyWorksOnDirectories(h, t, name, root, func(backend *Mock, f p9.File, shouldSucceed bool) error { + if shouldSucceed { + backend.EXPECT().Mknod("new-block-device", p9.FileMode(0), uint32(1), uint32(2), p9.UID(3), p9.GID(4)).Times(1) + } + _, err := f.Mknod("new-block-device", 0, 1, 2, 3, 4) + return err + }) + }) + } +} + +// concurrentFn is a specification of a concurrent operation. This is used to +// drive the concurrency tests below. +type concurrentFn struct { + name string + match func(p9.FileMode) bool + op func(h *Harness, backend *Mock, f p9.File, callback func()) +} + +func concurrentTest(t *testing.T, name string, fn1, fn2 concurrentFn, sameDir, expectedOkay bool) { var ( - // Sentinel error. - sentinelErr = syscall.Errno(4383) - - // Backend mocks. - a = &AttachMock{} - sf = &FileMock{} - d = &FileMock{} - f = &FileMock{} - s = &FileMock{} - - // Client Files for the above. - sfFile p9.File + names1 []string + names2 []string ) + if sameDir { + // Use the same file one directory up. + names1, names2 = []string{"one", name}, []string{"one", name} + } else { + // For different directories, just use siblings. + names1, names2 = []string{"one", name}, []string{"three", name} + } - testSteps := []struct { - name string - fn func(*p9.Client) error - want error - }{ - { - name: "bad-attach", - want: sentinelErr, - fn: func(c *p9.Client) error { - a.File = nil - a.Err = sentinelErr - _, err := c.Attach("") - return err + t.Run(fmt.Sprintf("%s(%v)+%s(%v)", fn1.name, names1, fn2.name, names2), func(t *testing.T) { + h, c := NewHarness(t) + defer h.Finish() + + _, root := newRoot(h, c) + defer root.Close() + + // Walk to both files as given. + _, f1, err := root.Walk(names1) + if err != nil { + t.Fatalf("error walking, got %v, want nil", err) + } + defer f1.Close() + b1 := h.Pop(f1) + _, f2, err := root.Walk(names2) + if err != nil { + t.Fatalf("error walking, got %v, want nil", err) + } + defer f2.Close() + b2 := h.Pop(f2) + + // Are these a good match for the current test case? + if !fn1.match(b1.Attr.Mode) { + t.SkipNow() + } + if !fn2.match(b2.Attr.Mode) { + t.SkipNow() + } + + // Construct our "concurrency creator". + in1 := make(chan struct{}, 1) + in2 := make(chan struct{}, 1) + var top sync.WaitGroup + var fns sync.WaitGroup + defer top.Wait() + top.Add(2) // Accounting for below. + defer fns.Done() + fns.Add(1) // See line above; released before top.Wait. + go func() { + defer top.Done() + fn1.op(h, b1, f1, func() { + in1 <- struct{}{} + fns.Wait() + }) + }() + go func() { + defer top.Done() + fn2.op(h, b2, f2, func() { + in2 <- struct{}{} + fns.Wait() + }) + }() + + // Compute a reasonable timeout. If we expect the operation to hang, + // give it 10 milliseconds before we assert that it's fine. After all, + // there will be a lot of these tests. If we don't expect it to hang, + // give it a full minute, since the machine could be slow. + timeout := 10 * time.Millisecond + if expectedOkay { + timeout = 1 * time.Minute + } + + // Read the first channel. + var second chan struct{} + select { + case <-in1: + second = in2 + case <-in2: + second = in1 + } + + // Catch concurrency. + select { + case <-second: + // We finished successful. Is this good? Depends on the + // expected result. + if !expectedOkay { + t.Errorf("%q and %q proceeded concurrently!", fn1.name, fn2.name) + } + case <-time.After(timeout): + // Great, things did not proceed concurrently. Is that what we + // expected? + if expectedOkay { + t.Errorf("%q and %q hung concurrently!", fn1.name, fn2.name) + } + } + }) +} + +func randomFileName() string { + return fmt.Sprintf("%x", rand.Int63()) +} + +func TestConcurrency(t *testing.T) { + readExclusive := []concurrentFn{ + { + // N.B. We can't explicitly check WalkGetAttr behavior, + // but we rely on the fact that the internal code paths + // are the same. + name: "walk", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + // See the documentation of WalkCallback. + // Because walk is actually implemented by the + // mock, we need a special place for this + // callback. + // + // Note that a clone actually locks the parent + // node. So we walk from this node to test + // concurrent operations appropriately. + backend.WalkCallback = func() error { + callback() + return nil + } + f.Walk([]string{randomFileName()}) // Won't exist. }, }, { - name: "attach", - fn: func(c *p9.Client) error { - a.Called = false - a.File = sf - a.Err = nil - // The attached root must have a valid mode. - sf.GetAttrMock.Attr = p9.Attr{Mode: p9.ModeDirectory} - sf.GetAttrMock.Valid = p9.AttrMask{Mode: true} - var err error - sfFile, err = c.Attach("") - if !a.Called { - t.Errorf("Attach never Called?") - } - return err + name: "fsync", + match: func(mode p9.FileMode) bool { return p9.CanOpen(mode) }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Open(gomock.Any()) + backend.EXPECT().FSync().Do(func() { + callback() + }) + f.Open(p9.ReadOnly) // Required. + f.FSync() }, }, { - name: "bad-walk", - want: sentinelErr, - fn: func(c *p9.Client) error { - // Walk only called when WalkGetAttr not available. - sf.WalkGetAttrMock.Err = syscall.ENOSYS - sf.WalkMock.File = d - sf.WalkMock.Err = sentinelErr - _, _, err := sfFile.Walk([]string{"foo", "bar"}) - return err + name: "readdir", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Open(gomock.Any()) + backend.EXPECT().Readdir(gomock.Any(), gomock.Any()).Do(func(uint64, uint32) { + callback() + }) + f.Open(p9.ReadOnly) // Required. + f.Readdir(0, 1) }, }, { - name: "walk-to-dir", - fn: func(c *p9.Client) error { - // Walk only called when WalkGetAttr not available. - sf.WalkGetAttrMock.Err = syscall.ENOSYS - sf.WalkMock.Called = false - sf.WalkMock.Names = nil - sf.WalkMock.File = d - sf.WalkMock.Err = nil - sf.WalkMock.QIDs = []p9.QID{{Type: 1}} - // All intermediate values must be directories. - d.WalkGetAttrMock.Err = syscall.ENOSYS - d.WalkMock.Called = false - d.WalkMock.Names = nil - d.WalkMock.File = d // Walk to self. - d.WalkMock.Err = nil - d.WalkMock.QIDs = []p9.QID{{Type: 1}} - d.GetAttrMock.Attr = p9.Attr{Mode: p9.ModeDirectory} - d.GetAttrMock.Valid = p9.AttrMask{Mode: true} - var qids []p9.QID - var err error - qids, _, err = sfFile.Walk([]string{"foo", "bar"}) - if !sf.WalkMock.Called { - t.Errorf("Walk never Called?") - } - if !d.GetAttrMock.Called { - t.Errorf("GetAttr never Called?") - } - if !reflect.DeepEqual(sf.WalkMock.Names, []string{"foo"}) { - t.Errorf("got names %v wanted []{foo}", sf.WalkMock.Names) - } - if !reflect.DeepEqual(d.WalkMock.Names, []string{"bar"}) { - t.Errorf("got names %v wanted []{bar}", d.WalkMock.Names) - } - if len(qids) != 2 || qids[len(qids)-1].Type != 1 { - t.Errorf("got qids %v wanted []{..., {Type: 1}}", qids) - } - return err + name: "readlink", + match: func(mode p9.FileMode) bool { return mode.IsSymlink() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Readlink().Do(func() { + callback() + }) + f.Readlink() }, }, { - name: "walkgetattr-to-dir", - fn: func(c *p9.Client) error { - sf.WalkGetAttrMock.Called = false - sf.WalkGetAttrMock.Names = nil - sf.WalkGetAttrMock.File = d - sf.WalkGetAttrMock.Err = nil - sf.WalkGetAttrMock.QIDs = []p9.QID{{Type: 1}} - sf.WalkGetAttrMock.Attr = p9.Attr{Mode: p9.ModeDirectory, UID: 1} - sf.WalkGetAttrMock.Valid = p9.AttrMask{Mode: true} - // See above. - d.WalkGetAttrMock.Called = false - d.WalkGetAttrMock.Names = nil - d.WalkGetAttrMock.File = d // Walk to self. - d.WalkGetAttrMock.Err = nil - d.WalkGetAttrMock.QIDs = []p9.QID{{Type: 1}} - d.WalkGetAttrMock.Attr = p9.Attr{Mode: p9.ModeDirectory, UID: 1} - d.WalkGetAttrMock.Valid = p9.AttrMask{Mode: true} - var qids []p9.QID - var err error - var mask p9.AttrMask - var attr p9.Attr - qids, _, mask, attr, err = sfFile.WalkGetAttr([]string{"foo", "bar"}) - if !sf.WalkGetAttrMock.Called { - t.Errorf("Walk never Called?") - } - if !reflect.DeepEqual(sf.WalkGetAttrMock.Names, []string{"foo"}) { - t.Errorf("got names %v wanted []{foo}", sf.WalkGetAttrMock.Names) - } - if !reflect.DeepEqual(d.WalkGetAttrMock.Names, []string{"bar"}) { - t.Errorf("got names %v wanted []{bar}", d.WalkGetAttrMock.Names) - } - if len(qids) != 2 || qids[len(qids)-1].Type != 1 { - t.Errorf("got qids %v wanted []{..., {Type: 1}}", qids) - } - if !reflect.DeepEqual(attr, sf.WalkGetAttrMock.Attr) { - t.Errorf("got attrs %s wanted %s", attr, sf.WalkGetAttrMock.Attr) - } - if !reflect.DeepEqual(mask, sf.WalkGetAttrMock.Valid) { - t.Errorf("got mask %s wanted %s", mask, sf.WalkGetAttrMock.Valid) - } - return err + name: "connect", + match: func(mode p9.FileMode) bool { return mode.IsSocket() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Connect(gomock.Any()).Do(func(p9.ConnectFlags) { + callback() + }) + f.Connect(0) }, }, { - name: "walk-to-file", - fn: func(c *p9.Client) error { - // Basic sanity check is done in walk-to-dir. - // - // Here we just create basic file FIDs to use. - sf.WalkMock.File = f - sf.WalkMock.Err = nil - var err error - _, _, err = sfFile.Walk(nil) - return err + name: "open", + match: func(mode p9.FileMode) bool { return p9.CanOpen(mode) }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Open(gomock.Any()).Do(func(p9.OpenFlags) { + callback() + }) + f.Open(p9.ReadOnly) }, }, { - name: "walk-to-symlink", - fn: func(c *p9.Client) error { - // See note in walk-to-file. - sf.WalkMock.File = s - sf.WalkMock.Err = nil - var err error - _, _, err = sfFile.Walk(nil) - return err + name: "flush", + match: func(mode p9.FileMode) bool { return true }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Flush().Do(func() { + callback() + }) + f.Flush() + }, + }, + } + writeExclusive := []concurrentFn{ + { + // N.B. We can't really check getattr. But this is an + // extremely low-risk function, it seems likely that + // this check is paranoid anyways. + name: "setattr", + match: func(mode p9.FileMode) bool { return true }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().SetAttr(gomock.Any(), gomock.Any()).Do(func(p9.SetAttrMask, p9.SetAttr) { + callback() + }) + f.SetAttr(p9.SetAttrMask{}, p9.SetAttr{}) }, }, { - name: "bad-statfs", - want: sentinelErr, - fn: func(c *p9.Client) error { - sf.StatFSMock.Err = sentinelErr - _, err := sfFile.StatFS() - return err + name: "unlinkAt", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().UnlinkAt(gomock.Any(), gomock.Any()).Do(func(string, uint32) { + callback() + }) + f.UnlinkAt(randomFileName(), 0) }, }, { - name: "statfs", - fn: func(c *p9.Client) error { - sf.StatFSMock.Called = false - sf.StatFSMock.Stat = p9.FSStat{Type: 1} - sf.StatFSMock.Err = nil - stat, err := sfFile.StatFS() - if !sf.StatFSMock.Called { - t.Errorf("StatfS never Called?") - } - if stat.Type != 1 { - t.Errorf("got stat %v wanted {Type: 1}", stat) - } - return err + name: "mknod", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Mknod(gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any()).Do(func(string, p9.FileMode, uint32, uint32, p9.UID, p9.GID) { + callback() + }) + f.Mknod(randomFileName(), 0, 0, 0, 0, 0) + }, + }, + { + name: "link", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Link(gomock.Any(), gomock.Any()).Do(func(p9.File, string) { + callback() + }) + f.Link(f, randomFileName()) + }, + }, + { + name: "symlink", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Symlink(gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any()).Do(func(string, string, p9.UID, p9.GID) { + callback() + }) + f.Symlink(randomFileName(), randomFileName(), 0, 0) + }, + }, + { + name: "mkdir", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().Mkdir(gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any()).Do(func(string, p9.FileMode, p9.UID, p9.GID) { + callback() + }) + f.Mkdir(randomFileName(), 0, 0, 0) + }, + }, + { + name: "create", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + // Return an error for the creation operation, as this is the simplest. + backend.EXPECT().Create(gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any()).Return(nil, nil, p9.QID{}, uint32(0), syscall.EINVAL).Do(func(string, p9.OpenFlags, p9.FileMode, p9.UID, p9.GID) { + callback() + }) + f.Create(randomFileName(), p9.ReadOnly, 0, 0, 0) }, }, } + globalExclusive := []concurrentFn{ + { + name: "remove", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + // Remove operates on a locked parent. So we + // add a child, walk to it and call remove. + // Note that because this operation can operate + // concurrently with itself, we need to + // generate a random file name. + randomFile := randomFileName() + backend.AddChild(randomFile, h.NewFile()) + defer backend.RemoveChild(randomFile) + _, file, err := f.Walk([]string{randomFile}) + if err != nil { + h.t.Fatalf("walk got %v, want nil", err) + } - // First, create a new server and connection. - serverSocket, clientSocket, err := unet.SocketPair(false) - if err != nil { - t.Fatalf("socketpair got err %v wanted nil", err) - } - defer clientSocket.Close() - server := p9.NewServer(a) - go server.Handle(serverSocket) - client, err := p9.NewClient(clientSocket, 1024*1024 /* 1M message size */, p9.HighestVersionString()) - if err != nil { - t.Fatalf("new client got err %v, wanted nil", err) - } + // Remove is automatically translated to the parent. + backend.EXPECT().UnlinkAt(gomock.Any(), gomock.Any()).Do(func(string, uint32) { + callback() + }) - // Now, run through each of the test steps. - for _, step := range testSteps { - err := step.fn(client) - if err != step.want { - // Don't fail, just note this one step failed. - t.Errorf("step %q got %v wanted %v", step.name, err, step.want) - } - } -} + // Remove is also a close. + file.(deprecatedRemover).Remove() + }, + }, + { + name: "rename", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + // Similarly to remove, because we need to + // operate on a child, we allow a walk. + randomFile := randomFileName() + backend.AddChild(randomFile, h.NewFile()) + defer backend.RemoveChild(randomFile) + _, file, err := f.Walk([]string{randomFile}) + if err != nil { + h.t.Fatalf("walk got %v, want nil", err) + } + defer file.Close() + fileBackend := h.Pop(file) -func BenchmarkClient(b *testing.B) { - // Backend mock. - a := &AttachMock{ - File: &FileMock{ - ReadAtMock: ReadAtMock{N: 1}, + // Rename is automatically translated to the parent. + backend.EXPECT().RenameAt(gomock.Any(), gomock.Any(), gomock.Any()).Do(func(string, p9.File, string) { + callback() + }) + + // Attempt the rename. + fileBackend.EXPECT().Renamed(gomock.Any(), gomock.Any()) + file.Rename(f, randomFileName()) + }, }, - } + { + name: "renameAt", + match: func(mode p9.FileMode) bool { return mode.IsDir() }, + op: func(h *Harness, backend *Mock, f p9.File, callback func()) { + backend.EXPECT().RenameAt(gomock.Any(), gomock.Any(), gomock.Any()).Do(func(string, p9.File, string) { + callback() + }) - // First, create a new server and connection. - serverSocket, clientSocket, err := unet.SocketPair(false) - if err != nil { - b.Fatalf("socketpair got err %v wanted nil", err) - } - defer clientSocket.Close() - server := p9.NewServer(a) - go server.Handle(serverSocket) - client, err := p9.NewClient(clientSocket, 1024*1024 /* 1M message size */, p9.HighestVersionString()) - if err != nil { - b.Fatalf("new client got %v, expected nil", err) + // Attempt the rename. There are no active fids + // with this name, so we don't need to expect + // Renamed hooks on anything. + f.RenameAt(randomFileName(), f, randomFileName()) + }, + }, } - // Attach to the server. - f, err := client.Attach("") - if err != nil { - b.Fatalf("error during attach, got %v wanted nil", err) + for _, fn1 := range readExclusive { + for _, fn2 := range readExclusive { + for name := range newTypeMap(nil) { + // Everything should be able to proceed in parallel. + concurrentTest(t, name, fn1, fn2, true, true) + concurrentTest(t, name, fn1, fn2, false, true) + } + } } - // Open the file. - if _, _, _, err := f.Open(p9.ReadOnly); err != nil { - b.Fatalf("error during open, got %v wanted nil", err) + for _, fn1 := range append(readExclusive, writeExclusive...) { + for _, fn2 := range writeExclusive { + for name := range newTypeMap(nil) { + // Only cross-directory functions should proceed in parallel. + concurrentTest(t, name, fn1, fn2, true, false) + concurrentTest(t, name, fn1, fn2, false, true) + } + } } - // Reset the clock. - b.ResetTimer() - - // Do N reads. - var buf [1]byte - for i := 0; i < b.N; i++ { - _, err := f.ReadAt(buf[:], 0) - if err != nil { - b.Fatalf("error during read %d, got %v wanted nil", i, err) + for _, fn1 := range append(append(readExclusive, writeExclusive...), globalExclusive...) { + for _, fn2 := range globalExclusive { + for name := range newTypeMap(nil) { + // Nothing should be able to run in parallel. + concurrentTest(t, name, fn1, fn2, true, false) + concurrentTest(t, name, fn1, fn2, false, false) + } } } } diff --git a/pkg/p9/p9test/mocks.go b/pkg/p9/p9test/mocks.go deleted file mode 100644 index 9a8c14975..000000000 --- a/pkg/p9/p9test/mocks.go +++ /dev/null @@ -1,489 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package p9test - -import ( - "gvisor.googlesource.com/gvisor/pkg/fd" - "gvisor.googlesource.com/gvisor/pkg/p9" -) - -// StatFSMock mocks p9.File.StatFS. -type StatFSMock struct { - Called bool - - // Return. - Stat p9.FSStat - Err error -} - -// StatFS implements p9.File.StatFS. -func (f *StatFSMock) StatFS() (p9.FSStat, error) { - f.Called = true - return f.Stat, f.Err -} - -// GetAttrMock mocks p9.File.GetAttr. -type GetAttrMock struct { - Called bool - - // Args. - Req p9.AttrMask - - // Return. - QID p9.QID - Valid p9.AttrMask - Attr p9.Attr - Err error -} - -// GetAttr implements p9.File.GetAttr. -func (g *GetAttrMock) GetAttr(req p9.AttrMask) (p9.QID, p9.AttrMask, p9.Attr, error) { - g.Called, g.Req = true, req - return g.QID, g.Valid, g.Attr, g.Err -} - -// WalkGetAttrMock mocks p9.File.WalkGetAttr. -type WalkGetAttrMock struct { - Called bool - - // Args. - Names []string - - // Return. - QIDs []p9.QID - File p9.File - Valid p9.AttrMask - Attr p9.Attr - Err error -} - -// WalkGetAttr implements p9.File.WalkGetAttr. -func (w *WalkGetAttrMock) WalkGetAttr(names []string) ([]p9.QID, p9.File, p9.AttrMask, p9.Attr, error) { - w.Called = true - w.Names = append(w.Names, names...) - return w.QIDs, w.File, w.Valid, w.Attr, w.Err -} - -// SetAttrMock mocks p9.File.SetAttr. -type SetAttrMock struct { - Called bool - - // Args. - Valid p9.SetAttrMask - Attr p9.SetAttr - - // Return. - Err error -} - -// SetAttr implements p9.File.SetAttr. -func (s *SetAttrMock) SetAttr(valid p9.SetAttrMask, attr p9.SetAttr) error { - s.Called, s.Valid, s.Attr = true, valid, attr - return s.Err -} - -// RemoveMock mocks p9.File.Remove. -type RemoveMock struct { - Called bool - - // Return. - Err error -} - -// Remove implements p9.File.Remove. -func (r *RemoveMock) Remove() error { - r.Called = true - return r.Err -} - -// OpenMock mocks p9.File.Open. -type OpenMock struct { - Called bool - - // Args. - Flags p9.OpenFlags - - // Return. - File *fd.FD - QID p9.QID - IOUnit uint32 - Err error -} - -// Open implements p9.File.Open. -func (o *OpenMock) Open(flags p9.OpenFlags) (*fd.FD, p9.QID, uint32, error) { - o.Called, o.Flags = true, flags - return o.File, o.QID, o.IOUnit, o.Err -} - -// ReadAtMock mocks p9.File.ReadAt. -type ReadAtMock struct { - Called bool - - // Args. - P []byte - Offset uint64 - - // Return. - N int - Err error -} - -// ReadAt implements p9.File.ReadAt. -func (r *ReadAtMock) ReadAt(p []byte, offset uint64) (int, error) { - r.Called, r.P, r.Offset = true, p, offset - return r.N, r.Err -} - -// WriteAtMock mocks p9.File.WriteAt. -type WriteAtMock struct { - Called bool - - // Args. - P []byte - Offset uint64 - - // Return. - N int - Err error -} - -// WriteAt implements p9.File.WriteAt. -func (w *WriteAtMock) WriteAt(p []byte, offset uint64) (int, error) { - w.Called, w.P, w.Offset = true, p, offset - return w.N, w.Err -} - -// FSyncMock mocks p9.File.FSync. -type FSyncMock struct { - Called bool - - // Return. - Err error -} - -// FSync implements p9.File.FSync. -func (f *FSyncMock) FSync() error { - f.Called = true - return f.Err -} - -// MkdirMock mocks p9.File.Mkdir. -type MkdirMock struct { - Called bool - - // Args. - Name string - Permissions p9.FileMode - UID p9.UID - GID p9.GID - - // Return. - QID p9.QID - Err error -} - -// Mkdir implements p9.File.Mkdir. -func (s *MkdirMock) Mkdir(name string, permissions p9.FileMode, uid p9.UID, gid p9.GID) (p9.QID, error) { - s.Called, s.Name, s.Permissions, s.UID, s.GID = true, name, permissions, uid, gid - return s.QID, s.Err -} - -// SymlinkMock mocks p9.File.Symlink. -type SymlinkMock struct { - Called bool - - // Args. - Oldname string - Newname string - UID p9.UID - GID p9.GID - - // Return. - QID p9.QID - Err error -} - -// Symlink implements p9.File.Symlink. -func (s *SymlinkMock) Symlink(oldname string, newname string, uid p9.UID, gid p9.GID) (p9.QID, error) { - s.Called, s.Oldname, s.Newname, s.UID, s.GID = true, oldname, newname, uid, gid - return s.QID, s.Err -} - -// MknodMock mocks p9.File.Mknod. -type MknodMock struct { - Called bool - - // Args. - Name string - Permissions p9.FileMode - Major uint32 - Minor uint32 - UID p9.UID - GID p9.GID - - // Return. - QID p9.QID - Err error -} - -// Mknod implements p9.File.Mknod. -func (m *MknodMock) Mknod(name string, permissions p9.FileMode, major uint32, minor uint32, uid p9.UID, gid p9.GID) (p9.QID, error) { - m.Called, m.Name, m.Permissions, m.Major, m.Minor, m.UID, m.GID = true, name, permissions, major, minor, uid, gid - return m.QID, m.Err -} - -// UnlinkAtMock mocks p9.File.UnlinkAt. -type UnlinkAtMock struct { - Called bool - - // Args. - Name string - Flags uint32 - - // Return. - Err error -} - -// UnlinkAt implements p9.File.UnlinkAt. -func (u *UnlinkAtMock) UnlinkAt(name string, flags uint32) error { - u.Called, u.Name, u.Flags = true, name, flags - return u.Err -} - -// ReaddirMock mocks p9.File.Readdir. -type ReaddirMock struct { - Called bool - - // Args. - Offset uint64 - Count uint32 - - // Return. - Dirents []p9.Dirent - Err error -} - -// Readdir implements p9.File.Readdir. -func (r *ReaddirMock) Readdir(offset uint64, count uint32) ([]p9.Dirent, error) { - r.Called, r.Offset, r.Count = true, offset, count - return r.Dirents, r.Err -} - -// ReadlinkMock mocks p9.File.Readlink. -type ReadlinkMock struct { - Called bool - - // Return. - Target string - Err error -} - -// Readlink implements p9.File.Readlink. -func (r *ReadlinkMock) Readlink() (string, error) { - r.Called = true - return r.Target, r.Err -} - -// AttachMock mocks p9.Attacher.Attach. -type AttachMock struct { - Called bool - - // Return. - File p9.File - Err error -} - -// Attach implements p9.Attacher.Attach. -func (a *AttachMock) Attach() (p9.File, error) { - a.Called = true - return a.File, a.Err -} - -// WalkMock mocks p9.File.Walk. -type WalkMock struct { - Called bool - - // Args. - Names []string - - // Return. - QIDs []p9.QID - File p9.File - Err error -} - -// Walk implements p9.File.Walk. -func (w *WalkMock) Walk(names []string) ([]p9.QID, p9.File, error) { - w.Called = true - w.Names = append(w.Names, names...) - return w.QIDs, w.File, w.Err -} - -// RenameMock mocks p9.File.Rename. -type RenameMock struct { - Called bool - - // Args. - Directory p9.File - Name string - - // Return. - Err error -} - -// Rename implements p9.File.Rename. -func (r *RenameMock) Rename(directory p9.File, name string) error { - r.Called, r.Directory, r.Name = true, directory, name - return r.Err -} - -// CloseMock mocks p9.File.Close. -type CloseMock struct { - Called bool - - // Return. - Err error -} - -// Close implements p9.File.Close. -func (d *CloseMock) Close() error { - d.Called = true - return d.Err -} - -// CreateMock mocks p9.File.Create. -type CreateMock struct { - Called bool - - // Args. - Name string - Flags p9.OpenFlags - Permissions p9.FileMode - UID p9.UID - GID p9.GID - - // Return. - HostFile *fd.FD - File p9.File - QID p9.QID - IOUnit uint32 - Err error -} - -// Create implements p9.File.Create. -func (c *CreateMock) Create(name string, flags p9.OpenFlags, permissions p9.FileMode, uid p9.UID, gid p9.GID) (*fd.FD, p9.File, p9.QID, uint32, error) { - c.Called, c.Name, c.Flags, c.Permissions, c.UID, c.GID = true, name, flags, permissions, uid, gid - return c.HostFile, c.File, c.QID, c.IOUnit, c.Err -} - -// LinkMock mocks p9.File.Link. -type LinkMock struct { - Called bool - - // Args. - Target p9.File - Newname string - - // Return. - Err error -} - -// Link implements p9.File.Link. -func (l *LinkMock) Link(target p9.File, newname string) error { - l.Called, l.Target, l.Newname = true, target, newname - return l.Err -} - -// RenameAtMock mocks p9.File.RenameAt. -type RenameAtMock struct { - Called bool - - // Args. - Oldname string - Newdir p9.File - Newname string - - // Return. - Err error -} - -// RenameAt implements p9.File.RenameAt. -func (r *RenameAtMock) RenameAt(oldname string, newdir p9.File, newname string) error { - r.Called, r.Oldname, r.Newdir, r.Newname = true, oldname, newdir, newname - return r.Err -} - -// FlushMock mocks p9.File.Flush. -type FlushMock struct { - Called bool - - // Return. - Err error -} - -// Flush implements p9.File.Flush. -func (f *FlushMock) Flush() error { - return f.Err -} - -// ConnectMock mocks p9.File.Connect. -type ConnectMock struct { - Called bool - - // Args. - Flags p9.ConnectFlags - - // Return. - File *fd.FD - Err error -} - -// Connect implements p9.File.Connect. -func (o *ConnectMock) Connect(flags p9.ConnectFlags) (*fd.FD, error) { - o.Called, o.Flags = true, flags - return o.File, o.Err -} - -// FileMock mocks p9.File. -type FileMock struct { - WalkMock - WalkGetAttrMock - StatFSMock - GetAttrMock - SetAttrMock - RemoveMock - RenameMock - CloseMock - OpenMock - ReadAtMock - WriteAtMock - FSyncMock - CreateMock - MkdirMock - SymlinkMock - LinkMock - MknodMock - RenameAtMock - UnlinkAtMock - ReaddirMock - ReadlinkMock - FlushMock - ConnectMock -} - -var ( - _ p9.File = &FileMock{} -) diff --git a/pkg/p9/p9test/p9test.go b/pkg/p9/p9test/p9test.go new file mode 100644 index 000000000..417b55950 --- /dev/null +++ b/pkg/p9/p9test/p9test.go @@ -0,0 +1,329 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package p9test provides standard mocks for p9. +package p9test + +import ( + "fmt" + "sync" + "sync/atomic" + "syscall" + "testing" + + "github.com/golang/mock/gomock" + "gvisor.googlesource.com/gvisor/pkg/p9" + "gvisor.googlesource.com/gvisor/pkg/unet" +) + +// Harness is an attacher mock. +type Harness struct { + t *testing.T + mockCtrl *gomock.Controller + Attacher *MockAttacher + wg sync.WaitGroup + clientSocket *unet.Socket + mu sync.Mutex + created []*Mock +} + +// globalPath is a QID.Path Generator. +var globalPath uint64 + +// MakePath returns a globally unique path. +func MakePath() uint64 { + return atomic.AddUint64(&globalPath, 1) +} + +// Generator is a function that generates a new file. +type Generator func(parent *Mock) *Mock + +// Mock is a common mock element. +type Mock struct { + p9.DefaultWalkGetAttr + *MockFile + parent *Mock + closed bool + harness *Harness + QID p9.QID + Attr p9.Attr + children map[string]Generator + + // WalkCallback is a special function that will be called from within + // the walk context. This is needed for the concurrent tests within + // this package. + WalkCallback func() error +} + +// globalMu protects the children maps in all mocks. Note that this is not a +// particularly elegant solution, but because the test has walks from the root +// through to final nodes, we must share maps below, and it's easiest to simply +// protect against concurrent access globally. +var globalMu sync.RWMutex + +// AddChild adds a new child to the Mock. +func (m *Mock) AddChild(name string, generator Generator) { + globalMu.Lock() + defer globalMu.Unlock() + m.children[name] = generator +} + +// RemoveChild removes the child with the given name. +func (m *Mock) RemoveChild(name string) { + globalMu.Lock() + defer globalMu.Unlock() + delete(m.children, name) +} + +// Matches implements gomock.Matcher.Matches. +func (m *Mock) Matches(x interface{}) bool { + if om, ok := x.(*Mock); ok { + return m.QID.Path == om.QID.Path + } + return false +} + +// String implements gomock.Matcher.String. +func (m *Mock) String() string { + return fmt.Sprintf("Mock{Mode: 0x%x, QID.Path: %d}", m.Attr.Mode, m.QID.Path) +} + +// GetAttr returns the current attributes. +func (m *Mock) GetAttr(mask p9.AttrMask) (p9.QID, p9.AttrMask, p9.Attr, error) { + return m.QID, p9.AttrMaskAll(), m.Attr, nil +} + +// Walk supports clone and walking in directories. +func (m *Mock) Walk(names []string) ([]p9.QID, p9.File, error) { + if m.WalkCallback != nil { + if err := m.WalkCallback(); err != nil { + return nil, nil, err + } + } + if len(names) == 0 { + // Clone the file appropriately. + nm := m.harness.NewMock(m.parent, m.QID.Path, m.Attr) + nm.children = m.children // Inherit children. + return []p9.QID{nm.QID}, nm, nil + } else if len(names) != 1 { + m.harness.t.Fail() // Should not happen. + return nil, nil, syscall.EINVAL + } + + if m.Attr.Mode.IsDir() { + globalMu.RLock() + defer globalMu.RUnlock() + if fn, ok := m.children[names[0]]; ok { + // Generate the child. + nm := fn(m) + return []p9.QID{nm.QID}, nm, nil + } + // No child found. + return nil, nil, syscall.ENOENT + } + + // Call the underlying mock. + return m.MockFile.Walk(names) +} + +// WalkGetAttr calls the default implementation; this is a client-side optimization. +func (m *Mock) WalkGetAttr(names []string) ([]p9.QID, p9.File, p9.AttrMask, p9.Attr, error) { + return m.DefaultWalkGetAttr.WalkGetAttr(names) +} + +// Pop pops off the most recently created Mock and assert that this mock +// represents the same file passed in. If nil is passed in, no check is +// performed. +// +// Precondition: there must be at least one Mock or this will panic. +func (h *Harness) Pop(clientFile p9.File) *Mock { + h.mu.Lock() + defer h.mu.Unlock() + + if clientFile == nil { + // If no clientFile is provided, then we always return the last + // created file. The caller can safely use this as long as + // there is no concurrency. + m := h.created[len(h.created)-1] + h.created = h.created[:len(h.created)-1] + return m + } + + qid, _, _, err := clientFile.GetAttr(p9.AttrMaskAll()) + if err != nil { + // We do not expect this to happen. + panic(fmt.Sprintf("err during Pop: %v", err)) + } + + // Find the relevant file in our created list. We must scan the last + // from back to front to ensure that we favor the most recently + // generated file. + for i := len(h.created) - 1; i >= 0; i-- { + m := h.created[i] + if qid.Path == m.QID.Path { + // Copy and truncate. + copy(h.created[i:], h.created[i+1:]) + h.created = h.created[:len(h.created)-1] + return m + } + } + + // Unable to find relevant file. + panic(fmt.Sprintf("unable to locate file with QID %+v", qid.Path)) +} + +// NewMock returns a new base file. +func (h *Harness) NewMock(parent *Mock, path uint64, attr p9.Attr) *Mock { + m := &Mock{ + MockFile: NewMockFile(h.mockCtrl), + parent: parent, + harness: h, + QID: p9.QID{ + Type: p9.QIDType((attr.Mode & p9.FileModeMask) >> 12), + Path: path, + }, + Attr: attr, + } + + // Always ensure Close is after the parent's close. Note that this + // can't be done via a straight-forward After call, because the parent + // might change after initial creation. We ensure that this is true at + // close time. + m.EXPECT().Close().Return(nil).Times(1).Do(func() { + if m.parent != nil && m.parent.closed { + h.t.FailNow() + } + // Note that this should not be racy, as this operation should + // be protected by the Times(1) above first. + m.closed = true + }) + + // Remember what was created. + h.mu.Lock() + defer h.mu.Unlock() + h.created = append(h.created, m) + + return m +} + +// NewFile returns a new file mock. +// +// Note that ReadAt and WriteAt must be mocked separately. +func (h *Harness) NewFile() Generator { + return func(parent *Mock) *Mock { + return h.NewMock(parent, MakePath(), p9.Attr{Mode: p9.ModeRegular}) + } +} + +// NewDirectory returns a new mock directory. +// +// Note that Mkdir, Link, Mknod, RenameAt, UnlinkAt and Readdir must be mocked +// separately. Walk is provided and children may be manipulated via AddChild +// and RemoveChild. After calling Walk remotely, one can use Pop to find the +// corresponding backend mock on the server side. +func (h *Harness) NewDirectory(contents map[string]Generator) Generator { + return func(parent *Mock) *Mock { + m := h.NewMock(parent, MakePath(), p9.Attr{Mode: p9.ModeDirectory}) + m.children = contents // Save contents. + return m + } +} + +// NewSymlink returns a new mock directory. +// +// Note that Readlink must be mocked separately. +func (h *Harness) NewSymlink() Generator { + return func(parent *Mock) *Mock { + return h.NewMock(parent, MakePath(), p9.Attr{Mode: p9.ModeSymlink}) + } +} + +// NewBlockDevice returns a new mock block device. +func (h *Harness) NewBlockDevice() Generator { + return func(parent *Mock) *Mock { + return h.NewMock(parent, MakePath(), p9.Attr{Mode: p9.ModeBlockDevice}) + } +} + +// NewCharacterDevice returns a new mock character device. +func (h *Harness) NewCharacterDevice() Generator { + return func(parent *Mock) *Mock { + return h.NewMock(parent, MakePath(), p9.Attr{Mode: p9.ModeCharacterDevice}) + } +} + +// NewNamedPipe returns a new mock named pipe. +func (h *Harness) NewNamedPipe() Generator { + return func(parent *Mock) *Mock { + return h.NewMock(parent, MakePath(), p9.Attr{Mode: p9.ModeNamedPipe}) + } +} + +// NewSocket returns a new mock socket. +func (h *Harness) NewSocket() Generator { + return func(parent *Mock) *Mock { + return h.NewMock(parent, MakePath(), p9.Attr{Mode: p9.ModeSocket}) + } +} + +// Finish completes all checks and shuts down the server. +func (h *Harness) Finish() { + h.clientSocket.Close() + h.wg.Wait() + h.mockCtrl.Finish() +} + +// NewHarness creates and returns a new test server. +// +// It should always be used as: +// +// h, c := NewHarness(t) +// defer h.Finish() +// +func NewHarness(t *testing.T) (*Harness, *p9.Client) { + // Create the mock. + mockCtrl := gomock.NewController(t) + h := &Harness{ + t: t, + mockCtrl: mockCtrl, + Attacher: NewMockAttacher(mockCtrl), + } + + // Make socket pair. + serverSocket, clientSocket, err := unet.SocketPair(false) + if err != nil { + t.Fatalf("socketpair got err %v wanted nil", err) + } + + // Start the server, synchronized on exit. + server := p9.NewServer(h.Attacher) + h.wg.Add(1) + go func() { + defer h.wg.Done() + server.Handle(serverSocket) + }() + + // Create the client. + client, err := p9.NewClient(clientSocket, 1024, p9.HighestVersionString()) + if err != nil { + serverSocket.Close() + clientSocket.Close() + t.Fatalf("new client got %v, expected nil", err) + return nil, nil // Never hit. + } + + // Capture the client socket. + h.clientSocket = clientSocket + return h, client +} diff --git a/pkg/p9/path_tree.go b/pkg/p9/path_tree.go new file mode 100644 index 000000000..97f90bcd5 --- /dev/null +++ b/pkg/p9/path_tree.go @@ -0,0 +1,109 @@ +// Copyright 2018 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package p9 + +import ( + "fmt" + "sync" +) + +// pathNode is a single node in a path traversal. +// +// These are shared by all fidRefs that point to the same path. +// +// These are not synchronized because we allow certain operations (file walk) +// to proceed without having to acquire a write lock. The lock in this +// structure exists to synchronize high-level, semantic operations, such as the +// simultaneous creation and deletion of a file. +// +// (+) below is the path component string. +type pathNode struct { + mu sync.RWMutex // See above. + fidRefs sync.Map // => map[*fidRef]string(+) + children sync.Map // => map[string(+)]*pathNode + count int64 +} + +// pathNodeFor returns the path node for the given name, or a new one. +// +// Precondition: mu must be held in a readable fashion. +func (p *pathNode) pathNodeFor(name string) *pathNode { + // Load the existing path node. + if pn, ok := p.children.Load(name); ok { + return pn.(*pathNode) + } + + // Create a new pathNode for shared use. + pn, _ := p.children.LoadOrStore(name, new(pathNode)) + return pn.(*pathNode) +} + +// nameFor returns the name for the given fidRef. +// +// Precondition: mu must be held in a readable fashion. +func (p *pathNode) nameFor(ref *fidRef) string { + if s, ok := p.fidRefs.Load(ref); ok { + return s.(string) + } + + // This should not happen, don't proceed. + panic(fmt.Sprintf("expected name for %+v, none found", ref)) +} + +// addChild adds a child to the given pathNode. +// +// This applies only to an individual fidRef. +// +// Precondition: mu must be held in a writable fashion. +func (p *pathNode) addChild(ref *fidRef, name string) { + if s, ok := p.fidRefs.Load(ref); ok { + // This should not happen, don't proceed. + panic(fmt.Sprintf("unexpected fidRef %+v with path %q, wanted %q", ref, s, name)) + } + + p.fidRefs.Store(ref, name) +} + +// removeChild removes the given child. +// +// This applies only to an individual fidRef. +// +// Precondition: mu must be held in a writable fashion. +func (p *pathNode) removeChild(ref *fidRef) { + p.fidRefs.Delete(ref) +} + +// removeWithName removes all references with the given name. +// +// The original pathNode is returned by this function, and removed from this +// pathNode. Any operations on the removed tree must use this value. +// +// The provided function is executed after removal. +// +// Precondition: mu must be held in a writable fashion. +func (p *pathNode) removeWithName(name string, fn func(ref *fidRef)) *pathNode { + p.fidRefs.Range(func(key, value interface{}) bool { + if value.(string) == name { + p.fidRefs.Delete(key) + fn(key.(*fidRef)) + } + return true + }) + + // Return the original path node. + origPathNode := p.pathNodeFor(name) + p.children.Delete(name) + return origPathNode +} diff --git a/pkg/p9/server.go b/pkg/p9/server.go index 5c7cb18c8..3ef151595 100644 --- a/pkg/p9/server.go +++ b/pkg/p9/server.go @@ -15,6 +15,8 @@ package p9 import ( + "io" + "runtime/debug" "sync" "sync/atomic" "syscall" @@ -27,6 +29,19 @@ import ( type Server struct { // attacher provides the attach function. attacher Attacher + + // pathTree is the full set of paths opened on this server. + // + // These may be across different connections, but rename operations + // must be serialized globally for safely. There is a single pathTree + // for the entire server, and not per connection. + pathTree pathNode + + // renameMu is a global lock protecting rename operations. With this + // lock, we can be certain that any given rename operation can safely + // acquire two path nodes in any order, as all other concurrent + // operations acquire at most a single node. + renameMu sync.RWMutex } // NewServer returns a new server. @@ -81,6 +96,9 @@ type connState struct { // fidRef wraps a node and tracks references. type fidRef struct { + // server is the associated server. + server *Server + // file is the associated File. file File @@ -97,13 +115,39 @@ type fidRef struct { // This is updated in handlers.go. opened bool - // walkable indicates this fidRef may be walked. - walkable bool + // mode is the fidRef's mode from the walk. Only the type bits are + // valid, the permissions may change. This is used to sanity check + // operations on this element, and prevent walks across + // non-directories. + mode FileMode // openFlags is the mode used in the open. // // This is updated in handlers.go. openFlags OpenFlags + + // pathNode is the current pathNode for this FID. + pathNode *pathNode + + // parent is the parent fidRef. We hold on to a parent reference to + // ensure that hooks, such as Renamed, can be executed safely by the + // server code. + // + // Note that parent cannot be changed without holding both the global + // rename lock and a writable lock on the associated pathNode for this + // fidRef. Holding either of these locks is sufficient to examine + // parent safely. + // + // The parent will be nil for root fidRefs, and non-nil otherwise. The + // method maybeParent can be used to return a cyclical reference, and + // isRoot should be used to check for root over looking at parent + // directly. + parent *fidRef + + // deleted indicates that the backing file has been deleted. We stop + // many operations at the API level if they are incompatible with a + // file that has already been unlinked. + deleted uint32 } // OpenFlags returns the flags the file was opened with and true iff the fid was opened previously. @@ -113,13 +157,146 @@ func (f *fidRef) OpenFlags() (OpenFlags, bool) { return f.openFlags, f.opened } +// IncRef increases the references on a fid. +func (f *fidRef) IncRef() { + atomic.AddInt64(&f.refs, 1) +} + // DecRef should be called when you're finished with a fid. func (f *fidRef) DecRef() { if atomic.AddInt64(&f.refs, -1) == 0 { f.file.Close() + + // Drop the parent reference. + // + // Since this fidRef is guaranteed to be non-discoverable when + // the references reach zero, we don't need to worry about + // clearing the parent. + if f.parent != nil { + // If we've been previously deleted, this removing this + // ref is a no-op. That's expected. + f.parent.pathNode.removeChild(f) + f.parent.DecRef() + } } } +// isDeleted returns true if this fidRef has been deleted. +func (f *fidRef) isDeleted() bool { + return atomic.LoadUint32(&f.deleted) != 0 +} + +// isRoot indicates whether this is a root fid. +func (f *fidRef) isRoot() bool { + return f.parent == nil +} + +// maybeParent returns a cyclic reference for roots, and the parent otherwise. +func (f *fidRef) maybeParent() *fidRef { + if f.parent != nil { + return f.parent + } + return f // Root has itself. +} + +// notifyDelete marks all fidRefs as deleted. +// +// Precondition: the write lock must be held on the given pathNode. +func notifyDelete(pn *pathNode) { + // Call on all local references. + pn.fidRefs.Range(func(key, _ interface{}) bool { + ref := key.(*fidRef) + atomic.StoreUint32(&ref.deleted, 1) + return true + }) + + // Call on all subtrees. + pn.children.Range(func(_, value interface{}) bool { + notifyDelete(value.(*pathNode)) + return true + }) +} + +// markChildDeleted marks all children below the given name as deleted. +// +// Precondition: this must be called via safelyWrite or safelyGlobal. +func (f *fidRef) markChildDeleted(name string) { + origPathNode := f.pathNode.removeWithName(name, func(ref *fidRef) { + atomic.StoreUint32(&ref.deleted, 1) + }) + + // Mark everything below as deleted. + notifyDelete(origPathNode) +} + +// notifyNameChange calls the relevant Renamed method on all nodes in the path, +// recursively. Note that this applies only for subtrees, as these +// notifications do not apply to the actual file whose name has changed. +// +// Precondition: the write lock must be held on the given pathNode. +func notifyNameChange(pn *pathNode) { + // Call on all local references. + pn.fidRefs.Range(func(key, value interface{}) bool { + ref := key.(*fidRef) + name := value.(string) + ref.file.Renamed(ref.parent.file, name) + return true + }) + + // Call on all subtrees. + pn.children.Range(func(_, value interface{}) bool { + notifyNameChange(value.(*pathNode)) + return true + }) +} + +// renameChildTo renames the given child to the target. +// +// Precondition: this must be called via safelyGlobal. +func (f *fidRef) renameChildTo(oldName string, target *fidRef, newName string) { + target.markChildDeleted(newName) + origPathNode := f.pathNode.removeWithName(oldName, func(ref *fidRef) { + ref.parent.DecRef() // Drop original reference. + ref.parent = target // Change parent. + ref.parent.IncRef() // Acquire new one. + target.pathNode.addChild(ref, newName) + ref.file.Renamed(target.file, newName) + }) + + // Replace the previous (now deleted) path node. + f.pathNode.children.Store(newName, origPathNode) + + // Call Renamed on everything above. + notifyNameChange(origPathNode) +} + +// safelyRead executes the given operation with the local path node locked. +// This implies that paths will not change during the operation. +func (f *fidRef) safelyRead(fn func() error) (err error) { + f.server.renameMu.RLock() + defer f.server.renameMu.RUnlock() + f.pathNode.mu.RLock() + defer f.pathNode.mu.RUnlock() + return fn() +} + +// safelyWrite executes the given operation with the local path node locked in +// a writable fashion. This implies some paths may change. +func (f *fidRef) safelyWrite(fn func() error) (err error) { + f.server.renameMu.RLock() + defer f.server.renameMu.RUnlock() + f.pathNode.mu.Lock() + defer f.pathNode.mu.Unlock() + return fn() +} + +// safelyGlobal executes the given operation with the global path lock held. +func (f *fidRef) safelyGlobal(fn func() error) (err error) { + f.server.renameMu.Lock() + defer f.server.renameMu.Unlock() + return fn() +} + // LookupFID finds the given FID. // // You should call fid.DecRef when you are finished using the fid. @@ -128,7 +305,7 @@ func (cs *connState) LookupFID(fid FID) (*fidRef, bool) { defer cs.fidMu.Unlock() fidRef, ok := cs.fids[fid] if ok { - atomic.AddInt64(&fidRef.refs, 1) + fidRef.IncRef() return fidRef, true } return nil, false @@ -145,7 +322,7 @@ func (cs *connState) InsertFID(fid FID, newRef *fidRef) { if ok { defer origRef.DecRef() } - atomic.AddInt64(&newRef.refs, 1) + newRef.IncRef() cs.fids[fid] = newRef } @@ -229,10 +406,9 @@ func (cs *connState) handleRequest() { cs.recvDone <- nil // Deal with other errors. - if err != nil { + if err != nil && err != io.EOF { // If it's not a connection error, but some other protocol error, // we can send a response immediately. - log.Debugf("err [%05d] %v", tag, err) cs.sendMu.Lock() err := send(cs.conn, tag, newErr(err)) cs.sendMu.Unlock() @@ -243,12 +419,38 @@ func (cs *connState) handleRequest() { // Try to start the tag. if !cs.StartTag(tag) { // Nothing we can do at this point; client is bogus. + log.Debugf("no valid tag [%05d]", tag) cs.sendDone <- ErrNoValidMessage return } // Handle the message. - var r message + var r message // r is the response. + defer func() { + if r == nil { + // Don't allow a panic to propagate. + recover() + + // Include a useful log message. + log.Warningf("panic in handler: %s", debug.Stack()) + + // Wrap in an EFAULT error; we don't really have a + // better way to describe this kind of error. It will + // usually manifest as a result of the test framework. + r = newErr(syscall.EFAULT) + } + + // Clear the tag before sending. That's because as soon as this + // hits the wire, the client can legally send another message + // with the same tag. + cs.ClearTag(tag) + + // Send back the result. + cs.sendMu.Lock() + err = send(cs.conn, tag, r) + cs.sendMu.Unlock() + cs.sendDone <- err + }() if handler, ok := m.(handler); ok { // Call the message handler. r = handler.handle(cs) @@ -256,18 +458,6 @@ func (cs *connState) handleRequest() { // Produce an ENOSYS error. r = newErr(syscall.ENOSYS) } - - // Clear the tag before sending. That's because as soon - // as this hits the wire, the client can legally send - // another message with the same tag. - cs.ClearTag(tag) - - // Send back the result. - cs.sendMu.Lock() - err = send(cs.conn, tag, r) - cs.sendMu.Unlock() - cs.sendDone <- err - return } func (cs *connState) handleRequests() { diff --git a/pkg/p9/transport.go b/pkg/p9/transport.go index 97396806c..bafb377de 100644 --- a/pkg/p9/transport.go +++ b/pkg/p9/transport.go @@ -167,7 +167,7 @@ func recv(s *unet.Socket, msize uint32, lookup lookupTagAndType) (Tag, message, r.EnableFDs(1) n, err := r.ReadVec([][]byte{hdr[:]}) - if err != nil { + if err != nil && (n == 0 || err != io.EOF) { r.CloseFDs() return NoTag, nil, ErrSocket{err} } @@ -189,10 +189,8 @@ func recv(s *unet.Socket, msize uint32, lookup lookupTagAndType) (Tag, message, // Continuing reading for a short header. for n < int(headerLength) { cur, err := r.ReadVec([][]byte{hdr[n:]}) - if err != nil { + if err != nil && (cur == 0 || err != io.EOF) { return NoTag, nil, ErrSocket{err} - } else if cur == 0 { - return NoTag, nil, ErrSocket{io.EOF} } n += cur } @@ -296,10 +294,8 @@ func recv(s *unet.Socket, msize uint32, lookup lookupTagAndType) (Tag, message, r := s.Reader(true) for n := 0; n < int(remaining); { cur, err := r.ReadVec(vecs) - if err != nil { + if err != nil && (cur == 0 || err != io.EOF) { return NoTag, nil, ErrSocket{err} - } else if cur == 0 { - return NoTag, nil, ErrSocket{io.EOF} } n += cur diff --git a/pkg/rand/BUILD b/pkg/rand/BUILD index 97b9ba3ff..0c9efc709 100644 --- a/pkg/rand/BUILD +++ b/pkg/rand/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "rand", srcs = [ diff --git a/pkg/seccomp/BUILD b/pkg/seccomp/BUILD index 1975d17a6..657f923ed 100644 --- a/pkg/seccomp/BUILD +++ b/pkg/seccomp/BUILD @@ -1,8 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//go:def.bzl", "go_binary", "go_embed_data") +package(licenses = ["notice"]) # Apache 2.0 + go_binary( name = "victim", testonly = 1, diff --git a/pkg/secio/BUILD b/pkg/secio/BUILD index 0ed38c64a..29f751725 100644 --- a/pkg/secio/BUILD +++ b/pkg/secio/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "secio", srcs = [ diff --git a/pkg/sentry/arch/BUILD b/pkg/sentry/arch/BUILD index 314b3e962..9bf04360a 100644 --- a/pkg/sentry/arch/BUILD +++ b/pkg/sentry/arch/BUILD @@ -1,6 +1,7 @@ +load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") + package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") load("//tools/go_stateify:defs.bzl", "go_library") go_library( diff --git a/pkg/sentry/context/BUILD b/pkg/sentry/context/BUILD index 2a7a6df23..02d24defd 100644 --- a/pkg/sentry/context/BUILD +++ b/pkg/sentry/context/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "context", srcs = ["context.go"], diff --git a/pkg/sentry/control/BUILD b/pkg/sentry/control/BUILD index fbdde0721..c3b682d6f 100644 --- a/pkg/sentry/control/BUILD +++ b/pkg/sentry/control/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "control", srcs = [ diff --git a/pkg/sentry/device/BUILD b/pkg/sentry/device/BUILD index 69c99b0b3..bebdb2939 100644 --- a/pkg/sentry/device/BUILD +++ b/pkg/sentry/device/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "device", srcs = ["device.go"], diff --git a/pkg/sentry/fs/anon/BUILD b/pkg/sentry/fs/anon/BUILD index ff4ab850a..4bd912e95 100644 --- a/pkg/sentry/fs/anon/BUILD +++ b/pkg/sentry/fs/anon/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "anon", srcs = [ diff --git a/pkg/sentry/fs/gofer/BUILD b/pkg/sentry/fs/gofer/BUILD index cef01829a..c9e531e40 100644 --- a/pkg/sentry/fs/gofer/BUILD +++ b/pkg/sentry/fs/gofer/BUILD @@ -56,14 +56,10 @@ go_test( srcs = ["gofer_test.go"], embed = [":gofer"], deps = [ - "//pkg/log", "//pkg/p9", "//pkg/p9/p9test", "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", - "//pkg/sentry/kernel/time", - "//pkg/sentry/usermem", - "//pkg/unet", ], ) diff --git a/pkg/sentry/fs/gofer/context_file.go b/pkg/sentry/fs/gofer/context_file.go index a0265c2aa..455953237 100644 --- a/pkg/sentry/fs/gofer/context_file.go +++ b/pkg/sentry/fs/gofer/context_file.go @@ -58,13 +58,6 @@ func (c *contextFile) setAttr(ctx context.Context, valid p9.SetAttrMask, attr p9 return c.file.SetAttr(valid, attr) } -func (c *contextFile) remove(ctx context.Context) error { - ctx.UninterruptibleSleepStart(false) - defer ctx.UninterruptibleSleepFinish(false) - - return c.file.Remove() -} - func (c *contextFile) rename(ctx context.Context, directory contextFile, name string) error { ctx.UninterruptibleSleepStart(false) defer ctx.UninterruptibleSleepFinish(false) diff --git a/pkg/sentry/fs/gofer/gofer_test.go b/pkg/sentry/fs/gofer/gofer_test.go index 3190d1e18..b450778ca 100644 --- a/pkg/sentry/fs/gofer/gofer_test.go +++ b/pkg/sentry/fs/gofer/gofer_test.go @@ -16,110 +16,102 @@ package gofer import ( "fmt" - "io" "syscall" "testing" "time" - "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/p9" "gvisor.googlesource.com/gvisor/pkg/p9/p9test" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/unet" ) -// goodMockFile returns a file that can be Walk'ed to and created. -func goodMockFile(mode p9.FileMode, size uint64) *p9test.FileMock { - return &p9test.FileMock{ - GetAttrMock: p9test.GetAttrMock{ - Attr: p9.Attr{Mode: mode, Size: size, RDev: 0}, - Valid: p9.AttrMaskAll(), - }, - } -} - -func newClosedSocket() (*unet.Socket, error) { - fd, err := syscall.Socket(syscall.AF_UNIX, syscall.SOCK_STREAM, 0) - if err != nil { - return nil, err - } - - s, err := unet.NewSocket(fd) - if err != nil { - syscall.Close(fd) - return nil, err - } - - return s, s.Close() -} - -// root returns a p9 file mock and an fs.InodeOperations created from that file. Any -// functions performed on fs.InodeOperations will use the p9 file mock. -func root(ctx context.Context, cp cachePolicy, mode p9.FileMode, size uint64) (*p9test.FileMock, *fs.Inode, error) { - sock, err := newClosedSocket() - if err != nil { - return nil, nil, err - } - - // Construct a dummy session that we can destruct. - s := &session{ - conn: sock, - mounter: fs.RootOwner, - cachePolicy: cp, - client: &p9.Client{}, - } - - rootFile := goodMockFile(mode, size) - sattr, rootInodeOperations := newInodeOperations(ctx, s, contextFile{file: rootFile}, p9.QID{}, rootFile.GetAttrMock.Valid, rootFile.GetAttrMock.Attr, false /* socket */) - m := fs.NewMountSource(s, &filesystem{}, fs.MountSourceFlags{}) - return rootFile, fs.NewInode(rootInodeOperations, m, sattr), nil +// rootTest runs a test with a p9 mock and an fs.InodeOperations created from +// the attached root directory. The root file will be closed and client +// disconnected, but additional files must be closed manually. +func rootTest(t *testing.T, name string, cp cachePolicy, fn func(context.Context, *p9test.Harness, *p9test.Mock, *fs.Inode)) { + t.Run(name, func(t *testing.T) { + h, c := p9test.NewHarness(t) + defer h.Finish() + + // Create a new root. Note that we pass an empty, but non-nil + // map here. This allows tests to extend the root children + // dynamically. + root := h.NewDirectory(map[string]p9test.Generator{})(nil) + + // Return this as the root. + h.Attacher.EXPECT().Attach().Return(root, nil).Times(1) + + // ... and open via the client. + rootFile, err := c.Attach("/") + if err != nil { + t.Fatalf("unable to attach: %v", err) + } + defer rootFile.Close() + + // Wrap an a session. + s := &session{ + mounter: fs.RootOwner, + cachePolicy: cp, + client: c, + } + + // ... and an INode, with only the mode being explicitly valid for now. + ctx := contexttest.Context(t) + sattr, rootInodeOperations := newInodeOperations(ctx, s, contextFile{ + file: rootFile, + }, root.QID, p9.AttrMaskAll(), root.Attr, false /* socket */) + m := fs.NewMountSource(s, &filesystem{}, fs.MountSourceFlags{}) + rootInode := fs.NewInode(rootInodeOperations, m, sattr) + + // Ensure that the cache is fully invalidated, so that any + // close actions actually take place before the full harness is + // torn down. + defer m.FlushDirentRefs() + + // Execute the test. + fn(ctx, h, root, rootInode) + }) } func TestLookup(t *testing.T) { - // Test parameters. type lookupTest struct { // Name of the test. name string - // Function input parameters. - fileName string - // Expected return value. want error } tests := []lookupTest{ { - name: "mock Walk passes (function succeeds)", - fileName: "ppp", - want: nil, + name: "mock Walk passes (function succeeds)", + want: nil, }, { - name: "mock Walk fails (function fails)", - fileName: "ppp", - want: syscall.ENOENT, + name: "mock Walk fails (function fails)", + want: syscall.ENOENT, }, } - ctx := contexttest.Context(t) + const file = "file" // The walked target file. + for _, test := range tests { - t.Run(test.name, func(t *testing.T) { - // Set up mock. - rootFile, rootInode, err := root(ctx, cacheNone, p9.PermissionsMask, 0) - if err != nil { - t.Fatalf("error creating root: %v", err) + rootTest(t, test.name, cacheNone, func(ctx context.Context, h *p9test.Harness, rootFile *p9test.Mock, rootInode *fs.Inode) { + // Setup the appropriate result. + rootFile.WalkCallback = func() error { + return test.want + } + if test.want == nil { + // Set the contents of the root. We expect a + // normal file generator for ppp above. This is + // overriden by setting WalkErr in the mock. + rootFile.AddChild(file, h.NewFile()) } - - rootFile.WalkGetAttrMock.QIDs = []p9.QID{{}} - rootFile.WalkGetAttrMock.Err = test.want - rootFile.WalkGetAttrMock.File = goodMockFile(p9.PermissionsMask, 0) // Call function. - dirent, err := rootInode.Lookup(ctx, test.fileName) + dirent, err := rootInode.Lookup(ctx, file) // Unwrap the InodeOperations. var newInodeOperations fs.InodeOperations @@ -138,19 +130,12 @@ func TestLookup(t *testing.T) { if err == nil && newInodeOperations == nil { t.Errorf("Lookup got non-nil err and non-nil node, wanted at least one non-nil") } - - // Check mock parameters. - if !rootFile.WalkGetAttrMock.Called { - t.Errorf("GetAttr not called; error: %v", err) - } else if rootFile.WalkGetAttrMock.Names[0] != test.fileName { - t.Errorf("file name not set") - } }) } } func TestRevalidation(t *testing.T) { - tests := []struct { + type revalidationTest struct { cachePolicy cachePolicy // Whether dirent should be reloaded before any modifications. @@ -167,7 +152,9 @@ func TestRevalidation(t *testing.T) { // Whether dirent should be reloaded after the remote has // removed the file. postRemovalWantReload bool - }{ + } + + tests := []revalidationTest{ { // Policy cacheNone causes Revalidate to always return // true. @@ -208,67 +195,83 @@ func TestRevalidation(t *testing.T) { }, } - ctx := contexttest.Context(t) + const file = "file" // The file walked below. + for _, test := range tests { name := fmt.Sprintf("cachepolicy=%s", test.cachePolicy) - t.Run(name, func(t *testing.T) { - // Set up mock. - rootFile, rootInode, err := root(ctx, test.cachePolicy, p9.ModeDirectory|p9.PermissionsMask, 0) - if err != nil { - t.Fatalf("error creating root: %v", err) - } - + rootTest(t, name, test.cachePolicy, func(ctx context.Context, h *p9test.Harness, rootFile *p9test.Mock, rootInode *fs.Inode) { + // Wrap in a dirent object. rootDir := fs.NewDirent(rootInode, "root") - // Create a mock file that we will walk to from the root. - const ( - name = "foo" - mode = p9.PermissionsMask - ) - file := goodMockFile(mode, 0) - file.GetAttrMock.Valid = p9.AttrMaskAll() - - // Tell the root mock how to walk to this file. - rootFile.WalkGetAttrMock.QIDs = []p9.QID{{}} - rootFile.WalkGetAttrMock.File = file - rootFile.WalkGetAttrMock.Attr = file.GetAttrMock.Attr - rootFile.WalkGetAttrMock.Valid = file.GetAttrMock.Valid + // Create a mock file a child of the root. We save when + // this is generated, so that when the time changed, we + // can update the original entry. + var origMocks []*p9test.Mock + rootFile.AddChild(file, func(parent *p9test.Mock) *p9test.Mock { + // Regular a regular file that has a consistent + // path number. This might be used by + // validation so we don't change it. + m := h.NewMock(parent, 0, p9.Attr{ + Mode: p9.ModeRegular, + }) + origMocks = append(origMocks, m) + return m + }) // Do the walk. - dirent, err := rootDir.Walk(ctx, rootDir, name) + dirent, err := rootDir.Walk(ctx, rootDir, file) if err != nil { - t.Fatalf("Lookup(%q) failed: %v", name, err) + t.Fatalf("Lookup failed: %v", err) } - // Walk again. Depending on the cache policy, we may get a new - // dirent. - newDirent, err := rootDir.Walk(ctx, rootDir, name) + // We must release the dirent, of the test will fail + // with a reference leak. This is tracked by p9test. + defer dirent.DecRef() + + // Walk again. Depending on the cache policy, we may + // get a new dirent. + newDirent, err := rootDir.Walk(ctx, rootDir, file) if err != nil { - t.Fatalf("Lookup(%q) failed: %v", name, err) + t.Fatalf("Lookup failed: %v", err) } if test.preModificationWantReload && dirent == newDirent { - t.Errorf("Lookup(%q) with cachePolicy=%s got old dirent %v, wanted a new dirent", name, test.cachePolicy, dirent) + t.Errorf("Lookup with cachePolicy=%s got old dirent %+v, wanted a new dirent", test.cachePolicy, dirent) } if !test.preModificationWantReload && dirent != newDirent { - t.Errorf("Lookup(%q) with cachePolicy=%s got new dirent %v, wanted old dirent %v", name, test.cachePolicy, newDirent, dirent) + t.Errorf("Lookup with cachePolicy=%s got new dirent %+v, wanted old dirent %+v", test.cachePolicy, newDirent, dirent) } + newDirent.DecRef() // See above. - // Modify the underlying mocked file's modification time. + // Modify the underlying mocked file's modification + // time for the next walk that occurs. nowSeconds := time.Now().Unix() - rootFile.WalkGetAttrMock.Attr.MTimeSeconds = uint64(nowSeconds) - file.GetAttrMock.Attr.MTimeSeconds = uint64(nowSeconds) + rootFile.AddChild(file, func(parent *p9test.Mock) *p9test.Mock { + // Ensure that the path is the same as above, + // but we change only the modification time of + // the file. + return h.NewMock(parent, 0, p9.Attr{ + Mode: p9.ModeRegular, + MTimeSeconds: uint64(nowSeconds), + }) + }) + + // We also modify the original time, so that GetAttr + // behaves as expected for the caching case. + for _, m := range origMocks { + m.Attr.MTimeSeconds = uint64(nowSeconds) + } - // Walk again. Depending on the cache policy, we may get a new - // dirent. - newDirent, err = rootDir.Walk(ctx, rootDir, name) + // Walk again. Depending on the cache policy, we may + // get a new dirent. + newDirent, err = rootDir.Walk(ctx, rootDir, file) if err != nil { - t.Fatalf("Lookup(%q) failed: %v", name, err) + t.Fatalf("Lookup failed: %v", err) } if test.postModificationWantReload && dirent == newDirent { - t.Errorf("Lookup(%q) with cachePolicy=%s got old dirent %v, wanted a new dirent", name, test.cachePolicy, dirent) + t.Errorf("Lookup with cachePolicy=%s got old dirent, wanted a new dirent", test.cachePolicy) } if !test.postModificationWantReload && dirent != newDirent { - t.Errorf("Lookup(%q) with cachePolicy=%s got new dirent %v, wanted old dirent %v", name, test.cachePolicy, newDirent, dirent) + t.Errorf("Lookup with cachePolicy=%s got new dirent, wanted old dirent", test.cachePolicy) } uattrs, err := newDirent.Inode.UnstableAttr(ctx) if err != nil { @@ -276,660 +279,25 @@ func TestRevalidation(t *testing.T) { } gotModTimeSeconds := uattrs.ModificationTime.Seconds() if test.postModificationWantUpdatedAttrs && gotModTimeSeconds != nowSeconds { - t.Fatalf("Lookup(%q) with cachePolicy=%s got new modification time %v, wanted %v", name, test.cachePolicy, gotModTimeSeconds, nowSeconds) + t.Fatalf("Lookup with cachePolicy=%s got new modification time %v, wanted %v", test.cachePolicy, gotModTimeSeconds, nowSeconds) } + newDirent.DecRef() // See above. - // Make WalkGetAttr return ENOENT. This simulates - // removing the file from the remote fs. - rootFile.WalkGetAttrMock = p9test.WalkGetAttrMock{ - Err: syscall.ENOENT, - } + // Remove the file from the remote fs, subsequent walks + // should now fail to find anything. + rootFile.RemoveChild(file) // Walk again. Depending on the cache policy, we may // get ENOENT. - newDirent, err = rootDir.Walk(ctx, rootDir, name) + newDirent, err = rootDir.Walk(ctx, rootDir, file) if test.postRemovalWantReload && err == nil { - t.Errorf("Lookup(%q) with cachePolicy=%s got nil error, wanted ENOENT", name, test.cachePolicy) + t.Errorf("Lookup with cachePolicy=%s got nil error, wanted ENOENT", test.cachePolicy) } if !test.postRemovalWantReload && (err != nil || dirent != newDirent) { - t.Errorf("Lookup(%q) with cachePolicy=%s got new dirent %v and error %v, wanted old dirent %v and nil error", name, test.cachePolicy, newDirent, err, dirent) - } - }) - } -} - -func TestSetTimestamps(t *testing.T) { - // Test parameters. - type setTimestampsTest struct { - // Name of the test. - name string - - // Function input parameters. - ts fs.TimeSpec - } - - ctx := contexttest.Context(t) - now := ktime.NowFromContext(ctx) - tests := []setTimestampsTest{ - { - name: "mock SetAttr passes (function succeeds)", - ts: fs.TimeSpec{ - ATime: now, - MTime: now, - }, - }, - { - name: "mock SetAttr passes, times are 0 (function succeeds)", - ts: fs.TimeSpec{}, - }, - { - name: "mock SetAttr passes, times are 0 and not system time (function succeeds)", - ts: fs.TimeSpec{ - ATimeSetSystemTime: false, - MTimeSetSystemTime: false, - }, - }, - { - name: "mock SetAttr passes, times are set to system time (function succeeds)", - ts: fs.TimeSpec{ - ATimeSetSystemTime: true, - MTimeSetSystemTime: true, - }, - }, - { - name: "mock SetAttr passes, times are omitted (function succeeds)", - ts: fs.TimeSpec{ - ATimeOmit: true, - MTimeOmit: true, - }, - }, - } - - for _, test := range tests { - t.Run(test.name, func(t *testing.T) { - // Set up mock. - rootFile, rootInode, err := root(ctx, cacheNone, p9.PermissionsMask, 0) - if err != nil { - t.Fatalf("error creating root: %v", err) - } - - // Call function. - err = rootInode.SetTimestamps(ctx, nil /* Dirent */, test.ts) - - // Check return values. - if err != nil { - t.Errorf("SetTimestamps failed: got error %v, want nil", err) - } - - // Check mock parameters. - if !(test.ts.ATimeOmit && test.ts.MTimeOmit) && !rootFile.SetAttrMock.Called { - t.Errorf("TestSetTimestamps failed: SetAttr not called") - return - } - - // Check what was passed to the mock function. - attr := rootFile.SetAttrMock.Attr - atimeGiven := ktime.FromUnix(int64(attr.ATimeSeconds), int64(attr.ATimeNanoSeconds)) - if test.ts.ATimeOmit { - if rootFile.SetAttrMock.Valid.ATime { - t.Errorf("ATime got set true in mask, wanted false") - } - } else { - if got, want := rootFile.SetAttrMock.Valid.ATimeNotSystemTime, !test.ts.ATimeSetSystemTime; got != want { - t.Errorf("got ATimeNotSystemTime %v, want %v", got, want) - } - if !test.ts.ATimeSetSystemTime && !test.ts.ATime.Equal(atimeGiven) { - t.Errorf("ATime got %v, want %v", atimeGiven, test.ts.ATime) - } - } - - mtimeGiven := ktime.FromUnix(int64(attr.MTimeSeconds), int64(attr.MTimeNanoSeconds)) - if test.ts.MTimeOmit { - if rootFile.SetAttrMock.Valid.MTime { - t.Errorf("MTime got set true in mask, wanted false") - } - } else { - if got, want := rootFile.SetAttrMock.Valid.MTimeNotSystemTime, !test.ts.MTimeSetSystemTime; got != want { - t.Errorf("got MTimeNotSystemTime %v, want %v", got, want) - } - if !test.ts.MTimeSetSystemTime && !test.ts.MTime.Equal(mtimeGiven) { - t.Errorf("MTime got %v, want %v", mtimeGiven, test.ts.MTime) - } - } - }) - } -} - -func TestSetPermissions(t *testing.T) { - // Test parameters. - type setPermissionsTest struct { - // Name of the test. - name string - - // SetPermissions input parameters. - perms fs.FilePermissions - - // Error that SetAttr mock should return. - setAttrErr error - - // Expected return value. - want bool - } - - tests := []setPermissionsTest{ - { - name: "SetAttr mock succeeds (function succeeds)", - perms: fs.FilePermissions{User: fs.PermMask{Read: true, Write: true, Execute: true}}, - want: true, - setAttrErr: nil, - }, - { - name: "SetAttr mock fails (function fails)", - perms: fs.FilePermissions{User: fs.PermMask{Read: true, Write: true}}, - want: false, - setAttrErr: syscall.ENOENT, - }, - } - - ctx := contexttest.Context(t) - for _, test := range tests { - t.Run(test.name, func(t *testing.T) { - // Set up mock. - rootFile, rootInode, err := root(ctx, cacheNone, 0, 0) - if err != nil { - t.Fatalf("error creating root: %v", err) - } - rootFile.SetAttrMock.Err = test.setAttrErr - - ok := rootInode.SetPermissions(ctx, nil /* Dirent */, test.perms) - - // Check return value. - if ok != test.want { - t.Errorf("SetPermissions got %v, want %v", ok, test.want) - } - - // Check mock parameters. - pattr := rootFile.SetAttrMock.Attr - if !rootFile.SetAttrMock.Called { - t.Errorf("SetAttr not called") - return - } - if !rootFile.SetAttrMock.Valid.Permissions { - t.Errorf("SetAttr did not get right request (got false, expected SetAttrMask.Permissions true)") - } - if got := fs.FilePermsFromP9(pattr.Permissions); got != test.perms { - t.Errorf("SetAttr did not get right permissions -- got %v, want %v", got, test.perms) - } - }) - } -} - -func TestClose(t *testing.T) { - ctx := contexttest.Context(t) - // Set up mock. - rootFile, rootInode, err := root(ctx, cacheNone, p9.PermissionsMask, 0) - if err != nil { - t.Fatalf("error creating root: %v", err) - } - - // Call function. - rootInode.InodeOperations.Release(ctx) - - // Check mock parameters. - if !rootFile.CloseMock.Called { - t.Errorf("TestClose failed: Close not called") - } -} - -func TestRename(t *testing.T) { - // Test parameters. - type renameTest struct { - // Name of the test. - name string - - // Input parameters. - newParent *fs.Inode - newName string - - // Rename mock parameters. - renameErr error - renameCalled bool - - // Error want to return given the parameters. (Same as what - // we expect and tell rename to return.) - want error - } - ctx := contexttest.Context(t) - rootFile, rootInode, err := root(ctx, cacheNone, p9.PermissionsMask, 0) - if err != nil { - t.Fatalf("error creating root: %v", err) - } - - tests := []renameTest{ - { - name: "mock Rename succeeds (function succeeds)", - newParent: rootInode, - newName: "foo2", - want: nil, - renameErr: nil, - renameCalled: true, - }, - { - name: "mock Rename fails (function fails)", - newParent: rootInode, - newName: "foo2", - want: syscall.ENOENT, - renameErr: syscall.ENOENT, - renameCalled: true, - }, - { - name: "newParent is not inodeOperations but should be (function fails)", - newParent: fs.NewMockInode(ctx, fs.NewMockMountSource(nil), fs.StableAttr{Type: fs.Directory}), - newName: "foo2", - want: syscall.EXDEV, - renameErr: nil, - renameCalled: false, - }, - } - - for _, test := range tests { - t.Run(test.name, func(t *testing.T) { - mockFile := goodMockFile(p9.PermissionsMask, 0) - rootFile.WalkGetAttrMock.QIDs = []p9.QID{{}} - rootFile.WalkGetAttrMock.File = mockFile - - dirent, err := rootInode.Lookup(ctx, "foo") - if err != nil { - t.Fatalf("root.Walk failed: %v", err) - } - mockFile.RenameMock.Err = test.renameErr - mockFile.RenameMock.Called = false - - // Use a dummy oldParent to acquire write access to that directory. - oldParent := &inodeOperations{ - readdirCache: fs.NewSortedDentryMap(nil), - } - oldInode := fs.NewInode(oldParent, fs.NewMockMountSource(nil), fs.StableAttr{Type: fs.Directory}) - - // Call function. - err = dirent.Inode.InodeOperations.Rename(ctx, oldInode, "", test.newParent, test.newName) - - // Check return value. - if err != test.want { - t.Errorf("Rename got %v, want %v", err, test.want) - } - - // Check mock parameters. - if got, want := mockFile.RenameMock.Called, test.renameCalled; got != want { - t.Errorf("renameCalled got %v want %v", got, want) - } - }) - } -} - -// This file is read from in TestPreadv. -type readAtFileFake struct { - p9test.FileMock - - // Parameters for faking ReadAt. - FileLength int - Err error - ChunkSize int - Called bool - LengthRead int -} - -func (r *readAtFileFake) ReadAt(p []byte, offset uint64) (int, error) { - r.Called = true - log.Warningf("ReadAt fake: length read so far = %d, len(p) = %d, offset = %d", r.LengthRead, len(p), offset) - if int(offset) != r.LengthRead { - return 0, fmt.Errorf("offset got %d; expected %d", offset, r.LengthRead) - } - - if r.Err != nil { - return 0, r.Err - } - - if r.LengthRead >= r.FileLength { - return 0, io.EOF - } - - // Read at most ChunkSize and read at most what's left in the file. - toBeRead := len(p) - if r.LengthRead+toBeRead >= r.FileLength { - toBeRead = r.FileLength - int(offset) - } - if toBeRead > r.ChunkSize { - toBeRead = r.ChunkSize - } - - r.LengthRead += toBeRead - if r.LengthRead == r.FileLength { - return toBeRead, io.EOF - } - return toBeRead, nil -} - -func TestPreadv(t *testing.T) { - // Test parameters. - type preadvTest struct { - // Name of the test. - name string - - // Mock parameters - mode p9.FileMode - - // Buffer to read into. - buffer [512]byte - sliceSize int - - // How much readAt returns at a time. - chunkSize int - - // Whether or not we expect ReadAt to be called. - readAtCalled bool - readAtErr error - - // Expected return values. - want error - } - - tests := []preadvTest{ - { - name: "fake ReadAt succeeds, 512 bytes requested, 512 byte chunks (function succeeds)", - want: nil, - readAtErr: nil, - mode: p9.PermissionsMask, - readAtCalled: true, - sliceSize: 512, - chunkSize: 512, - }, - { - name: "fake ReadAt succeeds, 512 bytes requested, 200 byte chunks (function succeeds)", - want: nil, - readAtErr: nil, - mode: p9.PermissionsMask, - readAtCalled: true, - sliceSize: 512, - chunkSize: 200, - }, - { - name: "fake ReadAt succeeds, 0 bytes requested (function succeeds)", - want: nil, - readAtErr: nil, - mode: p9.PermissionsMask, - readAtCalled: false, - sliceSize: 0, - chunkSize: 100, - }, - { - name: "fake ReadAt returns 0 bytes and EOF (function fails)", - want: io.EOF, - readAtErr: io.EOF, - mode: p9.PermissionsMask, - readAtCalled: true, - sliceSize: 512, - chunkSize: 512, - }, - } - - ctx := contexttest.Context(t) - for _, test := range tests { - t.Run(test.name, func(t *testing.T) { - // Set up mock. - rootFile, rootInode, err := root(ctx, cacheNone, test.mode, 1024) - if err != nil { - t.Fatalf("error creating root: %v", err) - } - - // Set up the read buffer. - dst := usermem.BytesIOSequence(test.buffer[:test.sliceSize]) - - // This file will be read from. - openFile := &readAtFileFake{ - Err: test.readAtErr, - FileLength: test.sliceSize, - ChunkSize: test.chunkSize, - } - rootFile.WalkGetAttrMock.File = openFile - rootFile.WalkGetAttrMock.Attr.Mode = test.mode - rootFile.WalkGetAttrMock.Valid.Mode = true - - f := NewFile( - ctx, - fs.NewDirent(rootInode, ""), - "", - fs.FileFlags{Read: true}, - rootInode.InodeOperations.(*inodeOperations), - &handles{File: contextFile{file: openFile}}, - ) - - // Call function. - _, err = f.Preadv(ctx, dst, 0) - - // Check return value. - if err != test.want { - t.Errorf("Preadv got %v, want %v", err, test.want) - } - - // Check mock parameters. - if test.readAtCalled != openFile.Called { - t.Errorf("ReadAt called: %v, but expected opposite", openFile.Called) - } - }) - } -} - -func TestReadlink(t *testing.T) { - // Test parameters. - type readlinkTest struct { - // Name of the test. - name string - - // Mock parameters - mode p9.FileMode - - // Whether or not we expect ReadAt to be called and what error - // it shall return. - readlinkCalled bool - readlinkErr error - - // Expected return values. - want error - } - - tests := []readlinkTest{ - { - name: "file is not symlink (function fails)", - want: syscall.ENOLINK, - mode: p9.PermissionsMask, - readlinkCalled: false, - readlinkErr: nil, - }, - { - name: "mock Readlink succeeds (function succeeds)", - want: nil, - mode: p9.PermissionsMask | p9.ModeSymlink, - readlinkCalled: true, - readlinkErr: nil, - }, - { - name: "mock Readlink fails (function fails)", - want: syscall.ENOENT, - mode: p9.PermissionsMask | p9.ModeSymlink, - readlinkCalled: true, - readlinkErr: syscall.ENOENT, - }, - } - - ctx := contexttest.Context(t) - for _, test := range tests { - t.Run(test.name, func(t *testing.T) { - // Set up mock. - rootFile, rootInode, err := root(ctx, cacheNone, test.mode, 0) - if err != nil { - t.Fatalf("error creating root: %v", err) - } - - openFile := goodMockFile(test.mode, 0) - rootFile.WalkMock.File = openFile - rootFile.ReadlinkMock.Err = test.readlinkErr - - // Call function. - _, err = rootInode.Readlink(ctx) - - // Check return value. - if err != test.want { - t.Errorf("Readlink got %v, want %v", err, test.want) - } - - // Check mock parameters. - if test.readlinkCalled && !rootFile.ReadlinkMock.Called { - t.Errorf("Readlink not called") - } - }) - } -} - -// This file is write from in TestPwritev. -type writeAtFileFake struct { - p9test.FileMock - - // Parameters for faking WriteAt. - Err error - ChunkSize int - Called bool - LengthWritten int -} - -func (r *writeAtFileFake) WriteAt(p []byte, offset uint64) (int, error) { - r.Called = true - log.Warningf("WriteAt fake: length written so far = %d, len(p) = %d, offset = %d", r.LengthWritten, len(p), offset) - if int(offset) != r.LengthWritten { - return 0, fmt.Errorf("offset got %d; want %d", offset, r.LengthWritten) - } - - if r.Err != nil { - return 0, r.Err - } - - // Write at most ChunkSize. - toBeWritten := len(p) - if toBeWritten > r.ChunkSize { - toBeWritten = r.ChunkSize - } - r.LengthWritten += toBeWritten - return toBeWritten, nil -} - -func TestPwritev(t *testing.T) { - // Test parameters. - type pwritevTest struct { - // Name of the test. - name string - - // Mock parameters - mode p9.FileMode - - allowWrite bool - - // Buffer to write into. - buffer [512]byte - sliceSize int - chunkSize int - - // Whether or not we expect writeAt to be called. - writeAtCalled bool - writeAtErr error - - // Expected return values. - want error - } - - tests := []pwritevTest{ - { - name: "fake writeAt succeeds, one chunk (function succeeds)", - want: nil, - writeAtErr: nil, - mode: p9.PermissionsMask, - allowWrite: true, - writeAtCalled: true, - sliceSize: 512, - chunkSize: 512, - }, - { - name: "fake writeAt fails, short write (function fails)", - want: io.ErrShortWrite, - writeAtErr: nil, - mode: p9.PermissionsMask, - allowWrite: true, - writeAtCalled: true, - sliceSize: 512, - chunkSize: 200, - }, - { - name: "fake writeAt succeeds, len 0 (function succeeds)", - want: nil, - writeAtErr: nil, - mode: p9.PermissionsMask, - allowWrite: true, - writeAtCalled: false, - sliceSize: 0, - chunkSize: 0, - }, - { - name: "writeAt can still write despite file permissions read only (function succeeds)", - want: nil, - writeAtErr: nil, - mode: p9.PermissionsMask, - allowWrite: false, - writeAtCalled: true, - sliceSize: 512, - chunkSize: 512, - }, - } - - ctx := contexttest.Context(t) - for _, test := range tests { - t.Run(test.name, func(t *testing.T) { - // Set up mock. - _, rootInode, err := root(ctx, cacheNone, test.mode, 0) - if err != nil { - t.Fatalf("error creating root: %v", err) - } - - src := usermem.BytesIOSequence(test.buffer[:test.sliceSize]) - - // This is the file that will be used for writing. - openFile := &writeAtFileFake{ - Err: test.writeAtErr, - ChunkSize: test.chunkSize, - } - - f := NewFile( - ctx, - fs.NewDirent(rootInode, ""), - "", - fs.FileFlags{Write: true}, - rootInode.InodeOperations.(*inodeOperations), - &handles{File: contextFile{file: openFile}}, - ) - - // Call function. - _, err = f.Pwritev(ctx, src, 0) - - // Check return value. - if err != test.want { - t.Errorf("Pwritev got %v, want %v", err, test.want) - } - - // Check mock parameters. - if test.writeAtCalled != openFile.Called { - t.Errorf("WriteAt called: %v, but expected opposite", openFile.Called) - return + t.Errorf("Lookup with cachePolicy=%s got new dirent and error %v, wanted old dirent and nil error", test.cachePolicy, err) } - if openFile.Called && test.writeAtErr != nil && openFile.LengthWritten != test.sliceSize { - t.Errorf("wrote %d bytes, expected %d bytes written", openFile.LengthWritten, test.sliceSize) + if err == nil { + newDirent.DecRef() // See above. } }) } diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index 7552216f3..f76a83cd9 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -91,10 +91,6 @@ func (e *endpointMaps) get(key device.MultiDeviceKey) transport.BoundEndpoint { type session struct { refs.AtomicRefCount - // conn is a unet.Socket that wraps the readFD/writeFD mount option, - // see fs/gofer/fs.go. - conn *unet.Socket `state:"nosave"` - // msize is the value of the msize mount option, see fs/gofer/fs.go. msize uint32 `state:"wait"` @@ -142,7 +138,7 @@ type session struct { // Destroy tears down the session. func (s *session) Destroy() { - s.conn.Close() + s.client.Close() } // Revalidate implements MountSource.Revalidate. @@ -235,7 +231,6 @@ func Root(ctx context.Context, dev string, filesystem fs.Filesystem, superBlockF // Construct the session. s := &session{ connID: dev, - conn: conn, msize: o.msize, version: o.version, cachePolicy: o.policy, @@ -252,7 +247,7 @@ func Root(ctx context.Context, dev string, filesystem fs.Filesystem, superBlockF m := fs.NewMountSource(s, filesystem, superBlockFlags) // Send the Tversion request. - s.client, err = p9.NewClient(s.conn, s.msize, s.version) + s.client, err = p9.NewClient(conn, s.msize, s.version) if err != nil { // Drop our reference on the session, it needs to be torn down. s.DecRef() diff --git a/pkg/sentry/fs/gofer/session_state.go b/pkg/sentry/fs/gofer/session_state.go index f657135fc..d9fd7a221 100644 --- a/pkg/sentry/fs/gofer/session_state.go +++ b/pkg/sentry/fs/gofer/session_state.go @@ -84,13 +84,13 @@ func (s *session) afterLoad() { } // Manually restore the connection. - s.conn, err = unet.NewSocket(opts.fd) + conn, err := unet.NewSocket(opts.fd) if err != nil { panic(fmt.Sprintf("failed to create Socket for FD %d: %v", opts.fd, err)) } // Manually restore the client. - s.client, err = p9.NewClient(s.conn, s.msize, s.version) + s.client, err = p9.NewClient(conn, s.msize, s.version) if err != nil { panic(fmt.Sprintf("failed to connect client to server: %v", err)) } diff --git a/pkg/sentry/fs/proc/device/BUILD b/pkg/sentry/fs/proc/device/BUILD index 34582f275..ff7dacf07 100644 --- a/pkg/sentry/fs/proc/device/BUILD +++ b/pkg/sentry/fs/proc/device/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "device", srcs = ["device.go"], diff --git a/pkg/sentry/hostcpu/BUILD b/pkg/sentry/hostcpu/BUILD index f362d15c8..33197cf14 100644 --- a/pkg/sentry/hostcpu/BUILD +++ b/pkg/sentry/hostcpu/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "hostcpu", srcs = [ diff --git a/pkg/sentry/kernel/kdefs/BUILD b/pkg/sentry/kernel/kdefs/BUILD index fe6fa2260..3f8fa206c 100644 --- a/pkg/sentry/kernel/kdefs/BUILD +++ b/pkg/sentry/kernel/kdefs/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "kdefs", srcs = ["kdefs.go"], diff --git a/pkg/sentry/kernel/memevent/BUILD b/pkg/sentry/kernel/memevent/BUILD index 66899910c..e903badd3 100644 --- a/pkg/sentry/kernel/memevent/BUILD +++ b/pkg/sentry/kernel/memevent/BUILD @@ -1,8 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "memevent", srcs = ["memory_events.go"], diff --git a/pkg/sentry/kernel/sched/BUILD b/pkg/sentry/kernel/sched/BUILD index 125792f39..52e226a39 100644 --- a/pkg/sentry/kernel/sched/BUILD +++ b/pkg/sentry/kernel/sched/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "sched", srcs = [ diff --git a/pkg/sentry/loader/BUILD b/pkg/sentry/loader/BUILD index 0beb4561b..83cad186a 100644 --- a/pkg/sentry/loader/BUILD +++ b/pkg/sentry/loader/BUILD @@ -1,6 +1,7 @@ +load("@io_bazel_rules_go//go:def.bzl", "go_embed_data") + package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_embed_data") load("//tools/go_stateify:defs.bzl", "go_library") go_embed_data( diff --git a/pkg/sentry/memutil/BUILD b/pkg/sentry/memutil/BUILD index 341b30b98..88738d65d 100644 --- a/pkg/sentry/memutil/BUILD +++ b/pkg/sentry/memutil/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "memutil", srcs = [ diff --git a/pkg/sentry/platform/interrupt/BUILD b/pkg/sentry/platform/interrupt/BUILD index 35121321a..dbafa3204 100644 --- a/pkg/sentry/platform/interrupt/BUILD +++ b/pkg/sentry/platform/interrupt/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "interrupt", srcs = [ diff --git a/pkg/sentry/platform/kvm/BUILD b/pkg/sentry/platform/kvm/BUILD index 4ef9e20d7..1b71e629f 100644 --- a/pkg/sentry/platform/kvm/BUILD +++ b/pkg/sentry/platform/kvm/BUILD @@ -1,6 +1,7 @@ +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") + package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/sentry/platform/kvm/testutil/BUILD b/pkg/sentry/platform/kvm/testutil/BUILD index e779e3893..1dffe94a4 100644 --- a/pkg/sentry/platform/kvm/testutil/BUILD +++ b/pkg/sentry/platform/kvm/testutil/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "testutil", testonly = 1, diff --git a/pkg/sentry/platform/procid/BUILD b/pkg/sentry/platform/procid/BUILD index ba68d48f4..d3398d1e8 100644 --- a/pkg/sentry/platform/procid/BUILD +++ b/pkg/sentry/platform/procid/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "procid", srcs = [ diff --git a/pkg/sentry/platform/ptrace/BUILD b/pkg/sentry/platform/ptrace/BUILD index debae058b..2eb354ad4 100644 --- a/pkg/sentry/platform/ptrace/BUILD +++ b/pkg/sentry/platform/ptrace/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "ptrace", srcs = [ diff --git a/pkg/sentry/platform/ring0/BUILD b/pkg/sentry/platform/ring0/BUILD index 2485eb2eb..c35d49f2d 100644 --- a/pkg/sentry/platform/ring0/BUILD +++ b/pkg/sentry/platform/ring0/BUILD @@ -1,6 +1,7 @@ +load("//tools/go_stateify:defs.bzl", "go_library") + package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library") load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") go_template( diff --git a/pkg/sentry/platform/ring0/gen_offsets/BUILD b/pkg/sentry/platform/ring0/gen_offsets/BUILD index 3bce56985..b76d7974e 100644 --- a/pkg/sentry/platform/ring0/gen_offsets/BUILD +++ b/pkg/sentry/platform/ring0/gen_offsets/BUILD @@ -1,6 +1,7 @@ +load("@io_bazel_rules_go//go:def.bzl", "go_binary") + package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_binary") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/sentry/platform/ring0/pagetables/BUILD b/pkg/sentry/platform/ring0/pagetables/BUILD index 7a86e2234..de1b920af 100644 --- a/pkg/sentry/platform/ring0/pagetables/BUILD +++ b/pkg/sentry/platform/ring0/pagetables/BUILD @@ -1,6 +1,7 @@ +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") + package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") go_template( diff --git a/pkg/sentry/platform/safecopy/BUILD b/pkg/sentry/platform/safecopy/BUILD index 7dcf6e561..614d9e21e 100644 --- a/pkg/sentry/platform/safecopy/BUILD +++ b/pkg/sentry/platform/safecopy/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "safecopy", srcs = [ diff --git a/pkg/sentry/safemem/BUILD b/pkg/sentry/safemem/BUILD index e96509ce1..87a9bff12 100644 --- a/pkg/sentry/safemem/BUILD +++ b/pkg/sentry/safemem/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "safemem", srcs = [ diff --git a/pkg/sentry/sighandling/BUILD b/pkg/sentry/sighandling/BUILD index 751176747..41313d334 100644 --- a/pkg/sentry/sighandling/BUILD +++ b/pkg/sentry/sighandling/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "sighandling", srcs = [ diff --git a/pkg/sentry/socket/rpcinet/BUILD b/pkg/sentry/socket/rpcinet/BUILD index 38fa54283..06e121946 100644 --- a/pkg/sentry/socket/rpcinet/BUILD +++ b/pkg/sentry/socket/rpcinet/BUILD @@ -1,8 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "rpcinet", srcs = [ diff --git a/pkg/sentry/socket/rpcinet/conn/BUILD b/pkg/sentry/socket/rpcinet/conn/BUILD index c51ca14b1..a16977f29 100644 --- a/pkg/sentry/socket/rpcinet/conn/BUILD +++ b/pkg/sentry/socket/rpcinet/conn/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # BSD - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # BSD + go_library( name = "conn", srcs = ["conn.go"], diff --git a/pkg/sentry/socket/rpcinet/notifier/BUILD b/pkg/sentry/socket/rpcinet/notifier/BUILD index 2ae902b3f..2bab01774 100644 --- a/pkg/sentry/socket/rpcinet/notifier/BUILD +++ b/pkg/sentry/socket/rpcinet/notifier/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # BSD - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # BSD + go_library( name = "notifier", srcs = ["notifier.go"], diff --git a/pkg/sentry/state/BUILD b/pkg/sentry/state/BUILD index a57a8298e..f1f6fdb7d 100644 --- a/pkg/sentry/state/BUILD +++ b/pkg/sentry/state/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "state", srcs = [ diff --git a/pkg/sentry/strace/BUILD b/pkg/sentry/strace/BUILD index 674554081..52c7f325c 100644 --- a/pkg/sentry/strace/BUILD +++ b/pkg/sentry/strace/BUILD @@ -1,8 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "strace", srcs = [ diff --git a/pkg/sentry/syscalls/BUILD b/pkg/sentry/syscalls/BUILD index 2a9f0915e..35192ff49 100644 --- a/pkg/sentry/syscalls/BUILD +++ b/pkg/sentry/syscalls/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "syscalls", srcs = [ diff --git a/pkg/sentry/time/BUILD b/pkg/sentry/time/BUILD index 9452787fb..5dadb8a2d 100644 --- a/pkg/sentry/time/BUILD +++ b/pkg/sentry/time/BUILD @@ -1,6 +1,7 @@ +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") + package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/sentry/unimpl/BUILD b/pkg/sentry/unimpl/BUILD index 63da5e81f..42e24ace5 100644 --- a/pkg/sentry/unimpl/BUILD +++ b/pkg/sentry/unimpl/BUILD @@ -1,8 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") +package(licenses = ["notice"]) # Apache 2.0 + proto_library( name = "unimplemented_syscall_proto", srcs = ["unimplemented_syscall.proto"], diff --git a/pkg/sentry/uniqueid/BUILD b/pkg/sentry/uniqueid/BUILD index 68b82af47..0929497c3 100644 --- a/pkg/sentry/uniqueid/BUILD +++ b/pkg/sentry/uniqueid/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "uniqueid", srcs = ["context.go"], diff --git a/pkg/sentry/watchdog/BUILD b/pkg/sentry/watchdog/BUILD index 13bc33eb1..b2c687b20 100644 --- a/pkg/sentry/watchdog/BUILD +++ b/pkg/sentry/watchdog/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "watchdog", srcs = ["watchdog.go"], diff --git a/pkg/sleep/BUILD b/pkg/sleep/BUILD index 05e4ca540..338fd9336 100644 --- a/pkg/sleep/BUILD +++ b/pkg/sleep/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "sleep", srcs = [ diff --git a/pkg/state/BUILD b/pkg/state/BUILD index 6a5b2d4ff..dd0f250fa 100644 --- a/pkg/state/BUILD +++ b/pkg/state/BUILD @@ -1,7 +1,8 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") + +package(licenses = ["notice"]) # Apache 2.0 + load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/state/statefile/BUILD b/pkg/state/statefile/BUILD index 6be78dc9b..66c8f3807 100644 --- a/pkg/state/statefile/BUILD +++ b/pkg/state/statefile/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "statefile", srcs = ["statefile.go"], diff --git a/pkg/sync/atomicptrtest/BUILD b/pkg/sync/atomicptrtest/BUILD index 4fa959df0..9cb7f66fe 100644 --- a/pkg/sync/atomicptrtest/BUILD +++ b/pkg/sync/atomicptrtest/BUILD @@ -1,6 +1,7 @@ +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") + package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/sync/seqatomictest/BUILD b/pkg/sync/seqatomictest/BUILD index 07b4f85ab..54f8e59b1 100644 --- a/pkg/sync/seqatomictest/BUILD +++ b/pkg/sync/seqatomictest/BUILD @@ -1,6 +1,7 @@ +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") + package(licenses = ["notice"]) # Apache 2.0 -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") go_template_instance( diff --git a/pkg/syserr/BUILD b/pkg/syserr/BUILD index 5dd2e90bb..30ae20772 100644 --- a/pkg/syserr/BUILD +++ b/pkg/syserr/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "syserr", srcs = [ diff --git a/pkg/syserror/BUILD b/pkg/syserror/BUILD index e050c2043..d4c6da97a 100644 --- a/pkg/syserror/BUILD +++ b/pkg/syserror/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "syserror", srcs = ["syserror.go"], diff --git a/pkg/tcpip/adapters/gonet/BUILD b/pkg/tcpip/adapters/gonet/BUILD index bf618831a..723ad668f 100644 --- a/pkg/tcpip/adapters/gonet/BUILD +++ b/pkg/tcpip/adapters/gonet/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "gonet", srcs = ["gonet.go"], diff --git a/pkg/tcpip/checker/BUILD b/pkg/tcpip/checker/BUILD index e8a524918..a1de808b9 100644 --- a/pkg/tcpip/checker/BUILD +++ b/pkg/tcpip/checker/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "checker", testonly = 1, diff --git a/pkg/tcpip/link/channel/BUILD b/pkg/tcpip/link/channel/BUILD index 9a6f49c45..25f6c1457 100644 --- a/pkg/tcpip/link/channel/BUILD +++ b/pkg/tcpip/link/channel/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "channel", srcs = ["channel.go"], diff --git a/pkg/tcpip/link/fdbased/BUILD b/pkg/tcpip/link/fdbased/BUILD index 6e75e9f47..94391433c 100644 --- a/pkg/tcpip/link/fdbased/BUILD +++ b/pkg/tcpip/link/fdbased/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "fdbased", srcs = ["endpoint.go"], diff --git a/pkg/tcpip/link/loopback/BUILD b/pkg/tcpip/link/loopback/BUILD index cc4247ffd..a46ba7f11 100644 --- a/pkg/tcpip/link/loopback/BUILD +++ b/pkg/tcpip/link/loopback/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "loopback", srcs = ["loopback.go"], diff --git a/pkg/tcpip/link/rawfile/BUILD b/pkg/tcpip/link/rawfile/BUILD index 10b35a37e..829ea7c42 100644 --- a/pkg/tcpip/link/rawfile/BUILD +++ b/pkg/tcpip/link/rawfile/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "rawfile", srcs = [ diff --git a/pkg/tcpip/link/sharedmem/BUILD b/pkg/tcpip/link/sharedmem/BUILD index 5390257c5..d7f1e66ef 100644 --- a/pkg/tcpip/link/sharedmem/BUILD +++ b/pkg/tcpip/link/sharedmem/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "sharedmem", srcs = [ diff --git a/pkg/tcpip/link/sharedmem/pipe/BUILD b/pkg/tcpip/link/sharedmem/pipe/BUILD index ff798ae6f..12e813509 100644 --- a/pkg/tcpip/link/sharedmem/pipe/BUILD +++ b/pkg/tcpip/link/sharedmem/pipe/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "pipe", srcs = [ diff --git a/pkg/tcpip/link/sharedmem/queue/BUILD b/pkg/tcpip/link/sharedmem/queue/BUILD index c4a7879c4..661037bb2 100644 --- a/pkg/tcpip/link/sharedmem/queue/BUILD +++ b/pkg/tcpip/link/sharedmem/queue/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "queue", srcs = [ diff --git a/pkg/tcpip/link/sniffer/BUILD b/pkg/tcpip/link/sniffer/BUILD index 7155aea66..52e237c25 100644 --- a/pkg/tcpip/link/sniffer/BUILD +++ b/pkg/tcpip/link/sniffer/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "sniffer", srcs = [ diff --git a/pkg/tcpip/link/tun/BUILD b/pkg/tcpip/link/tun/BUILD index a8bb03661..5ec01cec9 100644 --- a/pkg/tcpip/link/tun/BUILD +++ b/pkg/tcpip/link/tun/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "tun", srcs = ["tun_unsafe.go"], diff --git a/pkg/tcpip/link/waitable/BUILD b/pkg/tcpip/link/waitable/BUILD index 7582df32e..ba495c437 100644 --- a/pkg/tcpip/link/waitable/BUILD +++ b/pkg/tcpip/link/waitable/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "waitable", srcs = [ diff --git a/pkg/tcpip/network/BUILD b/pkg/tcpip/network/BUILD index 25a3c98b6..a2a07f533 100644 --- a/pkg/tcpip/network/BUILD +++ b/pkg/tcpip/network/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_test( name = "ip_test", size = "small", diff --git a/pkg/tcpip/network/arp/BUILD b/pkg/tcpip/network/arp/BUILD index 44f2b66e5..f6fb7daf7 100644 --- a/pkg/tcpip/network/arp/BUILD +++ b/pkg/tcpip/network/arp/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "arp", srcs = ["arp.go"], diff --git a/pkg/tcpip/network/hash/BUILD b/pkg/tcpip/network/hash/BUILD index 1c22c52fc..401dce646 100644 --- a/pkg/tcpip/network/hash/BUILD +++ b/pkg/tcpip/network/hash/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "hash", srcs = ["hash.go"], diff --git a/pkg/tcpip/network/ipv4/BUILD b/pkg/tcpip/network/ipv4/BUILD index 90d65d531..e72317e9f 100644 --- a/pkg/tcpip/network/ipv4/BUILD +++ b/pkg/tcpip/network/ipv4/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "ipv4", srcs = [ diff --git a/pkg/tcpip/network/ipv6/BUILD b/pkg/tcpip/network/ipv6/BUILD index 2f19a659e..808c37df3 100644 --- a/pkg/tcpip/network/ipv6/BUILD +++ b/pkg/tcpip/network/ipv6/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "ipv6", srcs = [ diff --git a/pkg/tcpip/ports/BUILD b/pkg/tcpip/ports/BUILD index 3c3374275..c69fc0744 100644 --- a/pkg/tcpip/ports/BUILD +++ b/pkg/tcpip/ports/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "ports", srcs = ["ports.go"], diff --git a/pkg/tcpip/sample/tun_tcp_connect/BUILD b/pkg/tcpip/sample/tun_tcp_connect/BUILD index 21d32245d..32baf2115 100644 --- a/pkg/tcpip/sample/tun_tcp_connect/BUILD +++ b/pkg/tcpip/sample/tun_tcp_connect/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_binary") +package(licenses = ["notice"]) # Apache 2.0 + go_binary( name = "tun_tcp_connect", srcs = ["main.go"], diff --git a/pkg/tcpip/sample/tun_tcp_echo/BUILD b/pkg/tcpip/sample/tun_tcp_echo/BUILD index d7402aaa2..760445843 100644 --- a/pkg/tcpip/sample/tun_tcp_echo/BUILD +++ b/pkg/tcpip/sample/tun_tcp_echo/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_binary") +package(licenses = ["notice"]) # Apache 2.0 + go_binary( name = "tun_tcp_echo", srcs = ["main.go"], diff --git a/pkg/tcpip/transport/tcp/testing/context/BUILD b/pkg/tcpip/transport/tcp/testing/context/BUILD index 7a95594ef..814e5c1ea 100644 --- a/pkg/tcpip/transport/tcp/testing/context/BUILD +++ b/pkg/tcpip/transport/tcp/testing/context/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "context", testonly = 1, diff --git a/pkg/tcpip/transport/tcpconntrack/BUILD b/pkg/tcpip/transport/tcpconntrack/BUILD index 46da3e6f1..ac1a94d4d 100644 --- a/pkg/tcpip/transport/tcpconntrack/BUILD +++ b/pkg/tcpip/transport/tcpconntrack/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "tcpconntrack", srcs = ["tcp_conntrack.go"], diff --git a/pkg/tmutex/BUILD b/pkg/tmutex/BUILD index d18338fff..c20df7005 100644 --- a/pkg/tmutex/BUILD +++ b/pkg/tmutex/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "tmutex", srcs = ["tmutex.go"], diff --git a/pkg/unet/BUILD b/pkg/unet/BUILD index acdfd7cb6..f90e43c89 100644 --- a/pkg/unet/BUILD +++ b/pkg/unet/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "unet", srcs = [ diff --git a/pkg/urpc/BUILD b/pkg/urpc/BUILD index d32c57d1a..21008cf6c 100644 --- a/pkg/urpc/BUILD +++ b/pkg/urpc/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "urpc", srcs = ["urpc.go"], diff --git a/pkg/waiter/fdnotifier/BUILD b/pkg/waiter/fdnotifier/BUILD index 4e582755d..af6baa303 100644 --- a/pkg/waiter/fdnotifier/BUILD +++ b/pkg/waiter/fdnotifier/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("//tools/go_stateify:defs.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "fdnotifier", srcs = [ diff --git a/runsc/boot/BUILD b/runsc/boot/BUILD index 04cc0e854..07afce807 100644 --- a/runsc/boot/BUILD +++ b/runsc/boot/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "boot", srcs = [ diff --git a/runsc/boot/filter/BUILD b/runsc/boot/filter/BUILD index 48f2c8024..004222242 100644 --- a/runsc/boot/filter/BUILD +++ b/runsc/boot/filter/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "filter", srcs = [ diff --git a/runsc/cgroup/BUILD b/runsc/cgroup/BUILD index 10a8e5feb..bf2f373a9 100644 --- a/runsc/cgroup/BUILD +++ b/runsc/cgroup/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "cgroup", srcs = ["cgroup.go"], diff --git a/runsc/cmd/BUILD b/runsc/cmd/BUILD index 7040eb4ec..394bb0e1f 100644 --- a/runsc/cmd/BUILD +++ b/runsc/cmd/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "cmd", srcs = [ diff --git a/runsc/console/BUILD b/runsc/console/BUILD index fa1a7d430..ff4ccff69 100644 --- a/runsc/console/BUILD +++ b/runsc/console/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "console", srcs = ["console.go"], diff --git a/runsc/container/BUILD b/runsc/container/BUILD index f4c6f1525..bdd93aaba 100644 --- a/runsc/container/BUILD +++ b/runsc/container/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_binary", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "container", srcs = [ diff --git a/runsc/fsgofer/BUILD b/runsc/fsgofer/BUILD index 24e172f48..f28e4fa77 100644 --- a/runsc/fsgofer/BUILD +++ b/runsc/fsgofer/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "fsgofer", srcs = [ diff --git a/runsc/fsgofer/filter/BUILD b/runsc/fsgofer/filter/BUILD index 40f4f2205..c7848d10c 100644 --- a/runsc/fsgofer/filter/BUILD +++ b/runsc/fsgofer/filter/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "filter", srcs = [ diff --git a/runsc/fsgofer/fsgofer.go b/runsc/fsgofer/fsgofer.go index e03bb7752..fd913831a 100644 --- a/runsc/fsgofer/fsgofer.go +++ b/runsc/fsgofer/fsgofer.go @@ -26,7 +26,6 @@ import ( "math" "os" "path" - "strings" "sync" "syscall" @@ -181,18 +180,6 @@ func (a *attachPoint) makeQID(stat syscall.Stat_t) p9.QID { } } -func isNameValid(name string) bool { - if name == "" || name == "." || name == ".." { - log.Warningf("Invalid name: %s", name) - return false - } - if strings.IndexByte(name, '/') >= 0 { - log.Warningf("Invalid name: %s", name) - return false - } - return true -} - // localFile implements p9.File wrapping a local file. The underlying file // is opened during Walk() and stored in 'controlFile' to be used with other // operations. The mode in which the file is opened varies depending on the @@ -228,11 +215,7 @@ type localFile struct { // attachPoint is the attachPoint that serves this localFile. attachPoint *attachPoint - // mu protects 'hostPath' when file is renamed. - mu sync.Mutex - - // TODO: hostPath is not safe to use as path needs to be walked - // everytime (and can change underneath us). Remove all usages. + // hostPath will be safely updated by the Renamed hook. hostPath string // controlFile is opened when localFile is created and it's never nil. @@ -246,6 +229,7 @@ type localFile struct { // if localFile isn't opened. mode p9.OpenFlags + // ft is the fileType for this file. ft fileType // readDirMu protects against concurrent Readdir calls. @@ -296,10 +280,7 @@ func openAnyFile(parent *localFile, name string) (*os.File, string, error) { return nil, "", extractErrno(err) } - parent.mu.Lock() - defer parent.mu.Unlock() newPath := path.Join(parent.hostPath, name) - return os.NewFile(uintptr(fd), newPath), newPath, nil } @@ -382,13 +363,10 @@ func (l *localFile) Open(mode p9.OpenFlags) (*fd.FD, p9.QID, uint32, error) { log.Debugf("Open reopening file, mode: %v, %q", mode, l.controlFile.Name()) var err error - l.mu.Lock() newFile, err = os.OpenFile(l.hostPath, openFlags|mode.OSFlags(), 0) if err != nil { - l.mu.Unlock() return nil, p9.QID{}, 0, extractErrno(err) } - l.mu.Unlock() } stat, err := stat(int(newFile.Fd())) @@ -418,9 +396,6 @@ func (l *localFile) Create(name string, mode p9.OpenFlags, perm p9.FileMode, uid } return nil, nil, p9.QID{}, 0, syscall.EBADF } - if !isNameValid(name) { - return nil, nil, p9.QID{}, 0, syscall.EINVAL - } // Use a single file for both 'controlFile' and 'openedFile'. Mode must include read for control // and whichever else was requested by caller. Note that resulting file might have a wider mode @@ -452,9 +427,6 @@ func (l *localFile) Create(name string, mode p9.OpenFlags, perm p9.FileMode, uid return nil, nil, p9.QID{}, 0, extractErrno(err) } - l.mu.Lock() - defer l.mu.Unlock() - cPath := path.Join(l.hostPath, name) f := os.NewFile(uintptr(fd), cPath) c := &localFile{ @@ -477,10 +449,6 @@ func (l *localFile) Mkdir(name string, perm p9.FileMode, uid p9.UID, gid p9.GID) return p9.QID{}, syscall.EBADF } - if !isNameValid(name) { - return p9.QID{}, syscall.EINVAL - } - if err := syscall.Mkdirat(l.controlFD(), name, uint32(perm.Permissions())); err != nil { return p9.QID{}, extractErrno(err) } @@ -517,9 +485,6 @@ func (l *localFile) Walk(names []string) ([]p9.QID, p9.File, error) { return nil, nil, extractErrno(err) } - l.mu.Lock() - defer l.mu.Unlock() - c := &localFile{ attachPoint: l.attachPoint, hostPath: l.hostPath, @@ -532,10 +497,6 @@ func (l *localFile) Walk(names []string) ([]p9.QID, p9.File, error) { var qids []p9.QID last := l for _, name := range names { - if !isNameValid(name) { - return nil, nil, syscall.EINVAL - } - f, path, err := openAnyFile(last, name) if err != nil { return nil, nil, extractErrno(err) @@ -761,15 +722,15 @@ func (l *localFile) SetAttr(valid p9.SetAttrMask, attr p9.SetAttr) error { return err } -// Remove implements p9.File. -// -// This is deprecated in favor of UnlinkAt. -func (*localFile) Remove() error { - return syscall.ENOSYS +// Rename implements p9.File; this should never be called. +func (l *localFile) Rename(p9.File, string) error { + panic("rename called directly") } -// Rename implements p9.File. -func (l *localFile) Rename(directory p9.File, name string) error { +// RenameAt implements p9.File.RenameAt. +// +// TODO: change to renameat(2). +func (l *localFile) RenameAt(oldName string, directory p9.File, newName string) error { conf := l.attachPoint.conf if conf.ROMount { if conf.PanicOnWrite { @@ -777,34 +738,16 @@ func (l *localFile) Rename(directory p9.File, name string) error { } return syscall.EBADF } - if !isNameValid(name) { - return syscall.EINVAL - } - - l.mu.Lock() - defer l.mu.Unlock() - // TODO: change to renameat(2) - parent := directory.(*localFile) - newPath := path.Join(parent.hostPath, name) - if err := syscall.Rename(l.hostPath, newPath); err != nil { + newParent := directory.(*localFile) + oldPath := path.Join(l.hostPath, oldName) + newPath := path.Join(newParent.hostPath, newName) + if err := syscall.Rename(oldPath, newPath); err != nil { return extractErrno(err) } - - // Update path on success. - // TODO: this doesn't cover cases where any of the - // parents have been renamed. - l.hostPath = newPath return nil } -// RenameAt implements p9.File.RenameAt. -// -// Code still uses [deprecated] Rename(). -func (*localFile) RenameAt(_ string, _ p9.File, _ string) error { - return syscall.ENOSYS -} - // ReadAt implements p9.File. func (l *localFile) ReadAt(p []byte, offset uint64) (int, error) { if l.mode != p9.ReadOnly && l.mode != p9.ReadWrite { @@ -848,9 +791,6 @@ func (l *localFile) Symlink(target, newName string, uid p9.UID, gid p9.GID) (p9. } return p9.QID{}, syscall.EBADF } - if !isNameValid(newName) { - return p9.QID{}, syscall.EINVAL - } if err := unix.Symlinkat(target, l.controlFD(), newName); err != nil { return p9.QID{}, extractErrno(err) @@ -882,9 +822,6 @@ func (l *localFile) Link(target p9.File, newName string) error { } return syscall.EBADF } - if !isNameValid(newName) { - return syscall.EINVAL - } targetFile := target.(*localFile) if err := unix.Linkat(targetFile.controlFD(), "", l.controlFD(), newName, linux.AT_EMPTY_PATH); err != nil { @@ -909,9 +846,7 @@ func (l *localFile) UnlinkAt(name string, flags uint32) error { } return syscall.EBADF } - if !isNameValid(name) { - return syscall.EINVAL - } + if err := unix.Unlinkat(l.controlFD(), name, int(flags)); err != nil { return extractErrno(err) } @@ -1000,6 +935,11 @@ func (l *localFile) Close() error { return err } +// Renamed implements p9.Renamed. +func (l *localFile) Renamed(newDir p9.File, newName string) { + l.hostPath = path.Join(newDir.(*localFile).hostPath, newName) +} + // extractErrno tries to determine the errno. func extractErrno(err error) syscall.Errno { if err == nil { diff --git a/runsc/fsgofer/fsgofer_test.go b/runsc/fsgofer/fsgofer_test.go index 48860f952..34033245b 100644 --- a/runsc/fsgofer/fsgofer_test.go +++ b/runsc/fsgofer/fsgofer_test.go @@ -415,22 +415,22 @@ func TestLink(t *testing.T) { func TestROMountChecks(t *testing.T) { runCustom(t, allTypes, roConfs, func(t *testing.T, s state) { - if _, _, _, _, err := s.file.Create("..", p9.ReadWrite, 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EBADF { + if _, _, _, _, err := s.file.Create("some_file", p9.ReadWrite, 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EBADF { t.Errorf("%v: Create() should have failed, got: %v, expected: syscall.EBADF", s, err) } - if _, err := s.file.Mkdir("..", 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EBADF { + if _, err := s.file.Mkdir("some_dir", 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EBADF { t.Errorf("%v: MkDir() should have failed, got: %v, expected: syscall.EBADF", s, err) } - if err := s.file.Rename(s.file, ".."); err != syscall.EBADF { + if err := s.file.RenameAt("some_file", s.file, "other_file"); err != syscall.EBADF { t.Errorf("%v: Rename() should have failed, got: %v, expected: syscall.EBADF", s, err) } - if _, err := s.file.Symlink("some_place", "..", p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EBADF { + if _, err := s.file.Symlink("some_place", "some_symlink", p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EBADF { t.Errorf("%v: Symlink() should have failed, got: %v, expected: syscall.EBADF", s, err) } - if err := s.file.UnlinkAt("..", 0); err != syscall.EBADF { + if err := s.file.UnlinkAt("some_file", 0); err != syscall.EBADF { t.Errorf("%v: UnlinkAt() should have failed, got: %v, expected: syscall.EBADF", s, err) } - if err := s.file.Link(s.file, ".."); err != syscall.EBADF { + if err := s.file.Link(s.file, "some_link"); err != syscall.EBADF { t.Errorf("%v: Link() should have failed, got: %v, expected: syscall.EBADF", s, err) } @@ -445,12 +445,12 @@ func TestROMountChecks(t *testing.T) { func TestROMountPanics(t *testing.T) { conf := Config{ROMount: true, PanicOnWrite: true} runCustom(t, allTypes, []Config{conf}, func(t *testing.T, s state) { - assertPanic(t, func() { s.file.Create("..", p9.ReadWrite, 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())) }) - assertPanic(t, func() { s.file.Mkdir("..", 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())) }) - assertPanic(t, func() { s.file.Rename(s.file, "..") }) - assertPanic(t, func() { s.file.Symlink("some_place", "..", p9.UID(os.Getuid()), p9.GID(os.Getgid())) }) - assertPanic(t, func() { s.file.UnlinkAt("..", 0) }) - assertPanic(t, func() { s.file.Link(s.file, "..") }) + assertPanic(t, func() { s.file.Create("some_file", p9.ReadWrite, 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())) }) + assertPanic(t, func() { s.file.Mkdir("some_dir", 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())) }) + assertPanic(t, func() { s.file.RenameAt("some_file", s.file, "other_file") }) + assertPanic(t, func() { s.file.Symlink("some_place", "some_symlink", p9.UID(os.Getuid()), p9.GID(os.Getgid())) }) + assertPanic(t, func() { s.file.UnlinkAt("some_file", 0) }) + assertPanic(t, func() { s.file.Link(s.file, "some_link") }) valid := p9.SetAttrMask{Size: true} attr := p9.SetAttr{Size: 0} @@ -458,60 +458,6 @@ func TestROMountPanics(t *testing.T) { }) } -func TestInvalidName(t *testing.T) { - runCustom(t, []fileType{regular}, rwConfs, func(t *testing.T, s state) { - if _, _, _, _, err := s.file.Create("..", p9.ReadWrite, 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EINVAL { - t.Errorf("%v: Create() should have failed, got: %v, expected: syscall.EINVAL", s, err) - } - if _, _, err := s.file.Walk([]string{".."}); err != syscall.EINVAL { - t.Errorf("%v: Walk() should have failed, got: %v, expected: syscall.EINVAL", s, err) - } - if _, err := s.file.Mkdir("..", 0777, p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EINVAL { - t.Errorf("%v: MkDir() should have failed, got: %v, expected: syscall.EINVAL", s, err) - } - if err := s.file.Rename(s.file, ".."); err != syscall.EINVAL { - t.Errorf("%v: Rename() should have failed, got: %v, expected: syscall.EINVAL", s, err) - } - if _, err := s.file.Symlink("some_place", "..", p9.UID(os.Getuid()), p9.GID(os.Getgid())); err != syscall.EINVAL { - t.Errorf("%v: Symlink() should have failed, got: %v, expected: syscall.EINVAL", s, err) - } - if err := s.file.UnlinkAt("..", 0); err != syscall.EINVAL { - t.Errorf("%v: UnlinkAt() should have failed, got: %v, expected: syscall.EINVAL", s, err) - } - if err := s.file.Link(s.file, ".."); err != syscall.EINVAL { - t.Errorf("%v: Link() should have failed, got: %v, expected: syscall.EINVAL", s, err) - } - }) -} - -func TestIsNameValid(t *testing.T) { - valid := []string{ - "name", - "123", - "!@#$%^&*()", - ".name", - "..name", - "...", - } - for _, s := range valid { - if got := isNameValid(s); !got { - t.Errorf("isNameValid(%s) failed, got: %v, expected: true", s, got) - } - } - invalid := []string{ - ".", - "..", - "name/name", - "/name", - "name/", - } - for _, s := range invalid { - if got := isNameValid(s); got { - t.Errorf("isNameValid(%s) failed, got: %v, expected: false", s, got) - } - } -} - func TestWalkNotFound(t *testing.T) { runCustom(t, []fileType{directory}, allConfs, func(t *testing.T, s state) { if _, _, err := s.file.Walk([]string{"nobody-here"}); err != syscall.ENOENT { diff --git a/runsc/sandbox/BUILD b/runsc/sandbox/BUILD index eb9c4cd76..d6043bcf7 100644 --- a/runsc/sandbox/BUILD +++ b/runsc/sandbox/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "sandbox", srcs = [ diff --git a/runsc/specutils/BUILD b/runsc/specutils/BUILD index e73b2293f..a1e5da3f5 100644 --- a/runsc/specutils/BUILD +++ b/runsc/specutils/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "specutils", srcs = [ diff --git a/runsc/test/image/BUILD b/runsc/test/image/BUILD index c41161d50..22b3ebd2a 100644 --- a/runsc/test/image/BUILD +++ b/runsc/test/image/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_test( name = "image_test", size = "large", diff --git a/runsc/test/integration/BUILD b/runsc/test/integration/BUILD index 726ebf49e..e7204dc66 100644 --- a/runsc/test/integration/BUILD +++ b/runsc/test/integration/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_test( name = "integration_test", size = "large", diff --git a/runsc/test/root/BUILD b/runsc/test/root/BUILD index c69249b52..c2567ef23 100644 --- a/runsc/test/root/BUILD +++ b/runsc/test/root/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "root", srcs = ["root.go"], diff --git a/runsc/test/testutil/BUILD b/runsc/test/testutil/BUILD index da2535bfa..128bd80fb 100644 --- a/runsc/test/testutil/BUILD +++ b/runsc/test/testutil/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "testutil", srcs = [ diff --git a/runsc/tools/dockercfg/BUILD b/runsc/tools/dockercfg/BUILD index 5abb0c90a..a80b3abab 100644 --- a/runsc/tools/dockercfg/BUILD +++ b/runsc/tools/dockercfg/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_binary") +package(licenses = ["notice"]) # Apache 2.0 + go_binary( name = "dockercfg", srcs = ["dockercfg.go"], diff --git a/tools/go_generics/BUILD b/tools/go_generics/BUILD index 1afc58625..22c2e62c3 100644 --- a/tools/go_generics/BUILD +++ b/tools/go_generics/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_binary") +package(licenses = ["notice"]) # Apache 2.0 + go_binary( name = "go_generics", srcs = [ diff --git a/tools/go_generics/globals/BUILD b/tools/go_generics/globals/BUILD index a238becab..c26ac56d2 100644 --- a/tools/go_generics/globals/BUILD +++ b/tools/go_generics/globals/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_library") +package(licenses = ["notice"]) # Apache 2.0 + go_library( name = "globals", srcs = [ diff --git a/tools/go_generics/rules_tests/BUILD b/tools/go_generics/rules_tests/BUILD index 2d9a6fa9d..23b2d656d 100644 --- a/tools/go_generics/rules_tests/BUILD +++ b/tools/go_generics/rules_tests/BUILD @@ -1,6 +1,7 @@ +load("@io_bazel_rules_go//go:def.bzl", "go_test") + package(licenses = ["notice"]) # Apache 2.0 -load("@io_bazel_rules_go//go:def.bzl", "go_test") load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") go_template_instance( diff --git a/tools/go_stateify/BUILD b/tools/go_stateify/BUILD index edbeb4e2d..68d37f5d7 100644 --- a/tools/go_stateify/BUILD +++ b/tools/go_stateify/BUILD @@ -1,7 +1,7 @@ -package(licenses = ["notice"]) # Apache 2.0 - load("@io_bazel_rules_go//go:def.bzl", "go_binary") +package(licenses = ["notice"]) # Apache 2.0 + go_binary( name = "stateify", srcs = ["main.go"], -- cgit v1.2.3 From 46603b569c3ab20f45cf1b651d1fd3d2dda33243 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Tue, 23 Oct 2018 14:17:47 -0700 Subject: Fix panic on creation of zero-len shm segments. Attempting to create a zero-len shm segment causes a panic since we try to allocate a zero-len filemem region. The existing code had a guard to disallow this, but the check didn't encode the fact that requesting a private segment implies a segment creation regardless of whether IPC_CREAT is explicitly specified. PiperOrigin-RevId: 218405743 Change-Id: I30aef1232b2125ebba50333a73352c2f907977da --- pkg/sentry/kernel/shm/shm.go | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 8d0d14e45..2feffe612 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -101,9 +101,12 @@ func (r *Registry) findByKey(key int32) *Shm { // FindOrCreate looks up or creates a segment in the registry. It's functionally // analogous to open(2). func (r *Registry) FindOrCreate(ctx context.Context, pid, key int32, size uint64, mode linux.FileMode, private, create, exclusive bool) (*Shm, error) { - if create && (size < linux.SHMMIN || size > linux.SHMMAX) { + if (create || private) && (size < linux.SHMMIN || size > linux.SHMMAX) { // "A new segment was to be created and size is less than SHMMIN or // greater than SHMMAX." - man shmget(2) + // + // Note that 'private' always implies the creation of a new segment + // whether IPC_CREAT is specified or not. return nil, syserror.EINVAL } -- cgit v1.2.3 From 0e277a39c8b6f905e289b75e8ad0594e6b3562ca Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Thu, 1 Nov 2018 15:53:25 -0700 Subject: Prevent premature destruction of shm segments. Shm segments can be marked for lazy destruction via shmctl(IPC_RMID), which destroys a segment once it is no longer attached to any processes. We were unconditionally decrementing the segment refcount on shmctl(IPC_RMID) which allowed a user to force a segment to be destroyed by repeatedly calling shmctl(IPC_RMID), with outstanding memory maps to the segment. This is problematic because the memory released by a segment destroyed this way can be reused by a different process while remaining accessible by the process with outstanding maps to the segment. PiperOrigin-RevId: 219713660 Change-Id: I443ab838322b4fb418ed87b2722c3413ead21845 --- pkg/sentry/kernel/shm/shm.go | 13 +++++++++++-- pkg/sentry/syscalls/linux/sys_shm.go | 2 +- 2 files changed, 12 insertions(+), 3 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 2feffe612..f760f5f76 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -575,10 +575,19 @@ func (s *Shm) destroy() { func (s *Shm) MarkDestroyed() { s.mu.Lock() defer s.mu.Unlock() + // Prevent the segment from being found in the registry. s.key = linux.IPC_PRIVATE - s.pendingDestruction = true - s.DecRef() + + // Only drop the segment's self-reference once, when destruction is + // requested. Otherwise, repeated calls shmctl(IPC_RMID) would force a + // segment to be destroyed prematurely, potentially with active maps to the + // segment's address range. Remaining references are dropped when the + // segment is detached or unmaped. + if !s.pendingDestruction { + s.pendingDestruction = true + s.DecRef() + } } // checkOwnership verifies whether a segment may be accessed by ctx as an diff --git a/pkg/sentry/syscalls/linux/sys_shm.go b/pkg/sentry/syscalls/linux/sys_shm.go index 5f887523a..8753c2e58 100644 --- a/pkg/sentry/syscalls/linux/sys_shm.go +++ b/pkg/sentry/syscalls/linux/sys_shm.go @@ -144,7 +144,7 @@ func Shmctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal return 0, nil, nil case linux.SHM_LOCK, linux.SHM_UNLOCK: - // We currently do not support memmory locking anywhere. + // We currently do not support memory locking anywhere. // mlock(2)/munlock(2) are currently stubbed out as no-ops so do the // same here. t.Kernel().EmitUnimplementedEvent(t) -- cgit v1.2.3 From 5a0be6fa203273d1e4ab06a206eaffeca5724533 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Thu, 8 Nov 2018 11:08:41 -0800 Subject: Create stubs for syscalls upto Linux 4.4. Create syscall stubs for missing syscalls upto Linux 4.4 and advertise a kernel version of 4.4. PiperOrigin-RevId: 220667680 Change-Id: Idbdccde538faabf16debc22f492dd053a8af0ba7 --- pkg/sentry/fs/inode.go | 5 +---- pkg/sentry/fs/proc/README.md | 2 +- pkg/sentry/fs/proc/fds.go | 5 ----- pkg/sentry/fs/proc/task.go | 2 +- pkg/sentry/kernel/auth/user_namespace.go | 5 +---- pkg/sentry/kernel/version.go | 2 +- pkg/sentry/syscalls/linux/linux64.go | 21 +++++++++++++++++---- pkg/sentry/syscalls/linux/sys_file.go | 3 --- pkg/sentry/syscalls/linux/sys_thread.go | 4 ++-- 9 files changed, 24 insertions(+), 25 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index 95769ccf8..38b140bd2 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -439,10 +439,7 @@ func (i *Inode) CheckOwnership(ctx context.Context) bool { // CheckCapability checks whether `ctx` has capability `cp` with respect to // operations on this Inode. // -// Compare Linux's kernel/capability.c:capable_wrt_inode_uidgid(). Note that -// this function didn't exist in Linux 3.11.10, but was added by upstream -// 23adbe12ef7d "fs,userns: Change inode_capable to capable_wrt_inode_uidgid" -// to fix local privilege escalation CVE-2014-4014. +// Compare Linux's kernel/capability.c:capable_wrt_inode_uidgid(). func (i *Inode) CheckCapability(ctx context.Context, cp linux.Capability) bool { uattr, err := i.UnstableAttr(ctx) if err != nil { diff --git a/pkg/sentry/fs/proc/README.md b/pkg/sentry/fs/proc/README.md index e1ed88512..686d40f0c 100644 --- a/pkg/sentry/fs/proc/README.md +++ b/pkg/sentry/fs/proc/README.md @@ -223,7 +223,7 @@ Number of seconds idle | Always zero ```bash $ cat /proc/version -Linux version 3.11.10 #1 SMP Fri Nov 29 10:47:50 PST 2013 +Linux version 4.4 #1 SMP Sun Jan 10 15:06:54 PST 2016 ``` ## Process-specific data diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index 5ebb33703..5acbce75e 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -173,11 +173,6 @@ func (f *fdDir) Check(ctx context.Context, inode *fs.Inode, req fs.PermMask) boo if t := kernel.TaskFromContext(ctx); t != nil { // Allow access if the task trying to access it is in the // thread group corresponding to this directory. - // - // N.B. Technically, in Linux 3.11, this compares what would be - // the equivalent of task pointers. However, this was fixed - // later in 54708d2858e7 ("proc: actually make - // proc_fd_permission() thread-friendly"). if f.t.ThreadGroup() == t.ThreadGroup() { return true } diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 404faea0a..9f13ff91c 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -76,7 +76,7 @@ func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace "gid_map": newGIDMap(t, msrc), // TODO: This is incorrect for /proc/[pid]/task/[tid]/io, i.e. if // showSubtasks is false: - // http://lxr.free-electrons.com/source/fs/proc/base.c?v=3.11#L2980 + // https://elixir.bootlin.com/linux/v4.4/source/fs/proc/base.c#L3154 "io": newIO(t, msrc), "maps": newMaps(t, msrc), "mountinfo": seqfile.NewSeqFileInode(t, &mountInfoFile{t: t}, msrc), diff --git a/pkg/sentry/kernel/auth/user_namespace.go b/pkg/sentry/kernel/auth/user_namespace.go index 5bb9c44c0..30957bb9a 100644 --- a/pkg/sentry/kernel/auth/user_namespace.go +++ b/pkg/sentry/kernel/auth/user_namespace.go @@ -49,10 +49,7 @@ type UserNamespace struct { gidMapFromParent idMapSet gidMapToParent idMapSet - // TODO: Consider supporting disabling setgroups(2), which "was - // added in Linux 3.19, but was backported to many earlier stable kernel - // series, because it addresses a security issue" - user_namespaces(7). (It - // was not backported to 3.11.10, which we are currently imitating.) + // TODO: Support disabling setgroups(2). } // NewRootUserNamespace returns a UserNamespace that is appropriate for a diff --git a/pkg/sentry/kernel/version.go b/pkg/sentry/kernel/version.go index 72bb0f93c..8d2f14209 100644 --- a/pkg/sentry/kernel/version.go +++ b/pkg/sentry/kernel/version.go @@ -19,7 +19,7 @@ type Version struct { // Operating system name (e.g. "Linux"). Sysname string - // Operating system release (e.g. "3.11.10-amd64"). + // Operating system release (e.g. "4.4-amd64"). Release string // Operating system version. On Linux this takes the shape diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index 11bf81f88..13084c0ef 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -32,15 +32,19 @@ import ( const _AUDIT_ARCH_X86_64 = 0xc000003e // AMD64 is a table of Linux amd64 syscall API with the corresponding syscall -// numbers from Linux 3.11. The entries commented out are those syscalls we +// numbers from Linux 4.4. The entries commented out are those syscalls we // don't currently support. var AMD64 = &kernel.SyscallTable{ OS: abi.Linux, Arch: arch.AMD64, Version: kernel.Version{ + // Version 4.4 is chosen as a stable, longterm version of Linux, which + // guides the interface provided by this syscall table. The build + // version is that for a clean build with default kernel config, at 5 + // minutes after v4.4 was tagged. Sysname: "Linux", - Release: "3.11.10", - Version: "#1 SMP Fri Nov 29 10:47:50 PST 2013", + Release: "4.4", + Version: "#1 SMP Sun Jan 10 15:06:54 PST 2016", }, AuditNumber: _AUDIT_ARCH_X86_64, Table: map[uintptr]kernel.SyscallFn{ @@ -358,9 +362,18 @@ var AMD64 = &kernel.SyscallTable{ // 311: ProcessVmWritev, TODO may require cap_sys_ptrace 312: syscalls.CapError(linux.CAP_SYS_PTRACE), // Kcmp, requires cap_sys_ptrace 313: syscalls.CapError(linux.CAP_SYS_MODULE), // FinitModule, requires cap_sys_module - // "Backports." + // 314: SchedSetattr, TODO, we have no scheduler + // 315: SchedGetattr, TODO, we have no scheduler + // 316: Renameat2, TODO 317: Seccomp, 318: GetRandom, + // 319: MemfdCreate, TODO + 320: syscalls.CapError(linux.CAP_SYS_BOOT), // KexecFileLoad, infeasible to support + 321: syscalls.CapError(linux.CAP_SYS_ADMIN), // Bpf, requires cap_sys_admin for all commands + // 322: Execveat, TODO + // 323: Userfaultfd, TODO + // 324: Membarrier, TODO + 325: syscalls.Error(nil), // Mlock2, TODO }, Emulate: map[usermem.Addr]uintptr{ diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index a70f35be0..89d21dd98 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -1140,9 +1140,6 @@ func mayLinkAt(t *kernel.Task, target *fs.Inode) error { // always enabled, and thus imposes the following restrictions on hard // links. - // Technically Linux is more restrictive in 3.11.10 (requires CAP_FOWNER in - // root user namespace); this is from the later f2ca379642d7 "namei: permit - // linking with CAP_FOWNER in userns". if target.CheckOwnership(t) { // fs/namei.c:may_linkat: "Source inode owner (or CAP_FOWNER) // can hardlink all they like." diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index 820ca680e..9eed613a1 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -159,8 +159,8 @@ func clone(t *kernel.Task, flags int, stack usermem.Addr, parentTID usermem.Addr } // Clone implements linux syscall clone(2). -// sys_clone has so many flavors. We implement the default one in the -// current linux 3.11 x86_64: +// sys_clone has so many flavors. We implement the default one in linux 3.11 +// x86_64: // sys_clone(clone_flags, newsp, parent_tidptr, child_tidptr, tls_val) func Clone(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { flags := int(args[0].Int()) -- cgit v1.2.3 From f7aa9371247a3e7d8c490ac0fd4c4f3ff6de2017 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Thu, 15 Nov 2018 15:13:52 -0800 Subject: Advertise vsyscall support via /proc//maps. Also update test utilities for probing vsyscall support and add a metric to see if vsyscalls are actually used in sandboxes. PiperOrigin-RevId: 221698834 Change-Id: I57870ecc33ea8c864bd7437833f21aa1e8117477 --- pkg/sentry/kernel/BUILD | 1 + pkg/sentry/kernel/task_syscall.go | 5 +++++ pkg/sentry/mm/proc_pid_maps.go | 16 ++++++++++++++++ 3 files changed, 22 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 389824b25..10d7b97c2 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -131,6 +131,7 @@ go_library( "//pkg/cpuid", "//pkg/eventchannel", "//pkg/log", + "//pkg/metric", "//pkg/refs", "//pkg/secio", "//pkg/sentry/arch", diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index 0318adb35..2a39ebc68 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -21,6 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/bits" + "gvisor.googlesource.com/gvisor/pkg/metric" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -60,6 +61,8 @@ const ( ERESTART_RESTARTBLOCK = SyscallRestartErrno(516) ) +var vsyscallCount = metric.MustCreateNewUint64Metric("/kernel/vsyscall_count", false /* sync */, "Number of times vsyscalls were invoked by the application") + // Error implements error.Error. func (e SyscallRestartErrno) Error() string { // Descriptions are borrowed from strace. @@ -325,6 +328,8 @@ func (*runSyscallExit) execute(t *Task) taskRunState { // indicated by an execution fault at address addr. doVsyscall returns the // task's next run state. func (t *Task) doVsyscall(addr usermem.Addr, sysno uintptr) taskRunState { + vsyscallCount.Increment() + // Grab the caller up front, to make sure there's a sensible stack. caller := t.Arch().Native(uintptr(0)) if _, err := t.CopyIn(usermem.Addr(t.Arch().Stack()), caller); err != nil { diff --git a/pkg/sentry/mm/proc_pid_maps.go b/pkg/sentry/mm/proc_pid_maps.go index 0bf1cdb51..247ee45ef 100644 --- a/pkg/sentry/mm/proc_pid_maps.go +++ b/pkg/sentry/mm/proc_pid_maps.go @@ -53,6 +53,22 @@ func (mm *MemoryManager) ReadSeqFileData(ctx context.Context, handle seqfile.Seq Handle: &vmaAddr, }) } + + // We always emulate vsyscall, so advertise it here. Everything about a + // vsyscall region is static, so just hard code the maps entry since we + // don't have a real vma backing it. The vsyscall region is at the end of + // the virtual address space so nothing should be mapped after it (if + // something is really mapped in the tiny ~10 MiB segment afterwards, we'll + // get the sorting on the maps file wrong at worst; but that's not possible + // on any current platform). + // + // Artifically adjust the seqfile handle so we only output vsyscall entry once. + if vsyscallEnd := usermem.Addr(0xffffffffff601000); start != vsyscallEnd { + data = append(data, seqfile.SeqData{ + Buf: []byte("ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0 [vsyscall]\n"), + Handle: &vsyscallEnd, + }) + } return data, 1 } -- cgit v1.2.3 From bb9a2bb62ed37f9b29c7ab4418b8b90417d1b2a2 Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Fri, 16 Nov 2018 12:16:37 -0800 Subject: Update futex to use usermem abstractions. This eliminates the indirection that existed in task_futex. PiperOrigin-RevId: 221832498 Change-Id: Ifb4c926d493913aa6694e193deae91616a29f042 --- pkg/sentry/kernel/futex/BUILD | 3 + pkg/sentry/kernel/futex/futex.go | 155 ++++++++++++++++++++++++--------- pkg/sentry/kernel/futex/futex_test.go | 29 +++--- pkg/sentry/kernel/task_exit.go | 2 +- pkg/sentry/kernel/task_futex.go | 127 +++------------------------ pkg/sentry/mm/syscalls.go | 5 +- pkg/sentry/syscalls/linux/sys_futex.go | 28 +++--- 7 files changed, 160 insertions(+), 189 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index e13fcb5ff..afd35985f 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -36,7 +36,9 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex", visibility = ["//pkg/sentry:internal"], deps = [ + "//pkg/abi/linux", "//pkg/sentry/memmap", + "//pkg/sentry/usermem", "//pkg/syserror", ], ) @@ -46,4 +48,5 @@ go_test( size = "small", srcs = ["futex_test.go"], embed = [":futex"], + deps = ["//pkg/sentry/usermem"], ) diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index ea69d433b..b3e628fd4 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -20,7 +20,9 @@ package futex import ( "sync" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -81,8 +83,8 @@ func (k *Key) clone() Key { } // Preconditions: k.Kind == KindPrivate or KindSharedPrivate. -func (k *Key) addr() uintptr { - return uintptr(k.Offset) +func (k *Key) addr() usermem.Addr { + return usermem.Addr(k.Offset) } // matches returns true if a wakeup on k2 should wake a waiter waiting on k. @@ -91,23 +93,13 @@ func (k *Key) matches(k2 *Key) bool { return k.Kind == k2.Kind && k.Mappable == k2.Mappable && k.Offset == k2.Offset } -// Checker abstracts memory accesses. This is useful because the "addresses" -// used in this package may not be real addresses (they could be indices of an -// array, for example), or they could be mapped via some special mechanism. -// -// TODO: Replace this with usermem.IO. -type Checker interface { - // Check should validate that given address contains the given value. - // If it does not contain the value, syserror.EAGAIN must be returned. - // Any other error may be returned, which will be propagated. - Check(addr uintptr, val uint32) error - - // Op should atomically perform the operation encoded in op on the data - // pointed to by addr, then apply the comparison encoded in op to the - // original value at addr, returning the result. - // Note that op is an opaque operation whose behaviour is defined - // outside of the futex manager. - Op(addr uintptr, op uint32) (bool, error) +// Target abstracts memory accesses and keys. +type Target interface { + // SwapUint32 gives access to usermem.SwapUint32. + SwapUint32(addr usermem.Addr, new uint32) (uint32, error) + + // CompareAndSwap gives access to usermem.CompareAndSwapUint32. + CompareAndSwapUint32(addr usermem.Addr, old, new uint32) (uint32, error) // GetSharedKey returns a Key with kind KindSharedPrivate or // KindSharedMappable corresponding to the memory mapped at address addr. @@ -115,7 +107,84 @@ type Checker interface { // If GetSharedKey returns a Key with a non-nil MappingIdentity, a // reference is held on the MappingIdentity, which must be dropped by the // caller when the Key is no longer in use. - GetSharedKey(addr uintptr) (Key, error) + GetSharedKey(addr usermem.Addr) (Key, error) +} + +// check performs a basic equality check on the given address. +func check(t Target, addr usermem.Addr, val uint32) error { + prev, err := t.CompareAndSwapUint32(addr, val, val) + if err != nil { + return err + } + if prev != val { + return syserror.EAGAIN + } + return nil +} + +// atomicOp performs a complex operation on the given address. +func atomicOp(t Target, addr usermem.Addr, opIn uint32) (bool, error) { + opType := (opIn >> 28) & 0xf + cmp := (opIn >> 24) & 0xf + opArg := (opIn >> 12) & 0xfff + cmpArg := opIn & 0xfff + + if opType&linux.FUTEX_OP_OPARG_SHIFT != 0 { + opArg = 1 << opArg + opType &^= linux.FUTEX_OP_OPARG_SHIFT // Clear flag. + } + + var ( + oldVal uint32 + err error + ) + if opType == linux.FUTEX_OP_SET { + oldVal, err = t.SwapUint32(addr, opArg) + } else { + for { + oldVal, err = t.CompareAndSwapUint32(addr, 0, 0) + if err != nil { + break + } + var newVal uint32 + switch opType { + case linux.FUTEX_OP_ADD: + newVal = oldVal + opArg + case linux.FUTEX_OP_OR: + newVal = oldVal | opArg + case linux.FUTEX_OP_ANDN: + newVal = oldVal &^ opArg + case linux.FUTEX_OP_XOR: + newVal = oldVal ^ opArg + default: + return false, syserror.ENOSYS + } + prev, err := t.CompareAndSwapUint32(addr, oldVal, newVal) + if err != nil { + break + } + if prev == oldVal { + break // Success. + } + } + } + + switch cmp { + case linux.FUTEX_OP_CMP_EQ: + return oldVal == cmpArg, nil + case linux.FUTEX_OP_CMP_NE: + return oldVal != cmpArg, nil + case linux.FUTEX_OP_CMP_LT: + return oldVal < cmpArg, nil + case linux.FUTEX_OP_CMP_LE: + return oldVal <= cmpArg, nil + case linux.FUTEX_OP_CMP_GT: + return oldVal > cmpArg, nil + case linux.FUTEX_OP_CMP_GE: + return oldVal >= cmpArg, nil + default: + return false, syserror.ENOSYS + } } // Waiter is the struct which gets enqueued into buckets for wake up routines @@ -243,7 +312,7 @@ const ( ) // getKey returns a Key representing address addr in c. -func getKey(c Checker, addr uintptr, private bool) (Key, error) { +func getKey(t Target, addr usermem.Addr, private bool) (Key, error) { // Ensure the address is aligned. // It must be a DWORD boundary. if addr&0x3 != 0 { @@ -252,11 +321,11 @@ func getKey(c Checker, addr uintptr, private bool) (Key, error) { if private { return Key{Kind: KindPrivate, Offset: uint64(addr)}, nil } - return c.GetSharedKey(addr) + return t.GetSharedKey(addr) } // bucketIndexForAddr returns the index into Manager.buckets for addr. -func bucketIndexForAddr(addr uintptr) uintptr { +func bucketIndexForAddr(addr usermem.Addr) uintptr { // - The bottom 2 bits of addr must be 0, per getKey. // // - On amd64, the top 16 bits of addr (bits 48-63) must be equal to bit 47 @@ -277,8 +346,8 @@ func bucketIndexForAddr(addr uintptr) uintptr { // is also why h1 and h2 are grouped separately; for "(addr >> 2) + ... + // (addr >> 42)" without any additional grouping, the compiler puts all 4 // additions in the critical path. - h1 := (addr >> 2) + (addr >> 12) + (addr >> 22) - h2 := (addr >> 32) + (addr >> 42) + h1 := uintptr(addr>>2) + uintptr(addr>>12) + uintptr(addr>>22) + h2 := uintptr(addr>>32) + uintptr(addr>>42) return (h1 + h2) % bucketCount } @@ -363,9 +432,9 @@ func (m *Manager) lockBuckets(k1, k2 *Key) (*bucket, *bucket) { // Wake wakes up to n waiters matching the bitmask on the given addr. // The number of waiters woken is returned. -func (m *Manager) Wake(c Checker, addr uintptr, private bool, bitmask uint32, n int) (int, error) { +func (m *Manager) Wake(t Target, addr usermem.Addr, private bool, bitmask uint32, n int) (int, error) { // This function is very hot; avoid defer. - k, err := getKey(c, addr, private) + k, err := getKey(t, addr, private) if err != nil { return 0, err } @@ -378,13 +447,13 @@ func (m *Manager) Wake(c Checker, addr uintptr, private bool, bitmask uint32, n return r, nil } -func (m *Manager) doRequeue(c Checker, addr, naddr uintptr, private bool, checkval bool, val uint32, nwake int, nreq int) (int, error) { - k1, err := getKey(c, addr, private) +func (m *Manager) doRequeue(t Target, addr, naddr usermem.Addr, private bool, checkval bool, val uint32, nwake int, nreq int) (int, error) { + k1, err := getKey(t, addr, private) if err != nil { return 0, err } defer k1.release() - k2, err := getKey(c, naddr, private) + k2, err := getKey(t, naddr, private) if err != nil { return 0, err } @@ -397,7 +466,7 @@ func (m *Manager) doRequeue(c Checker, addr, naddr uintptr, private bool, checkv } if checkval { - if err := c.Check(addr, val); err != nil { + if err := check(t, addr, val); err != nil { return 0, err } } @@ -413,28 +482,28 @@ func (m *Manager) doRequeue(c Checker, addr, naddr uintptr, private bool, checkv // Requeue wakes up to nwake waiters on the given addr, and unconditionally // requeues up to nreq waiters on naddr. -func (m *Manager) Requeue(c Checker, addr, naddr uintptr, private bool, nwake int, nreq int) (int, error) { - return m.doRequeue(c, addr, naddr, private, false, 0, nwake, nreq) +func (m *Manager) Requeue(t Target, addr, naddr usermem.Addr, private bool, nwake int, nreq int) (int, error) { + return m.doRequeue(t, addr, naddr, private, false, 0, nwake, nreq) } -// RequeueCmp atomically checks that the addr contains val (via the Checker), +// RequeueCmp atomically checks that the addr contains val (via the Target), // wakes up to nwake waiters on addr and then unconditionally requeues nreq // waiters on naddr. -func (m *Manager) RequeueCmp(c Checker, addr, naddr uintptr, private bool, val uint32, nwake int, nreq int) (int, error) { - return m.doRequeue(c, addr, naddr, private, true, val, nwake, nreq) +func (m *Manager) RequeueCmp(t Target, addr, naddr usermem.Addr, private bool, val uint32, nwake int, nreq int) (int, error) { + return m.doRequeue(t, addr, naddr, private, true, val, nwake, nreq) } // WakeOp atomically applies op to the memory address addr2, wakes up to nwake1 // waiters unconditionally from addr1, and, based on the original value at addr2 // and a comparison encoded in op, wakes up to nwake2 waiters from addr2. // It returns the total number of waiters woken. -func (m *Manager) WakeOp(c Checker, addr1, addr2 uintptr, private bool, nwake1 int, nwake2 int, op uint32) (int, error) { - k1, err := getKey(c, addr1, private) +func (m *Manager) WakeOp(t Target, addr1, addr2 usermem.Addr, private bool, nwake1 int, nwake2 int, op uint32) (int, error) { + k1, err := getKey(t, addr1, private) if err != nil { return 0, err } defer k1.release() - k2, err := getKey(c, addr2, private) + k2, err := getKey(t, addr2, private) if err != nil { return 0, err } @@ -447,7 +516,7 @@ func (m *Manager) WakeOp(c Checker, addr1, addr2 uintptr, private bool, nwake1 i } done := 0 - cond, err := c.Op(addr2, op) + cond, err := atomicOp(t, addr2, op) if err != nil { return 0, err } @@ -468,8 +537,8 @@ func (m *Manager) WakeOp(c Checker, addr1, addr2 uintptr, private bool, nwake1 i // enqueues w to be woken by a send to w.C. If WaitPrepare returns nil, the // Waiter must be subsequently removed by calling WaitComplete, whether or not // a wakeup is received on w.C. -func (m *Manager) WaitPrepare(w *Waiter, c Checker, addr uintptr, private bool, val uint32, bitmask uint32) error { - k, err := getKey(c, addr, private) +func (m *Manager) WaitPrepare(w *Waiter, t Target, addr usermem.Addr, private bool, val uint32, bitmask uint32) error { + k, err := getKey(t, addr, private) if err != nil { return err } @@ -487,7 +556,7 @@ func (m *Manager) WaitPrepare(w *Waiter, c Checker, addr uintptr, private bool, // This function is very hot; avoid defer. // Perform our atomic check. - if err := c.Check(addr, val); err != nil { + if err := check(t, addr, val); err != nil { b.mu.Unlock() w.key.release() return err diff --git a/pkg/sentry/kernel/futex/futex_test.go b/pkg/sentry/kernel/futex/futex_test.go index ea506a29b..a7ab9f229 100644 --- a/pkg/sentry/kernel/futex/futex_test.go +++ b/pkg/sentry/kernel/futex/futex_test.go @@ -22,9 +22,11 @@ import ( "syscall" "testing" "unsafe" + + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// testData implements the Checker interface, and allows us to +// testData implements the Target interface, and allows us to // treat the address passed for futex operations as an index in // a byte slice for testing simplicity. type testData []byte @@ -35,18 +37,19 @@ func newTestData(size uint) testData { return make([]byte, size) } -func (t testData) Check(addr uintptr, val uint32) error { - if val != atomic.LoadUint32((*uint32)(unsafe.Pointer(&t[addr]))) { - return syscall.EAGAIN - } - return nil +func (t testData) SwapUint32(addr usermem.Addr, new uint32) (uint32, error) { + val := atomic.SwapUint32((*uint32)(unsafe.Pointer(&t[addr])), new) + return val, nil } -func (t testData) Op(addr uintptr, val uint32) (bool, error) { - return val == 0, nil +func (t testData) CompareAndSwapUint32(addr usermem.Addr, old, new uint32) (uint32, error) { + if atomic.CompareAndSwapUint32((*uint32)(unsafe.Pointer(&t[addr])), old, new) { + return old, nil + } + return atomic.LoadUint32((*uint32)(unsafe.Pointer(&t[addr]))), nil } -func (t testData) GetSharedKey(addr uintptr) (Key, error) { +func (t testData) GetSharedKey(addr usermem.Addr) (Key, error) { return Key{ Kind: KindSharedMappable, Offset: uint64(addr), @@ -60,9 +63,9 @@ func futexKind(private bool) string { return "shared" } -func newPreparedTestWaiter(t *testing.T, m *Manager, c Checker, addr uintptr, private bool, val uint32, bitmask uint32) *Waiter { +func newPreparedTestWaiter(t *testing.T, m *Manager, ta Target, addr usermem.Addr, private bool, val uint32, bitmask uint32) *Waiter { w := NewWaiter() - if err := m.WaitPrepare(w, c, addr, private, val, bitmask); err != nil { + if err := m.WaitPrepare(w, ta, addr, private, val, bitmask); err != nil { t.Fatalf("WaitPrepare failed: %v", err) } return w @@ -450,12 +453,12 @@ const ( // Beyond being used as a Locker, this is a simple mechanism for // changing the underlying values for simpler tests. type testMutex struct { - a uintptr + a usermem.Addr d testData m *Manager } -func newTestMutex(addr uintptr, d testData, m *Manager) *testMutex { +func newTestMutex(addr usermem.Addr, d testData, m *Manager) *testMutex { return &testMutex{a: addr, d: d, m: m} } diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index 44fbb487c..791cc9831 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -247,7 +247,7 @@ func (*runExitMain) execute(t *Task) taskRunState { t.tg.signalHandlers.mu.Unlock() if !signaled { if _, err := t.CopyOut(t.cleartid, ThreadID(0)); err == nil { - t.Futex().Wake(t.FutexChecker(), uintptr(t.cleartid), false, ^uint32(0), 1) + t.Futex().Wake(t, t.cleartid, false, ^uint32(0), 1) } // If the CopyOut fails, there's nothing we can do. } diff --git a/pkg/sentry/kernel/task_futex.go b/pkg/sentry/kernel/task_futex.go index 5a11ca3df..921f7bdbc 100644 --- a/pkg/sentry/kernel/task_futex.go +++ b/pkg/sentry/kernel/task_futex.go @@ -15,10 +15,8 @@ package kernel import ( - "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" ) // Futex returns t's futex manager. @@ -29,120 +27,21 @@ func (t *Task) Futex() *futex.Manager { return t.tc.fu } -// FutexChecker returns a futex.Checker that interprets addresses in t's -// address space. -// -// Preconditions: All uses of the returned futex.Checker must be on the task -// goroutine. -func (t *Task) FutexChecker() futex.Checker { - return futexChecker{t} -} - -type futexChecker struct { - t *Task +// SwapUint32 implements futex.Target.SwapUint32. +func (t *Task) SwapUint32(addr usermem.Addr, new uint32) (uint32, error) { + return t.MemoryManager().SwapUint32(t, addr, new, usermem.IOOpts{ + AddressSpaceActive: true, + }) } -// Check implements futex.Checker.Check. -func (f futexChecker) Check(addr uintptr, val uint32) error { - // FIXME - in := f.t.CopyScratchBuffer(4) - _, err := f.t.CopyInBytes(usermem.Addr(addr), in) - if err != nil { - return err - } - nval := usermem.ByteOrder.Uint32(in) - if val != nval { - return syserror.EAGAIN - } - return nil -} - -func (f futexChecker) atomicOp(addr uintptr, op func(uint32) uint32) (uint32, error) { - // FIXME - in := f.t.CopyScratchBuffer(4) - _, err := f.t.CopyInBytes(usermem.Addr(addr), in) - if err != nil { - return 0, err - } - o := usermem.ByteOrder.Uint32(in) - mm := f.t.MemoryManager() - for { - n := op(o) - r, err := mm.CompareAndSwapUint32(f.t, usermem.Addr(addr), o, n, usermem.IOOpts{ - AddressSpaceActive: true, - }) - if err != nil { - return 0, err - } - - if r == o { - return o, nil - } - o = r - } -} - -// Op implements futex.Checker.Op, interpreting opIn consistently with Linux. -func (f futexChecker) Op(addr uintptr, opIn uint32) (bool, error) { - op := (opIn >> 28) & 0xf - cmp := (opIn >> 24) & 0xf - opArg := (opIn >> 12) & 0xfff - cmpArg := opIn & 0xfff - - if op&linux.FUTEX_OP_OPARG_SHIFT != 0 { - opArg = 1 << opArg - op &^= linux.FUTEX_OP_OPARG_SHIFT // clear flag - } - - var oldVal uint32 - var err error - switch op { - case linux.FUTEX_OP_SET: - oldVal, err = f.t.MemoryManager().SwapUint32(f.t, usermem.Addr(addr), opArg, usermem.IOOpts{ - AddressSpaceActive: true, - }) - case linux.FUTEX_OP_ADD: - oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { - return a + opArg - }) - case linux.FUTEX_OP_OR: - oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { - return a | opArg - }) - case linux.FUTEX_OP_ANDN: - oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { - return a &^ opArg - }) - case linux.FUTEX_OP_XOR: - oldVal, err = f.atomicOp(addr, func(a uint32) uint32 { - return a ^ opArg - }) - default: - return false, syserror.ENOSYS - } - if err != nil { - return false, err - } - - switch cmp { - case linux.FUTEX_OP_CMP_EQ: - return oldVal == cmpArg, nil - case linux.FUTEX_OP_CMP_NE: - return oldVal != cmpArg, nil - case linux.FUTEX_OP_CMP_LT: - return oldVal < cmpArg, nil - case linux.FUTEX_OP_CMP_LE: - return oldVal <= cmpArg, nil - case linux.FUTEX_OP_CMP_GT: - return oldVal > cmpArg, nil - case linux.FUTEX_OP_CMP_GE: - return oldVal >= cmpArg, nil - default: - return false, syserror.ENOSYS - } +// CompareAndSwapUint32 implemets futex.Target.CompareAndSwapUint32. +func (t *Task) CompareAndSwapUint32(addr usermem.Addr, old, new uint32) (uint32, error) { + return t.MemoryManager().CompareAndSwapUint32(t, addr, old, new, usermem.IOOpts{ + AddressSpaceActive: true, + }) } -// GetSharedKey implements futex.Checker.GetSharedKey. -func (f futexChecker) GetSharedKey(addr uintptr) (futex.Key, error) { - return f.t.MemoryManager().GetSharedFutexKey(f.t, usermem.Addr(addr)) +// GetSharedKey implements futex.Target.GetSharedKey. +func (t *Task) GetSharedKey(addr usermem.Addr) (futex.Key, error) { + return t.MemoryManager().GetSharedFutexKey(t, addr) } diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index a721cc456..9519c7390 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -794,10 +794,9 @@ func (mm *MemoryManager) Sync(ctx context.Context, addr usermem.Addr, length uin return nil } -// GetSharedFutexKey is used by kernel.futexChecker.GetSharedKey to implement -// futex.Checker.GetSharedKey. +// GetSharedFutexKey is used by kernel.Task.GetSharedKey. func (mm *MemoryManager) GetSharedFutexKey(ctx context.Context, addr usermem.Addr) (futex.Key, error) { - ar, ok := addr.ToRange(4) // sizeof(int32) + ar, ok := addr.ToRange(4) // sizeof(int32). if !ok { return futex.Key{}, syserror.EFAULT } diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index cf04428bc..7a1d396ec 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -21,6 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -32,8 +33,7 @@ type futexWaitRestartBlock struct { duration time.Duration // addr stored as uint64 since uintptr is not save-able. - addr uint64 - + addr uint64 private bool val uint32 mask uint32 @@ -41,7 +41,7 @@ type futexWaitRestartBlock struct { // Restart implements kernel.SyscallRestartBlock.Restart. func (f *futexWaitRestartBlock) Restart(t *kernel.Task) (uintptr, error) { - return futexWaitDuration(t, f.duration, false, uintptr(f.addr), f.private, f.val, f.mask) + return futexWaitDuration(t, f.duration, false, usermem.Addr(f.addr), f.private, f.val, f.mask) } // futexWaitAbsolute performs a FUTEX_WAIT_BITSET, blocking until the wait is @@ -51,9 +51,9 @@ func (f *futexWaitRestartBlock) Restart(t *kernel.Task) (uintptr, error) { // // If blocking is interrupted, the syscall is restarted with the original // arguments. -func futexWaitAbsolute(t *kernel.Task, clockRealtime bool, ts linux.Timespec, forever bool, addr uintptr, private bool, val, mask uint32) (uintptr, error) { +func futexWaitAbsolute(t *kernel.Task, clockRealtime bool, ts linux.Timespec, forever bool, addr usermem.Addr, private bool, val, mask uint32) (uintptr, error) { w := t.FutexWaiter() - err := t.Futex().WaitPrepare(w, t.FutexChecker(), addr, private, val, mask) + err := t.Futex().WaitPrepare(w, t, addr, private, val, mask) if err != nil { return 0, err } @@ -87,9 +87,9 @@ func futexWaitAbsolute(t *kernel.Task, clockRealtime bool, ts linux.Timespec, fo // syscall. If forever is true, the syscall is restarted with the original // arguments. If forever is false, duration is a relative timeout and the // syscall is restarted with the remaining timeout. -func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, addr uintptr, private bool, val, mask uint32) (uintptr, error) { +func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, addr usermem.Addr, private bool, val, mask uint32) (uintptr, error) { w := t.FutexWaiter() - err := t.Futex().WaitPrepare(w, t.FutexChecker(), addr, private, val, mask) + err := t.Futex().WaitPrepare(w, t, addr, private, val, mask) if err != nil { return 0, err } @@ -128,16 +128,14 @@ func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, add // It provides a method for a program to wait for a value at a given address to // change, and a method to wake up anyone waiting on a particular address. func Futex(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { - uaddr := args[0].Pointer() + addr := args[0].Pointer() futexOp := args[1].Int() val := int(args[2].Int()) nreq := int(args[3].Int()) timeout := args[3].Pointer() - uaddr2 := args[4].Pointer() + naddr := args[4].Pointer() val3 := args[5].Int() - addr := uintptr(uaddr) - naddr := uintptr(uaddr2) cmd := futexOp &^ (linux.FUTEX_PRIVATE_FLAG | linux.FUTEX_CLOCK_REALTIME) private := (futexOp & linux.FUTEX_PRIVATE_FLAG) != 0 clockRealtime := (futexOp & linux.FUTEX_CLOCK_REALTIME) == linux.FUTEX_CLOCK_REALTIME @@ -188,23 +186,23 @@ func Futex(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall if mask == 0 { return 0, nil, syserror.EINVAL } - n, err := t.Futex().Wake(t.FutexChecker(), addr, private, mask, val) + n, err := t.Futex().Wake(t, addr, private, mask, val) return uintptr(n), nil, err case linux.FUTEX_REQUEUE: - n, err := t.Futex().Requeue(t.FutexChecker(), addr, naddr, private, val, nreq) + n, err := t.Futex().Requeue(t, addr, naddr, private, val, nreq) return uintptr(n), nil, err case linux.FUTEX_CMP_REQUEUE: // 'val3' contains the value to be checked at 'addr' and // 'val' is the number of waiters that should be woken up. nval := uint32(val3) - n, err := t.Futex().RequeueCmp(t.FutexChecker(), addr, naddr, private, nval, val, nreq) + n, err := t.Futex().RequeueCmp(t, addr, naddr, private, nval, val, nreq) return uintptr(n), nil, err case linux.FUTEX_WAKE_OP: op := uint32(val3) - n, err := t.Futex().WakeOp(t.FutexChecker(), addr, naddr, private, val, nreq, op) + n, err := t.Futex().WakeOp(t, addr, naddr, private, val, nreq, op) return uintptr(n), nil, err case linux.FUTEX_LOCK_PI, linux.FUTEX_UNLOCK_PI, linux.FUTEX_TRYLOCK_PI, linux.FUTEX_WAIT_REQUEUE_PI, linux.FUTEX_CMP_REQUEUE_PI: -- cgit v1.2.3 From 8b314b0bf402da58f90ccaac852a880d375f0885 Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Tue, 20 Nov 2018 15:07:12 -0800 Subject: Fix recursive read lock taken on TaskSet SyncSyscallFiltersToThreadGroup and Task.TheadID() both acquired TaskSet RWLock in R mode and could deadlock if a writer comes in between. PiperOrigin-RevId: 222313551 Change-Id: I4221057d8d46fec544cbfa55765c9a284fe7ebfa --- pkg/sentry/kernel/seccomp.go | 48 +++++++++++--------------------- pkg/sentry/kernel/task.go | 2 +- pkg/sentry/syscalls/linux/sys_seccomp.go | 7 +---- 3 files changed, 19 insertions(+), 38 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/seccomp.go b/pkg/sentry/kernel/seccomp.go index 37dd3e4c9..433b900c7 100644 --- a/pkg/sentry/kernel/seccomp.go +++ b/pkg/sentry/kernel/seccomp.go @@ -179,20 +179,19 @@ func (t *Task) evaluateSyscallFilters(sysno int32, args arch.SyscallArguments, i // AppendSyscallFilter adds BPF program p as a system call filter. // // Preconditions: The caller must be running on the task goroutine. -func (t *Task) AppendSyscallFilter(p bpf.Program) error { +func (t *Task) AppendSyscallFilter(p bpf.Program, syncAll bool) error { + // While syscallFilters are an atomic.Value we must take the mutex to prevent + // our read-copy-update from happening while another task is syncing syscall + // filters to us, this keeps the filters in a consistent state. + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + // Cap the combined length of all syscall filters (plus a penalty of 4 - // instructions per filter beyond the first) to - // maxSyscallFilterInstructions. (This restriction is inherited from - // Linux.) + // instructions per filter beyond the first) to maxSyscallFilterInstructions. + // This restriction is inherited from Linux. totalLength := p.Length() var newFilters []bpf.Program - // While syscallFilters are an atomic.Value we must take the mutex to - // prevent our read-copy-update from happening while another task - // is syncing syscall filters to us, this keeps the filters in a - // consistent state. - t.mu.Lock() - defer t.mu.Unlock() if sf := t.syscallFilters.Load(); sf != nil { oldFilters := sf.([]bpf.Program) for _, f := range oldFilters { @@ -207,31 +206,18 @@ func (t *Task) AppendSyscallFilter(p bpf.Program) error { newFilters = append(newFilters, p) t.syscallFilters.Store(newFilters) - return nil -} -// SyncSyscallFiltersToThreadGroup will copy this task's filters to all other -// threads in our thread group. -func (t *Task) SyncSyscallFiltersToThreadGroup() error { - f := t.syscallFilters.Load() - - t.tg.pidns.owner.mu.RLock() - defer t.tg.pidns.owner.mu.RUnlock() - - // Note: No new privs is always assumed to be set. - for ot := t.tg.tasks.Front(); ot != nil; ot = ot.Next() { - if ot.ThreadID() != t.ThreadID() { - // We must take the other task's mutex to prevent it from - // appending to its own syscall filters while we're syncing. - ot.mu.Lock() - var copiedFilters []bpf.Program - if f != nil { - copiedFilters = append(copiedFilters, f.([]bpf.Program)...) + if syncAll { + // Note: No new privs is always assumed to be set. + for ot := t.tg.tasks.Front(); ot != nil; ot = ot.Next() { + if ot != t { + var copiedFilters []bpf.Program + copiedFilters = append(copiedFilters, newFilters...) + ot.syscallFilters.Store(copiedFilters) } - ot.syscallFilters.Store(copiedFilters) - ot.mu.Unlock() } } + return nil } diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 73ba8bee9..2982bc5d1 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -392,7 +392,7 @@ type Task struct { // syscallFilters is all seccomp-bpf syscall filters applicable to the // task, in the order in which they were installed. The type of the atomic - // is []bpf.Program. Writing needs to be protected by mu. + // is []bpf.Program. Writing needs to be protected by the signal mutex. // // syscallFilters is owned by the task goroutine. syscallFilters atomic.Value `state:".([]bpf.Program)"` diff --git a/pkg/sentry/syscalls/linux/sys_seccomp.go b/pkg/sentry/syscalls/linux/sys_seccomp.go index 969acaa36..f08fdf5cb 100644 --- a/pkg/sentry/syscalls/linux/sys_seccomp.go +++ b/pkg/sentry/syscalls/linux/sys_seccomp.go @@ -68,12 +68,7 @@ func seccomp(t *kernel.Task, mode, flags uint64, addr usermem.Addr) error { return syscall.EINVAL } - err = t.AppendSyscallFilter(compiledFilter) - if err == nil && tsync { - // Now we must copy this seccomp program to all other threads. - err = t.SyncSyscallFiltersToThreadGroup() - } - return err + return t.AppendSyscallFilter(compiledFilter, tsync) } // Seccomp implements linux syscall seccomp(2). -- cgit v1.2.3 From eaac94d91c28b745c51c33dd352ed9bfdd671b8c Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Tue, 20 Nov 2018 22:55:41 -0800 Subject: Use RET_KILL_PROCESS if available in kernel RET_KILL_THREAD doesn't work well for Go because it will kill only the offending thread and leave the process hanging. RET_TRAP can be masked out and it's not guaranteed to kill the process. RET_KILL_PROCESS is available since 4.14. For older kernel, continue to use RET_TRAP as this is the best option (likely to kill process, easy to debug). PiperOrigin-RevId: 222357867 Change-Id: Icc1d7d731274b16c2125b7a1ba4f7883fbdb2cbd --- pkg/abi/linux/seccomp.go | 12 +++--- pkg/seccomp/seccomp.go | 52 ++++++++++++++++++++++---- pkg/seccomp/seccomp_test.go | 20 +++++++--- pkg/seccomp/seccomp_test_victim.go | 2 +- pkg/seccomp/seccomp_unsafe.go | 30 ++++++++++++--- pkg/sentry/kernel/seccomp.go | 4 +- pkg/sentry/platform/ptrace/subprocess_linux.go | 2 +- runsc/boot/filter/filter.go | 3 +- runsc/fsgofer/filter/filter.go | 3 +- 9 files changed, 95 insertions(+), 33 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/seccomp.go b/pkg/abi/linux/seccomp.go index 9963ceeba..5ec01cc4a 100644 --- a/pkg/abi/linux/seccomp.go +++ b/pkg/abi/linux/seccomp.go @@ -19,17 +19,19 @@ const ( SECCOMP_MODE_NONE = 0 SECCOMP_MODE_FILTER = 2 - SECCOMP_RET_KILL = 0x00000000 - SECCOMP_RET_TRAP = 0x00030000 - SECCOMP_RET_ERRNO = 0x00050000 - SECCOMP_RET_TRACE = 0x7ff00000 - SECCOMP_RET_ALLOW = 0x7fff0000 + SECCOMP_RET_KILL_PROCESS = 0x80000000 + SECCOMP_RET_KILL_THREAD = 0x00000000 + SECCOMP_RET_TRAP = 0x00030000 + SECCOMP_RET_ERRNO = 0x00050000 + SECCOMP_RET_TRACE = 0x7ff00000 + SECCOMP_RET_ALLOW = 0x7fff0000 SECCOMP_RET_ACTION = 0x7fff0000 SECCOMP_RET_DATA = 0x0000ffff SECCOMP_SET_MODE_FILTER = 1 SECCOMP_FILTER_FLAG_TSYNC = 1 + SECCOMP_GET_ACTION_AVAIL = 2 ) const ( diff --git a/pkg/seccomp/seccomp.go b/pkg/seccomp/seccomp.go index 1dfbf749e..9d714d02d 100644 --- a/pkg/seccomp/seccomp.go +++ b/pkg/seccomp/seccomp.go @@ -33,17 +33,42 @@ const ( defaultLabel = "default_action" ) +func actionName(a uint32) string { + switch a { + case linux.SECCOMP_RET_KILL_PROCESS: + return "kill process" + case linux.SECCOMP_RET_TRAP: + return "trap" + } + panic(fmt.Sprintf("invalid action: %d", a)) +} + // Install generates BPF code based on the set of syscalls provided. It only -// allows syscalls that conform to the specification and generates SIGSYS -// trap unless kill is set. +// allows syscalls that conform to the specification. Syscalls that violate the +// specification will trigger RET_KILL_PROCESS, except for the cases below. +// +// RET_TRAP is used in violations, instead of RET_KILL_PROCESS, in the +// following cases: +// 1. Kernel doesn't support RET_KILL_PROCESS: RET_KILL_THREAD only kills the +// offending thread and often keeps the sentry hanging. +// 2. Debug: RET_TRAP generates a panic followed by a stack trace which is +// much easier to debug then RET_KILL_PROCESS which can't be caught. // -// This is a convenience wrapper around BuildProgram and SetFilter. -func Install(rules SyscallRules, kill bool) error { - log.Infof("Installing seccomp filters for %d syscalls (kill=%t)", len(rules), kill) - defaultAction := uint32(linux.SECCOMP_RET_TRAP) - if kill { - defaultAction = uint32(linux.SECCOMP_RET_KILL) +// Be aware that RET_TRAP sends SIGSYS to the process and it may be ignored, +// making it possible for the process to continue running after a violation. +// However, it will leave a SECCOMP audit event trail behind. In any case, the +// syscall is still blocked from executing. +func Install(rules SyscallRules) error { + defaultAction, err := defaultAction() + if err != nil { + return err } + + // Uncomment to get stack trace when there is a violation. + // defaultAction = uint32(linux.SECCOMP_RET_TRAP) + + log.Infof("Installing seccomp filters for %d syscalls (action=%s)", len(rules), actionName(defaultAction)) + instrs, err := BuildProgram([]RuleSet{ RuleSet{ Rules: rules, @@ -70,6 +95,17 @@ func Install(rules SyscallRules, kill bool) error { return nil } +func defaultAction() (uint32, error) { + available, err := isKillProcessAvailable() + if err != nil { + return 0, err + } + if available { + return uint32(linux.SECCOMP_RET_KILL_PROCESS), nil + } + return uint32(linux.SECCOMP_RET_TRAP), nil +} + // RuleSet is a set of rules and associated action. type RuleSet struct { Rules SyscallRules diff --git a/pkg/seccomp/seccomp_test.go b/pkg/seccomp/seccomp_test.go index 226f30b7b..f2b903e42 100644 --- a/pkg/seccomp/seccomp_test.go +++ b/pkg/seccomp/seccomp_test.go @@ -121,7 +121,7 @@ func TestBasic(t *testing.T) { Action: linux.SECCOMP_RET_TRAP, }, }, - defaultAction: linux.SECCOMP_RET_KILL, + defaultAction: linux.SECCOMP_RET_KILL_THREAD, specs: []spec{ { desc: "Multiple rulesets allowed (1a)", @@ -141,7 +141,7 @@ func TestBasic(t *testing.T) { { desc: "Multiple rulesets allowed (2)", data: seccompData{nr: 0, arch: linux.AUDIT_ARCH_X86_64}, - want: linux.SECCOMP_RET_KILL, + want: linux.SECCOMP_RET_KILL_THREAD, }, }, }, @@ -431,15 +431,23 @@ func TestRealDeal(t *testing.T) { t.Errorf("victim was not killed as expected, output: %s", out) continue } + // Depending on kernel version, either RET_TRAP or RET_KILL_PROCESS is + // used. RET_TRAP dumps reason for exit in output, while RET_KILL_PROCESS + // returns SIGSYS as exit status. + if !strings.Contains(string(out), test.want) && + !strings.Contains(err.Error(), test.want) { + t.Errorf("Victim error is wrong, got: %v, err: %v, want: %v", string(out), err, test.want) + continue + } } else { if err != nil { t.Errorf("victim failed to execute, err: %v", err) continue } - } - if !strings.Contains(string(out), test.want) { - t.Errorf("Victim output is wrong, got: %v, want: %v", err, test.want) - continue + if !strings.Contains(string(out), test.want) { + t.Errorf("Victim output is wrong, got: %v, want: %v", string(out), test.want) + continue + } } } } diff --git a/pkg/seccomp/seccomp_test_victim.go b/pkg/seccomp/seccomp_test_victim.go index 007038273..dd5ed0041 100644 --- a/pkg/seccomp/seccomp_test_victim.go +++ b/pkg/seccomp/seccomp_test_victim.go @@ -106,7 +106,7 @@ func main() { } } - if err := seccomp.Install(syscalls, false); err != nil { + if err := seccomp.Install(syscalls); err != nil { fmt.Printf("Failed to install seccomp: %v", err) os.Exit(1) } diff --git a/pkg/seccomp/seccomp_unsafe.go b/pkg/seccomp/seccomp_unsafe.go index dd009221a..a31c6471d 100644 --- a/pkg/seccomp/seccomp_unsafe.go +++ b/pkg/seccomp/seccomp_unsafe.go @@ -36,22 +36,40 @@ type sockFprog struct { // //go:nosplit func SetFilter(instrs []linux.BPFInstruction) syscall.Errno { - // SYS_SECCOMP is not available in syscall package. - const SYS_SECCOMP = 317 - // PR_SET_NO_NEW_PRIVS is required in order to enable seccomp. See seccomp(2) for details. if _, _, errno := syscall.RawSyscall(syscall.SYS_PRCTL, linux.PR_SET_NO_NEW_PRIVS, 1, 0); errno != 0 { return errno } - // TODO: Use SECCOMP_FILTER_FLAG_KILL_PROCESS when available. sockProg := sockFprog{ Len: uint16(len(instrs)), Filter: (*linux.BPFInstruction)(unsafe.Pointer(&instrs[0])), } - if _, _, errno := syscall.RawSyscall(SYS_SECCOMP, linux.SECCOMP_SET_MODE_FILTER, linux.SECCOMP_FILTER_FLAG_TSYNC, uintptr(unsafe.Pointer(&sockProg))); errno != 0 { - return errno + return seccomp(linux.SECCOMP_SET_MODE_FILTER, linux.SECCOMP_FILTER_FLAG_TSYNC, unsafe.Pointer(&sockProg)) +} + +func isKillProcessAvailable() (bool, error) { + action := uint32(linux.SECCOMP_RET_KILL_PROCESS) + if errno := seccomp(linux.SECCOMP_GET_ACTION_AVAIL, 0, unsafe.Pointer(&action)); errno != 0 { + // EINVAL: SECCOMP_GET_ACTION_AVAIL not in this kernel yet. + // EOPNOTSUPP: SECCOMP_RET_KILL_PROCESS not supported. + if errno == syscall.EINVAL || errno == syscall.EOPNOTSUPP { + return false, nil + } + return false, errno } + return true, nil +} +// seccomp calls seccomp(2). This is safe to call from an afterFork context. +// +//go:nosplit +func seccomp(op, flags uint32, ptr unsafe.Pointer) syscall.Errno { + // SYS_SECCOMP is not available in syscall package. + const SYS_SECCOMP = 317 + + if _, _, errno := syscall.RawSyscall(SYS_SECCOMP, uintptr(op), uintptr(flags), uintptr(ptr)); errno != 0 { + return errno + } return 0 } diff --git a/pkg/sentry/kernel/seccomp.go b/pkg/sentry/kernel/seccomp.go index 433b900c7..d6dc45bbd 100644 --- a/pkg/sentry/kernel/seccomp.go +++ b/pkg/sentry/kernel/seccomp.go @@ -117,7 +117,7 @@ func (t *Task) checkSeccompSyscall(sysno int32, args arch.SyscallArguments, ip u // "Results in the system call being executed." return seccompResultAllow - case linux.SECCOMP_RET_KILL: + case linux.SECCOMP_RET_KILL_THREAD: // "Results in the task exiting immediately without executing the // system call. The exit status of the task will be SIGSYS, not // SIGKILL." @@ -155,7 +155,7 @@ func (t *Task) evaluateSyscallFilters(sysno int32, args arch.SyscallArguments, i thisRet, err := bpf.Exec(f.([]bpf.Program)[i], input) if err != nil { t.Debugf("seccomp-bpf filter %d returned error: %v", i, err) - thisRet = linux.SECCOMP_RET_KILL + thisRet = linux.SECCOMP_RET_KILL_THREAD } // "If multiple filters exist, the return value for the evaluation of a // given system call will always use the highest precedent value." - diff --git a/pkg/sentry/platform/ptrace/subprocess_linux.go b/pkg/sentry/platform/ptrace/subprocess_linux.go index 885ba4b2e..25b8e8cb7 100644 --- a/pkg/sentry/platform/ptrace/subprocess_linux.go +++ b/pkg/sentry/platform/ptrace/subprocess_linux.go @@ -115,7 +115,7 @@ func createStub() (*thread, error) { var defaultAction uint32 if probeSeccomp() { log.Infof("Latest seccomp behavior found (kernel >= 4.8 likely)") - defaultAction = uint32(linux.SECCOMP_RET_KILL) + defaultAction = uint32(linux.SECCOMP_RET_KILL_THREAD) } else { // We must rely on SYSEMU behavior; tracing with SYSEMU is broken. log.Infof("Legacy seccomp behavior found (kernel < 4.8 likely)") diff --git a/runsc/boot/filter/filter.go b/runsc/boot/filter/filter.go index dc7294b1d..d69a6a2cc 100644 --- a/runsc/boot/filter/filter.go +++ b/runsc/boot/filter/filter.go @@ -57,8 +57,7 @@ func Install(opt Options) error { return fmt.Errorf("unknown platform type %T", p) } - // TODO: Set kill=true when SECCOMP_RET_KILL_PROCESS is supported. - return seccomp.Install(s, false) + return seccomp.Install(s) } // Report writes a warning message to the log. diff --git a/runsc/fsgofer/filter/filter.go b/runsc/fsgofer/filter/filter.go index f50b6bc87..c120d57a6 100644 --- a/runsc/fsgofer/filter/filter.go +++ b/runsc/fsgofer/filter/filter.go @@ -29,6 +29,5 @@ func Install() error { // when not enabled. s.Merge(instrumentationFilters()) - // TODO: Set kill=true when SECCOMP_RET_KILL_PROCESS is supported. - return seccomp.Install(s, false) + return seccomp.Install(s) } -- cgit v1.2.3 From 82719be42e636f86780d21b01e10ecb2c9a25e53 Mon Sep 17 00:00:00 2001 From: Brian Geffon Date: Tue, 4 Dec 2018 14:31:08 -0800 Subject: Max link traversals should be for an entire path. The number of symbolic links that are allowed to be followed are for a full path and not just a chain of symbolic links. PiperOrigin-RevId: 224047321 Change-Id: I5e3c4caf66a93c17eeddcc7f046d1e8bb9434a40 --- pkg/sentry/fs/copy_up_test.go | 3 ++- pkg/sentry/fs/host/fs.go | 3 ++- pkg/sentry/fs/host/fs_test.go | 3 ++- pkg/sentry/fs/inode_overlay_test.go | 3 ++- pkg/sentry/fs/mount_test.go | 10 +++++++--- pkg/sentry/fs/mounts.go | 22 ++++++++++++---------- pkg/sentry/fs/mounts_test.go | 6 ++++-- pkg/sentry/fs/ramfs/tree_test.go | 3 ++- pkg/sentry/kernel/kernel.go | 6 ++++-- pkg/sentry/kernel/task_context.go | 2 +- pkg/sentry/loader/elf.go | 2 +- pkg/sentry/loader/loader.go | 10 +++++----- pkg/sentry/socket/unix/unix.go | 6 ++++-- pkg/sentry/syscalls/linux/sys_file.go | 11 +++++++---- pkg/sentry/syscalls/linux/sys_thread.go | 3 ++- runsc/boot/fs.go | 12 ++++++++---- runsc/boot/loader_test.go | 3 ++- 17 files changed, 67 insertions(+), 41 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/copy_up_test.go b/pkg/sentry/fs/copy_up_test.go index 64f030f72..fcba14ed4 100644 --- a/pkg/sentry/fs/copy_up_test.go +++ b/pkg/sentry/fs/copy_up_test.go @@ -166,7 +166,8 @@ func makeOverlayTestFiles(t *testing.T) []*overlayTestFile { // Walk to all of the files in the overlay, open them readable. for _, f := range files { - d, err := mns.FindInode(ctx, mns.Root(), mns.Root(), f.name, 0) + maxTraversals := uint(0) + d, err := mns.FindInode(ctx, mns.Root(), mns.Root(), f.name, &maxTraversals) if err != nil { t.Fatalf("failed to find %q: %v", f.name, err) } diff --git a/pkg/sentry/fs/host/fs.go b/pkg/sentry/fs/host/fs.go index fec890964..54cbb94f9 100644 --- a/pkg/sentry/fs/host/fs.go +++ b/pkg/sentry/fs/host/fs.go @@ -170,7 +170,8 @@ func installWhitelist(ctx context.Context, m *fs.MountNamespace, paths []string) current := paths[i][:j] // Lookup the given component in the tree. - d, err := m.FindLink(ctx, root, nil, current, maxTraversals) + remainingTraversals := uint(maxTraversals) + d, err := m.FindLink(ctx, root, nil, current, &remainingTraversals) if err != nil { log.Warningf("populate failed for %q: %v", current, err) continue diff --git a/pkg/sentry/fs/host/fs_test.go b/pkg/sentry/fs/host/fs_test.go index e69559aac..44db61ecd 100644 --- a/pkg/sentry/fs/host/fs_test.go +++ b/pkg/sentry/fs/host/fs_test.go @@ -150,7 +150,8 @@ func allPaths(ctx context.Context, t *testing.T, m *fs.MountNamespace, base stri root := m.Root() defer root.DecRef() - d, err := m.FindLink(ctx, root, nil, base, 1) + maxTraversals := uint(1) + d, err := m.FindLink(ctx, root, nil, base, &maxTraversals) if err != nil { t.Logf("FindLink failed for %q", base) return paths, err diff --git a/pkg/sentry/fs/inode_overlay_test.go b/pkg/sentry/fs/inode_overlay_test.go index bba20da14..acdb2b4f8 100644 --- a/pkg/sentry/fs/inode_overlay_test.go +++ b/pkg/sentry/fs/inode_overlay_test.go @@ -324,7 +324,8 @@ func TestCacheFlush(t *testing.T) { for _, fileName := range []string{upperFileName, lowerFileName} { // Walk to the file. - dirent, err := mns.FindInode(ctx, root, nil, fileName, 0) + maxTraversals := uint(0) + dirent, err := mns.FindInode(ctx, root, nil, fileName, &maxTraversals) if err != nil { t.Fatalf("FindInode(%q) failed: %v", fileName, err) } diff --git a/pkg/sentry/fs/mount_test.go b/pkg/sentry/fs/mount_test.go index a1c9f4f79..269d6b9da 100644 --- a/pkg/sentry/fs/mount_test.go +++ b/pkg/sentry/fs/mount_test.go @@ -115,8 +115,10 @@ func TestMountSourceParentChildRelationship(t *testing.T) { "/waldo", } + var maxTraversals uint for _, p := range paths { - d, err := mm.FindLink(ctx, rootDirent, nil, p, 0) + maxTraversals = 0 + d, err := mm.FindLink(ctx, rootDirent, nil, p, &maxTraversals) if err != nil { t.Fatalf("could not find path %q in mount manager: %v", p, err) } @@ -164,7 +166,8 @@ func TestMountSourceParentChildRelationship(t *testing.T) { } // "foo" mount should have two children: /foo/bar, and /foo/qux. - d, err := mm.FindLink(ctx, rootDirent, nil, "/foo", 0) + maxTraversals = 0 + d, err := mm.FindLink(ctx, rootDirent, nil, "/foo", &maxTraversals) if err != nil { t.Fatalf("could not find path %q in mount manager: %v", "/foo", err) } @@ -185,7 +188,8 @@ func TestMountSourceParentChildRelationship(t *testing.T) { } // "waldo" mount should have no submounts or children. - waldo, err := mm.FindLink(ctx, rootDirent, nil, "/waldo", 0) + maxTraversals = 0 + waldo, err := mm.FindLink(ctx, rootDirent, nil, "/waldo", &maxTraversals) if err != nil { t.Fatalf("could not find path %q in mount manager: %v", "/waldo", err) } diff --git a/pkg/sentry/fs/mounts.go b/pkg/sentry/fs/mounts.go index 7c5348cce..f6f7be0aa 100644 --- a/pkg/sentry/fs/mounts.go +++ b/pkg/sentry/fs/mounts.go @@ -350,7 +350,7 @@ func (mns *MountNamespace) Unmount(ctx context.Context, node *Dirent, detachOnly // // Precondition: root must be non-nil. // Precondition: the path must be non-empty. -func (mns *MountNamespace) FindLink(ctx context.Context, root, wd *Dirent, path string, maxTraversals uint) (*Dirent, error) { +func (mns *MountNamespace) FindLink(ctx context.Context, root, wd *Dirent, path string, remainingTraversals *uint) (*Dirent, error) { if root == nil { panic("MountNamespace.FindLink: root must not be nil") } @@ -419,7 +419,7 @@ func (mns *MountNamespace) FindLink(ctx context.Context, root, wd *Dirent, path // // See resolve for reference semantics; on err next // will have one dropped. - current, err = mns.resolve(ctx, root, next, maxTraversals) + current, err = mns.resolve(ctx, root, next, remainingTraversals) if err != nil { return nil, err } @@ -439,15 +439,15 @@ func (mns *MountNamespace) FindLink(ctx context.Context, root, wd *Dirent, path // FindInode is identical to FindLink except the return value is resolved. // //go:nosplit -func (mns *MountNamespace) FindInode(ctx context.Context, root, wd *Dirent, path string, maxTraversals uint) (*Dirent, error) { - d, err := mns.FindLink(ctx, root, wd, path, maxTraversals) +func (mns *MountNamespace) FindInode(ctx context.Context, root, wd *Dirent, path string, remainingTraversals *uint) (*Dirent, error) { + d, err := mns.FindLink(ctx, root, wd, path, remainingTraversals) if err != nil { return nil, err } // See resolve for reference semantics; on err d will have the // reference dropped. - return mns.resolve(ctx, root, d, maxTraversals) + return mns.resolve(ctx, root, d, remainingTraversals) } // resolve resolves the given link. @@ -458,14 +458,14 @@ func (mns *MountNamespace) FindInode(ctx context.Context, root, wd *Dirent, path // If not successful, a reference is _also_ dropped on the node and an error // returned. This is for convenience in using resolve directly as a return // value. -func (mns *MountNamespace) resolve(ctx context.Context, root, node *Dirent, maxTraversals uint) (*Dirent, error) { +func (mns *MountNamespace) resolve(ctx context.Context, root, node *Dirent, remainingTraversals *uint) (*Dirent, error) { // Resolve the path. target, err := node.Inode.Getlink(ctx) switch err { case nil: // Make sure we didn't exhaust the traversal budget. - if maxTraversals == 0 { + if *remainingTraversals == 0 { target.DecRef() return nil, syscall.ELOOP } @@ -481,7 +481,7 @@ func (mns *MountNamespace) resolve(ctx context.Context, root, node *Dirent, maxT defer node.DecRef() // See above. // First, check if we should traverse. - if maxTraversals == 0 { + if *remainingTraversals == 0 { return nil, syscall.ELOOP } @@ -492,7 +492,8 @@ func (mns *MountNamespace) resolve(ctx context.Context, root, node *Dirent, maxT } // Find the node; we resolve relative to the current symlink's parent. - d, err := mns.FindInode(ctx, root, node.parent, targetPath, maxTraversals-1) + *remainingTraversals-- + d, err := mns.FindInode(ctx, root, node.parent, targetPath, remainingTraversals) if err != nil { return nil, err } @@ -544,7 +545,8 @@ func (mns *MountNamespace) ResolveExecutablePath(ctx context.Context, wd, name s defer root.DecRef() for _, p := range paths { binPath := path.Join(p, name) - d, err := mns.FindInode(ctx, root, nil, binPath, linux.MaxSymlinkTraversals) + traversals := uint(linux.MaxSymlinkTraversals) + d, err := mns.FindInode(ctx, root, nil, binPath, &traversals) if err == syserror.ENOENT || err == syserror.EACCES { // Didn't find it here. continue diff --git a/pkg/sentry/fs/mounts_test.go b/pkg/sentry/fs/mounts_test.go index cc7c32c9b..2f7a1710f 100644 --- a/pkg/sentry/fs/mounts_test.go +++ b/pkg/sentry/fs/mounts_test.go @@ -77,7 +77,8 @@ func TestFindLink(t *testing.T) { {"bar", foo, "/foo/bar"}, } { wdPath, _ := tc.wd.FullName(root) - if d, err := mm.FindLink(ctx, root, tc.wd, tc.findPath, 0); err != nil { + maxTraversals := uint(0) + if d, err := mm.FindLink(ctx, root, tc.wd, tc.findPath, &maxTraversals); err != nil { t.Errorf("FindLink(%q, wd=%q) failed: %v", tc.findPath, wdPath, err) } else if got, _ := d.FullName(root); got != tc.wantPath { t.Errorf("FindLink(%q, wd=%q) got dirent %q, want %q", tc.findPath, wdPath, got, tc.wantPath) @@ -95,7 +96,8 @@ func TestFindLink(t *testing.T) { {"foo", foo}, } { wdPath, _ := tc.wd.FullName(root) - if _, err := mm.FindLink(ctx, root, tc.wd, tc.findPath, 0); err == nil { + maxTraversals := uint(0) + if _, err := mm.FindLink(ctx, root, tc.wd, tc.findPath, &maxTraversals); err == nil { t.Errorf("FindLink(%q, wd=%q) did not return error", tc.findPath, wdPath) } } diff --git a/pkg/sentry/fs/ramfs/tree_test.go b/pkg/sentry/fs/ramfs/tree_test.go index d5567d9e1..54df2143c 100644 --- a/pkg/sentry/fs/ramfs/tree_test.go +++ b/pkg/sentry/fs/ramfs/tree_test.go @@ -70,7 +70,8 @@ func TestMakeDirectoryTree(t *testing.T) { defer mm.DecRef() for _, p := range test.subdirs { - if _, err := mm.FindInode(ctx, root, nil, p, 0); err != nil { + maxTraversals := uint(0) + if _, err := mm.FindInode(ctx, root, nil, p, &maxTraversals); err != nil { t.Errorf("%s: failed to find node %s: %v", test.name, p, err) break } diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 17425e656..cb61e27f1 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -634,10 +634,11 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, ThreadID, args.Root = nil // Grab the working directory. + remainingTraversals := uint(args.MaxSymlinkTraversals) wd := root // Default. if args.WorkingDirectory != "" { var err error - wd, err = k.mounts.FindInode(ctx, root, nil, args.WorkingDirectory, args.MaxSymlinkTraversals) + wd, err = k.mounts.FindInode(ctx, root, nil, args.WorkingDirectory, &remainingTraversals) if err != nil { return nil, 0, fmt.Errorf("failed to find initial working directory %q: %v", args.WorkingDirectory, err) } @@ -656,7 +657,8 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, ThreadID, } // Create a fresh task context. - tc, err := k.LoadTaskImage(ctx, k.mounts, root, wd, args.MaxSymlinkTraversals, args.Filename, args.Argv, args.Envv, k.featureSet) + remainingTraversals = uint(args.MaxSymlinkTraversals) + tc, err := k.LoadTaskImage(ctx, k.mounts, root, wd, &remainingTraversals, args.Filename, args.Argv, args.Envv, k.featureSet) if err != nil { return nil, 0, err } diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index 45b8d2b04..aaff309f0 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -142,7 +142,7 @@ func (t *Task) Stack() *arch.Stack { // * argv: Binary argv // * envv: Binary envv // * fs: Binary FeatureSet -func (k *Kernel) LoadTaskImage(ctx context.Context, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals uint, filename string, argv, envv []string, fs *cpuid.FeatureSet) (*TaskContext, error) { +func (k *Kernel) LoadTaskImage(ctx context.Context, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, filename string, argv, envv []string, fs *cpuid.FeatureSet) (*TaskContext, error) { // Prepare a new user address space to load into. m := mm.NewMemoryManager(k) defer m.DecUsers(ctx) diff --git a/pkg/sentry/loader/elf.go b/pkg/sentry/loader/elf.go index 9b1e81dc9..385ad0102 100644 --- a/pkg/sentry/loader/elf.go +++ b/pkg/sentry/loader/elf.go @@ -610,7 +610,7 @@ func loadInterpreterELF(ctx context.Context, m *mm.MemoryManager, f *fs.File, in // // Preconditions: // * f is an ELF file -func loadELF(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals uint, fs *cpuid.FeatureSet, f *fs.File) (loadedELF, arch.Context, error) { +func loadELF(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, fs *cpuid.FeatureSet, f *fs.File) (loadedELF, arch.Context, error) { bin, ac, err := loadInitialELF(ctx, m, fs, f) if err != nil { ctx.Infof("Error loading binary: %v", err) diff --git a/pkg/sentry/loader/loader.go b/pkg/sentry/loader/loader.go index d1417c4f1..69a090844 100644 --- a/pkg/sentry/loader/loader.go +++ b/pkg/sentry/loader/loader.go @@ -55,7 +55,7 @@ func readFull(ctx context.Context, f *fs.File, dst usermem.IOSequence, offset in // installed in the Task FDMap. The caller takes ownership of both. // // name must be a readable, executable, regular file. -func openPath(ctx context.Context, mm *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals uint, name string) (*fs.Dirent, *fs.File, error) { +func openPath(ctx context.Context, mm *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, name string) (*fs.Dirent, *fs.File, error) { if name == "" { ctx.Infof("cannot open empty name") return nil, nil, syserror.ENOENT @@ -136,9 +136,9 @@ const ( // * arch.Context matching the binary arch // * fs.Dirent of the binary file // * Possibly updated argv -func loadPath(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals uint, fs *cpuid.FeatureSet, filename string, argv, envv []string) (loadedELF, arch.Context, *fs.Dirent, []string, error) { +func loadPath(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, root, wd *fs.Dirent, remainingTraversals *uint, fs *cpuid.FeatureSet, filename string, argv, envv []string) (loadedELF, arch.Context, *fs.Dirent, []string, error) { for i := 0; i < maxLoaderAttempts; i++ { - d, f, err := openPath(ctx, mounts, root, wd, maxTraversals, filename) + d, f, err := openPath(ctx, mounts, root, wd, remainingTraversals, filename) if err != nil { ctx.Infof("Error opening %s: %v", filename, err) return loadedELF{}, nil, nil, nil, err @@ -163,7 +163,7 @@ func loadPath(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespac switch { case bytes.Equal(hdr[:], []byte(elfMagic)): - loaded, ac, err := loadELF(ctx, m, mounts, root, wd, maxTraversals, fs, f) + loaded, ac, err := loadELF(ctx, m, mounts, root, wd, remainingTraversals, fs, f) if err != nil { ctx.Infof("Error loading ELF: %v", err) return loadedELF{}, nil, nil, nil, err @@ -196,7 +196,7 @@ func loadPath(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespac // Preconditions: // * The Task MemoryManager is empty. // * Load is called on the Task goroutine. -func Load(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals uint, fs *cpuid.FeatureSet, filename string, argv, envv []string, extraAuxv []arch.AuxEntry, vdso *VDSO) (abi.OS, arch.Context, string, error) { +func Load(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, fs *cpuid.FeatureSet, filename string, argv, envv []string, extraAuxv []arch.AuxEntry, vdso *VDSO) (abi.OS, arch.Context, string, error) { // Load the binary itself. loaded, ac, d, argv, err := loadPath(ctx, m, mounts, root, wd, maxTraversals, fs, filename, argv, envv) if err != nil { diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 334169372..4379486cf 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -266,7 +266,8 @@ func (s *SocketOperations) Bind(t *kernel.Task, sockaddr []byte) *syserr.Error { subPath = "/" } var err error - d, err = t.MountNamespace().FindInode(t, root, cwd, subPath, fs.DefaultTraversalLimit) + remainingTraversals := uint(fs.DefaultTraversalLimit) + d, err = t.MountNamespace().FindInode(t, root, cwd, subPath, &remainingTraversals) if err != nil { // No path available. return syserr.ErrNoSuchFile @@ -314,7 +315,8 @@ func extractEndpoint(t *kernel.Task, sockaddr []byte) (transport.BoundEndpoint, // Find the node in the filesystem. root := t.FSContext().RootDirectory() cwd := t.FSContext().WorkingDirectory() - d, e := t.MountNamespace().FindInode(t, root, cwd, path, fs.DefaultTraversalLimit) + remainingTraversals := uint(fs.DefaultTraversalLimit) + d, e := t.MountNamespace().FindInode(t, root, cwd, path, &remainingTraversals) cwd.DecRef() root.DecRef() if e != nil { diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 89d21dd98..37c90f6fd 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -92,10 +92,11 @@ func fileOpOn(t *kernel.Task, dirFD kdefs.FD, path string, resolve bool, fn func root := t.FSContext().RootDirectory() // Lookup the node. + remainingTraversals := uint(linux.MaxSymlinkTraversals) if resolve { - d, err = t.MountNamespace().FindInode(t, root, rel, path, linux.MaxSymlinkTraversals) + d, err = t.MountNamespace().FindInode(t, root, rel, path, &remainingTraversals) } else { - d, err = t.MountNamespace().FindLink(t, root, rel, path, linux.MaxSymlinkTraversals) + d, err = t.MountNamespace().FindLink(t, root, rel, path, &remainingTraversals) } root.DecRef() if wd != nil { @@ -312,7 +313,8 @@ func createAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, flags uint, mod fileFlags.LargeFile = true // Does this file exist already? - targetDirent, err := t.MountNamespace().FindInode(t, root, d, name, linux.MaxSymlinkTraversals) + remainingTraversals := uint(linux.MaxSymlinkTraversals) + targetDirent, err := t.MountNamespace().FindInode(t, root, d, name, &remainingTraversals) var newFile *fs.File switch err { case nil: @@ -997,7 +999,8 @@ func mkdirAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, mode linux.FileM } // Does this directory exist already? - f, err := t.MountNamespace().FindInode(t, root, d, name, linux.MaxSymlinkTraversals) + remainingTraversals := uint(linux.MaxSymlinkTraversals) + f, err := t.MountNamespace().FindInode(t, root, d, name, &remainingTraversals) switch err { case nil: // The directory existed. diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index 9eed613a1..c12693ee2 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -103,7 +103,8 @@ func Execve(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal defer wd.DecRef() // Load the new TaskContext. - tc, err := t.Kernel().LoadTaskImage(t, t.MountNamespace(), root, wd, linux.MaxSymlinkTraversals, filename, argv, envv, t.Arch().FeatureSet()) + maxTraversals := uint(linux.MaxSymlinkTraversals) + tc, err := t.Kernel().LoadTaskImage(t, t.MountNamespace(), root, wd, &maxTraversals, filename, argv, envv, t.Arch().FeatureSet()) if err != nil { return 0, nil, err } diff --git a/runsc/boot/fs.go b/runsc/boot/fs.go index 1e355fe4e..1e75b0efc 100644 --- a/runsc/boot/fs.go +++ b/runsc/boot/fs.go @@ -338,7 +338,8 @@ func mountSubmount(ctx context.Context, conf *Config, mns *fs.MountNamespace, ro } } - dirent, err := mns.FindInode(ctx, root, root, m.Destination, 0 /* maxTraversals */) + maxTraversals := uint(0) + dirent, err := mns.FindInode(ctx, root, root, m.Destination, &maxTraversals) if err != nil { return fmt.Errorf("failed to find mount destination %q: %v", m.Destination, err) } @@ -582,7 +583,8 @@ func setupContainerFS(procArgs *kernel.CreateProcessArgs, spec *specs.Spec, conf defer globalRoot.DecRef() // Create mount point for the container's rootfs. - contDir, err := mns.FindInode(ctx, globalRoot, nil, ChildContainersDir, 0 /* TraversalLimit */) + maxTraversals := uint(0) + contDir, err := mns.FindInode(ctx, globalRoot, nil, ChildContainersDir, &maxTraversals) if err != nil { return fmt.Errorf("couldn't find child container dir %q: %v", ChildContainersDir, err) } @@ -656,7 +658,8 @@ func destroyContainerFS(ctx context.Context, cid string, k *kernel.Kernel) error mnsRoot := mns.Root() defer mnsRoot.DecRef() containerRoot := path.Join(ChildContainersDir, cid) - containerRootDirent, err := mns.FindInode(ctx, mnsRoot, nil, containerRoot, 0 /* maxTraversals */) + maxTraversals := uint(0) + containerRootDirent, err := mns.FindInode(ctx, mnsRoot, nil, containerRoot, &maxTraversals) if err == syserror.ENOENT { // Container must have been destroyed already. That's fine. return nil @@ -691,7 +694,8 @@ func destroyContainerFS(ctx context.Context, cid string, k *kernel.Kernel) error // Get a reference to the parent directory and remove the root // container directory. - containersDirDirent, err := mns.FindInode(ctx, mnsRoot, nil, ChildContainersDir, 0 /* maxTraversals */) + maxTraversals = 0 + containersDirDirent, err := mns.FindInode(ctx, mnsRoot, nil, ChildContainersDir, &maxTraversals) if err != nil { return fmt.Errorf("error finding containers directory %q: %v", ChildContainersDir, err) } diff --git a/runsc/boot/loader_test.go b/runsc/boot/loader_test.go index d5cee5608..0ed3002e0 100644 --- a/runsc/boot/loader_test.go +++ b/runsc/boot/loader_test.go @@ -406,7 +406,8 @@ func TestCreateMountNamespace(t *testing.T) { root := mm.Root() defer root.DecRef() for _, p := range tc.expectedPaths { - if d, err := mm.FindInode(ctx, root, root, p, 0); err != nil { + maxTraversals := uint(0) + if d, err := mm.FindInode(ctx, root, root, p, &maxTraversals); err != nil { t.Errorf("expected path %v to exist with spec %v, but got error %v", p, tc.spec, err) } else { d.DecRef() -- cgit v1.2.3 From 685eaf119ffa6c44c4dcaec0e083bbdc0271231a Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Thu, 6 Dec 2018 11:14:57 -0800 Subject: Add counters for memory events. Also ensure an event is emitted at startup. PiperOrigin-RevId: 224372065 Change-Id: I5f642b6d6b13c6468ee8f794effe285fcbbf29cf --- pkg/sentry/kernel/memevent/BUILD | 1 + pkg/sentry/kernel/memevent/memory_events.go | 10 ++++++++++ 2 files changed, 11 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/memevent/BUILD b/pkg/sentry/kernel/memevent/BUILD index e903badd3..dfd8dd062 100644 --- a/pkg/sentry/kernel/memevent/BUILD +++ b/pkg/sentry/kernel/memevent/BUILD @@ -12,6 +12,7 @@ go_library( ":memory_events_go_proto", "//pkg/eventchannel", "//pkg/log", + "//pkg/metric", "//pkg/sentry/kernel", "//pkg/sentry/usage", ], diff --git a/pkg/sentry/kernel/memevent/memory_events.go b/pkg/sentry/kernel/memevent/memory_events.go index f05ef1b64..1a8e86827 100644 --- a/pkg/sentry/kernel/memevent/memory_events.go +++ b/pkg/sentry/kernel/memevent/memory_events.go @@ -22,11 +22,15 @@ import ( "gvisor.googlesource.com/gvisor/pkg/eventchannel" "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/metric" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" pb "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/memevent/memory_events_go_proto" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" ) +var totalTicks = metric.MustCreateNewUint64Metric("/memory_events/ticks", false /*sync*/, "Total number of memory event periods that have elapsed since startup.") +var totalEvents = metric.MustCreateNewUint64Metric("/memory_events/events", false /*sync*/, "Total number of memory events emitted.") + // MemoryEvents describes the configuration for the global memory event emitter. type MemoryEvents struct { k *kernel.Kernel @@ -71,6 +75,10 @@ func (m *MemoryEvents) Start() { func (m *MemoryEvents) run() { m.done.Add(1) + // Emit the first event immediately on startup. + totalTicks.Increment() + m.emit() + ticker := time.NewTicker(m.period) defer ticker.Stop() @@ -80,6 +88,7 @@ func (m *MemoryEvents) run() { m.done.Done() return case <-ticker.C: + totalTicks.Increment() m.emit() } } @@ -94,6 +103,7 @@ func (m *MemoryEvents) emit() { snapshot, _ := usage.MemoryAccounting.Copy() total := totalPlatform + snapshot.Mapped + totalEvents.Increment() eventchannel.Emit(&pb.MemoryUsageEvent{ Mapped: snapshot.Mapped, Total: total, -- cgit v1.2.3 From 99d595869332f817de8f570fae184658c513a43c Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Mon, 10 Dec 2018 12:36:27 -0800 Subject: Validate FS_BASE in Task.Clone arch_prctl already verified that the new FS_BASE was canonical, but Task.Clone did not. Centralize these checks in the arch packages. Failure to validate could cause an error in PTRACE_SET_REGS when we try to switch to the app. PiperOrigin-RevId: 224862398 Change-Id: Iefe63b3f9aa6c4810326b8936e501be3ec407f14 --- pkg/sentry/arch/arch.go | 6 ++++++ pkg/sentry/arch/arch_amd64.go | 16 ++++++++++++++++ pkg/sentry/arch/arch_x86.go | 10 ++++++++-- pkg/sentry/kernel/task_clone.go | 4 +++- pkg/sentry/platform/ptrace/subprocess.go | 4 ++-- pkg/sentry/syscalls/linux/sys_tls.go | 9 +++------ 6 files changed, 38 insertions(+), 11 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go index 575b7ba66..4cd7a9af5 100644 --- a/pkg/sentry/arch/arch.go +++ b/pkg/sentry/arch/arch.go @@ -115,6 +115,12 @@ type Context interface { // SetStack sets the current stack pointer. SetStack(value uintptr) + // TLS returns the current TLS pointer. + TLS() uintptr + + // SetTLS sets the current TLS pointer. Returns false if value is invalid. + SetTLS(value uintptr) bool + // SetRSEQInterruptedIP sets the register that contains the old IP when a // restartable sequence is interrupted. SetRSEQInterruptedIP(value uintptr) diff --git a/pkg/sentry/arch/arch_amd64.go b/pkg/sentry/arch/arch_amd64.go index bb80a7bed..2507774f7 100644 --- a/pkg/sentry/arch/arch_amd64.go +++ b/pkg/sentry/arch/arch_amd64.go @@ -158,6 +158,22 @@ func (c *context64) SetStack(value uintptr) { c.Regs.Rsp = uint64(value) } +// TLS returns the current TLS pointer. +func (c *context64) TLS() uintptr { + return uintptr(c.Regs.Fs_base) +} + +// SetTLS sets the current TLS pointer. Returns false if value is invalid. +func (c *context64) SetTLS(value uintptr) bool { + if !isValidSegmentBase(uint64(value)) { + return false + } + + c.Regs.Fs = 0 + c.Regs.Fs_base = uint64(value) + return true +} + // SetRSEQInterruptedIP implements Context.SetRSEQInterruptedIP. func (c *context64) SetRSEQInterruptedIP(value uintptr) { c.Regs.R10 = uint64(value) diff --git a/pkg/sentry/arch/arch_x86.go b/pkg/sentry/arch/arch_x86.go index 59bf89d99..e50a76083 100644 --- a/pkg/sentry/arch/arch_x86.go +++ b/pkg/sentry/arch/arch_x86.go @@ -353,10 +353,10 @@ func (s *State) PtraceSetRegs(src io.Reader) (int, error) { if !isUserSegmentSelector(regs.Ss) { return 0, syscall.EIO } - if regs.Fs_base >= uint64(maxAddr64) { + if !isValidSegmentBase(regs.Fs_base) { return 0, syscall.EIO } - if regs.Gs_base >= uint64(maxAddr64) { + if !isValidSegmentBase(regs.Gs_base) { return 0, syscall.EIO } // CS and SS are validated, but changes to them are otherwise silently @@ -389,6 +389,12 @@ func isUserSegmentSelector(reg uint64) bool { return reg&3 == 3 } +// isValidSegmentBase returns true if the given segment base specifies a +// canonical user address. +func isValidSegmentBase(reg uint64) bool { + return reg < uint64(maxAddr64) +} + // ptraceFPRegsSize is the size in bytes of Linux's user_i387_struct, the type // manipulated by PTRACE_GETFPREGS and PTRACE_SETFPREGS on x86. Equivalently, // ptraceFPRegsSize is the size in bytes of the x86 FXSAVE area. diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 755fe0370..b66fa34a9 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -210,7 +210,9 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { tc.Arch.SetStack(uintptr(opts.Stack)) } if opts.SetTLS { - tc.Arch.StateData().Regs.Fs_base = uint64(opts.TLS) + if !tc.Arch.SetTLS(uintptr(opts.TLS)) { + return 0, nil, syserror.EPERM + } } var fsc *FSContext diff --git a/pkg/sentry/platform/ptrace/subprocess.go b/pkg/sentry/platform/ptrace/subprocess.go index 5e56a1514..a9d083f5a 100644 --- a/pkg/sentry/platform/ptrace/subprocess.go +++ b/pkg/sentry/platform/ptrace/subprocess.go @@ -480,10 +480,10 @@ func (s *subprocess) switchToApp(c *context, ac arch.Context) bool { // Set registers. if err := t.setRegs(regs); err != nil { - panic(fmt.Sprintf("ptrace set regs failed: %v", err)) + panic(fmt.Sprintf("ptrace set regs (%+v) failed: %v", regs, err)) } if err := t.setFPRegs(fpState, uint64(fpLen), useXsave); err != nil { - panic(fmt.Sprintf("ptrace set fpregs failed: %v", err)) + panic(fmt.Sprintf("ptrace set fpregs (%+v) failed: %v", fpState, err)) } for { diff --git a/pkg/sentry/syscalls/linux/sys_tls.go b/pkg/sentry/syscalls/linux/sys_tls.go index 40e84825b..8ea78093b 100644 --- a/pkg/sentry/syscalls/linux/sys_tls.go +++ b/pkg/sentry/syscalls/linux/sys_tls.go @@ -22,7 +22,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) // ArchPrctl implements linux syscall arch_prctl(2). @@ -31,19 +30,17 @@ func ArchPrctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys switch args[0].Int() { case linux.ARCH_GET_FS: addr := args[1].Pointer() - _, err := t.CopyOut(addr, &t.Arch().StateData().Regs.Fs_base) + fsbase := t.Arch().TLS() + _, err := t.CopyOut(addr, uint64(fsbase)) if err != nil { return 0, nil, err } case linux.ARCH_SET_FS: fsbase := args[1].Uint64() - if _, ok := t.MemoryManager().CheckIORange(usermem.Addr(fsbase), 0); !ok { + if !t.Arch().SetTLS(uintptr(fsbase)) { return 0, nil, syscall.EPERM } - regs := &t.Arch().StateData().Regs - regs.Fs = 0 - regs.Fs_base = fsbase case linux.ARCH_GET_GS, linux.ARCH_SET_GS: t.Kernel().EmitUnimplementedEvent(t) -- cgit v1.2.3 From fc297702511edef4760c4f7a1d89cc6f02347d50 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Mon, 10 Dec 2018 12:47:20 -0800 Subject: Add type safety to shm ids and keys. PiperOrigin-RevId: 224864380 Change-Id: I49542279ad56bf15ba462d3de1ef2b157b31830a --- pkg/sentry/kernel/shm/shm.go | 24 +++++++++++++++--------- pkg/sentry/syscalls/linux/sys_shm.go | 8 ++++---- 2 files changed, 19 insertions(+), 13 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index f760f5f76..4343dee13 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -51,6 +51,12 @@ import ( "gvisor.googlesource.com/gvisor/pkg/syserror" ) +// Key represents a shm segment key. Analogous to a file name. +type Key int32 + +// ID represents the opaque handle for a shm segment. Analogous to an fd. +type ID int32 + // Registry tracks all shared memory segments in an IPC namespace. The registry // provides the mechanisms for creating and finding segments, and reporting // global shm parameters. @@ -63,33 +69,33 @@ type Registry struct { mu sync.Mutex `state:"nosave"` // shms maps segment ids to segments. Protected by mu. - shms map[int32]*Shm + shms map[ID]*Shm // Sum of the sizes of all existing segments rounded up to page size, in // units of page size. Protected by mu. totalPages uint64 // lastIDUsed is protected by mu. - lastIDUsed int32 + lastIDUsed ID } // NewRegistry creates a new shm registry. func NewRegistry(userNS *auth.UserNamespace) *Registry { return &Registry{ userNS: userNS, - shms: make(map[int32]*Shm), + shms: make(map[ID]*Shm), } } // FindByID looks up a segment given an ID. -func (r *Registry) FindByID(id int32) *Shm { +func (r *Registry) FindByID(id ID) *Shm { r.mu.Lock() defer r.mu.Unlock() return r.shms[id] } // Precondition: Caller must hold r.mu. -func (r *Registry) findByKey(key int32) *Shm { +func (r *Registry) findByKey(key Key) *Shm { for _, v := range r.shms { if v.key == key { return v @@ -100,7 +106,7 @@ func (r *Registry) findByKey(key int32) *Shm { // FindOrCreate looks up or creates a segment in the registry. It's functionally // analogous to open(2). -func (r *Registry) FindOrCreate(ctx context.Context, pid, key int32, size uint64, mode linux.FileMode, private, create, exclusive bool) (*Shm, error) { +func (r *Registry) FindOrCreate(ctx context.Context, pid int32, key Key, size uint64, mode linux.FileMode, private, create, exclusive bool) (*Shm, error) { if (create || private) && (size < linux.SHMMIN || size > linux.SHMMAX) { // "A new segment was to be created and size is less than SHMMIN or // greater than SHMMAX." - man shmget(2) @@ -178,7 +184,7 @@ func (r *Registry) FindOrCreate(ctx context.Context, pid, key int32, size uint64 } // newShm creates a new segment in the registry. -func (r *Registry) newShm(ctx context.Context, pid, key int32, creator fs.FileOwner, perms fs.FilePermissions, size uint64) (*Shm, error) { +func (r *Registry) newShm(ctx context.Context, pid int32, key Key, creator fs.FileOwner, perms fs.FilePermissions, size uint64) (*Shm, error) { p := platform.FromContext(ctx) if p == nil { panic(fmt.Sprintf("context.Context %T lacks non-nil value for key %T", ctx, platform.CtxPlatform)) @@ -289,7 +295,7 @@ type Shm struct { registry *Registry // ID is the kernel identifier for this segment. Immutable. - ID int32 + ID ID // creator is the user that created the segment. Immutable. creator fs.FileOwner @@ -309,7 +315,7 @@ type Shm struct { fr platform.FileRange // key is the public identifier for this segment. - key int32 + key Key // mu protects all fields below. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/syscalls/linux/sys_shm.go b/pkg/sentry/syscalls/linux/sys_shm.go index 8753c2e58..a0d3a73c5 100644 --- a/pkg/sentry/syscalls/linux/sys_shm.go +++ b/pkg/sentry/syscalls/linux/sys_shm.go @@ -24,7 +24,7 @@ import ( // Shmget implements shmget(2). func Shmget(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { - key := args[0].Int() + key := shm.Key(args[0].Int()) size := uint64(args[1].SizeT()) flag := args[2].Int() @@ -43,7 +43,7 @@ func Shmget(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal } // findSegment retrives a shm segment by the given id. -func findSegment(t *kernel.Task, id int32) (*shm.Shm, error) { +func findSegment(t *kernel.Task, id shm.ID) (*shm.Shm, error) { r := t.IPCNamespace().ShmRegistry() segment := r.FindByID(id) if segment == nil { @@ -55,7 +55,7 @@ func findSegment(t *kernel.Task, id int32) (*shm.Shm, error) { // Shmat implements shmat(2). func Shmat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { - id := args[0].Int() + id := shm.ID(args[0].Int()) addr := args[1].Pointer() flag := args[2].Int() @@ -86,7 +86,7 @@ func Shmdt(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall // Shmctl implements shmctl(2). func Shmctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { - id := args[0].Int() + id := shm.ID(args[0].Int()) cmd := args[1].Int() buf := args[2].Pointer() -- cgit v1.2.3 From 75e39eaa74c65b6f7cfb95addb6ac0cbcc7d951a Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Wed, 12 Dec 2018 13:09:10 -0800 Subject: Pass information about map writableness to filesystems. This is necessary to implement file seals for memfds. PiperOrigin-RevId: 225239394 Change-Id: Ib3f1ab31385afc4b24e96cd81a05ef1bebbcbb70 --- pkg/sentry/fs/binder/binder.go | 6 +- pkg/sentry/fs/copy_up.go | 8 +-- pkg/sentry/fs/fsutil/inode_cached.go | 12 ++-- pkg/sentry/fs/fsutil/inode_cached_test.go | 8 +-- pkg/sentry/fs/overlay.go | 18 +++--- pkg/sentry/fs/tmpfs/inode_file.go | 12 ++-- pkg/sentry/kernel/shm/shm.go | 6 +- pkg/sentry/memmap/mapping_set.go | 18 ++++-- pkg/sentry/memmap/mapping_set_test.go | 102 ++++++++++++++++++++++++++---- pkg/sentry/memmap/memmap.go | 17 +++-- pkg/sentry/mm/aio_context.go | 6 +- pkg/sentry/mm/lifecycle.go | 2 +- pkg/sentry/mm/mm.go | 4 ++ pkg/sentry/mm/special_mappable.go | 6 +- pkg/sentry/mm/syscalls.go | 4 +- pkg/sentry/mm/vma.go | 4 +- 16 files changed, 162 insertions(+), 71 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index 42b9e8b26..e642c7f22 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -302,7 +302,7 @@ func (bp *Proc) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArgum } // AddMapping implements memmap.Mappable.AddMapping. -func (bp *Proc) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) error { +func (bp *Proc) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64, _ bool) error { bp.mu.Lock() defer bp.mu.Unlock() if bp.mapped.Length() != 0 { @@ -320,12 +320,12 @@ func (bp *Proc) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar userm } // RemoveMapping implements memmap.Mappable.RemoveMapping. -func (bp *Proc) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) { +func (*Proc) RemoveMapping(context.Context, memmap.MappingSpace, usermem.AddrRange, uint64, bool) { // Nothing to do. Notably, we don't free bp.mapped to allow another mmap. } // CopyMapping implements memmap.Mappable.CopyMapping. -func (bp *Proc) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error { +func (bp *Proc) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64, _ bool) error { // Nothing to do. Notably, this is one case where CopyMapping isn't // equivalent to AddMapping, as AddMapping would return EBUSY. return nil diff --git a/pkg/sentry/fs/copy_up.go b/pkg/sentry/fs/copy_up.go index d65dc74bf..6d4ebaaa4 100644 --- a/pkg/sentry/fs/copy_up.go +++ b/pkg/sentry/fs/copy_up.go @@ -270,13 +270,13 @@ func copyUpLocked(ctx context.Context, parent *Dirent, next *Dirent) error { for seg := next.Inode.overlay.mappings.FirstSegment(); seg.Ok(); seg = seg.NextSegment() { added := make(memmap.MappingsOfRange) for m := range seg.Value() { - if err := upperMappable.AddMapping(ctx, m.MappingSpace, m.AddrRange, seg.Start()); err != nil { + if err := upperMappable.AddMapping(ctx, m.MappingSpace, m.AddrRange, seg.Start(), m.Writable); err != nil { for m := range added { - upperMappable.RemoveMapping(ctx, m.MappingSpace, m.AddrRange, seg.Start()) + upperMappable.RemoveMapping(ctx, m.MappingSpace, m.AddrRange, seg.Start(), m.Writable) } for mr, mappings := range allAdded { for m := range mappings { - upperMappable.RemoveMapping(ctx, m.MappingSpace, m.AddrRange, mr.Start) + upperMappable.RemoveMapping(ctx, m.MappingSpace, m.AddrRange, mr.Start, m.Writable) } } return err @@ -301,7 +301,7 @@ func copyUpLocked(ctx context.Context, parent *Dirent, next *Dirent) error { if lowerMappable != nil { for seg := next.Inode.overlay.mappings.FirstSegment(); seg.Ok(); seg = seg.NextSegment() { for m := range seg.Value() { - lowerMappable.RemoveMapping(ctx, m.MappingSpace, m.AddrRange, seg.Start()) + lowerMappable.RemoveMapping(ctx, m.MappingSpace, m.AddrRange, seg.Start(), m.Writable) } } } diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index b0af44ddd..707ca76d2 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -686,10 +686,10 @@ func (rw *inodeReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error } // AddMapping implements memmap.Mappable.AddMapping. -func (c *CachingInodeOperations) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) error { +func (c *CachingInodeOperations) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) error { // Hot path. Avoid defers. c.mapsMu.Lock() - mapped := c.mappings.AddMapping(ms, ar, offset) + mapped := c.mappings.AddMapping(ms, ar, offset, writable) // Do this unconditionally since whether we have c.backingFile.FD() >= 0 // can change across save/restore. for _, r := range mapped { @@ -705,10 +705,10 @@ func (c *CachingInodeOperations) AddMapping(ctx context.Context, ms memmap.Mappi } // RemoveMapping implements memmap.Mappable.RemoveMapping. -func (c *CachingInodeOperations) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) { +func (c *CachingInodeOperations) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) { // Hot path. Avoid defers. c.mapsMu.Lock() - unmapped := c.mappings.RemoveMapping(ms, ar, offset) + unmapped := c.mappings.RemoveMapping(ms, ar, offset, writable) for _, r := range unmapped { c.hostFileMapper.DecRefOn(r) } @@ -739,8 +739,8 @@ func (c *CachingInodeOperations) RemoveMapping(ctx context.Context, ms memmap.Ma } // CopyMapping implements memmap.Mappable.CopyMapping. -func (c *CachingInodeOperations) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error { - return c.AddMapping(ctx, ms, dstAR, offset) +func (c *CachingInodeOperations) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64, writable bool) error { + return c.AddMapping(ctx, ms, dstAR, offset, writable) } // Translate implements memmap.Mappable.Translate. diff --git a/pkg/sentry/fs/fsutil/inode_cached_test.go b/pkg/sentry/fs/fsutil/inode_cached_test.go index e388ec3d7..ce5201a40 100644 --- a/pkg/sentry/fs/fsutil/inode_cached_test.go +++ b/pkg/sentry/fs/fsutil/inode_cached_test.go @@ -305,7 +305,7 @@ func TestRead(t *testing.T) { // be cached. var ms noopMappingSpace ar := usermem.AddrRange{usermem.PageSize, 2 * usermem.PageSize} - if err := iops.AddMapping(ctx, ms, ar, usermem.PageSize); err != nil { + if err := iops.AddMapping(ctx, ms, ar, usermem.PageSize, true); err != nil { t.Fatalf("AddMapping got %v, want nil", err) } mr := memmap.MappableRange{usermem.PageSize, 2 * usermem.PageSize} @@ -334,7 +334,7 @@ func TestRead(t *testing.T) { // Delete the memory mapping and expect it to cause the cached page to be // uncached. - iops.RemoveMapping(ctx, ms, ar, usermem.PageSize) + iops.RemoveMapping(ctx, ms, ar, usermem.PageSize, true) if cached := iops.cache.Span(); cached != 0 { t.Fatalf("Span got %d, want 0", cached) } @@ -363,10 +363,10 @@ func TestWrite(t *testing.T) { // Translate to force them to be cached. var ms noopMappingSpace ar := usermem.AddrRange{usermem.PageSize, 3 * usermem.PageSize} - if err := iops.AddMapping(ctx, ms, ar, usermem.PageSize); err != nil { + if err := iops.AddMapping(ctx, ms, ar, usermem.PageSize, true); err != nil { t.Fatalf("AddMapping got %v, want nil", err) } - defer iops.RemoveMapping(ctx, ms, ar, usermem.PageSize) + defer iops.RemoveMapping(ctx, ms, ar, usermem.PageSize, true) mr := memmap.MappableRange{usermem.PageSize, 3 * usermem.PageSize} if _, err := iops.Translate(ctx, mr, mr, usermem.Read); err != nil { t.Fatalf("Translate got %v, want nil", err) diff --git a/pkg/sentry/fs/overlay.go b/pkg/sentry/fs/overlay.go index 8ace4ee64..f3e2d5cbe 100644 --- a/pkg/sentry/fs/overlay.go +++ b/pkg/sentry/fs/overlay.go @@ -259,32 +259,32 @@ func (o *overlayEntry) isMappableLocked() bool { } // AddMapping implements memmap.Mappable.AddMapping. -func (o *overlayEntry) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) error { +func (o *overlayEntry) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) error { o.mapsMu.Lock() defer o.mapsMu.Unlock() - if err := o.inodeLocked().Mappable().AddMapping(ctx, ms, ar, offset); err != nil { + if err := o.inodeLocked().Mappable().AddMapping(ctx, ms, ar, offset, writable); err != nil { return err } - o.mappings.AddMapping(ms, ar, offset) + o.mappings.AddMapping(ms, ar, offset, writable) return nil } // RemoveMapping implements memmap.Mappable.RemoveMapping. -func (o *overlayEntry) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) { +func (o *overlayEntry) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) { o.mapsMu.Lock() defer o.mapsMu.Unlock() - o.inodeLocked().Mappable().RemoveMapping(ctx, ms, ar, offset) - o.mappings.RemoveMapping(ms, ar, offset) + o.inodeLocked().Mappable().RemoveMapping(ctx, ms, ar, offset, writable) + o.mappings.RemoveMapping(ms, ar, offset, writable) } // CopyMapping implements memmap.Mappable.CopyMapping. -func (o *overlayEntry) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error { +func (o *overlayEntry) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64, writable bool) error { o.mapsMu.Lock() defer o.mapsMu.Unlock() - if err := o.inodeLocked().Mappable().CopyMapping(ctx, ms, srcAR, dstAR, offset); err != nil { + if err := o.inodeLocked().Mappable().CopyMapping(ctx, ms, srcAR, dstAR, offset, writable); err != nil { return err } - o.mappings.AddMapping(ms, dstAR, offset) + o.mappings.AddMapping(ms, dstAR, offset, writable) return nil } diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 42a7d7b9c..e0181c52c 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -426,23 +426,23 @@ func (rw *fileReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error) } // AddMapping implements memmap.Mappable.AddMapping. -func (f *fileInodeOperations) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) error { +func (f *fileInodeOperations) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) error { f.mapsMu.Lock() defer f.mapsMu.Unlock() - f.mappings.AddMapping(ms, ar, offset) + f.mappings.AddMapping(ms, ar, offset, writable) return nil } // RemoveMapping implements memmap.Mappable.RemoveMapping. -func (f *fileInodeOperations) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) { +func (f *fileInodeOperations) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) { f.mapsMu.Lock() defer f.mapsMu.Unlock() - f.mappings.RemoveMapping(ms, ar, offset) + f.mappings.RemoveMapping(ms, ar, offset, writable) } // CopyMapping implements memmap.Mappable.CopyMapping. -func (f *fileInodeOperations) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error { - return f.AddMapping(ctx, ms, dstAR, offset) +func (f *fileInodeOperations) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64, writable bool) error { + return f.AddMapping(ctx, ms, dstAR, offset, writable) } // Translate implements memmap.Mappable.Translate. diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 4343dee13..2f400cbba 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -375,7 +375,7 @@ func (s *Shm) Msync(context.Context, memmap.MappableRange) error { } // AddMapping implements memmap.Mappable.AddMapping. -func (s *Shm) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) error { +func (s *Shm) AddMapping(ctx context.Context, _ memmap.MappingSpace, _ usermem.AddrRange, _ uint64, _ bool) error { s.mu.Lock() defer s.mu.Unlock() s.attachTime = ktime.NowFromContext(ctx) @@ -390,7 +390,7 @@ func (s *Shm) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem } // RemoveMapping implements memmap.Mappable.RemoveMapping. -func (s *Shm) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) { +func (s *Shm) RemoveMapping(ctx context.Context, _ memmap.MappingSpace, _ usermem.AddrRange, _ uint64, _ bool) { s.mu.Lock() defer s.mu.Unlock() // TODO: RemoveMapping may be called during task exit, when ctx @@ -411,7 +411,7 @@ func (s *Shm) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar user } // CopyMapping implements memmap.Mappable.CopyMapping. -func (s *Shm) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error { +func (*Shm) CopyMapping(context.Context, memmap.MappingSpace, usermem.AddrRange, usermem.AddrRange, uint64, bool) error { return nil } diff --git a/pkg/sentry/memmap/mapping_set.go b/pkg/sentry/memmap/mapping_set.go index 33cf16f91..bd07e9aac 100644 --- a/pkg/sentry/memmap/mapping_set.go +++ b/pkg/sentry/memmap/mapping_set.go @@ -40,6 +40,7 @@ type MappingsOfRange map[MappingOfRange]struct{} type MappingOfRange struct { MappingSpace MappingSpace AddrRange usermem.AddrRange + Writable bool } func (r MappingOfRange) invalidate(opts InvalidateOpts) { @@ -92,6 +93,7 @@ func (mappingSetFunctions) Merge(r1 MappableRange, val1 MappingsOfRange, r2 Mapp Start: k1.AddrRange.End, End: k1.AddrRange.End + usermem.Addr(r2.Length()), }, + Writable: k1.Writable, } if _, ok := val2[k2]; !ok { return nil, false @@ -104,6 +106,7 @@ func (mappingSetFunctions) Merge(r1 MappableRange, val1 MappingsOfRange, r2 Mapp Start: k1.AddrRange.Start, End: k2.AddrRange.End, }, + Writable: k1.Writable, }] = struct{}{} } @@ -129,6 +132,7 @@ func (mappingSetFunctions) Split(r MappableRange, val MappingsOfRange, split uin Start: k.AddrRange.Start, End: k.AddrRange.Start + offset, }, + Writable: k.Writable, } m1[k1] = struct{}{} @@ -138,6 +142,7 @@ func (mappingSetFunctions) Split(r MappableRange, val MappingsOfRange, split uin Start: k.AddrRange.Start + offset, End: k.AddrRange.End, }, + Writable: k.Writable, } m2[k2] = struct{}{} } @@ -152,7 +157,7 @@ func (mappingSetFunctions) Split(r MappableRange, val MappingsOfRange, split uin // indicating that ms maps addresses [0x4000, 0x6000) to MappableRange [0x0, // 0x2000). Then for subsetRange = [0x1000, 0x2000), subsetMapping returns a // MappingOfRange for which AddrRange = [0x5000, 0x6000). -func subsetMapping(wholeRange, subsetRange MappableRange, ms MappingSpace, addr usermem.Addr) MappingOfRange { +func subsetMapping(wholeRange, subsetRange MappableRange, ms MappingSpace, addr usermem.Addr, writable bool) MappingOfRange { if !wholeRange.IsSupersetOf(subsetRange) { panic(fmt.Sprintf("%v is not a superset of %v", wholeRange, subsetRange)) } @@ -165,6 +170,7 @@ func subsetMapping(wholeRange, subsetRange MappableRange, ms MappingSpace, addr Start: start, End: start + usermem.Addr(subsetRange.Length()), }, + Writable: writable, } } @@ -172,7 +178,7 @@ func subsetMapping(wholeRange, subsetRange MappableRange, ms MappingSpace, addr // previously had no mappings. // // Preconditions: As for Mappable.AddMapping. -func (s *MappingSet) AddMapping(ms MappingSpace, ar usermem.AddrRange, offset uint64) []MappableRange { +func (s *MappingSet) AddMapping(ms MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) []MappableRange { mr := MappableRange{offset, offset + uint64(ar.Length())} var mapped []MappableRange seg, gap := s.Find(mr.Start) @@ -180,7 +186,7 @@ func (s *MappingSet) AddMapping(ms MappingSpace, ar usermem.AddrRange, offset ui switch { case seg.Ok() && seg.Start() < mr.End: seg = s.Isolate(seg, mr) - seg.Value()[subsetMapping(mr, seg.Range(), ms, ar.Start)] = struct{}{} + seg.Value()[subsetMapping(mr, seg.Range(), ms, ar.Start, writable)] = struct{}{} seg, gap = seg.NextNonEmpty() case gap.Ok() && gap.Start() < mr.End: @@ -199,7 +205,7 @@ func (s *MappingSet) AddMapping(ms MappingSpace, ar usermem.AddrRange, offset ui // MappableRanges that now have no mappings. // // Preconditions: As for Mappable.RemoveMapping. -func (s *MappingSet) RemoveMapping(ms MappingSpace, ar usermem.AddrRange, offset uint64) []MappableRange { +func (s *MappingSet) RemoveMapping(ms MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) []MappableRange { mr := MappableRange{offset, offset + uint64(ar.Length())} var unmapped []MappableRange @@ -213,7 +219,7 @@ func (s *MappingSet) RemoveMapping(ms MappingSpace, ar usermem.AddrRange, offset // Remove this part of the mapping. mappings := seg.Value() - delete(mappings, subsetMapping(mr, seg.Range(), ms, ar.Start)) + delete(mappings, subsetMapping(mr, seg.Range(), ms, ar.Start, writable)) if len(mappings) == 0 { unmapped = append(unmapped, seg.Range()) @@ -231,7 +237,7 @@ func (s *MappingSet) Invalidate(mr MappableRange, opts InvalidateOpts) { for seg := s.LowerBoundSegment(mr.Start); seg.Ok() && seg.Start() < mr.End; seg = seg.NextSegment() { segMR := seg.Range() for m := range seg.Value() { - region := subsetMapping(segMR, segMR.Intersect(mr), m.MappingSpace, m.AddrRange.Start) + region := subsetMapping(segMR, segMR.Intersect(mr), m.MappingSpace, m.AddrRange.Start, m.Writable) region.invalidate(opts) } } diff --git a/pkg/sentry/memmap/mapping_set_test.go b/pkg/sentry/memmap/mapping_set_test.go index 49ee34548..45d1d4688 100644 --- a/pkg/sentry/memmap/mapping_set_test.go +++ b/pkg/sentry/memmap/mapping_set_test.go @@ -40,7 +40,7 @@ func TestAddRemoveMapping(t *testing.T) { set := MappingSet{} ms := &testMappingSpace{} - mapped := set.AddMapping(ms, usermem.AddrRange{0x10000, 0x12000}, 0x1000) + mapped := set.AddMapping(ms, usermem.AddrRange{0x10000, 0x12000}, 0x1000, true) if got, want := mapped, []MappableRange{{0x1000, 0x3000}}; !reflect.DeepEqual(got, want) { t.Errorf("AddMapping: got %+v, wanted %+v", got, want) } @@ -49,7 +49,7 @@ func TestAddRemoveMapping(t *testing.T) { // [0x10000, 0x12000) => [0x1000, 0x3000) t.Log(&set) - mapped = set.AddMapping(ms, usermem.AddrRange{0x20000, 0x21000}, 0x2000) + mapped = set.AddMapping(ms, usermem.AddrRange{0x20000, 0x21000}, 0x2000, true) if len(mapped) != 0 { t.Errorf("AddMapping: got %+v, wanted []", mapped) } @@ -59,7 +59,7 @@ func TestAddRemoveMapping(t *testing.T) { // [0x11000, 0x12000) and [0x20000, 0x21000) => [0x2000, 0x3000) t.Log(&set) - mapped = set.AddMapping(ms, usermem.AddrRange{0x30000, 0x31000}, 0x4000) + mapped = set.AddMapping(ms, usermem.AddrRange{0x30000, 0x31000}, 0x4000, true) if got, want := mapped, []MappableRange{{0x4000, 0x5000}}; !reflect.DeepEqual(got, want) { t.Errorf("AddMapping: got %+v, wanted %+v", got, want) } @@ -70,7 +70,7 @@ func TestAddRemoveMapping(t *testing.T) { // [0x30000, 0x31000) => [0x4000, 0x5000) t.Log(&set) - mapped = set.AddMapping(ms, usermem.AddrRange{0x12000, 0x15000}, 0x3000) + mapped = set.AddMapping(ms, usermem.AddrRange{0x12000, 0x15000}, 0x3000, true) if got, want := mapped, []MappableRange{{0x3000, 0x4000}, {0x5000, 0x6000}}; !reflect.DeepEqual(got, want) { t.Errorf("AddMapping: got %+v, wanted %+v", got, want) } @@ -83,7 +83,7 @@ func TestAddRemoveMapping(t *testing.T) { // [0x14000, 0x15000) => [0x5000, 0x6000) t.Log(&set) - unmapped := set.RemoveMapping(ms, usermem.AddrRange{0x10000, 0x11000}, 0x1000) + unmapped := set.RemoveMapping(ms, usermem.AddrRange{0x10000, 0x11000}, 0x1000, true) if got, want := unmapped, []MappableRange{{0x1000, 0x2000}}; !reflect.DeepEqual(got, want) { t.Errorf("RemoveMapping: got %+v, wanted %+v", got, want) } @@ -95,7 +95,7 @@ func TestAddRemoveMapping(t *testing.T) { // [0x14000, 0x15000) => [0x5000, 0x6000) t.Log(&set) - unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x20000, 0x21000}, 0x2000) + unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x20000, 0x21000}, 0x2000, true) if len(unmapped) != 0 { t.Errorf("RemoveMapping: got %+v, wanted []", unmapped) } @@ -106,7 +106,7 @@ func TestAddRemoveMapping(t *testing.T) { // [0x14000, 0x15000) => [0x5000, 0x6000) t.Log(&set) - unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x11000, 0x15000}, 0x2000) + unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x11000, 0x15000}, 0x2000, true) if got, want := unmapped, []MappableRange{{0x2000, 0x4000}, {0x5000, 0x6000}}; !reflect.DeepEqual(got, want) { t.Errorf("RemoveMapping: got %+v, wanted %+v", got, want) } @@ -115,7 +115,7 @@ func TestAddRemoveMapping(t *testing.T) { // [0x30000, 0x31000) => [0x4000, 0x5000) t.Log(&set) - unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x30000, 0x31000}, 0x4000) + unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x30000, 0x31000}, 0x4000, true) if got, want := unmapped, []MappableRange{{0x4000, 0x5000}}; !reflect.DeepEqual(got, want) { t.Errorf("RemoveMapping: got %+v, wanted %+v", got, want) } @@ -125,7 +125,7 @@ func TestInvalidateWholeMapping(t *testing.T) { set := MappingSet{} ms := &testMappingSpace{} - set.AddMapping(ms, usermem.AddrRange{0x10000, 0x11000}, 0) + set.AddMapping(ms, usermem.AddrRange{0x10000, 0x11000}, 0, true) // Mappings: // [0x10000, 0x11000) => [0, 0x1000) t.Log(&set) @@ -139,7 +139,7 @@ func TestInvalidatePartialMapping(t *testing.T) { set := MappingSet{} ms := &testMappingSpace{} - set.AddMapping(ms, usermem.AddrRange{0x10000, 0x13000}, 0) + set.AddMapping(ms, usermem.AddrRange{0x10000, 0x13000}, 0, true) // Mappings: // [0x10000, 0x13000) => [0, 0x3000) t.Log(&set) @@ -153,8 +153,8 @@ func TestInvalidateMultipleMappings(t *testing.T) { set := MappingSet{} ms := &testMappingSpace{} - set.AddMapping(ms, usermem.AddrRange{0x10000, 0x11000}, 0) - set.AddMapping(ms, usermem.AddrRange{0x20000, 0x21000}, 0x2000) + set.AddMapping(ms, usermem.AddrRange{0x10000, 0x11000}, 0, true) + set.AddMapping(ms, usermem.AddrRange{0x20000, 0x21000}, 0x2000, true) // Mappings: // [0x10000, 0x11000) => [0, 0x1000) // [0x12000, 0x13000) => [0x2000, 0x3000) @@ -170,8 +170,8 @@ func TestInvalidateOverlappingMappings(t *testing.T) { ms1 := &testMappingSpace{} ms2 := &testMappingSpace{} - set.AddMapping(ms1, usermem.AddrRange{0x10000, 0x12000}, 0) - set.AddMapping(ms2, usermem.AddrRange{0x20000, 0x22000}, 0x1000) + set.AddMapping(ms1, usermem.AddrRange{0x10000, 0x12000}, 0, true) + set.AddMapping(ms2, usermem.AddrRange{0x20000, 0x22000}, 0x1000, true) // Mappings: // ms1:[0x10000, 0x12000) => [0, 0x2000) // ms2:[0x11000, 0x13000) => [0x1000, 0x3000) @@ -184,3 +184,77 @@ func TestInvalidateOverlappingMappings(t *testing.T) { t.Errorf("Invalidate: ms1: got %+v, wanted %+v", got, want) } } + +func TestMixedWritableMappings(t *testing.T) { + set := MappingSet{} + ms := &testMappingSpace{} + + mapped := set.AddMapping(ms, usermem.AddrRange{0x10000, 0x12000}, 0x1000, true) + if got, want := mapped, []MappableRange{{0x1000, 0x3000}}; !reflect.DeepEqual(got, want) { + t.Errorf("AddMapping: got %+v, wanted %+v", got, want) + } + + // Mappings: + // [0x10000, 0x12000) writable => [0x1000, 0x3000) + t.Log(&set) + + mapped = set.AddMapping(ms, usermem.AddrRange{0x20000, 0x22000}, 0x2000, false) + if got, want := mapped, []MappableRange{{0x3000, 0x4000}}; !reflect.DeepEqual(got, want) { + t.Errorf("AddMapping: got %+v, wanted %+v", got, want) + } + + // Mappings: + // [0x10000, 0x11000) writable => [0x1000, 0x2000) + // [0x11000, 0x12000) writable and [0x20000, 0x21000) readonly => [0x2000, 0x3000) + // [0x21000, 0x22000) readonly => [0x3000, 0x4000) + t.Log(&set) + + // Unmap should fail because we specified the readonly map address range, but + // asked to unmap a writable segment. + unmapped := set.RemoveMapping(ms, usermem.AddrRange{0x20000, 0x21000}, 0x2000, true) + if len(unmapped) != 0 { + t.Errorf("RemoveMapping: got %+v, wanted []", unmapped) + } + + // Readonly mapping removed, but writable mapping still exists in the range, + // so no mappable range fully unmapped. + unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x20000, 0x21000}, 0x2000, false) + if len(unmapped) != 0 { + t.Errorf("RemoveMapping: got %+v, wanted []", unmapped) + } + + // Mappings: + // [0x10000, 0x12000) writable => [0x1000, 0x3000) + // [0x21000, 0x22000) readonly => [0x3000, 0x4000) + t.Log(&set) + + unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x11000, 0x12000}, 0x2000, true) + if got, want := unmapped, []MappableRange{{0x2000, 0x3000}}; !reflect.DeepEqual(got, want) { + t.Errorf("RemoveMapping: got %+v, wanted %+v", got, want) + } + + // Mappings: + // [0x10000, 0x12000) writable => [0x1000, 0x3000) + // [0x21000, 0x22000) readonly => [0x3000, 0x4000) + t.Log(&set) + + // Unmap should fail since writable bit doesn't match. + unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x10000, 0x12000}, 0x1000, false) + if len(unmapped) != 0 { + t.Errorf("RemoveMapping: got %+v, wanted []", unmapped) + } + + unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x10000, 0x12000}, 0x1000, true) + if got, want := unmapped, []MappableRange{{0x1000, 0x2000}}; !reflect.DeepEqual(got, want) { + t.Errorf("RemoveMapping: got %+v, wanted %+v", got, want) + } + + // Mappings: + // [0x21000, 0x22000) readonly => [0x3000, 0x4000) + t.Log(&set) + + unmapped = set.RemoveMapping(ms, usermem.AddrRange{0x21000, 0x22000}, 0x3000, false) + if got, want := unmapped, []MappableRange{{0x3000, 0x4000}}; !reflect.DeepEqual(got, want) { + t.Errorf("RemoveMapping: got %+v, wanted %+v", got, want) + } +} diff --git a/pkg/sentry/memmap/memmap.go b/pkg/sentry/memmap/memmap.go index 05349a77f..28e2bed9b 100644 --- a/pkg/sentry/memmap/memmap.go +++ b/pkg/sentry/memmap/memmap.go @@ -36,16 +36,22 @@ type Mappable interface { // AddMapping notifies the Mappable of a mapping from addresses ar in ms to // offsets [offset, offset+ar.Length()) in this Mappable. // + // The writable flag indicates whether the backing data for a Mappable can + // be modified through the mapping. Effectively, this means a shared mapping + // where Translate may be called with at.Write == true. This is a property + // established at mapping creation and must remain constant throughout the + // lifetime of the mapping. + // // Preconditions: offset+ar.Length() does not overflow. - AddMapping(ctx context.Context, ms MappingSpace, ar usermem.AddrRange, offset uint64) error + AddMapping(ctx context.Context, ms MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) error // RemoveMapping notifies the Mappable of the removal of a mapping from // addresses ar in ms to offsets [offset, offset+ar.Length()) in this // Mappable. // // Preconditions: offset+ar.Length() does not overflow. The removed mapping - // must exist. - RemoveMapping(ctx context.Context, ms MappingSpace, ar usermem.AddrRange, offset uint64) + // must exist. writable must match the corresponding call to AddMapping. + RemoveMapping(ctx context.Context, ms MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) // CopyMapping notifies the Mappable of an attempt to copy a mapping in ms // from srcAR to dstAR. For most Mappables, this is equivalent to @@ -56,8 +62,9 @@ type Mappable interface { // MappingSpace; it is analogous to Linux's vm_operations_struct::mremap. // // Preconditions: offset+srcAR.Length() and offset+dstAR.Length() do not - // overflow. The mapping at srcAR must exist. - CopyMapping(ctx context.Context, ms MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error + // overflow. The mapping at srcAR must exist. writable must match the + // corresponding call to AddMapping. + CopyMapping(ctx context.Context, ms MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64, writable bool) error // Translate returns the Mappable's current mappings for at least the range // of offsets specified by required, and at most the range of offsets diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index 87942af0e..5e86d3b49 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -244,7 +244,7 @@ func (m *aioMappable) Msync(ctx context.Context, mr memmap.MappableRange) error } // AddMapping implements memmap.Mappable.AddMapping. -func (m *aioMappable) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) error { +func (m *aioMappable) AddMapping(_ context.Context, _ memmap.MappingSpace, ar usermem.AddrRange, offset uint64, _ bool) error { // Don't allow mappings to be expanded (in Linux, fs/aio.c:aio_ring_mmap() // sets VM_DONTEXPAND). if offset != 0 || uint64(ar.Length()) != aioRingBufferSize { @@ -254,11 +254,11 @@ func (m *aioMappable) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar } // RemoveMapping implements memmap.Mappable.RemoveMapping. -func (m *aioMappable) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) { +func (m *aioMappable) RemoveMapping(context.Context, memmap.MappingSpace, usermem.AddrRange, uint64, bool) { } // CopyMapping implements memmap.Mappable.CopyMapping. -func (m *aioMappable) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error { +func (m *aioMappable) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64, _ bool) error { // Don't allow mappings to be expanded (in Linux, fs/aio.c:aio_ring_mmap() // sets VM_DONTEXPAND). if offset != 0 || uint64(dstAR.Length()) != aioRingBufferSize { diff --git a/pkg/sentry/mm/lifecycle.go b/pkg/sentry/mm/lifecycle.go index b248b76e7..1613ce11d 100644 --- a/pkg/sentry/mm/lifecycle.go +++ b/pkg/sentry/mm/lifecycle.go @@ -81,7 +81,7 @@ func (mm *MemoryManager) Fork(ctx context.Context) (*MemoryManager, error) { vmaAR := srcvseg.Range() // Inform the Mappable, if any, of the new mapping. if vma.mappable != nil { - if err := vma.mappable.AddMapping(ctx, mm2, vmaAR, vma.off); err != nil { + if err := vma.mappable.AddMapping(ctx, mm2, vmaAR, vma.off, vma.isMappableAsWritable()); err != nil { mm2.removeVMAsLocked(ctx, mm2.applicationAddrRange()) return nil, err } diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index aab697f9e..b1e39e898 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -349,6 +349,10 @@ func (v *vma) loadRealPerms(b int) { } } +func (v *vma) isMappableAsWritable() bool { + return !v.private && v.maxPerms.Write +} + // pma represents a platform mapping area. // // +stateify savable diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index e511472f4..64d0dd3f6 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -76,16 +76,16 @@ func (m *SpecialMappable) Msync(ctx context.Context, mr memmap.MappableRange) er } // AddMapping implements memmap.Mappable.AddMapping. -func (m *SpecialMappable) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) error { +func (*SpecialMappable) AddMapping(context.Context, memmap.MappingSpace, usermem.AddrRange, uint64, bool) error { return nil } // RemoveMapping implements memmap.Mappable.RemoveMapping. -func (m *SpecialMappable) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64) { +func (*SpecialMappable) RemoveMapping(context.Context, memmap.MappingSpace, usermem.AddrRange, uint64, bool) { } // CopyMapping implements memmap.Mappable.CopyMapping. -func (m *SpecialMappable) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR usermem.AddrRange, offset uint64) error { +func (*SpecialMappable) CopyMapping(context.Context, memmap.MappingSpace, usermem.AddrRange, usermem.AddrRange, uint64, bool) error { return nil } diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index 1a46c2105..daaae4da1 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -443,7 +443,7 @@ func (mm *MemoryManager) MRemap(ctx context.Context, oldAddr usermem.Addr, oldSi return 0, syserror.EINVAL } // Inform the Mappable, if any, of the new mapping. - if err := vma.mappable.CopyMapping(ctx, mm, oldAR, newAR, vseg.mappableOffsetAt(oldAR.Start)); err != nil { + if err := vma.mappable.CopyMapping(ctx, mm, oldAR, newAR, vseg.mappableOffsetAt(oldAR.Start), vma.isMappableAsWritable()); err != nil { return 0, err } } @@ -498,7 +498,7 @@ func (mm *MemoryManager) MRemap(ctx context.Context, oldAddr usermem.Addr, oldSi // Now that pmas have been moved to newAR, we can notify vma.mappable that // oldAR is no longer mapped. if vma.mappable != nil { - vma.mappable.RemoveMapping(ctx, mm, oldAR, vma.off) + vma.mappable.RemoveMapping(ctx, mm, oldAR, vma.off, vma.isMappableAsWritable()) } return newAR.Start, nil diff --git a/pkg/sentry/mm/vma.go b/pkg/sentry/mm/vma.go index dafdbd0e4..5c2c802f6 100644 --- a/pkg/sentry/mm/vma.go +++ b/pkg/sentry/mm/vma.go @@ -65,7 +65,7 @@ func (mm *MemoryManager) createVMALocked(ctx context.Context, opts memmap.MMapOp // Inform the Mappable, if any, of the new mapping. if opts.Mappable != nil { - if err := opts.Mappable.AddMapping(ctx, mm, ar, opts.Offset); err != nil { + if err := opts.Mappable.AddMapping(ctx, mm, ar, opts.Offset, !opts.Private && opts.MaxPerms.Write); err != nil { return vmaIterator{}, usermem.AddrRange{}, err } } @@ -332,7 +332,7 @@ func (mm *MemoryManager) removeVMAsLocked(ctx context.Context, ar usermem.AddrRa vmaAR := vseg.Range() vma := vseg.ValuePtr() if vma.mappable != nil { - vma.mappable.RemoveMapping(ctx, mm, vmaAR, vma.off) + vma.mappable.RemoveMapping(ctx, mm, vmaAR, vma.off, vma.isMappableAsWritable()) } if vma.id != nil { vma.id.DecRef() -- cgit v1.2.3 From f93c288dd70846f335239e2d0cb351135a756f51 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Wed, 12 Dec 2018 13:17:46 -0800 Subject: Fix a data race on Shm.key. PiperOrigin-RevId: 225240907 Change-Id: Ie568ce3cd643f3e4a0eaa0444f4ed589dcf6031f --- pkg/sentry/kernel/shm/shm.go | 89 +++++++++++++++++++++++++++++--------------- 1 file changed, 59 insertions(+), 30 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 2f400cbba..96414d060 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -66,24 +66,30 @@ type Registry struct { // userNS owns the IPC namespace this registry belong to. Immutable. userNS *auth.UserNamespace + // mu protects all fields below. mu sync.Mutex `state:"nosave"` - // shms maps segment ids to segments. Protected by mu. + // shms maps segment ids to segments. shms map[ID]*Shm + // keysToShms maps segment keys to segments. + keysToShms map[Key]*Shm + // Sum of the sizes of all existing segments rounded up to page size, in - // units of page size. Protected by mu. + // units of page size. totalPages uint64 - // lastIDUsed is protected by mu. + // ID assigned to the last created segment. Used to quickly find the next + // unused ID. lastIDUsed ID } // NewRegistry creates a new shm registry. func NewRegistry(userNS *auth.UserNamespace) *Registry { return &Registry{ - userNS: userNS, - shms: make(map[ID]*Shm), + userNS: userNS, + shms: make(map[ID]*Shm), + keysToShms: make(map[Key]*Shm), } } @@ -94,14 +100,20 @@ func (r *Registry) FindByID(id ID) *Shm { return r.shms[id] } -// Precondition: Caller must hold r.mu. -func (r *Registry) findByKey(key Key) *Shm { - for _, v := range r.shms { - if v.key == key { - return v - } +// dissociateKey removes the association between a segment and its key, +// preventing it from being discovered in the registry. This doesn't necessarily +// mean the segment is about to be destroyed. This is analogous to unlinking a +// file; the segment can still be used by a process already referencing it, but +// cannot be discovered by a new process. +func (r *Registry) dissociateKey(s *Shm) { + r.mu.Lock() + defer r.mu.Unlock() + s.mu.Lock() + defer s.mu.Unlock() + if s.key != linux.IPC_PRIVATE { + delete(r.keysToShms, s.key) + s.key = linux.IPC_PRIVATE } - return nil } // FindOrCreate looks up or creates a segment in the registry. It's functionally @@ -127,7 +139,7 @@ func (r *Registry) FindOrCreate(ctx context.Context, pid int32, key Key, size ui if !private { // Look up an existing segment. - if shm := r.findByKey(key); shm != nil { + if shm := r.keysToShms[key]; shm != nil { shm.mu.Lock() defer shm.mu.Unlock() @@ -184,6 +196,8 @@ func (r *Registry) FindOrCreate(ctx context.Context, pid int32, key Key, size ui } // newShm creates a new segment in the registry. +// +// Precondition: Caller must hold r.mu. func (r *Registry) newShm(ctx context.Context, pid int32, key Key, creator fs.FileOwner, perms fs.FilePermissions, size uint64) (*Shm, error) { p := platform.FromContext(ctx) if p == nil { @@ -219,8 +233,10 @@ func (r *Registry) newShm(ctx context.Context, pid int32, key Key, creator fs.Fi } if r.shms[id] == nil { r.lastIDUsed = id - r.shms[id] = shm + shm.ID = id + r.shms[id] = shm + r.keysToShms[key] = shm r.totalPages += effectiveSize / usermem.PageSize @@ -258,13 +274,20 @@ func (r *Registry) ShmInfo() *linux.ShmInfo { } } -// remove unregisters a segment from this registry, preventing it from being -// discovered in the future. Caller is responsible for ensuring s is destroyed. +// remove deletes a segment from this registry, deaccounting the memory used by +// the segment. // -// Precondition: To preserve lock ordering, caller must not hold s.mu. +// Precondition: Must follow a call to r.dissociateKey(s). func (r *Registry) remove(s *Shm) { r.mu.Lock() defer r.mu.Unlock() + s.mu.Lock() + defer s.mu.Unlock() + + if s.key != linux.IPC_PRIVATE { + panic(fmt.Sprintf("Attempted to remove shm segment %+v from the registry whose key is still associated", s)) + } + delete(r.shms, s.ID) r.totalPages -= s.effectiveSize / usermem.PageSize } @@ -314,12 +337,12 @@ type Shm struct { // segment. Immutable. fr platform.FileRange - // key is the public identifier for this segment. - key Key - // mu protects all fields below. mu sync.Mutex `state:"nosave"` + // key is the public identifier for this segment. + key Key + // perms is the access permissions for the segment. perms fs.FilePermissions @@ -342,12 +365,14 @@ type Shm struct { // pendingDestruction indicates the segment was marked as destroyed through // shmctl(IPC_RMID). When marked as destroyed, the segment will not be found // in the registry and can no longer be attached. When the last user - // detaches from the segment, it is destroyed. Protected by mu. + // detaches from the segment, it is destroyed. pendingDestruction bool } // MappedName implements memmap.MappingIdentity.MappedName. func (s *Shm) MappedName(ctx context.Context) string { + s.mu.Lock() + defer s.mu.Unlock() return fmt.Sprintf("SYSV%08d", s.key) } @@ -364,6 +389,8 @@ func (s *Shm) InodeID() uint64 { } // DecRef overrides refs.RefCount.DecRef with a destructor. +// +// Precondition: Caller must not hold s.mu. func (s *Shm) DecRef() { s.DecRefWithDestructor(s.destroy) } @@ -572,28 +599,30 @@ func (s *Shm) Set(ctx context.Context, ds *linux.ShmidDS) error { } func (s *Shm) destroy() { - s.registry.remove(s) s.p.Memory().DecRef(s.fr) + s.registry.remove(s) } -// MarkDestroyed marks a shm for destruction. The shm is actually destroyed once -// it has no references. See shmctl(IPC_RMID). +// MarkDestroyed marks a segment for destruction. The segment is actually +// destroyed once it has no references. MarkDestroyed may be called multiple +// times, and is safe to call after a segment has already been destroyed. See +// shmctl(IPC_RMID). func (s *Shm) MarkDestroyed() { - s.mu.Lock() - defer s.mu.Unlock() - - // Prevent the segment from being found in the registry. - s.key = linux.IPC_PRIVATE + s.registry.dissociateKey(s) + s.mu.Lock() // Only drop the segment's self-reference once, when destruction is - // requested. Otherwise, repeated calls shmctl(IPC_RMID) would force a + // requested. Otherwise, repeated calls to shmctl(IPC_RMID) would force a // segment to be destroyed prematurely, potentially with active maps to the // segment's address range. Remaining references are dropped when the // segment is detached or unmaped. if !s.pendingDestruction { s.pendingDestruction = true + s.mu.Unlock() // Must release s.mu before calling s.DecRef. s.DecRef() + return } + s.mu.Unlock() } // checkOwnership verifies whether a segment may be accessed by ctx as an -- cgit v1.2.3 From ccce1d4281ce82fe551d7c8569fe3a545c62e296 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Wed, 12 Dec 2018 17:47:01 -0800 Subject: Filesystems shouldn't be saving references to Platform. Platform objects are not savable, storing references to them in filesystem datastructures would cause save to fail if someone actually passed in a Platform. Current implementations work because everywhere a Platform is expected, we currently pass in a Kernel object which embeds Platform and thus satisfies the interface. Eliminate this indirection and save pointers to Kernel directly. PiperOrigin-RevId: 225288336 Change-Id: Ica399ff43f425e15bc150a0d7102196c3d54a2ab --- pkg/sentry/context/contexttest/contexttest.go | 32 ++++++++++++++++------ pkg/sentry/fs/BUILD | 2 +- pkg/sentry/fs/ashmem/area.go | 8 +++--- pkg/sentry/fs/copy_up_test.go | 2 +- pkg/sentry/fs/dev/BUILD | 1 + pkg/sentry/fs/dev/dev.go | 4 +-- pkg/sentry/fs/file_overlay_test.go | 2 +- pkg/sentry/fs/inode_overlay_test.go | 2 +- pkg/sentry/fs/mounts_test.go | 2 +- pkg/sentry/fs/tmpfs/BUILD | 6 ++--- pkg/sentry/fs/tmpfs/file_test.go | 6 ++--- pkg/sentry/fs/tmpfs/fs.go | 4 +-- pkg/sentry/fs/tmpfs/inode_file.go | 20 +++++++------- pkg/sentry/fs/tmpfs/tmpfs.go | 14 +++++----- pkg/sentry/kernel/contexttest/BUILD | 17 ++++++++++++ pkg/sentry/kernel/contexttest/contexttest.go | 38 +++++++++++++++++++++++++++ 16 files changed, 116 insertions(+), 44 deletions(-) create mode 100644 pkg/sentry/kernel/contexttest/BUILD create mode 100644 pkg/sentry/kernel/contexttest/contexttest.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/context/contexttest/contexttest.go b/pkg/sentry/context/contexttest/contexttest.go index d2f084ed7..d5fd9f165 100644 --- a/pkg/sentry/context/contexttest/contexttest.go +++ b/pkg/sentry/context/contexttest/contexttest.go @@ -31,23 +31,30 @@ import ( // Context returns a Context that may be used in tests. Uses ptrace as the // platform.Platform. +// +// Note that some filesystems may require a minimal kernel for testing, which +// this test context does not provide. For such tests, see kernel/contexttest. func Context(tb testing.TB) context.Context { p, err := ptrace.New() if err != nil { tb.Fatal(err) } // Test usage of context.Background is fine. - return &testContext{ - Context: context.Background(), - l: limits.NewLimitSet(), - platform: p, + return &TestContext{ + Context: context.Background(), + l: limits.NewLimitSet(), + platform: p, + otherValues: make(map[interface{}]interface{}), } } -type testContext struct { +// TestContext represents a context with minimal functionality suitable for +// running tests. +type TestContext struct { context.Context - l *limits.LimitSet - platform platform.Platform + l *limits.LimitSet + platform platform.Platform + otherValues map[interface{}]interface{} } // globalUniqueID tracks incremental unique identifiers for tests. @@ -76,8 +83,14 @@ func (hostClock) Now() ktime.Time { return ktime.FromNanoseconds(time.Now().UnixNano()) } +// RegisterValue registers additional values with this test context. Useful for +// providing values from external packages that contexttest can't depend on. +func (t *TestContext) RegisterValue(key, value interface{}) { + t.otherValues[key] = value +} + // Value implements context.Context. -func (t *testContext) Value(key interface{}) interface{} { +func (t *TestContext) Value(key interface{}) interface{} { switch key { case limits.CtxLimits: return t.l @@ -92,6 +105,9 @@ func (t *testContext) Value(key interface{}) interface{} { case ktime.CtxRealtimeClock: return hostClock{} default: + if val, ok := t.otherValues[key]; ok { + return val + } return t.Context.Value(key) } } diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 548898aaa..0fe2b14bf 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -94,9 +94,9 @@ go_test( deps = [ ":fs", "//pkg/sentry/context", - "//pkg/sentry/context/contexttest", "//pkg/sentry/fs/ramfs/test", "//pkg/sentry/fs/tmpfs", + "//pkg/sentry/kernel/contexttest", "//pkg/sentry/usermem", "//pkg/syserror", ], diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index 5372875ac..d7dd2c084 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -23,8 +23,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tmpfs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" @@ -117,11 +117,11 @@ func (a *Area) ConfigureMMap(ctx context.Context, file *fs.File, opts *memmap.MM opts.MaxPerms = opts.MaxPerms.Intersect(a.perms) if a.tmpfsFile == nil { - p := platform.FromContext(ctx) - if p == nil { + k := kernel.KernelFromContext(ctx) + if k == nil { return syserror.ENOMEM } - tmpfsInodeOps := tmpfs.NewInMemoryFile(ctx, usage.Tmpfs, fs.UnstableAttr{}, p) + tmpfsInodeOps := tmpfs.NewInMemoryFile(ctx, usage.Tmpfs, fs.UnstableAttr{}, k) // This is not backed by a real filesystem, so we pass in nil. tmpfsInode := fs.NewInode(tmpfsInodeOps, fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}), fs.StableAttr{}) dirent := fs.NewDirent(tmpfsInode, namePrefix+"/"+a.name) diff --git a/pkg/sentry/fs/copy_up_test.go b/pkg/sentry/fs/copy_up_test.go index fcba14ed4..2b2f4bb8f 100644 --- a/pkg/sentry/fs/copy_up_test.go +++ b/pkg/sentry/fs/copy_up_test.go @@ -22,9 +22,9 @@ import ( "sync" "testing" - "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" _ "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tmpfs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) diff --git a/pkg/sentry/fs/dev/BUILD b/pkg/sentry/fs/dev/BUILD index fc069bb5f..b17b5202c 100644 --- a/pkg/sentry/fs/dev/BUILD +++ b/pkg/sentry/fs/dev/BUILD @@ -25,6 +25,7 @@ go_library( "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/ramfs", "//pkg/sentry/fs/tmpfs", + "//pkg/sentry/kernel", "//pkg/sentry/memmap", "//pkg/sentry/mm", "//pkg/sentry/platform", diff --git a/pkg/sentry/fs/dev/dev.go b/pkg/sentry/fs/dev/dev.go index 05a5005ad..3e127bf04 100644 --- a/pkg/sentry/fs/dev/dev.go +++ b/pkg/sentry/fs/dev/dev.go @@ -22,7 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/binder" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/tmpfs" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) @@ -86,7 +86,7 @@ func New(ctx context.Context, msrc *fs.MountSource, binderEnabled bool, ashmemEn "random": newCharacterDevice(newRandomDevice(ctx, fs.RootOwner, 0444), msrc), "urandom": newCharacterDevice(newRandomDevice(ctx, fs.RootOwner, 0444), msrc), - "shm": tmpfs.NewDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0777), msrc, platform.FromContext(ctx)), + "shm": tmpfs.NewDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0777), msrc, kernel.KernelFromContext(ctx)), // A devpts is typically mounted at /dev/pts to provide // pseudoterminal support. Place an empty directory there for diff --git a/pkg/sentry/fs/file_overlay_test.go b/pkg/sentry/fs/file_overlay_test.go index 11e4f7203..f121cbdda 100644 --- a/pkg/sentry/fs/file_overlay_test.go +++ b/pkg/sentry/fs/file_overlay_test.go @@ -19,9 +19,9 @@ import ( "testing" "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" ramfstest "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest" ) func TestReaddir(t *testing.T) { diff --git a/pkg/sentry/fs/inode_overlay_test.go b/pkg/sentry/fs/inode_overlay_test.go index acdb2b4f8..9e922d008 100644 --- a/pkg/sentry/fs/inode_overlay_test.go +++ b/pkg/sentry/fs/inode_overlay_test.go @@ -18,9 +18,9 @@ import ( "testing" "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" ramfstest "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest" "gvisor.googlesource.com/gvisor/pkg/syserror" ) diff --git a/pkg/sentry/fs/mounts_test.go b/pkg/sentry/fs/mounts_test.go index 2f7a1710f..7d682d99b 100644 --- a/pkg/sentry/fs/mounts_test.go +++ b/pkg/sentry/fs/mounts_test.go @@ -18,9 +18,9 @@ import ( "testing" "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" ramfstest "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest" ) // Creates a new MountNamespace with filesystem: diff --git a/pkg/sentry/fs/tmpfs/BUILD b/pkg/sentry/fs/tmpfs/BUILD index 277583113..9065cdd5d 100644 --- a/pkg/sentry/fs/tmpfs/BUILD +++ b/pkg/sentry/fs/tmpfs/BUILD @@ -20,10 +20,10 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/ramfs", + "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/pipe", "//pkg/sentry/memmap", - "//pkg/sentry/platform", "//pkg/sentry/safemem", "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usage", @@ -39,9 +39,9 @@ go_test( embed = [":tmpfs"], deps = [ "//pkg/sentry/context", - "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", - "//pkg/sentry/platform", + "//pkg/sentry/kernel", + "//pkg/sentry/kernel/contexttest", "//pkg/sentry/usage", "//pkg/sentry/usermem", ], diff --git a/pkg/sentry/fs/tmpfs/file_test.go b/pkg/sentry/fs/tmpfs/file_test.go index b5830d3df..02da9af82 100644 --- a/pkg/sentry/fs/tmpfs/file_test.go +++ b/pkg/sentry/fs/tmpfs/file_test.go @@ -19,16 +19,16 @@ import ( "testing" "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) func newFileInode(ctx context.Context) *fs.Inode { m := fs.NewCachingMountSource(&Filesystem{}, fs.MountSourceFlags{}) - iops := NewInMemoryFile(ctx, usage.Tmpfs, fs.WithCurrentTime(ctx, fs.UnstableAttr{}), platform.FromContext(ctx)) + iops := NewInMemoryFile(ctx, usage.Tmpfs, fs.WithCurrentTime(ctx, fs.UnstableAttr{}), kernel.KernelFromContext(ctx)) return fs.NewInode(iops, m, fs.StableAttr{ DeviceID: tmpfsDevice.DeviceID(), InodeID: tmpfsDevice.NextIno(), diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go index 3ac0c4dd4..88f85b85a 100644 --- a/pkg/sentry/fs/tmpfs/fs.go +++ b/pkg/sentry/fs/tmpfs/fs.go @@ -21,8 +21,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" ) const ( @@ -131,5 +131,5 @@ func (f *Filesystem) Mount(ctx context.Context, device string, flags fs.MountSou msrc := fs.NewCachingMountSource(f, flags) // Construct the tmpfs root. - return NewDir(ctx, nil, owner, perms, msrc, platform.FromContext(ctx)), nil + return NewDir(ctx, nil, owner, perms, msrc, kernel.KernelFromContext(ctx)), nil } diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index e0181c52c..ca2b4aabb 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -21,8 +21,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -52,8 +52,8 @@ type fileInodeOperations struct { fsutil.InodeNotSymlink `state:"nosave"` fsutil.NoopWriteOut `state:"nosave"` - // platform is used to allocate memory that stores the file's contents. - platform platform.Platform + // kernel is used to allocate platform memory that stores the file's contents. + kernel *kernel.Kernel // memUsage is the default memory usage that will be reported by this file. memUsage usage.MemoryKind @@ -84,12 +84,12 @@ type fileInodeOperations struct { } // NewInMemoryFile returns a new file backed by p.Memory(). -func NewInMemoryFile(ctx context.Context, usage usage.MemoryKind, uattr fs.UnstableAttr, p platform.Platform) fs.InodeOperations { +func NewInMemoryFile(ctx context.Context, usage usage.MemoryKind, uattr fs.UnstableAttr, k *kernel.Kernel) fs.InodeOperations { return &fileInodeOperations{ attr: fsutil.InMemoryAttributes{ Unstable: uattr, }, - platform: p, + kernel: k, memUsage: usage, } } @@ -98,7 +98,7 @@ func NewInMemoryFile(ctx context.Context, usage usage.MemoryKind, uattr fs.Unsta func (f *fileInodeOperations) Release(context.Context) { f.dataMu.Lock() defer f.dataMu.Unlock() - f.data.DropAll(f.platform.Memory()) + f.data.DropAll(f.kernel.Platform.Memory()) } // Mappable implements fs.InodeOperations.Mappable. @@ -212,7 +212,7 @@ func (f *fileInodeOperations) Truncate(ctx context.Context, inode *fs.Inode, siz // and can remove them. f.dataMu.Lock() defer f.dataMu.Unlock() - f.data.Truncate(uint64(size), f.platform.Memory()) + f.data.Truncate(uint64(size), f.kernel.Platform.Memory()) return nil } @@ -310,7 +310,7 @@ func (rw *fileReadWriter) ReadToBlocks(dsts safemem.BlockSeq) (uint64, error) { return 0, nil } - mem := rw.f.platform.Memory() + mem := rw.f.kernel.Platform.Memory() var done uint64 seg, gap := rw.f.data.Find(uint64(rw.offset)) for rw.offset < end { @@ -376,7 +376,7 @@ func (rw *fileReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error) } }() - mem := rw.f.platform.Memory() + mem := rw.f.kernel.Platform.Memory() // Page-aligned mr for when we need to allocate memory. RoundUp can't // overflow since end is an int64. pgstartaddr := usermem.Addr(rw.offset).RoundDown() @@ -465,7 +465,7 @@ func (f *fileInodeOperations) Translate(ctx context.Context, required, optional optional.End = pgend } - mem := f.platform.Memory() + mem := f.kernel.Platform.Memory() cerr := f.data.Fill(ctx, required, optional, mem, f.memUsage, func(_ context.Context, dsts safemem.BlockSeq, _ uint64) (uint64, error) { // Newly-allocated pages are zeroed, so we don't need to do anything. return dsts.NumBytes(), nil diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 91b782540..40a8c4b1e 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -20,8 +20,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/pipe" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -54,13 +54,13 @@ func rename(ctx context.Context, oldParent *fs.Inode, oldName string, newParent type Dir struct { ramfs.Dir - // platform is used to allocate storage for tmpfs Files. - platform platform.Platform + // kernel is used to allocate platform memory as storage for tmpfs Files. + kernel *kernel.Kernel } // NewDir returns a new directory. -func NewDir(ctx context.Context, contents map[string]*fs.Inode, owner fs.FileOwner, perms fs.FilePermissions, msrc *fs.MountSource, platform platform.Platform) *fs.Inode { - d := &Dir{platform: platform} +func NewDir(ctx context.Context, contents map[string]*fs.Inode, owner fs.FileOwner, perms fs.FilePermissions, msrc *fs.MountSource, kernel *kernel.Kernel) *fs.Inode { + d := &Dir{kernel: kernel} d.InitDir(ctx, contents, owner, perms) // Manually set the CreateOps. @@ -84,7 +84,7 @@ func (d *Dir) afterLoad() { func (d *Dir) newCreateOps() *ramfs.CreateOps { return &ramfs.CreateOps{ NewDir: func(ctx context.Context, dir *fs.Inode, perms fs.FilePermissions) (*fs.Inode, error) { - return NewDir(ctx, nil, fs.FileOwnerFromContext(ctx), perms, dir.MountSource, d.platform), nil + return NewDir(ctx, nil, fs.FileOwnerFromContext(ctx), perms, dir.MountSource, d.kernel), nil }, NewFile: func(ctx context.Context, dir *fs.Inode, perms fs.FilePermissions) (*fs.Inode, error) { uattr := fs.WithCurrentTime(ctx, fs.UnstableAttr{ @@ -93,7 +93,7 @@ func (d *Dir) newCreateOps() *ramfs.CreateOps { // Always start unlinked. Links: 0, }) - iops := NewInMemoryFile(ctx, usage.Tmpfs, uattr, d.platform) + iops := NewInMemoryFile(ctx, usage.Tmpfs, uattr, d.kernel) return fs.NewInode(iops, dir.MountSource, fs.StableAttr{ DeviceID: tmpfsDevice.DeviceID(), InodeID: tmpfsDevice.NextIno(), diff --git a/pkg/sentry/kernel/contexttest/BUILD b/pkg/sentry/kernel/contexttest/BUILD new file mode 100644 index 000000000..391986291 --- /dev/null +++ b/pkg/sentry/kernel/contexttest/BUILD @@ -0,0 +1,17 @@ +package(licenses = ["notice"]) # Apache 2.0 + +load("//tools/go_stateify:defs.bzl", "go_library") + +go_library( + name = "contexttest", + testonly = 1, + srcs = ["contexttest.go"], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/sentry/context", + "//pkg/sentry/context/contexttest", + "//pkg/sentry/kernel", + "//pkg/sentry/platform", + ], +) diff --git a/pkg/sentry/kernel/contexttest/contexttest.go b/pkg/sentry/kernel/contexttest/contexttest.go new file mode 100644 index 000000000..9eb18e7e8 --- /dev/null +++ b/pkg/sentry/kernel/contexttest/contexttest.go @@ -0,0 +1,38 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package contexttest provides a test context.Context which includes +// a dummy kernel pointing to a valid platform. +package contexttest + +import ( + "testing" + + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" +) + +// Context returns a Context that may be used in tests. Uses ptrace as the +// platform.Platform, and provides a stub kernel that only serves to point to +// the platform. +func Context(tb testing.TB) context.Context { + ctx := contexttest.Context(tb) + k := &kernel.Kernel{ + Platform: platform.FromContext(ctx), + } + ctx.(*contexttest.TestContext).RegisterValue(kernel.CtxKernel, k) + return ctx +} -- cgit v1.2.3 From 5d8cf31346376eb7c6a93bad3eab7666f145fa0e Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Fri, 14 Dec 2018 18:03:43 -0800 Subject: Move fdnotifier package to reduce internal confusion. PiperOrigin-RevId: 225632398 Change-Id: I909e7e2925aa369adc28e844c284d9a6108e85ce --- pkg/fdnotifier/BUILD | 14 +++ pkg/fdnotifier/fdnotifier.go | 200 ++++++++++++++++++++++++++++++++++ pkg/fdnotifier/poll_unsafe.go | 74 +++++++++++++ pkg/sentry/fs/fdpipe/BUILD | 4 +- pkg/sentry/fs/fdpipe/pipe.go | 2 +- pkg/sentry/fs/fdpipe/pipe_test.go | 2 +- pkg/sentry/fs/host/BUILD | 4 +- pkg/sentry/fs/host/descriptor.go | 2 +- pkg/sentry/fs/host/descriptor_test.go | 2 +- pkg/sentry/fs/host/file.go | 2 +- pkg/sentry/fs/host/socket.go | 2 +- pkg/sentry/fs/host/socket_test.go | 2 +- pkg/sentry/kernel/eventfd/BUILD | 2 +- pkg/sentry/kernel/eventfd/eventfd.go | 2 +- pkg/sentry/socket/hostinet/BUILD | 2 +- pkg/sentry/socket/hostinet/socket.go | 2 +- pkg/waiter/fdnotifier/BUILD | 14 --- pkg/waiter/fdnotifier/fdnotifier.go | 200 ---------------------------------- pkg/waiter/fdnotifier/poll_unsafe.go | 74 ------------- 19 files changed, 303 insertions(+), 303 deletions(-) create mode 100644 pkg/fdnotifier/BUILD create mode 100644 pkg/fdnotifier/fdnotifier.go create mode 100644 pkg/fdnotifier/poll_unsafe.go delete mode 100644 pkg/waiter/fdnotifier/BUILD delete mode 100644 pkg/waiter/fdnotifier/fdnotifier.go delete mode 100644 pkg/waiter/fdnotifier/poll_unsafe.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/fdnotifier/BUILD b/pkg/fdnotifier/BUILD new file mode 100644 index 000000000..27d378d5b --- /dev/null +++ b/pkg/fdnotifier/BUILD @@ -0,0 +1,14 @@ +load("//tools/go_stateify:defs.bzl", "go_library") + +package(licenses = ["notice"]) # Apache 2.0 + +go_library( + name = "fdnotifier", + srcs = [ + "fdnotifier.go", + "poll_unsafe.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/fdnotifier", + visibility = ["//:sandbox"], + deps = ["//pkg/waiter"], +) diff --git a/pkg/fdnotifier/fdnotifier.go b/pkg/fdnotifier/fdnotifier.go new file mode 100644 index 000000000..624b1a0c5 --- /dev/null +++ b/pkg/fdnotifier/fdnotifier.go @@ -0,0 +1,200 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package fdnotifier contains an adapter that translates IO events (e.g., a +// file became readable/writable) from native FDs to the notifications in the +// waiter package. It uses epoll in edge-triggered mode to receive notifications +// for registered FDs. +package fdnotifier + +import ( + "fmt" + "sync" + "syscall" + + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +type fdInfo struct { + queue *waiter.Queue + waiting bool +} + +// notifier holds all the state necessary to issue notifications when IO events +// occur in the observed FDs. +type notifier struct { + // epFD is the epoll file descriptor used to register for io + // notifications. + epFD int + + // mu protects fdMap. + mu sync.Mutex + + // fdMap maps file descriptors to their notification queues and waiting + // status. + fdMap map[int32]*fdInfo +} + +// newNotifier creates a new notifier object. +func newNotifier() (*notifier, error) { + epfd, err := syscall.EpollCreate1(0) + if err != nil { + return nil, err + } + + w := ¬ifier{ + epFD: epfd, + fdMap: make(map[int32]*fdInfo), + } + + go w.waitAndNotify() // S/R-SAFE: no waiter exists during save / load. + + return w, nil +} + +// waitFD waits on mask for fd. The fdMap mutex must be hold. +func (n *notifier) waitFD(fd int32, fi *fdInfo, mask waiter.EventMask) error { + if !fi.waiting && mask == 0 { + return nil + } + + e := syscall.EpollEvent{ + Events: uint32(mask) | -syscall.EPOLLET, + Fd: fd, + } + + switch { + case !fi.waiting && mask != 0: + if err := syscall.EpollCtl(n.epFD, syscall.EPOLL_CTL_ADD, int(fd), &e); err != nil { + return err + } + fi.waiting = true + case fi.waiting && mask == 0: + syscall.EpollCtl(n.epFD, syscall.EPOLL_CTL_DEL, int(fd), nil) + fi.waiting = false + case fi.waiting && mask != 0: + if err := syscall.EpollCtl(n.epFD, syscall.EPOLL_CTL_MOD, int(fd), &e); err != nil { + return err + } + } + + return nil +} + +// addFD adds an FD to the list of FDs observed by n. +func (n *notifier) addFD(fd int32, queue *waiter.Queue) { + n.mu.Lock() + defer n.mu.Unlock() + + // Panic if we're already notifying on this FD. + if _, ok := n.fdMap[fd]; ok { + panic(fmt.Sprintf("File descriptor %v added twice", fd)) + } + + // We have nothing to wait for at the moment. Just add it to the map. + n.fdMap[fd] = &fdInfo{queue: queue} +} + +// updateFD updates the set of events the fd needs to be notified on. +func (n *notifier) updateFD(fd int32) error { + n.mu.Lock() + defer n.mu.Unlock() + + if fi, ok := n.fdMap[fd]; ok { + return n.waitFD(fd, fi, fi.queue.Events()) + } + + return nil +} + +// RemoveFD removes an FD from the list of FDs observed by n. +func (n *notifier) removeFD(fd int32) { + n.mu.Lock() + defer n.mu.Unlock() + + // Remove from map, then from epoll object. + n.waitFD(fd, n.fdMap[fd], 0) + delete(n.fdMap, fd) +} + +// hasFD returns true if the fd is in the list of observed FDs. +func (n *notifier) hasFD(fd int32) bool { + n.mu.Lock() + defer n.mu.Unlock() + + _, ok := n.fdMap[fd] + return ok +} + +// waitAndNotify run is its own goroutine and loops waiting for io event +// notifications from the epoll object. Once notifications arrive, they are +// dispatched to the registered queue. +func (n *notifier) waitAndNotify() error { + e := make([]syscall.EpollEvent, 100) + for { + v, err := epollWait(n.epFD, e, -1) + if err == syscall.EINTR { + continue + } + + if err != nil { + return err + } + + n.mu.Lock() + for i := 0; i < v; i++ { + if fi, ok := n.fdMap[e[i].Fd]; ok { + fi.queue.Notify(waiter.EventMask(e[i].Events)) + } + } + n.mu.Unlock() + } +} + +var shared struct { + notifier *notifier + once sync.Once + initErr error +} + +// AddFD adds an FD to the list of observed FDs. +func AddFD(fd int32, queue *waiter.Queue) error { + shared.once.Do(func() { + shared.notifier, shared.initErr = newNotifier() + }) + + if shared.initErr != nil { + return shared.initErr + } + + shared.notifier.addFD(fd, queue) + return nil +} + +// UpdateFD updates the set of events the fd needs to be notified on. +func UpdateFD(fd int32) error { + return shared.notifier.updateFD(fd) +} + +// RemoveFD removes an FD from the list of observed FDs. +func RemoveFD(fd int32) { + shared.notifier.removeFD(fd) +} + +// HasFD returns true if the FD is in the list of observed FDs. +// +// This should only be used by tests to assert that FDs are correctly registered. +func HasFD(fd int32) bool { + return shared.notifier.hasFD(fd) +} diff --git a/pkg/fdnotifier/poll_unsafe.go b/pkg/fdnotifier/poll_unsafe.go new file mode 100644 index 000000000..8459d4c74 --- /dev/null +++ b/pkg/fdnotifier/poll_unsafe.go @@ -0,0 +1,74 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package fdnotifier + +import ( + "syscall" + "unsafe" + + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// NonBlockingPoll polls the given FD in non-blocking fashion. It is used just +// to query the FD's current state. +func NonBlockingPoll(fd int32, mask waiter.EventMask) waiter.EventMask { + e := struct { + fd int32 + events int16 + revents int16 + }{ + fd: fd, + events: int16(mask), + } + + for { + n, _, err := syscall.RawSyscall(syscall.SYS_POLL, uintptr(unsafe.Pointer(&e)), 1, 0) + // Interrupted by signal, try again. + if err == syscall.EINTR { + continue + } + // If an error occur we'll conservatively say the FD is ready for + // whatever is being checked. + if err != 0 { + return mask + } + + // If no FDs were returned, it wasn't ready for anything. + if n == 0 { + return 0 + } + + // Otherwise we got the ready events in the revents field. + return waiter.EventMask(e.revents) + } +} + +// epollWait performs a blocking wait on epfd. +// +// Preconditions: +// * len(events) > 0 +func epollWait(epfd int, events []syscall.EpollEvent, msec int) (int, error) { + if len(events) == 0 { + panic("Empty events passed to EpollWait") + } + + // We actually use epoll_pwait with NULL sigmask instead of epoll_wait + // since that is what the Go >= 1.11 runtime prefers. + r, _, e := syscall.Syscall6(syscall.SYS_EPOLL_PWAIT, uintptr(epfd), uintptr(unsafe.Pointer(&events[0])), uintptr(len(events)), uintptr(msec), 0, 0) + if e != 0 { + return 0, e + } + return int(r), nil +} diff --git a/pkg/sentry/fs/fdpipe/BUILD b/pkg/sentry/fs/fdpipe/BUILD index ffe4204bc..8a0937cda 100644 --- a/pkg/sentry/fs/fdpipe/BUILD +++ b/pkg/sentry/fs/fdpipe/BUILD @@ -14,6 +14,7 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/fd", + "//pkg/fdnotifier", "//pkg/log", "//pkg/secio", "//pkg/sentry/context", @@ -23,7 +24,6 @@ go_library( "//pkg/sentry/usermem", "//pkg/syserror", "//pkg/waiter", - "//pkg/waiter/fdnotifier", ], ) @@ -37,12 +37,12 @@ go_test( embed = [":fdpipe"], deps = [ "//pkg/fd", + "//pkg/fdnotifier", "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", "//pkg/sentry/usermem", "//pkg/syserror", - "//pkg/waiter/fdnotifier", "@com_github_google_uuid//:go_default_library", ], ) diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index bfafff5ec..e3b830747 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -21,6 +21,7 @@ import ( "syscall" "gvisor.googlesource.com/gvisor/pkg/fd" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/secio" "gvisor.googlesource.com/gvisor/pkg/sentry/context" @@ -30,7 +31,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/waiter" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) // pipeOperations are the fs.FileOperations of a host pipe. diff --git a/pkg/sentry/fs/fdpipe/pipe_test.go b/pkg/sentry/fs/fdpipe/pipe_test.go index d3f15be6b..7e3ee5257 100644 --- a/pkg/sentry/fs/fdpipe/pipe_test.go +++ b/pkg/sentry/fs/fdpipe/pipe_test.go @@ -22,11 +22,11 @@ import ( "testing" "gvisor.googlesource.com/gvisor/pkg/fd" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) func singlePipeFD() (int, error) { diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 73d9cc71a..6877eb161 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -27,6 +27,7 @@ go_library( deps = [ "//pkg/abi/linux", "//pkg/fd", + "//pkg/fdnotifier", "//pkg/log", "//pkg/refs", "//pkg/secio", @@ -51,7 +52,6 @@ go_library( "//pkg/tcpip", "//pkg/unet", "//pkg/waiter", - "//pkg/waiter/fdnotifier", ], ) @@ -68,6 +68,7 @@ go_test( embed = [":host"], deps = [ "//pkg/fd", + "//pkg/fdnotifier", "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", @@ -78,6 +79,5 @@ go_test( "//pkg/syserr", "//pkg/tcpip", "//pkg/waiter", - "//pkg/waiter/fdnotifier", ], ) diff --git a/pkg/sentry/fs/host/descriptor.go b/pkg/sentry/fs/host/descriptor.go index 7c9d2b299..554e1693a 100644 --- a/pkg/sentry/fs/host/descriptor.go +++ b/pkg/sentry/fs/host/descriptor.go @@ -19,9 +19,9 @@ import ( "path" "syscall" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/waiter" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) // descriptor wraps a host fd. diff --git a/pkg/sentry/fs/host/descriptor_test.go b/pkg/sentry/fs/host/descriptor_test.go index 6bc1bd2ae..5dec84ab2 100644 --- a/pkg/sentry/fs/host/descriptor_test.go +++ b/pkg/sentry/fs/host/descriptor_test.go @@ -20,8 +20,8 @@ import ( "syscall" "testing" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/waiter" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) func TestDescriptorRelease(t *testing.T) { diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index 975084c86..bc6ee7aa4 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -19,6 +19,7 @@ import ( "syscall" "gvisor.googlesource.com/gvisor/pkg/fd" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/secio" "gvisor.googlesource.com/gvisor/pkg/sentry/context" @@ -29,7 +30,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/waiter" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) // fileOperations implements fs.FileOperations for a host file descriptor. diff --git a/pkg/sentry/fs/host/socket.go b/pkg/sentry/fs/host/socket.go index b9e2aa705..be2c3581f 100644 --- a/pkg/sentry/fs/host/socket.go +++ b/pkg/sentry/fs/host/socket.go @@ -19,6 +19,7 @@ import ( "syscall" "gvisor.googlesource.com/gvisor/pkg/fd" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/sentry/context" @@ -32,7 +33,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/unet" "gvisor.googlesource.com/gvisor/pkg/waiter" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) // maxSendBufferSize is the maximum host send buffer size allowed for endpoint. diff --git a/pkg/sentry/fs/host/socket_test.go b/pkg/sentry/fs/host/socket_test.go index 6ddf63a6a..83e8e1b3c 100644 --- a/pkg/sentry/fs/host/socket_test.go +++ b/pkg/sentry/fs/host/socket_test.go @@ -20,6 +20,7 @@ import ( "testing" "gvisor.googlesource.com/gvisor/pkg/fd" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/socket" @@ -28,7 +29,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/tcpip" "gvisor.googlesource.com/gvisor/pkg/waiter" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) var ( diff --git a/pkg/sentry/kernel/eventfd/BUILD b/pkg/sentry/kernel/eventfd/BUILD index cc1120b4f..d96803fc9 100644 --- a/pkg/sentry/kernel/eventfd/BUILD +++ b/pkg/sentry/kernel/eventfd/BUILD @@ -9,6 +9,7 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/fdnotifier", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", @@ -16,7 +17,6 @@ go_library( "//pkg/sentry/usermem", "//pkg/syserror", "//pkg/waiter", - "//pkg/waiter/fdnotifier", ], ) diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index 26dc59a85..063a1d5f5 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -22,6 +22,7 @@ import ( "syscall" "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/anon" @@ -29,7 +30,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/waiter" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) // EventOperations represents an event with the semantics of Linux's file-based event diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD index c30220a46..b8dceb102 100644 --- a/pkg/sentry/socket/hostinet/BUILD +++ b/pkg/sentry/socket/hostinet/BUILD @@ -17,6 +17,7 @@ go_library( deps = [ "//pkg/abi/linux", "//pkg/binary", + "//pkg/fdnotifier", "//pkg/log", "//pkg/sentry/arch", "//pkg/sentry/context", @@ -34,6 +35,5 @@ go_library( "//pkg/syserr", "//pkg/syserror", "//pkg/waiter", - "//pkg/waiter/fdnotifier", ], ) diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index 34281cac0..f3ecb6dc3 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -19,6 +19,7 @@ import ( "syscall" "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/fdnotifier" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" @@ -32,7 +33,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/waiter" - "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier" ) const ( diff --git a/pkg/waiter/fdnotifier/BUILD b/pkg/waiter/fdnotifier/BUILD deleted file mode 100644 index af6baa303..000000000 --- a/pkg/waiter/fdnotifier/BUILD +++ /dev/null @@ -1,14 +0,0 @@ -load("//tools/go_stateify:defs.bzl", "go_library") - -package(licenses = ["notice"]) # Apache 2.0 - -go_library( - name = "fdnotifier", - srcs = [ - "fdnotifier.go", - "poll_unsafe.go", - ], - importpath = "gvisor.googlesource.com/gvisor/pkg/waiter/fdnotifier", - visibility = ["//:sandbox"], - deps = ["//pkg/waiter"], -) diff --git a/pkg/waiter/fdnotifier/fdnotifier.go b/pkg/waiter/fdnotifier/fdnotifier.go deleted file mode 100644 index 624b1a0c5..000000000 --- a/pkg/waiter/fdnotifier/fdnotifier.go +++ /dev/null @@ -1,200 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package fdnotifier contains an adapter that translates IO events (e.g., a -// file became readable/writable) from native FDs to the notifications in the -// waiter package. It uses epoll in edge-triggered mode to receive notifications -// for registered FDs. -package fdnotifier - -import ( - "fmt" - "sync" - "syscall" - - "gvisor.googlesource.com/gvisor/pkg/waiter" -) - -type fdInfo struct { - queue *waiter.Queue - waiting bool -} - -// notifier holds all the state necessary to issue notifications when IO events -// occur in the observed FDs. -type notifier struct { - // epFD is the epoll file descriptor used to register for io - // notifications. - epFD int - - // mu protects fdMap. - mu sync.Mutex - - // fdMap maps file descriptors to their notification queues and waiting - // status. - fdMap map[int32]*fdInfo -} - -// newNotifier creates a new notifier object. -func newNotifier() (*notifier, error) { - epfd, err := syscall.EpollCreate1(0) - if err != nil { - return nil, err - } - - w := ¬ifier{ - epFD: epfd, - fdMap: make(map[int32]*fdInfo), - } - - go w.waitAndNotify() // S/R-SAFE: no waiter exists during save / load. - - return w, nil -} - -// waitFD waits on mask for fd. The fdMap mutex must be hold. -func (n *notifier) waitFD(fd int32, fi *fdInfo, mask waiter.EventMask) error { - if !fi.waiting && mask == 0 { - return nil - } - - e := syscall.EpollEvent{ - Events: uint32(mask) | -syscall.EPOLLET, - Fd: fd, - } - - switch { - case !fi.waiting && mask != 0: - if err := syscall.EpollCtl(n.epFD, syscall.EPOLL_CTL_ADD, int(fd), &e); err != nil { - return err - } - fi.waiting = true - case fi.waiting && mask == 0: - syscall.EpollCtl(n.epFD, syscall.EPOLL_CTL_DEL, int(fd), nil) - fi.waiting = false - case fi.waiting && mask != 0: - if err := syscall.EpollCtl(n.epFD, syscall.EPOLL_CTL_MOD, int(fd), &e); err != nil { - return err - } - } - - return nil -} - -// addFD adds an FD to the list of FDs observed by n. -func (n *notifier) addFD(fd int32, queue *waiter.Queue) { - n.mu.Lock() - defer n.mu.Unlock() - - // Panic if we're already notifying on this FD. - if _, ok := n.fdMap[fd]; ok { - panic(fmt.Sprintf("File descriptor %v added twice", fd)) - } - - // We have nothing to wait for at the moment. Just add it to the map. - n.fdMap[fd] = &fdInfo{queue: queue} -} - -// updateFD updates the set of events the fd needs to be notified on. -func (n *notifier) updateFD(fd int32) error { - n.mu.Lock() - defer n.mu.Unlock() - - if fi, ok := n.fdMap[fd]; ok { - return n.waitFD(fd, fi, fi.queue.Events()) - } - - return nil -} - -// RemoveFD removes an FD from the list of FDs observed by n. -func (n *notifier) removeFD(fd int32) { - n.mu.Lock() - defer n.mu.Unlock() - - // Remove from map, then from epoll object. - n.waitFD(fd, n.fdMap[fd], 0) - delete(n.fdMap, fd) -} - -// hasFD returns true if the fd is in the list of observed FDs. -func (n *notifier) hasFD(fd int32) bool { - n.mu.Lock() - defer n.mu.Unlock() - - _, ok := n.fdMap[fd] - return ok -} - -// waitAndNotify run is its own goroutine and loops waiting for io event -// notifications from the epoll object. Once notifications arrive, they are -// dispatched to the registered queue. -func (n *notifier) waitAndNotify() error { - e := make([]syscall.EpollEvent, 100) - for { - v, err := epollWait(n.epFD, e, -1) - if err == syscall.EINTR { - continue - } - - if err != nil { - return err - } - - n.mu.Lock() - for i := 0; i < v; i++ { - if fi, ok := n.fdMap[e[i].Fd]; ok { - fi.queue.Notify(waiter.EventMask(e[i].Events)) - } - } - n.mu.Unlock() - } -} - -var shared struct { - notifier *notifier - once sync.Once - initErr error -} - -// AddFD adds an FD to the list of observed FDs. -func AddFD(fd int32, queue *waiter.Queue) error { - shared.once.Do(func() { - shared.notifier, shared.initErr = newNotifier() - }) - - if shared.initErr != nil { - return shared.initErr - } - - shared.notifier.addFD(fd, queue) - return nil -} - -// UpdateFD updates the set of events the fd needs to be notified on. -func UpdateFD(fd int32) error { - return shared.notifier.updateFD(fd) -} - -// RemoveFD removes an FD from the list of observed FDs. -func RemoveFD(fd int32) { - shared.notifier.removeFD(fd) -} - -// HasFD returns true if the FD is in the list of observed FDs. -// -// This should only be used by tests to assert that FDs are correctly registered. -func HasFD(fd int32) bool { - return shared.notifier.hasFD(fd) -} diff --git a/pkg/waiter/fdnotifier/poll_unsafe.go b/pkg/waiter/fdnotifier/poll_unsafe.go deleted file mode 100644 index 8459d4c74..000000000 --- a/pkg/waiter/fdnotifier/poll_unsafe.go +++ /dev/null @@ -1,74 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package fdnotifier - -import ( - "syscall" - "unsafe" - - "gvisor.googlesource.com/gvisor/pkg/waiter" -) - -// NonBlockingPoll polls the given FD in non-blocking fashion. It is used just -// to query the FD's current state. -func NonBlockingPoll(fd int32, mask waiter.EventMask) waiter.EventMask { - e := struct { - fd int32 - events int16 - revents int16 - }{ - fd: fd, - events: int16(mask), - } - - for { - n, _, err := syscall.RawSyscall(syscall.SYS_POLL, uintptr(unsafe.Pointer(&e)), 1, 0) - // Interrupted by signal, try again. - if err == syscall.EINTR { - continue - } - // If an error occur we'll conservatively say the FD is ready for - // whatever is being checked. - if err != 0 { - return mask - } - - // If no FDs were returned, it wasn't ready for anything. - if n == 0 { - return 0 - } - - // Otherwise we got the ready events in the revents field. - return waiter.EventMask(e.revents) - } -} - -// epollWait performs a blocking wait on epfd. -// -// Preconditions: -// * len(events) > 0 -func epollWait(epfd int, events []syscall.EpollEvent, msec int) (int, error) { - if len(events) == 0 { - panic("Empty events passed to EpollWait") - } - - // We actually use epoll_pwait with NULL sigmask instead of epoll_wait - // since that is what the Go >= 1.11 runtime prefers. - r, _, e := syscall.Syscall6(syscall.SYS_EPOLL_PWAIT, uintptr(epfd), uintptr(unsafe.Pointer(&events[0])), uintptr(len(events)), uintptr(msec), 0, 0) - if e != 0 { - return 0, e - } - return int(r), nil -} -- cgit v1.2.3 From 03226cd95055aee73d4e4dfcb4954490b4fd8a2d Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Tue, 18 Dec 2018 10:27:16 -0800 Subject: Add BPFAction type with Stringer PiperOrigin-RevId: 226018694 Change-Id: I98965e26fe565f37e98e5df5f997363ab273c91b --- pkg/abi/linux/seccomp.go | 48 ++++++++++++++++++----- pkg/seccomp/seccomp.go | 32 ++++++---------- pkg/seccomp/seccomp_test.go | 14 +++---- pkg/sentry/kernel/seccomp.go | 53 +++++++++----------------- pkg/sentry/kernel/task_syscall.go | 18 +++++---- pkg/sentry/platform/ptrace/subprocess_linux.go | 16 ++++---- test/syscalls/linux/seccomp.cc | 19 +++++++++ 7 files changed, 112 insertions(+), 88 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/seccomp.go b/pkg/abi/linux/seccomp.go index 785f2f284..8673a27bf 100644 --- a/pkg/abi/linux/seccomp.go +++ b/pkg/abi/linux/seccomp.go @@ -14,22 +14,52 @@ package linux +import "fmt" + // Seccomp constants taken from . const ( SECCOMP_MODE_NONE = 0 SECCOMP_MODE_FILTER = 2 - SECCOMP_RET_KILL_PROCESS = 0x80000000 - SECCOMP_RET_KILL_THREAD = 0x00000000 - SECCOMP_RET_TRAP = 0x00030000 - SECCOMP_RET_ERRNO = 0x00050000 - SECCOMP_RET_TRACE = 0x7ff00000 - SECCOMP_RET_ALLOW = 0x7fff0000 - - SECCOMP_RET_ACTION = 0x7fff0000 - SECCOMP_RET_DATA = 0x0000ffff + SECCOMP_RET_ACTION_FULL = 0xffff0000 + SECCOMP_RET_ACTION = 0x7fff0000 + SECCOMP_RET_DATA = 0x0000ffff SECCOMP_SET_MODE_FILTER = 1 SECCOMP_FILTER_FLAG_TSYNC = 1 SECCOMP_GET_ACTION_AVAIL = 2 ) + +type BPFAction uint32 + +const ( + SECCOMP_RET_KILL_PROCESS BPFAction = 0x80000000 + SECCOMP_RET_KILL_THREAD = 0x00000000 + SECCOMP_RET_TRAP = 0x00030000 + SECCOMP_RET_ERRNO = 0x00050000 + SECCOMP_RET_TRACE = 0x7ff00000 + SECCOMP_RET_ALLOW = 0x7fff0000 +) + +func (a BPFAction) String() string { + switch a & SECCOMP_RET_ACTION_FULL { + case SECCOMP_RET_KILL_PROCESS: + return "kill process" + case SECCOMP_RET_KILL_THREAD: + return "kill thread" + case SECCOMP_RET_TRAP: + return fmt.Sprintf("trap (%d)", a.Data()) + case SECCOMP_RET_ERRNO: + return fmt.Sprintf("errno (%d)", a.Data()) + case SECCOMP_RET_TRACE: + return fmt.Sprintf("trace (%d)", a.Data()) + case SECCOMP_RET_ALLOW: + return "allow" + } + return fmt.Sprintf("invalid action: %#x", a) +} + +// Data returns the SECCOMP_RET_DATA portion of the action. +func (a BPFAction) Data() uint16 { + return uint16(a & SECCOMP_RET_DATA) +} diff --git a/pkg/seccomp/seccomp.go b/pkg/seccomp/seccomp.go index 9d714d02d..ba2955752 100644 --- a/pkg/seccomp/seccomp.go +++ b/pkg/seccomp/seccomp.go @@ -33,16 +33,6 @@ const ( defaultLabel = "default_action" ) -func actionName(a uint32) string { - switch a { - case linux.SECCOMP_RET_KILL_PROCESS: - return "kill process" - case linux.SECCOMP_RET_TRAP: - return "trap" - } - panic(fmt.Sprintf("invalid action: %d", a)) -} - // Install generates BPF code based on the set of syscalls provided. It only // allows syscalls that conform to the specification. Syscalls that violate the // specification will trigger RET_KILL_PROCESS, except for the cases below. @@ -67,12 +57,12 @@ func Install(rules SyscallRules) error { // Uncomment to get stack trace when there is a violation. // defaultAction = uint32(linux.SECCOMP_RET_TRAP) - log.Infof("Installing seccomp filters for %d syscalls (action=%s)", len(rules), actionName(defaultAction)) + log.Infof("Installing seccomp filters for %d syscalls (action=%v)", len(rules), defaultAction) instrs, err := BuildProgram([]RuleSet{ RuleSet{ Rules: rules, - Action: uint32(linux.SECCOMP_RET_ALLOW), + Action: linux.SECCOMP_RET_ALLOW, }, }, defaultAction) if log.IsLogging(log.Debug) { @@ -95,21 +85,21 @@ func Install(rules SyscallRules) error { return nil } -func defaultAction() (uint32, error) { +func defaultAction() (linux.BPFAction, error) { available, err := isKillProcessAvailable() if err != nil { return 0, err } if available { - return uint32(linux.SECCOMP_RET_KILL_PROCESS), nil + return linux.SECCOMP_RET_KILL_PROCESS, nil } - return uint32(linux.SECCOMP_RET_TRAP), nil + return linux.SECCOMP_RET_TRAP, nil } // RuleSet is a set of rules and associated action. type RuleSet struct { Rules SyscallRules - Action uint32 + Action linux.BPFAction // Vsyscall indicates that a check is made for a function being called // from kernel mappings. This is where the vsyscall page is located @@ -127,7 +117,7 @@ var SyscallName = func(sysno uintptr) string { // BuildProgram builds a BPF program from the given map of actions to matching // SyscallRules. The single generated program covers all provided RuleSets. -func BuildProgram(rules []RuleSet, defaultAction uint32) ([]linux.BPFInstruction, error) { +func BuildProgram(rules []RuleSet, defaultAction linux.BPFAction) ([]linux.BPFInstruction, error) { program := bpf.NewProgramBuilder() // Be paranoid and check that syscall is done in the expected architecture. @@ -147,7 +137,7 @@ func BuildProgram(rules []RuleSet, defaultAction uint32) ([]linux.BPFInstruction if err := program.AddLabel(defaultLabel); err != nil { return nil, err } - program.AddStmt(bpf.Ret|bpf.K, defaultAction) + program.AddStmt(bpf.Ret|bpf.K, uint32(defaultAction)) return program.Instructions() } @@ -217,7 +207,7 @@ func checkArgsLabel(sysno uintptr) string { // not insert a jump to the default action at the end and it is the // responsibility of the caller to insert an appropriate jump after calling // this function. -func addSyscallArgsCheck(p *bpf.ProgramBuilder, rules []Rule, action uint32, ruleSetIdx int, sysno uintptr) error { +func addSyscallArgsCheck(p *bpf.ProgramBuilder, rules []Rule, action linux.BPFAction, ruleSetIdx int, sysno uintptr) error { for ruleidx, rule := range rules { labelled := false for i, arg := range rule { @@ -240,7 +230,7 @@ func addSyscallArgsCheck(p *bpf.ProgramBuilder, rules []Rule, action uint32, rul } // Matched, emit the given action. - p.AddStmt(bpf.Ret|bpf.K, action) + p.AddStmt(bpf.Ret|bpf.K, uint32(action)) // Label the end of the rule if necessary. This is added for // the jumps above when the argument check fails. @@ -319,7 +309,7 @@ func buildBSTProgram(n *node, rules []RuleSet, program *bpf.ProgramBuilder) erro // Emit matchers. if len(rs.Rules[sysno]) == 0 { // This is a blanket action. - program.AddStmt(bpf.Ret|bpf.K, rs.Action) + program.AddStmt(bpf.Ret|bpf.K, uint32(rs.Action)) emitted = true } else { // Add an argument check for these particular diff --git a/pkg/seccomp/seccomp_test.go b/pkg/seccomp/seccomp_test.go index f2b903e42..11ed90eb4 100644 --- a/pkg/seccomp/seccomp_test.go +++ b/pkg/seccomp/seccomp_test.go @@ -72,12 +72,12 @@ func TestBasic(t *testing.T) { data seccompData // want is the expected return value of the BPF program. - want uint32 + want linux.BPFAction } for _, test := range []struct { ruleSets []RuleSet - defaultAction uint32 + defaultAction linux.BPFAction specs []spec }{ { @@ -357,7 +357,7 @@ func TestBasic(t *testing.T) { t.Errorf("%s: bpf.Exec() got error: %v", spec.desc, err) continue } - if got != spec.want { + if got != uint32(spec.want) { t.Errorf("%s: bpd.Exec() = %d, want: %d", spec.desc, got, spec.want) } } @@ -380,9 +380,9 @@ func TestRandom(t *testing.T) { instrs, err := BuildProgram([]RuleSet{ RuleSet{ Rules: syscallRules, - Action: uint32(linux.SECCOMP_RET_ALLOW), + Action: linux.SECCOMP_RET_ALLOW, }, - }, uint32(linux.SECCOMP_RET_TRAP)) + }, linux.SECCOMP_RET_TRAP) if err != nil { t.Fatalf("buildProgram() got error: %v", err) } @@ -397,11 +397,11 @@ func TestRandom(t *testing.T) { t.Errorf("bpf.Exec() got error: %v, for syscall %d", err, i) continue } - want := uint32(linux.SECCOMP_RET_TRAP) + want := linux.SECCOMP_RET_TRAP if _, ok := syscallRules[uintptr(i)]; ok { want = linux.SECCOMP_RET_ALLOW } - if got != want { + if got != uint32(want) { t.Errorf("bpf.Exec() = %d, want: %d, for syscall %d", got, want, i) } } diff --git a/pkg/sentry/kernel/seccomp.go b/pkg/sentry/kernel/seccomp.go index d6dc45bbd..cec179246 100644 --- a/pkg/sentry/kernel/seccomp.go +++ b/pkg/sentry/kernel/seccomp.go @@ -27,24 +27,6 @@ import ( const maxSyscallFilterInstructions = 1 << 15 -type seccompResult int - -const ( - // seccompResultDeny indicates that a syscall should not be executed. - seccompResultDeny seccompResult = iota - - // seccompResultAllow indicates that a syscall should be executed. - seccompResultAllow - - // seccompResultKill indicates that the task should be killed immediately, - // with the exit status indicating that the task was killed by SIGSYS. - seccompResultKill - - // seccompResultTrace indicates that a ptracer was successfully notified as - // a result of a SECCOMP_RET_TRACE. - seccompResultTrace -) - // seccompData is equivalent to struct seccomp_data, which contains the data // passed to seccomp-bpf filters. type seccompData struct { @@ -83,48 +65,47 @@ func seccompSiginfo(t *Task, errno, sysno int32, ip usermem.Addr) *arch.SignalIn // in because vsyscalls do not use the values in t.Arch().) // // Preconditions: The caller must be running on the task goroutine. -func (t *Task) checkSeccompSyscall(sysno int32, args arch.SyscallArguments, ip usermem.Addr) seccompResult { - result := t.evaluateSyscallFilters(sysno, args, ip) - switch result & linux.SECCOMP_RET_ACTION { +func (t *Task) checkSeccompSyscall(sysno int32, args arch.SyscallArguments, ip usermem.Addr) linux.BPFAction { + result := linux.BPFAction(t.evaluateSyscallFilters(sysno, args, ip)) + action := result & linux.SECCOMP_RET_ACTION + switch action { case linux.SECCOMP_RET_TRAP: // "Results in the kernel sending a SIGSYS signal to the triggering // task without executing the system call. ... The SECCOMP_RET_DATA // portion of the return value will be passed as si_errno." - // Documentation/prctl/seccomp_filter.txt - t.SendSignal(seccompSiginfo(t, int32(result&linux.SECCOMP_RET_DATA), sysno, ip)) - return seccompResultDeny + t.SendSignal(seccompSiginfo(t, int32(result.Data()), sysno, ip)) case linux.SECCOMP_RET_ERRNO: // "Results in the lower 16-bits of the return value being passed to // userland as the errno without executing the system call." - t.Arch().SetReturn(-uintptr(result & linux.SECCOMP_RET_DATA)) - return seccompResultDeny + t.Arch().SetReturn(-uintptr(result.Data())) case linux.SECCOMP_RET_TRACE: // "When returned, this value will cause the kernel to attempt to // notify a ptrace()-based tracer prior to executing the system call. // If there is no tracer present, -ENOSYS is returned to userland and // the system call is not executed." - if t.ptraceSeccomp(uint16(result & linux.SECCOMP_RET_DATA)) { - return seccompResultTrace + if !t.ptraceSeccomp(result.Data()) { + // This useless-looking temporary is needed because Go. + tmp := uintptr(syscall.ENOSYS) + t.Arch().SetReturn(-tmp) + return linux.SECCOMP_RET_ERRNO } - // This useless-looking temporary is needed because Go. - tmp := uintptr(syscall.ENOSYS) - t.Arch().SetReturn(-tmp) - return seccompResultDeny case linux.SECCOMP_RET_ALLOW: // "Results in the system call being executed." - return seccompResultAllow case linux.SECCOMP_RET_KILL_THREAD: // "Results in the task exiting immediately without executing the // system call. The exit status of the task will be SIGSYS, not // SIGKILL." - fallthrough - default: // consistent with Linux - return seccompResultKill + + default: + // consistent with Linux + return linux.SECCOMP_RET_KILL_THREAD } + return action } func (t *Task) evaluateSyscallFilters(sysno int32, args arch.SyscallArguments, ip usermem.Addr) uint32 { @@ -155,7 +136,7 @@ func (t *Task) evaluateSyscallFilters(sysno int32, args arch.SyscallArguments, i thisRet, err := bpf.Exec(f.([]bpf.Program)[i], input) if err != nil { t.Debugf("seccomp-bpf filter %d returned error: %v", i, err) - thisRet = linux.SECCOMP_RET_KILL_THREAD + thisRet = uint32(linux.SECCOMP_RET_KILL_THREAD) } // "If multiple filters exist, the return value for the evaluation of a // given system call will always use the highest precedent value." - diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index 2a39ebc68..9e43f089a 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -199,16 +199,16 @@ func (t *Task) doSyscall() taskRunState { // is rare), not needed for correctness. if t.syscallFilters.Load() != nil { switch r := t.checkSeccompSyscall(int32(sysno), args, usermem.Addr(t.Arch().IP())); r { - case seccompResultDeny: + case linux.SECCOMP_RET_ERRNO, linux.SECCOMP_RET_TRAP: t.Debugf("Syscall %d: denied by seccomp", sysno) return (*runSyscallExit)(nil) - case seccompResultAllow: + case linux.SECCOMP_RET_ALLOW: // ok - case seccompResultKill: + case linux.SECCOMP_RET_KILL_THREAD: t.Debugf("Syscall %d: killed by seccomp", sysno) t.PrepareExit(ExitStatus{Signo: int(linux.SIGSYS)}) return (*runExit)(nil) - case seccompResultTrace: + case linux.SECCOMP_RET_TRACE: t.Debugf("Syscall %d: stopping for PTRACE_EVENT_SECCOMP", sysno) return (*runSyscallAfterPtraceEventSeccomp)(nil) default: @@ -345,14 +345,18 @@ func (t *Task) doVsyscall(addr usermem.Addr, sysno uintptr) taskRunState { args := t.Arch().SyscallArgs() if t.syscallFilters.Load() != nil { switch r := t.checkSeccompSyscall(int32(sysno), args, addr); r { - case seccompResultDeny: + case linux.SECCOMP_RET_ERRNO, linux.SECCOMP_RET_TRAP: t.Debugf("vsyscall %d, caller %x: denied by seccomp", sysno, t.Arch().Value(caller)) return (*runApp)(nil) - case seccompResultAllow: + case linux.SECCOMP_RET_ALLOW: // ok - case seccompResultTrace: + case linux.SECCOMP_RET_TRACE: t.Debugf("vsyscall %d, caller %x: stopping for PTRACE_EVENT_SECCOMP", sysno, t.Arch().Value(caller)) return &runVsyscallAfterPtraceEventSeccomp{addr, sysno, caller} + case linux.SECCOMP_RET_KILL_THREAD: + t.Debugf("vsyscall %d: killed by seccomp", sysno) + t.PrepareExit(ExitStatus{Signo: int(linux.SIGSYS)}) + return (*runExit)(nil) default: panic(fmt.Sprintf("Unknown seccomp result %d", r)) } diff --git a/pkg/sentry/platform/ptrace/subprocess_linux.go b/pkg/sentry/platform/ptrace/subprocess_linux.go index 25b8e8cb7..e2aab8135 100644 --- a/pkg/sentry/platform/ptrace/subprocess_linux.go +++ b/pkg/sentry/platform/ptrace/subprocess_linux.go @@ -38,7 +38,7 @@ const syscallEvent syscall.Signal = 0x80 // Precondition: the runtime OS thread must be locked. func probeSeccomp() bool { // Create a completely new, destroyable process. - t, err := attachedThread(0, uint32(linux.SECCOMP_RET_ERRNO)) + t, err := attachedThread(0, linux.SECCOMP_RET_ERRNO) if err != nil { panic(fmt.Sprintf("seccomp probe failed: %v", err)) } @@ -112,14 +112,14 @@ func createStub() (*thread, error) { // ptrace emulation check. This simplifies using SYSEMU, since seccomp // will never run for emulation. Seccomp will only run for injected // system calls, and thus we can use RET_KILL as our violation action. - var defaultAction uint32 + var defaultAction linux.BPFAction if probeSeccomp() { log.Infof("Latest seccomp behavior found (kernel >= 4.8 likely)") - defaultAction = uint32(linux.SECCOMP_RET_KILL_THREAD) + defaultAction = linux.SECCOMP_RET_KILL_THREAD } else { // We must rely on SYSEMU behavior; tracing with SYSEMU is broken. log.Infof("Legacy seccomp behavior found (kernel < 4.8 likely)") - defaultAction = uint32(linux.SECCOMP_RET_ALLOW) + defaultAction = linux.SECCOMP_RET_ALLOW } // When creating the new child process, we specify SIGKILL as the @@ -135,7 +135,7 @@ func createStub() (*thread, error) { // attachedThread returns a new attached thread. // // Precondition: the runtime OS thread must be locked. -func attachedThread(flags uintptr, defaultAction uint32) (*thread, error) { +func attachedThread(flags uintptr, defaultAction linux.BPFAction) (*thread, error) { // Create a BPF program that allows only the system calls needed by the // stub and all its children. This is used to create child stubs // (below), so we must include the ability to fork, but otherwise lock @@ -148,11 +148,11 @@ func attachedThread(flags uintptr, defaultAction uint32) (*thread, error) { syscall.SYS_TIME: {}, 309: {}, // SYS_GETCPU. }, - Action: uint32(linux.SECCOMP_RET_TRAP), + Action: linux.SECCOMP_RET_TRAP, Vsyscall: true, }, } - if defaultAction != uint32(linux.SECCOMP_RET_ALLOW) { + if defaultAction != linux.SECCOMP_RET_ALLOW { rules = append(rules, seccomp.RuleSet{ Rules: seccomp.SyscallRules{ syscall.SYS_CLONE: []seccomp.Rule{ @@ -191,7 +191,7 @@ func attachedThread(flags uintptr, defaultAction uint32) (*thread, error) { syscall.SYS_MMAP: {}, syscall.SYS_MUNMAP: {}, }, - Action: uint32(linux.SECCOMP_RET_ALLOW), + Action: linux.SECCOMP_RET_ALLOW, }) } instrs, err := seccomp.BuildProgram(rules, defaultAction) diff --git a/test/syscalls/linux/seccomp.cc b/test/syscalls/linux/seccomp.cc index d6ac166a4..ac416b75f 100644 --- a/test/syscalls/linux/seccomp.cc +++ b/test/syscalls/linux/seccomp.cc @@ -215,6 +215,25 @@ TEST(SeccompTest, SeccompAppliesToVsyscall) { << "status " << status; } +TEST(SeccompTest, RetKillVsyscallCausesDeathBySIGSYS) { + SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(IsVsyscallEnabled())); + + pid_t const pid = fork(); + if (pid == 0) { + // Register a signal handler for SIGSYS that we don't expect to be invoked. + RegisterSignalHandler( + SIGSYS, +[](int, siginfo_t*, void*) { _exit(1); }); + ApplySeccompFilter(SYS_time, SECCOMP_RET_KILL); + vsyscall_time(nullptr); // Should result in death. + TEST_CHECK_MSG(false, "Survived invocation of test syscall"); + } + ASSERT_THAT(pid, SyscallSucceeds()); + int status; + ASSERT_THAT(waitpid(pid, &status, 0), SyscallSucceedsWithValue(pid)); + EXPECT_TRUE(WIFSIGNALED(status) && WTERMSIG(status) == SIGSYS) + << "status " << status; +} + TEST(SeccompTest, RetTraceWithoutPtracerReturnsENOSYS) { pid_t const pid = fork(); if (pid == 0) { -- cgit v1.2.3 From bfa2f314ca05854b0d08aa2f5c2b93b16542d95f Mon Sep 17 00:00:00 2001 From: Brian Geffon Date: Wed, 26 Dec 2018 11:24:29 -0800 Subject: Add EventChannel messages for uncaught signals. PiperOrigin-RevId: 226936778 Change-Id: I2a6dda157c55d39d81e1b543ab11a58a0bfe5c05 --- pkg/sentry/kernel/BUILD | 18 ++++++++++++++++ pkg/sentry/kernel/task_signals.go | 19 +++++++++++++++++ pkg/sentry/kernel/uncaught_signal.proto | 37 +++++++++++++++++++++++++++++++++ 3 files changed, 74 insertions(+) create mode 100644 pkg/sentry/kernel/uncaught_signal.proto (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 10d7b97c2..490f674c0 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -1,3 +1,5 @@ +load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") + package(licenses = ["notice"]) # Apache 2.0 load("//tools/go_generics:defs.bzl", "go_template_instance") @@ -62,6 +64,21 @@ go_template_instance( }, ) +proto_library( + name = "uncaught_signal_proto", + srcs = ["uncaught_signal.proto"], + visibility = ["//visibility:public"], + deps = ["//pkg/sentry/arch:registers_proto"], +) + +go_proto_library( + name = "uncaught_signal_go_proto", + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/uncaught_signal_go_proto", + proto = ":uncaught_signal_proto", + visibility = ["//visibility:public"], + deps = ["//pkg/sentry/arch:registers_go_proto"], +) + go_library( name = "kernel", srcs = [ @@ -122,6 +139,7 @@ go_library( ], visibility = ["//:sandbox"], deps = [ + ":uncaught_signal_go_proto", "//pkg/abi", "//pkg/abi/linux", "//pkg/amutex", diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index fe24f7542..583acddb1 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -22,8 +22,10 @@ import ( "time" "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/eventchannel" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + ucspb "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/uncaught_signal_go_proto" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -184,6 +186,23 @@ func (t *Task) deliverSignal(info *arch.SignalInfo, act arch.SignalAct) taskRunS case SignalActionTerm, SignalActionCore: // "Default action is to terminate the process." - signal(7) t.Debugf("Signal %d: terminating thread group", info.Signo) + + // Emit an event channel messages related to this uncaught signal. + ucs := &ucspb.UncaughtSignal{ + Tid: int32(t.Kernel().TaskSet().Root.IDOfTask(t)), + Pid: int32(t.Kernel().TaskSet().Root.IDOfThreadGroup(t.ThreadGroup())), + Registers: t.Arch().StateData().Proto(), + SignalNumber: info.Signo, + } + + // Attach an fault address if appropriate. + switch linux.Signal(info.Signo) { + case linux.SIGSEGV, linux.SIGFPE, linux.SIGILL, linux.SIGTRAP, linux.SIGBUS: + ucs.FaultAddr = info.Addr() + } + + eventchannel.Emit(ucs) + t.PrepareGroupExit(ExitStatus{Signo: int(info.Signo)}) return (*runExit)(nil) diff --git a/pkg/sentry/kernel/uncaught_signal.proto b/pkg/sentry/kernel/uncaught_signal.proto new file mode 100644 index 000000000..c7f6a1978 --- /dev/null +++ b/pkg/sentry/kernel/uncaught_signal.proto @@ -0,0 +1,37 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +syntax = "proto3"; + +package gvisor; + +import "pkg/sentry/arch/registers.proto"; + +message UncaughtSignal { + // Thread ID. + int32 tid = 1; + + // Process ID. + int32 pid = 2; + + // Registers at the time of the fault or signal. + Registers registers = 3; + + // Signal number. + int32 signal_number = 4; + + // The memory location which caused the fault (set if applicable, 0 + // otherwise). This will be set for SIGILL, SIGFPE, SIGSEGV, and SIGBUS. + uint64 fault_addr = 5; +} -- cgit v1.2.3 From f95b94fbe3e557b16ed2b78c87e8936c0aeab6c5 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Tue, 8 Jan 2019 12:51:04 -0800 Subject: Grant no initial capabilities to non-root UIDs. See modified comment in auth.NewUserCredentials(); compare to the behavior of setresuid(2) as implemented by //pkg/sentry/kernel/task_identity.go:kernel.Task.setKUIDsUncheckedLocked(). PiperOrigin-RevId: 228381765 Change-Id: I45238777c8f63fcf41b99fce3969caaf682fe408 --- pkg/sentry/kernel/auth/credentials.go | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go index de33f1953..a843b9aab 100644 --- a/pkg/sentry/kernel/auth/credentials.go +++ b/pkg/sentry/kernel/auth/credentials.go @@ -119,19 +119,24 @@ func NewUserCredentials(kuid KUID, kgid KGID, extraKGIDs []KGID, capabilities *T // Set additional GIDs. creds.ExtraKGIDs = append(creds.ExtraKGIDs, extraKGIDs...) - // Set capabilities. If capabilities aren't specified, we default to - // all capabilities. + // Set capabilities. if capabilities != nil { creds.PermittedCaps = capabilities.PermittedCaps creds.EffectiveCaps = capabilities.EffectiveCaps creds.BoundingCaps = capabilities.BoundingCaps creds.InheritableCaps = capabilities.InheritableCaps - // // TODO: Support ambient capabilities. + // TODO: Support ambient capabilities. } else { - // If no capabilities are specified, grant the same capabilities - // that NewRootCredentials does. - creds.PermittedCaps = AllCapabilities - creds.EffectiveCaps = AllCapabilities + // If no capabilities are specified, grant capabilities consistent with + // setresuid + setresgid from NewRootCredentials to the given uid and + // gid. + if kuid == RootKUID { + creds.PermittedCaps = AllCapabilities + creds.EffectiveCaps = AllCapabilities + } else { + creds.PermittedCaps = 0 + creds.EffectiveCaps = 0 + } creds.BoundingCaps = AllCapabilities } -- cgit v1.2.3 From 3676b7ff1ca07e9fec1e380a0c2068390ce5d8de Mon Sep 17 00:00:00 2001 From: Brian Geffon Date: Tue, 8 Jan 2019 12:56:59 -0800 Subject: Improve loader related error messages returned to users. PiperOrigin-RevId: 228382827 Change-Id: Ica1d30e0df826bdd77f180a5092b2b735ea5c804 --- pkg/sentry/kernel/BUILD | 1 + pkg/sentry/kernel/kernel.go | 7 ++++--- pkg/sentry/kernel/task_context.go | 13 +++++++------ pkg/sentry/loader/BUILD | 1 + pkg/sentry/loader/loader.go | 28 +++++++++++----------------- pkg/sentry/syscalls/linux/sys_thread.go | 6 +++--- 6 files changed, 27 insertions(+), 29 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 490f674c0..7d41626dc 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -184,6 +184,7 @@ go_library( "//pkg/state", "//pkg/state/statefile", "//pkg/sync", + "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/stack", diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index cb61e27f1..43e9823cb 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -31,6 +31,7 @@ package kernel import ( + "errors" "fmt" "io" "path/filepath" @@ -658,9 +659,9 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, ThreadID, // Create a fresh task context. remainingTraversals = uint(args.MaxSymlinkTraversals) - tc, err := k.LoadTaskImage(ctx, k.mounts, root, wd, &remainingTraversals, args.Filename, args.Argv, args.Envv, k.featureSet) - if err != nil { - return nil, 0, err + tc, se := k.LoadTaskImage(ctx, k.mounts, root, wd, &remainingTraversals, args.Filename, args.Argv, args.Envv, k.featureSet) + if se != nil { + return nil, 0, errors.New(se.String()) } // Take a reference on the FDMap, which will be transferred to diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index aaff309f0..ee3e49d17 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -15,9 +15,9 @@ package kernel import ( - "errors" "fmt" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/cpuid" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" @@ -26,10 +26,10 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/loader" "gvisor.googlesource.com/gvisor/pkg/sentry/mm" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserr" ) -// ErrNoSyscalls is returned if there is no syscall table. -var ErrNoSyscalls = errors.New("no syscall table found") +var errNoSyscalls = syserr.New("no syscall table found", linux.ENOEXEC) // Auxmap contains miscellaneous data for the task. type Auxmap map[string]interface{} @@ -142,7 +142,7 @@ func (t *Task) Stack() *arch.Stack { // * argv: Binary argv // * envv: Binary envv // * fs: Binary FeatureSet -func (k *Kernel) LoadTaskImage(ctx context.Context, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, filename string, argv, envv []string, fs *cpuid.FeatureSet) (*TaskContext, error) { +func (k *Kernel) LoadTaskImage(ctx context.Context, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, filename string, argv, envv []string, fs *cpuid.FeatureSet) (*TaskContext, *syserr.Error) { // Prepare a new user address space to load into. m := mm.NewMemoryManager(k) defer m.DecUsers(ctx) @@ -155,8 +155,9 @@ func (k *Kernel) LoadTaskImage(ctx context.Context, mounts *fs.MountNamespace, r // Lookup our new syscall table. st, ok := LookupSyscallTable(os, ac.Arch()) if !ok { - // No syscall table found. Yikes. - return nil, ErrNoSyscalls + // No syscall table found. This means that the ELF binary does not match + // the architecture. + return nil, errNoSyscalls } if !m.IncUsers() { diff --git a/pkg/sentry/loader/BUILD b/pkg/sentry/loader/BUILD index 83cad186a..24e734b49 100644 --- a/pkg/sentry/loader/BUILD +++ b/pkg/sentry/loader/BUILD @@ -44,6 +44,7 @@ go_library( "//pkg/sentry/uniqueid", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/syserr", "//pkg/syserror", "//pkg/waiter", ], diff --git a/pkg/sentry/loader/loader.go b/pkg/sentry/loader/loader.go index e955502e3..deb8892f6 100644 --- a/pkg/sentry/loader/loader.go +++ b/pkg/sentry/loader/loader.go @@ -17,6 +17,7 @@ package loader import ( "bytes" + "fmt" "io" "path" @@ -30,6 +31,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/mm" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserr" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -196,20 +198,18 @@ func loadPath(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespac // Preconditions: // * The Task MemoryManager is empty. // * Load is called on the Task goroutine. -func Load(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, fs *cpuid.FeatureSet, filename string, argv, envv []string, extraAuxv []arch.AuxEntry, vdso *VDSO) (abi.OS, arch.Context, string, error) { +func Load(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, fs *cpuid.FeatureSet, filename string, argv, envv []string, extraAuxv []arch.AuxEntry, vdso *VDSO) (abi.OS, arch.Context, string, *syserr.Error) { // Load the binary itself. loaded, ac, d, argv, err := loadPath(ctx, m, mounts, root, wd, maxTraversals, fs, filename, argv) if err != nil { - ctx.Infof("Failed to load %s: %v", filename, err) - return 0, nil, "", err + return 0, nil, "", syserr.NewDynamic(fmt.Sprintf("Failed to load %s: %v", filename, err), syserr.FromError(err).ToLinux()) } defer d.DecRef() // Load the VDSO. vdsoAddr, err := loadVDSO(ctx, m, vdso, loaded) if err != nil { - ctx.Infof("Error loading VDSO: %v", err) - return 0, nil, "", err + return 0, nil, "", syserr.NewDynamic(fmt.Sprintf("Error loading VDSO: %v", err), syserr.FromError(err).ToLinux()) } // Setup the heap. brk starts at the next page after the end of the @@ -217,35 +217,30 @@ func Load(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, r // loaded.end is available for its use. e, ok := loaded.end.RoundUp() if !ok { - ctx.Warningf("brk overflows: %#x", loaded.end) - return 0, nil, "", syserror.ENOEXEC + return 0, nil, "", syserr.NewDynamic(fmt.Sprintf("brk overflows: %#x", loaded.end), linux.ENOEXEC) } m.BrkSetup(ctx, e) // Allocate our stack. stack, err := allocStack(ctx, m, ac) if err != nil { - ctx.Infof("Failed to allocate stack: %v", err) - return 0, nil, "", err + return 0, nil, "", syserr.NewDynamic(fmt.Sprintf("Failed to allocate stack: %v", err), syserr.FromError(err).ToLinux()) } // Push the original filename to the stack, for AT_EXECFN. execfn, err := stack.Push(filename) if err != nil { - ctx.Infof("Failed to push exec filename: %v", err) - return 0, nil, "", err + return 0, nil, "", syserr.NewDynamic(fmt.Sprintf("Failed to push exec filename: %v", err), syserr.FromError(err).ToLinux()) } // Push 16 random bytes on the stack which AT_RANDOM will point to. var b [16]byte if _, err := rand.Read(b[:]); err != nil { - ctx.Infof("Failed to read random bytes: %v", err) - return 0, nil, "", err + return 0, nil, "", syserr.NewDynamic(fmt.Sprintf("Failed to read random bytes: %v", err), syserr.FromError(err).ToLinux()) } random, err := stack.Push(b) if err != nil { - ctx.Infof("Failed to push random bytes: %v", err) - return 0, nil, "", err + return 0, nil, "", syserr.NewDynamic(fmt.Sprintf("Failed to push random bytes: %v", err), syserr.FromError(err).ToLinux()) } c := auth.CredentialsFromContext(ctx) @@ -266,8 +261,7 @@ func Load(ctx context.Context, m *mm.MemoryManager, mounts *fs.MountNamespace, r sl, err := stack.Load(argv, envv, auxv) if err != nil { - ctx.Infof("Failed to load stack: %v", err) - return 0, nil, "", err + return 0, nil, "", syserr.NewDynamic(fmt.Sprintf("Failed to load stack: %v", err), syserr.FromError(err).ToLinux()) } m.SetArgvStart(sl.ArgvStart) diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index c12693ee2..61cafefb9 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -104,9 +104,9 @@ func Execve(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal // Load the new TaskContext. maxTraversals := uint(linux.MaxSymlinkTraversals) - tc, err := t.Kernel().LoadTaskImage(t, t.MountNamespace(), root, wd, &maxTraversals, filename, argv, envv, t.Arch().FeatureSet()) - if err != nil { - return 0, nil, err + tc, se := t.Kernel().LoadTaskImage(t, t.MountNamespace(), root, wd, &maxTraversals, filename, argv, envv, t.Arch().FeatureSet()) + if se != nil { + return 0, nil, se.ToError() } ctrl, err := t.Execve(tc) -- cgit v1.2.3 From 9270d940eb1a6e31587c34f4644189f3b2c002e1 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Wed, 9 Jan 2019 17:53:31 -0800 Subject: Minor memevent fixes. - Call MemoryEvents.done.Add(1) outside of MemoryEvents.run() so that if MemoryEvents.Stop() => MemoryEvents.done.Wait() is called before the goroutine starts running, it still waits for the goroutine to stop. - Use defer to call MemoryEvents.done.Done() in MemoryEvents.run() so that it's called even if the goroutine panics. PiperOrigin-RevId: 228623307 Change-Id: I1b0459e7999606c1a1a271b16092b1ca87005015 --- pkg/sentry/kernel/memevent/memory_events.go | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/memevent/memory_events.go b/pkg/sentry/kernel/memevent/memory_events.go index 1a8e86827..b6283c5d1 100644 --- a/pkg/sentry/kernel/memevent/memory_events.go +++ b/pkg/sentry/kernel/memevent/memory_events.go @@ -69,11 +69,12 @@ func (m *MemoryEvents) Start() { if m.period == 0 { return } + m.done.Add(1) go m.run() // S/R-SAFE: doesn't interact with saved state. } func (m *MemoryEvents) run() { - m.done.Add(1) + defer m.done.Done() // Emit the first event immediately on startup. totalTicks.Increment() @@ -85,7 +86,6 @@ func (m *MemoryEvents) run() { for { select { case <-m.stop: - m.done.Done() return case <-ticker.C: totalTicks.Increment() -- cgit v1.2.3 From dc8450b5676d4c4ac9bcfa23cabd862e0060527d Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Mon, 14 Jan 2019 20:33:29 -0800 Subject: Remove fs.Handle, ramfs.Entry, and all the DeprecatedFileOperations. More helper structs have been added to the fsutil package to make it easier to implement fs.InodeOperations and fs.FileOperations. PiperOrigin-RevId: 229305982 Change-Id: Ib6f8d3862f4216745116857913dbfa351530223b --- pkg/abi/linux/fs.go | 2 + pkg/sentry/fs/BUILD | 3 +- pkg/sentry/fs/anon/anon.go | 16 +- pkg/sentry/fs/ashmem/BUILD | 1 + pkg/sentry/fs/ashmem/area.go | 16 +- pkg/sentry/fs/ashmem/device.go | 133 +-------- pkg/sentry/fs/attr.go | 47 +++ pkg/sentry/fs/binder/BUILD | 2 +- pkg/sentry/fs/binder/binder.go | 135 +-------- pkg/sentry/fs/dev/BUILD | 1 + pkg/sentry/fs/dev/dev.go | 29 +- pkg/sentry/fs/dev/fs.go | 2 + pkg/sentry/fs/dev/full.go | 55 ++-- pkg/sentry/fs/dev/null.go | 88 ++++-- pkg/sentry/fs/dev/random.go | 55 ++-- pkg/sentry/fs/dirent.go | 8 +- pkg/sentry/fs/fdpipe/pipe.go | 14 +- pkg/sentry/fs/file_operations.go | 2 +- pkg/sentry/fs/file_overlay_test.go | 7 +- pkg/sentry/fs/filetest/filetest.go | 16 +- pkg/sentry/fs/fsutil/BUILD | 15 - pkg/sentry/fs/fsutil/file.go | 226 ++++++++++----- pkg/sentry/fs/fsutil/fsutil.go | 2 - pkg/sentry/fs/fsutil/handle.go | 128 --------- pkg/sentry/fs/fsutil/handle_test.go | 227 --------------- pkg/sentry/fs/fsutil/inode.go | 409 ++++++++++++++------------ pkg/sentry/fs/fsutil/inode_cached_test.go | 14 +- pkg/sentry/fs/gofer/file.go | 2 +- pkg/sentry/fs/gofer/fs.go | 14 +- pkg/sentry/fs/gofer/inode.go | 1 - pkg/sentry/fs/host/file.go | 4 +- pkg/sentry/fs/host/fs.go | 2 + pkg/sentry/fs/host/inode.go | 1 - pkg/sentry/fs/inode.go | 29 +- pkg/sentry/fs/inode_operations.go | 80 ------ pkg/sentry/fs/inode_overlay.go | 13 - pkg/sentry/fs/inode_overlay_test.go | 62 +++- pkg/sentry/fs/mock.go | 11 - pkg/sentry/fs/mount.go | 14 +- pkg/sentry/fs/mounts_test.go | 15 +- pkg/sentry/fs/proc/BUILD | 4 +- pkg/sentry/fs/proc/cpuinfo.go | 41 +-- pkg/sentry/fs/proc/exec_args.go | 57 +++- pkg/sentry/fs/proc/fds.go | 138 +++++---- pkg/sentry/fs/proc/file.go | 58 ---- pkg/sentry/fs/proc/inode.go | 96 +++++++ pkg/sentry/fs/proc/net.go | 55 ++-- pkg/sentry/fs/proc/proc.go | 152 +++++----- pkg/sentry/fs/proc/rpcinet_proc.go | 246 ++++++++-------- pkg/sentry/fs/proc/seqfile/BUILD | 7 +- pkg/sentry/fs/proc/seqfile/seqfile.go | 133 ++++++--- pkg/sentry/fs/proc/seqfile/seqfile_test.go | 45 +-- pkg/sentry/fs/proc/sys.go | 116 +++++--- pkg/sentry/fs/proc/sys_net.go | 325 +++++++++++++-------- pkg/sentry/fs/proc/sys_net_state.go | 17 +- pkg/sentry/fs/proc/sys_net_test.go | 28 +- pkg/sentry/fs/proc/task.go | 264 +++++++++++------ pkg/sentry/fs/proc/uid_gid_map.go | 17 +- pkg/sentry/fs/proc/uptime.go | 40 ++- pkg/sentry/fs/ramfs/BUILD | 6 +- pkg/sentry/fs/ramfs/dir.go | 223 +++++++++++---- pkg/sentry/fs/ramfs/file.go | 150 ---------- pkg/sentry/fs/ramfs/ramfs.go | 441 ----------------------------- pkg/sentry/fs/ramfs/socket.go | 48 +++- pkg/sentry/fs/ramfs/symlink.go | 67 +++-- pkg/sentry/fs/ramfs/test/BUILD | 16 -- pkg/sentry/fs/ramfs/test/test.go | 46 --- pkg/sentry/fs/ramfs/tree.go | 3 +- pkg/sentry/fs/ramfs/tree_test.go | 2 +- pkg/sentry/fs/sys/BUILD | 3 +- pkg/sentry/fs/sys/devices.go | 51 ++-- pkg/sentry/fs/sys/fs.go | 2 + pkg/sentry/fs/sys/sys.go | 10 +- pkg/sentry/fs/timerfd/timerfd.go | 12 +- pkg/sentry/fs/tmpfs/BUILD | 2 + pkg/sentry/fs/tmpfs/file_regular.go | 14 +- pkg/sentry/fs/tmpfs/file_test.go | 6 +- pkg/sentry/fs/tmpfs/fs.go | 2 + pkg/sentry/fs/tmpfs/inode_file.go | 112 ++++---- pkg/sentry/fs/tmpfs/tmpfs.go | 164 +++++++++-- pkg/sentry/fs/tty/BUILD | 2 - pkg/sentry/fs/tty/dir.go | 108 ++----- pkg/sentry/fs/tty/inode.go | 145 ---------- pkg/sentry/fs/tty/master.go | 23 +- pkg/sentry/fs/tty/slave.go | 25 +- pkg/sentry/kernel/epoll/epoll.go | 12 +- pkg/sentry/kernel/eventfd/eventfd.go | 14 +- pkg/sentry/kernel/pipe/node.go | 40 +-- pkg/sentry/kernel/pipe/node_test.go | 36 +-- pkg/sentry/kernel/pipe/pipe.go | 31 +- pkg/sentry/kernel/pipe/reader_writer.go | 10 +- pkg/sentry/loader/vdso.go | 48 ++-- pkg/sentry/socket/epsocket/epsocket.go | 10 +- pkg/sentry/socket/hostinet/socket.go | 10 +- pkg/sentry/socket/netlink/socket.go | 10 +- pkg/sentry/socket/rpcinet/socket.go | 10 +- pkg/sentry/socket/socket.go | 18 +- pkg/sentry/socket/unix/unix.go | 10 +- runsc/boot/fs.go | 4 +- test/syscalls/linux/proc.cc | 5 + 100 files changed, 2547 insertions(+), 3144 deletions(-) delete mode 100644 pkg/sentry/fs/fsutil/handle.go delete mode 100644 pkg/sentry/fs/fsutil/handle_test.go delete mode 100644 pkg/sentry/fs/proc/file.go create mode 100644 pkg/sentry/fs/proc/inode.go delete mode 100644 pkg/sentry/fs/ramfs/file.go delete mode 100644 pkg/sentry/fs/ramfs/ramfs.go delete mode 100644 pkg/sentry/fs/ramfs/test/BUILD delete mode 100644 pkg/sentry/fs/ramfs/test/test.go delete mode 100644 pkg/sentry/fs/tty/inode.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/fs.go b/pkg/abi/linux/fs.go index 0b1c9f3db..a9f2ba132 100644 --- a/pkg/abi/linux/fs.go +++ b/pkg/abi/linux/fs.go @@ -22,8 +22,10 @@ const ( DEVPTS_SUPER_MAGIC = 0x00001cd1 OVERLAYFS_SUPER_MAGIC = 0x794c7630 PIPEFS_MAGIC = 0x50495045 + PROC_SUPER_MAGIC = 0x9fa0 RAMFS_MAGIC = 0x09041934 SOCKFS_MAGIC = 0x534F434B + SYSFS_MAGIC = 0x62656572 TMPFS_MAGIC = 0x01021994 V9FS_MAGIC = 0x01021997 ) diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 0fe2b14bf..6f368b0da 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -94,7 +94,8 @@ go_test( deps = [ ":fs", "//pkg/sentry/context", - "//pkg/sentry/fs/ramfs/test", + "//pkg/sentry/fs/fsutil", + "//pkg/sentry/fs/ramfs", "//pkg/sentry/fs/tmpfs", "//pkg/sentry/kernel/contexttest", "//pkg/sentry/usermem", diff --git a/pkg/sentry/fs/anon/anon.go b/pkg/sentry/fs/anon/anon.go index 743cf511f..a5e8c4f0d 100644 --- a/pkg/sentry/fs/anon/anon.go +++ b/pkg/sentry/fs/anon/anon.go @@ -28,16 +28,12 @@ import ( // with any real filesystem. Some types depend on completely pseudo // "anon" inodes (eventfds, epollfds, etc). func NewInode(ctx context.Context) *fs.Inode { - return fs.NewInode(fsutil.NewSimpleInodeOperations(fsutil.InodeSimpleAttributes{ - FSType: linux.ANON_INODE_FS_MAGIC, - UAttr: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: fs.FileOwnerFromContext(ctx), - Perms: fs.FilePermissions{ - User: fs.PermMask{Read: true, Write: true}, - }, - Links: 1, - }), - }), fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}), fs.StableAttr{ + iops := &fsutil.SimpleFileInode{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, fs.RootOwner, fs.FilePermissions{ + User: fs.PermMask{Read: true, Write: true}, + }, linux.ANON_INODE_FS_MAGIC), + } + return fs.NewInode(iops, fs.NewPseudoMountSource(), fs.StableAttr{ Type: fs.Anonymous, DeviceID: PseudoDevice.DeviceID(), InodeID: PseudoDevice.NextIno(), diff --git a/pkg/sentry/fs/ashmem/BUILD b/pkg/sentry/fs/ashmem/BUILD index 44ef82e64..2463111a8 100644 --- a/pkg/sentry/fs/ashmem/BUILD +++ b/pkg/sentry/fs/ashmem/BUILD @@ -28,6 +28,7 @@ go_library( "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/syserror", + "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index d7dd2c084..7c1b11464 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -28,6 +28,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) const ( @@ -42,9 +43,10 @@ const ( // // +stateify savable type Area struct { - fsutil.NoFsync `state:"nosave"` - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` ad *Device @@ -98,11 +100,6 @@ func (a *Area) Write(ctx context.Context, file *fs.File, src usermem.IOSequence, return 0, syserror.ENOSYS } -// Flush implements fs.FileOperations.Flush. -func (a *Area) Flush(ctx context.Context, file *fs.File) error { - return nil -} - // ConfigureMMap implements fs.FileOperations.ConfigureMMap. func (a *Area) ConfigureMMap(ctx context.Context, file *fs.File, opts *memmap.MMapOpts) error { a.mu.Lock() @@ -122,8 +119,7 @@ func (a *Area) ConfigureMMap(ctx context.Context, file *fs.File, opts *memmap.MM return syserror.ENOMEM } tmpfsInodeOps := tmpfs.NewInMemoryFile(ctx, usage.Tmpfs, fs.UnstableAttr{}, k) - // This is not backed by a real filesystem, so we pass in nil. - tmpfsInode := fs.NewInode(tmpfsInodeOps, fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}), fs.StableAttr{}) + tmpfsInode := fs.NewInode(tmpfsInodeOps, fs.NewPseudoMountSource(), fs.StableAttr{}) dirent := fs.NewDirent(tmpfsInode, namePrefix+"/"+a.name) tmpfsFile, err := tmpfsInode.GetFile(ctx, dirent, fs.FileFlags{Read: true, Write: true}) // Drop the extra reference on the Dirent. diff --git a/pkg/sentry/fs/ashmem/device.go b/pkg/sentry/fs/ashmem/device.go index 962da141b..5369d1b0d 100644 --- a/pkg/sentry/fs/ashmem/device.go +++ b/pkg/sentry/fs/ashmem/device.go @@ -16,49 +16,40 @@ package ashmem import ( - "sync" - + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" ) // Device implements fs.InodeOperations. // // +stateify savable type Device struct { - fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopTruncate `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` - fsutil.InodeNotRenameable `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotSocket `state:"nosave"` fsutil.InodeNotSymlink `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoMappable `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` - mu sync.Mutex `state:"nosave"` - unstable fs.UnstableAttr + fsutil.InodeSimpleAttributes } +var _ fs.InodeOperations = (*Device)(nil) + // NewDevice creates and intializes a Device structure. func NewDevice(ctx context.Context, owner fs.FileOwner, fp fs.FilePermissions) *Device { return &Device{ - unstable: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: owner, - Perms: fp, - Links: 1, - }), + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, fp, linux.ANON_INODE_FS_MAGIC), } } -// Release implements fs.InodeOperations.Release. -func (ad *Device) Release(context.Context) {} - // GetFile implements fs.InodeOperations.GetFile. func (ad *Device) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { return fs.NewFile(ctx, d, flags, &Area{ @@ -67,105 +58,3 @@ func (ad *Device) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) perms: usermem.AnyAccess, }), nil } - -// UnstableAttr implements fs.InodeOperations.UnstableAttr. -func (ad *Device) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - ad.mu.Lock() - defer ad.mu.Unlock() - return ad.unstable, nil -} - -// Check implements fs.InodeOperations.Check. -func (ad *Device) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { - return fs.ContextCanAccessFile(ctx, inode, p) -} - -// SetPermissions implements fs.InodeOperations.SetPermissions. -func (ad *Device) SetPermissions(ctx context.Context, inode *fs.Inode, fp fs.FilePermissions) bool { - ad.mu.Lock() - defer ad.mu.Unlock() - ad.unstable.Perms = fp - ad.unstable.StatusChangeTime = time.NowFromContext(ctx) - return true -} - -// SetOwner implements fs.InodeOperations.SetOwner. -func (ad *Device) SetOwner(ctx context.Context, inode *fs.Inode, owner fs.FileOwner) error { - ad.mu.Lock() - defer ad.mu.Unlock() - if owner.UID.Ok() { - ad.unstable.Owner.UID = owner.UID - } - if owner.GID.Ok() { - ad.unstable.Owner.GID = owner.GID - } - return nil -} - -// SetTimestamps implements fs.InodeOperations.SetTimestamps. -func (ad *Device) SetTimestamps(ctx context.Context, inode *fs.Inode, ts fs.TimeSpec) error { - if ts.ATimeOmit && ts.MTimeOmit { - return nil - } - - ad.mu.Lock() - defer ad.mu.Unlock() - - now := time.NowFromContext(ctx) - if !ts.ATimeOmit { - if ts.ATimeSetSystemTime { - ad.unstable.AccessTime = now - } else { - ad.unstable.AccessTime = ts.ATime - } - } - if !ts.MTimeOmit { - if ts.MTimeSetSystemTime { - ad.unstable.ModificationTime = now - } else { - ad.unstable.ModificationTime = ts.MTime - } - } - ad.unstable.StatusChangeTime = now - return nil -} - -// Truncate implements fs.InodeOperations.WriteOut. -// -// Ignored by ashmem. -func (ad *Device) Truncate(ctx context.Context, inode *fs.Inode, size int64) error { - return nil -} - -// AddLink implements fs.InodeOperations.AddLink. -// -// Ashmem doesn't support links, no-op. -func (ad *Device) AddLink() {} - -// DropLink implements fs.InodeOperations.DropLink. -// -// Ashmem doesn't support links, no-op. -func (ad *Device) DropLink() {} - -// NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange. -func (ad *Device) NotifyStatusChange(ctx context.Context) { - ad.mu.Lock() - defer ad.mu.Unlock() - now := time.NowFromContext(ctx) - ad.unstable.ModificationTime = now - ad.unstable.StatusChangeTime = now -} - -// IsVirtual implements fs.InodeOperations.IsVirtual. -// -// Ashmem is virtual. -func (ad *Device) IsVirtual() bool { - return true -} - -// StatFS implements fs.InodeOperations.StatFS. -// -// Ashmem doesn't support querying for filesystem info. -func (ad *Device) StatFS(context.Context) (fs.Info, error) { - return fs.Info{}, syserror.ENOSYS -} diff --git a/pkg/sentry/fs/attr.go b/pkg/sentry/fs/attr.go index 59e060e3c..3523b068a 100644 --- a/pkg/sentry/fs/attr.go +++ b/pkg/sentry/fs/attr.go @@ -180,6 +180,53 @@ type UnstableAttr struct { Links uint64 } +// SetOwner sets the owner and group if they are valid. +// +// This method is NOT thread-safe. Callers must prevent concurrent calls. +func (ua *UnstableAttr) SetOwner(ctx context.Context, owner FileOwner) { + if owner.UID.Ok() { + ua.Owner.UID = owner.UID + } + if owner.GID.Ok() { + ua.Owner.GID = owner.GID + } + ua.StatusChangeTime = ktime.NowFromContext(ctx) +} + +// SetPermissions sets the permissions. +// +// This method is NOT thread-safe. Callers must prevent concurrent calls. +func (ua *UnstableAttr) SetPermissions(ctx context.Context, p FilePermissions) { + ua.Perms = p + ua.StatusChangeTime = ktime.NowFromContext(ctx) +} + +// SetTimestamps sets the timestamps according to the TimeSpec. +// +// This method is NOT thread-safe. Callers must prevent concurrent calls. +func (ua *UnstableAttr) SetTimestamps(ctx context.Context, ts TimeSpec) { + if ts.ATimeOmit && ts.MTimeOmit { + return + } + + now := ktime.NowFromContext(ctx) + if !ts.ATimeOmit { + if ts.ATimeSetSystemTime { + ua.AccessTime = now + } else { + ua.AccessTime = ts.ATime + } + } + if !ts.MTimeOmit { + if ts.MTimeSetSystemTime { + ua.ModificationTime = now + } else { + ua.ModificationTime = ts.MTime + } + } + ua.StatusChangeTime = now +} + // WithCurrentTime returns u with AccessTime == ModificationTime == current time. func WithCurrentTime(ctx context.Context, u UnstableAttr) UnstableAttr { t := ktime.NowFromContext(ctx) diff --git a/pkg/sentry/fs/binder/BUILD b/pkg/sentry/fs/binder/BUILD index a077b91d2..27155819e 100644 --- a/pkg/sentry/fs/binder/BUILD +++ b/pkg/sentry/fs/binder/BUILD @@ -16,11 +16,11 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel", - "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/syserror", + "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index e642c7f22..19cd55e65 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -24,12 +24,12 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) const ( @@ -43,34 +43,29 @@ const ( // // +stateify savable type Device struct { + fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopTruncate `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` - fsutil.InodeNotRenameable `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotSocket `state:"nosave"` fsutil.InodeNotSymlink `state:"nosave"` - fsutil.NoMappable `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` - fsutil.DeprecatedFileOperations `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` - // mu protects unstable. - mu sync.Mutex `state:"nosave"` - unstable fs.UnstableAttr + fsutil.InodeSimpleAttributes } +var _ fs.InodeOperations = (*Device)(nil) + // NewDevice creates and intializes a Device structure. func NewDevice(ctx context.Context, owner fs.FileOwner, fp fs.FilePermissions) *Device { return &Device{ - unstable: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: owner, - Perms: fp, - Links: 1, - }), + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, fp, 0), } } -// Release implements fs.InodeOperations.Release. -func (bd *Device) Release(context.Context) {} - // GetFile implements fs.InodeOperations.GetFile. // // TODO: Add functionality to GetFile: Additional fields will be @@ -85,115 +80,13 @@ func (bd *Device) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) }), nil } -// UnstableAttr implements fs.InodeOperations.UnstableAttr. -func (bd *Device) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - bd.mu.Lock() - defer bd.mu.Unlock() - return bd.unstable, nil -} - -// Check implements fs.InodeOperations.Check. -func (bd *Device) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { - return fs.ContextCanAccessFile(ctx, inode, p) -} - -// SetPermissions implements fs.InodeOperations.SetPermissions. -func (bd *Device) SetPermissions(ctx context.Context, inode *fs.Inode, fp fs.FilePermissions) bool { - bd.mu.Lock() - defer bd.mu.Unlock() - bd.unstable.Perms = fp - bd.unstable.StatusChangeTime = time.NowFromContext(ctx) - return true -} - -// SetOwner implements fs.InodeOperations.SetOwner. -func (bd *Device) SetOwner(ctx context.Context, inode *fs.Inode, owner fs.FileOwner) error { - bd.mu.Lock() - defer bd.mu.Unlock() - if owner.UID.Ok() { - bd.unstable.Owner.UID = owner.UID - } - if owner.GID.Ok() { - bd.unstable.Owner.GID = owner.GID - } - return nil -} - -// SetTimestamps implements fs.InodeOperations.SetTimestamps. -func (bd *Device) SetTimestamps(ctx context.Context, inode *fs.Inode, ts fs.TimeSpec) error { - if ts.ATimeOmit && ts.MTimeOmit { - return nil - } - - bd.mu.Lock() - defer bd.mu.Unlock() - - now := time.NowFromContext(ctx) - if !ts.ATimeOmit { - if ts.ATimeSetSystemTime { - bd.unstable.AccessTime = now - } else { - bd.unstable.AccessTime = ts.ATime - } - } - if !ts.MTimeOmit { - if ts.MTimeSetSystemTime { - bd.unstable.ModificationTime = now - } else { - bd.unstable.ModificationTime = ts.MTime - } - } - bd.unstable.StatusChangeTime = now - return nil -} - -// Truncate implements fs.InodeOperations.WriteOut. -// -// Ignored for a character device, such as Binder. -func (bd *Device) Truncate(ctx context.Context, inode *fs.Inode, size int64) error { - return nil -} - -// AddLink implements fs.InodeOperations.AddLink. -// -// Binder doesn't support links, no-op. -func (bd *Device) AddLink() {} - -// DropLink implements fs.InodeOperations.DropLink. -// -// Binder doesn't support links, no-op. -func (bd *Device) DropLink() {} - -// NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange. -func (bd *Device) NotifyStatusChange(ctx context.Context) { - bd.mu.Lock() - defer bd.mu.Unlock() - now := time.NowFromContext(ctx) - bd.unstable.ModificationTime = now - bd.unstable.StatusChangeTime = now -} - -// IsVirtual implements fs.InodeOperations.IsVirtual. -// -// Binder is virtual. -func (bd *Device) IsVirtual() bool { - return true -} - -// StatFS implements fs.InodeOperations.StatFS. -// -// Binder doesn't support querying for filesystem info. -func (bd *Device) StatFS(context.Context) (fs.Info, error) { - return fs.Info{}, syserror.ENOSYS -} - // Proc implements fs.FileOperations and fs.IoctlGetter. // // +stateify savable type Proc struct { - fsutil.NoFsync `state:"nosave"` - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` bd *Device task *kernel.Task diff --git a/pkg/sentry/fs/dev/BUILD b/pkg/sentry/fs/dev/BUILD index b17b5202c..b9cfae05f 100644 --- a/pkg/sentry/fs/dev/BUILD +++ b/pkg/sentry/fs/dev/BUILD @@ -32,5 +32,6 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usermem", "//pkg/syserror", + "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/dev/dev.go b/pkg/sentry/fs/dev/dev.go index 3e127bf04..f8e8099f7 100644 --- a/pkg/sentry/fs/dev/dev.go +++ b/pkg/sentry/fs/dev/dev.go @@ -16,6 +16,8 @@ package dev import ( + "math" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ashmem" @@ -26,13 +28,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// Dev is the root node. -// -// +stateify savable -type Dev struct { - ramfs.Dir -} - func newCharacterDevice(iops fs.InodeOperations, msrc *fs.MountSource) *fs.Inode { return fs.NewInode(iops, msrc, fs.StableAttr{ DeviceID: devDevice.DeviceID(), @@ -43,8 +38,7 @@ func newCharacterDevice(iops fs.InodeOperations, msrc *fs.MountSource) *fs.Inode } func newDirectory(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - iops := &ramfs.Dir{} - iops.InitDir(ctx, map[string]*fs.Inode{}, fs.RootOwner, fs.FilePermsFromMode(0555)) + iops := ramfs.NewDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) return fs.NewInode(iops, msrc, fs.StableAttr{ DeviceID: devDevice.DeviceID(), InodeID: devDevice.NextIno(), @@ -54,8 +48,7 @@ func newDirectory(ctx context.Context, msrc *fs.MountSource) *fs.Inode { } func newSymlink(ctx context.Context, target string, msrc *fs.MountSource) *fs.Inode { - iops := &ramfs.Symlink{} - iops.InitSymlink(ctx, fs.RootOwner, target) + iops := ramfs.NewSymlink(ctx, fs.RootOwner, target) return fs.NewInode(iops, msrc, fs.StableAttr{ DeviceID: devDevice.DeviceID(), InodeID: devDevice.NextIno(), @@ -66,8 +59,6 @@ func newSymlink(ctx context.Context, target string, msrc *fs.MountSource) *fs.In // New returns the root node of a device filesystem. func New(ctx context.Context, msrc *fs.MountSource, binderEnabled bool, ashmemEnabled bool) *fs.Inode { - d := &Dev{} - contents := map[string]*fs.Inode{ "fd": newSymlink(ctx, "/proc/self/fd", msrc), "stdin": newSymlink(ctx, "/proc/self/fd/0", msrc), @@ -114,11 +105,19 @@ func New(ctx context.Context, msrc *fs.MountSource, binderEnabled bool, ashmemEn contents["ashmem"] = newCharacterDevice(ashmem, msrc) } - d.InitDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) - return fs.NewInode(d, msrc, fs.StableAttr{ + iops := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return fs.NewInode(iops, msrc, fs.StableAttr{ DeviceID: devDevice.DeviceID(), InodeID: devDevice.NextIno(), BlockSize: usermem.PageSize, Type: fs.Directory, }) } + +// readZeros implements fs.FileOperations.Read with infinite null bytes. +type readZeros struct{} + +// Read implements fs.FileOperations.Read. +func (readZeros) Read(ctx context.Context, file *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { + return dst.ZeroOut(ctx, math.MaxInt64) +} diff --git a/pkg/sentry/fs/dev/fs.go b/pkg/sentry/fs/dev/fs.go index d96f4f423..abfe689f0 100644 --- a/pkg/sentry/fs/dev/fs.go +++ b/pkg/sentry/fs/dev/fs.go @@ -33,6 +33,8 @@ const ashmemEnabledKey = "ashmem_enabled" // +stateify savable type filesystem struct{} +var _ fs.Filesystem = (*filesystem)(nil) + func init() { fs.RegisterFilesystem(&filesystem{}) } diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index eeda646ab..cbdd40161 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -15,41 +15,64 @@ package dev import ( - "math" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // fullDevice is used to implement /dev/full. // // +stateify savable type fullDevice struct { - ramfs.Entry + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopTruncate `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes } +var _ fs.InodeOperations = (*fullDevice)(nil) + func newFullDevice(ctx context.Context, owner fs.FileOwner, mode linux.FileMode) *fullDevice { - f := &fullDevice{} - f.InitEntry(ctx, owner, fs.FilePermsFromMode(mode)) + f := &fullDevice{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, fs.FilePermsFromMode(mode), linux.TMPFS_MAGIC), + } return f } -// DeprecatedPwritev implements fs.InodeOperations.DeprecatedPwritev by -// returining ENOSPC. -func (f *fullDevice) DeprecatedPwritev(_ context.Context, _ usermem.IOSequence, _ int64) (int64, error) { - return 0, syserror.ENOSPC +// GetFile implements fs.InodeOperations.GetFile. +func (f *fullDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + flags.Pread = true + return fs.NewFile(ctx, dirent, flags, &fullFileOperations{}), nil } -// DeprecatedPreadv implements fs.InodeOperations.DeprecatedPreadv. -func (f *fullDevice) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, _ int64) (int64, error) { - return dst.ZeroOut(ctx, math.MaxInt64) +// +stateify savable +type fullFileOperations struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + readZeros `state:"nosave"` } -// Truncate should be simply ignored for character devices on linux. -func (f *fullDevice) Truncate(context.Context, *fs.Inode, int64) error { - return nil +var _ fs.FileOperations = (*fullFileOperations)(nil) + +// Write implements FileOperations.Write. +func (fullFileOperations) Write(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, syserror.ENOSPC } diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 68090f353..73fd09058 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -15,78 +15,104 @@ package dev import ( - "io" - "math" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/mm" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // +stateify savable type nullDevice struct { - ramfs.Entry + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopTruncate `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes } +var _ fs.InodeOperations = (*nullDevice)(nil) + func newNullDevice(ctx context.Context, owner fs.FileOwner, mode linux.FileMode) *nullDevice { - n := &nullDevice{} - n.InitEntry(ctx, owner, fs.FilePermsFromMode(mode)) + n := &nullDevice{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, fs.FilePermsFromMode(mode), linux.TMPFS_MAGIC), + } return n } -// DeprecatedPreadv reads data from the device. -func (n *nullDevice) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - return 0, io.EOF -} +// GetFile implements fs.FileOperations.GetFile. +func (n *nullDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + flags.Pread = true + flags.Pwrite = true -// DeprecatedPwritev discards writes. -func (n *nullDevice) DeprecatedPwritev(_ context.Context, src usermem.IOSequence, offset int64) (int64, error) { - return src.NumBytes(), nil + return fs.NewFile(ctx, dirent, flags, &nullFileOperations{}), nil } -// Truncate should be simply ignored for character devices on linux. -func (n *nullDevice) Truncate(context.Context, *fs.Inode, int64) error { - return nil +// +stateify savable +type nullFileOperations struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRead `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` } +var _ fs.FileOperations = (*nullFileOperations)(nil) + // +stateify savable type zeroDevice struct { nullDevice } +var _ fs.InodeOperations = (*zeroDevice)(nil) + func newZeroDevice(ctx context.Context, owner fs.FileOwner, mode linux.FileMode) *zeroDevice { - zd := &zeroDevice{} - zd.InitEntry(ctx, owner, fs.FilePermsFromMode(mode)) + zd := &zeroDevice{ + nullDevice: nullDevice{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, fs.FilePermsFromMode(mode), linux.TMPFS_MAGIC), + }, + } return zd } -// DeprecatedPreadv implements fs.InodeOperations.DeprecatedPreadv. -func (zd *zeroDevice) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - return dst.ZeroOut(ctx, math.MaxInt64) -} - -// GetFile overrides ramfs.Entry.GetFile and returns a zeroFile instead. +// GetFile implements fs.FileOperations.GetFile. func (zd *zeroDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { - // Allow pread(2) and pwrite(2) on this file. flags.Pread = true flags.Pwrite = true - return fs.NewFile(ctx, dirent, flags, &zeroFileOperations{ - FileOperations: &fsutil.Handle{HandleOperations: dirent.Inode.HandleOps()}, - }), nil + return fs.NewFile(ctx, dirent, flags, &zeroFileOperations{}), nil } // +stateify savable type zeroFileOperations struct { - fs.FileOperations + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + readZeros `state:"nosave"` } +var _ fs.FileOperations = (*zeroFileOperations)(nil) + // ConfigureMMap implements fs.FileOperations.ConfigureMMap. func (*zeroFileOperations) ConfigureMMap(ctx context.Context, file *fs.File, opts *memmap.MMapOpts) error { m, err := mm.NewSharedAnonMappable(opts.Length, platform.FromContext(ctx)) diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index 33e4913e4..837b7793a 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -19,37 +19,58 @@ import ( "gvisor.googlesource.com/gvisor/pkg/rand" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // +stateify savable type randomDevice struct { - ramfs.Entry + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopTruncate `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes } +var _ fs.InodeOperations = (*randomDevice)(nil) + func newRandomDevice(ctx context.Context, owner fs.FileOwner, mode linux.FileMode) *randomDevice { - r := &randomDevice{} - r.InitEntry(ctx, owner, fs.FilePermsFromMode(mode)) + r := &randomDevice{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, fs.FilePermsFromMode(mode), linux.TMPFS_MAGIC), + } return r } -// DeprecatedPreadv reads random data. -func (*randomDevice) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - return dst.CopyOutFrom(ctx, safemem.FromIOReader{rand.Reader}) +// GetFile implements fs.InodeOperations.GetFile. +func (randomDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &randomFileOperations{}), nil } -// DeprecatedPwritev implements fs.HandleOperations.DeprecatedPwritev. -func (*randomDevice) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { - // On Linux, "Writing to /dev/random or /dev/urandom will update the - // entropy pool with the data written, but this will not result in a higher - // entropy count" - random(4). We don't need to support this, but we do - // need to support the write, so just make it a no-op a la /dev/null. - return src.NumBytes(), nil +// +stateify savable +type randomFileOperations struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` } -// Truncate should be simply ignored for character devices on linux. -func (r *randomDevice) Truncate(context.Context, *fs.Inode, int64) error { - return nil +var _ fs.FileOperations = (*randomFileOperations)(nil) + +// Read implements fs.FileOperations.Read. +func (randomFileOperations) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, _ int64) (int64, error) { + return dst.CopyOutFrom(ctx, safemem.FromIOReader{rand.Reader}) } diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go index c4918a11b..d6a19dc81 100644 --- a/pkg/sentry/fs/dirent.go +++ b/pkg/sentry/fs/dirent.go @@ -837,8 +837,8 @@ func (d *Dirent) CreateFifo(ctx context.Context, root *Dirent, name string, perm }) } -// getDotAttrs returns the DentAttrs corresponding to "." and ".." directories. -func (d *Dirent) getDotAttrs(root *Dirent) (DentAttr, DentAttr) { +// GetDotAttrs returns the DentAttrs corresponding to "." and ".." directories. +func (d *Dirent) GetDotAttrs(root *Dirent) (DentAttr, DentAttr) { // Get '.'. sattr := d.Inode.StableAttr dot := DentAttr{ @@ -870,7 +870,7 @@ func (d *Dirent) readdirFrozen(root *Dirent, offset int64, dirCtx *DirCtx) (int6 // Collect attrs for "." and "..". attrs := make(map[string]DentAttr) names := []string{".", ".."} - attrs["."], attrs[".."] = d.getDotAttrs(root) + attrs["."], attrs[".."] = d.GetDotAttrs(root) // Get info from all children. d.mu.Lock() @@ -965,7 +965,7 @@ func direntReaddir(ctx context.Context, d *Dirent, it DirIterator, root *Dirent, } // Collect attrs for "." and "..". - dot, dotdot := d.getDotAttrs(root) + dot, dotdot := d.GetDotAttrs(root) // Emit "." and ".." if the offset is low enough. if offset == 0 { diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index e3b830747..b4d11cb45 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -37,13 +37,13 @@ import ( // // +stateify savable type pipeOperations struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` - fsutil.NoIoctl `state:"nosave"` - waiter.Queue `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + waiter.Queue `state:"nosave"` // flags are the flags used to open the pipe. flags fs.FileFlags `state:".(fs.FileFlags)"` diff --git a/pkg/sentry/fs/file_operations.go b/pkg/sentry/fs/file_operations.go index 28e8e233d..81c6e2b5d 100644 --- a/pkg/sentry/fs/file_operations.go +++ b/pkg/sentry/fs/file_operations.go @@ -91,7 +91,7 @@ type FileOperations interface { Flush(ctx context.Context, file *File) error // ConfigureMMap mutates opts to implement mmap(2) for the file. Most - // implementations can either embed fsutil.NoMMap (if they don't support + // implementations can either embed fsutil.FileNoMMap (if they don't support // memory mapping) or call fsutil.GenericConfigureMMap with the appropriate // memmap.Mappable. ConfigureMMap(ctx context.Context, file *File, opts *memmap.MMapOpts) error diff --git a/pkg/sentry/fs/file_overlay_test.go b/pkg/sentry/fs/file_overlay_test.go index f121cbdda..a4ac58763 100644 --- a/pkg/sentry/fs/file_overlay_test.go +++ b/pkg/sentry/fs/file_overlay_test.go @@ -20,7 +20,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - ramfstest "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest" ) @@ -135,7 +136,7 @@ func TestReaddirRevalidation(t *testing.T) { // Get a handle to the dirent in the upper filesystem so that we can // modify it without going through the dirent. - upperDir := upper.InodeOperations.(*dir).InodeOperations.(*ramfstest.Dir) + upperDir := upper.InodeOperations.(*dir).InodeOperations.(*ramfs.Dir) // Check that overlay returns the files from both upper and lower. openDir, err := overlay.GetFile(ctx, fs.NewDirent(overlay, "stub"), fs.FileFlags{Read: true}) @@ -155,7 +156,7 @@ func TestReaddirRevalidation(t *testing.T) { if err := upperDir.Remove(ctx, upper, "a"); err != nil { t.Fatalf("error removing child: %v", err) } - upperDir.AddChild(ctx, "c", fs.NewInode(ramfstest.NewFile(ctx, fs.FilePermissions{}), + upperDir.AddChild(ctx, "c", fs.NewInode(fsutil.NewSimpleFileInode(ctx, fs.RootOwner, fs.FilePermissions{}, 0), upper.MountSource, fs.StableAttr{Type: fs.RegularFile})) // Seek to beginning of the directory and do the readdir again. diff --git a/pkg/sentry/fs/filetest/filetest.go b/pkg/sentry/fs/filetest/filetest.go index 65ca196d9..40d84d9f2 100644 --- a/pkg/sentry/fs/filetest/filetest.go +++ b/pkg/sentry/fs/filetest/filetest.go @@ -31,14 +31,14 @@ import ( // TestFileOperations is an implementation of the File interface. It provides all // required methods. type TestFileOperations struct { - fsutil.NoopRelease `state:"nosave"` - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` - fsutil.NoIoctl `state:"nosave"` - waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } // NewTestFile creates and initializes a new test file. diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index 6834e1272..4965e1a5f 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -67,7 +67,6 @@ go_library( "frame_ref_set.go", "frame_ref_set_impl.go", "fsutil.go", - "handle.go", "host_file_mapper.go", "host_file_mapper_state.go", "host_file_mapper_unsafe.go", @@ -96,20 +95,6 @@ go_library( ], ) -go_test( - name = "fsutil_x_test", - size = "small", - srcs = ["handle_test.go"], - deps = [ - ":fsutil", - "//pkg/sentry/context", - "//pkg/sentry/context/contexttest", - "//pkg/sentry/fs", - "//pkg/sentry/fs/ramfs/test", - "//pkg/sentry/usermem", - ], -) - go_test( name = "fsutil_test", size = "small", diff --git a/pkg/sentry/fs/fsutil/file.go b/pkg/sentry/fs/fsutil/file.go index 46db2e51c..0970f782b 100644 --- a/pkg/sentry/fs/fsutil/file.go +++ b/pkg/sentry/fs/fsutil/file.go @@ -24,12 +24,12 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// NoopRelease implements FileOperations.Release for files that have no +// FileNoopRelease implements fs.FileOperations.Release for files that have no // resources to release. -type NoopRelease struct{} +type FileNoopRelease struct{} // Release is a no-op. -func (NoopRelease) Release() {} +func (FileNoopRelease) Release() {} // SeekWithDirCursor is used to implement fs.FileOperations.Seek. If dirCursor // is not nil and the seek was on a directory, the cursor will be updated. @@ -127,71 +127,81 @@ func SeekWithDirCursor(ctx context.Context, file *fs.File, whence fs.SeekWhence, return current, syserror.EINVAL } -// GenericSeek implements FileOperations.Seek for files that use a generic -// seek implementation. -type GenericSeek struct{} +// FileGenericSeek implements fs.FileOperations.Seek for files that use a +// generic seek implementation. +type FileGenericSeek struct{} // Seek implements fs.FileOperations.Seek. -func (GenericSeek) Seek(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64) (int64, error) { +func (FileGenericSeek) Seek(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64) (int64, error) { return SeekWithDirCursor(ctx, file, whence, offset, nil) } -// ZeroSeek implements FileOperations.Seek for files that maintain a constant -// zero-value offset and require a no-op Seek. -type ZeroSeek struct{} +// FileZeroSeek implements fs.FileOperations.Seek for files that maintain a +// constant zero-value offset and require a no-op Seek. +type FileZeroSeek struct{} -// Seek implements FileOperations.Seek. -func (ZeroSeek) Seek(context.Context, *fs.File, fs.SeekWhence, int64) (int64, error) { +// Seek implements fs.FileOperations.Seek. +func (FileZeroSeek) Seek(context.Context, *fs.File, fs.SeekWhence, int64) (int64, error) { return 0, nil } -// PipeSeek implements FileOperations.Seek and can be used for files that behave -// like pipes (seeking is not supported). -type PipeSeek struct{} +// FileNoSeek implements fs.FileOperations.Seek to return EINVAL. +type FileNoSeek struct{} + +// Seek implements fs.FileOperations.Seek. +func (FileNoSeek) Seek(context.Context, *fs.File, fs.SeekWhence, int64) (int64, error) { + return 0, syserror.EINVAL +} -// Seek implements FileOperations.Seek. -func (PipeSeek) Seek(context.Context, *fs.File, fs.SeekWhence, int64) (int64, error) { +// FilePipeSeek implements fs.FileOperations.Seek and can be used for files +// that behave like pipes (seeking is not supported). +type FilePipeSeek struct{} + +// Seek implements fs.FileOperations.Seek. +func (FilePipeSeek) Seek(context.Context, *fs.File, fs.SeekWhence, int64) (int64, error) { return 0, syserror.ESPIPE } -// NotDirReaddir implements FileOperations.Readdir for non-directories. -type NotDirReaddir struct{} +// FileNotDirReaddir implements fs.FileOperations.Readdir for non-directories. +type FileNotDirReaddir struct{} -// Readdir implements FileOperations.NotDirReaddir. -func (NotDirReaddir) Readdir(context.Context, *fs.File, fs.DentrySerializer) (int64, error) { +// Readdir implements fs.FileOperations.FileNotDirReaddir. +func (FileNotDirReaddir) Readdir(context.Context, *fs.File, fs.DentrySerializer) (int64, error) { return 0, syserror.ENOTDIR } -// NoFsync implements FileOperations.Fsync for files that don't support syncing. -type NoFsync struct{} +// FileNoFsync implements fs.FileOperations.Fsync for files that don't support +// syncing. +type FileNoFsync struct{} -// Fsync implements FileOperations.Fsync. -func (NoFsync) Fsync(context.Context, *fs.File, int64, int64, fs.SyncType) error { +// Fsync implements fs.FileOperations.Fsync. +func (FileNoFsync) Fsync(context.Context, *fs.File, int64, int64, fs.SyncType) error { return syserror.EINVAL } -// NoopFsync implements FileOperations.Fsync for files that don't need to synced. -type NoopFsync struct{} +// FileNoopFsync implements fs.FileOperations.Fsync for files that don't need +// to synced. +type FileNoopFsync struct{} -// Fsync implements FileOperations.Fsync. -func (NoopFsync) Fsync(context.Context, *fs.File, int64, int64, fs.SyncType) error { +// Fsync implements fs.FileOperations.Fsync. +func (FileNoopFsync) Fsync(context.Context, *fs.File, int64, int64, fs.SyncType) error { return nil } -// NoopFlush implements FileOperations.Flush as a no-op. -type NoopFlush struct{} +// FileNoopFlush implements fs.FileOperations.Flush as a no-op. +type FileNoopFlush struct{} -// Flush implements FileOperations.Flush. -func (NoopFlush) Flush(context.Context, *fs.File) error { +// Flush implements fs.FileOperations.Flush. +func (FileNoopFlush) Flush(context.Context, *fs.File) error { return nil } -// NoMMap implements fs.FileOperations.Mappable for files that cannot +// FileNoMMap implements fs.FileOperations.Mappable for files that cannot // be memory mapped. -type NoMMap struct{} +type FileNoMMap struct{} // ConfigureMMap implements fs.FileOperations.ConfigureMMap. -func (NoMMap) ConfigureMMap(context.Context, *fs.File, *memmap.MMapOpts) error { +func (FileNoMMap) ConfigureMMap(context.Context, *fs.File, *memmap.MMapOpts) error { return syserror.ENODEV } @@ -204,26 +214,43 @@ func GenericConfigureMMap(file *fs.File, m memmap.Mappable, opts *memmap.MMapOpt return nil } -// NoIoctl implements fs.FileOperations.Ioctl for files that don't implement -// the ioctl syscall. -type NoIoctl struct{} +// FileNoIoctl implements fs.FileOperations.Ioctl for files that don't +// implement the ioctl syscall. +type FileNoIoctl struct{} // Ioctl implements fs.FileOperations.Ioctl. -func (NoIoctl) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { +func (FileNoIoctl) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { return 0, syserror.ENOTTY } -// DirFileOperations implements FileOperations for directories. +// DirFileOperations implements most of fs.FileOperations for directories, +// except for Readdir which the embedding type must implement. +type DirFileOperations struct { + waiter.AlwaysReady + FileGenericSeek + FileNoFsync + FileNoIoctl + FileNoMMap + FileNoopFlush + FileNoopRelease +} + +// Read implements fs.FileOperations.Read +func (*DirFileOperations) Read(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, syserror.EISDIR +} + +// Write implements fs.FileOperations.Write. +func (*DirFileOperations) Write(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, syserror.EISDIR +} + +// StaticDirFileOperations implements fs.FileOperations for directories with +// static children. // // +stateify savable -type DirFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - NoopRelease `state:"nosave"` - GenericSeek `state:"nosave"` - NoFsync `state:"nosave"` - NoopFlush `state:"nosave"` - NoMMap `state:"nosave"` - NoIoctl `state:"nosave"` +type StaticDirFileOperations struct { + DirFileOperations // dentryMap is a SortedDentryMap used to implement Readdir. dentryMap *fs.SortedDentryMap @@ -233,37 +260,106 @@ type DirFileOperations struct { dirCursor string } -// NewDirFileOperations returns a new DirFileOperations that will iterate the -// given denty map. -func NewDirFileOperations(dentries *fs.SortedDentryMap) *DirFileOperations { - return &DirFileOperations{ +// NewStaticDirFileOperations returns a new StaticDirFileOperations that will +// iterate the given denty map. +func NewStaticDirFileOperations(dentries *fs.SortedDentryMap) *StaticDirFileOperations { + return &StaticDirFileOperations{ dentryMap: dentries, } } // IterateDir implements DirIterator.IterateDir. -func (dfo *DirFileOperations) IterateDir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { - n, err := fs.GenericReaddir(dirCtx, dfo.dentryMap) +func (sdfo *StaticDirFileOperations) IterateDir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { + n, err := fs.GenericReaddir(dirCtx, sdfo.dentryMap) return offset + n, err } -// Readdir implements FileOperations.Readdir. -func (dfo *DirFileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { +// Readdir implements fs.FileOperations.Readdir. +func (sdfo *StaticDirFileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { root := fs.RootFromContext(ctx) defer root.DecRef() dirCtx := &fs.DirCtx{ Serializer: serializer, - DirCursor: &dfo.dirCursor, + DirCursor: &sdfo.dirCursor, } - return fs.DirentReaddir(ctx, file.Dirent, dfo, root, dirCtx, file.Offset()) + return fs.DirentReaddir(ctx, file.Dirent, sdfo, root, dirCtx, file.Offset()) } -// Read implements FileOperations.Read -func (*DirFileOperations) Read(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { - return 0, syserror.EISDIR +// NoReadWriteFile is a file that does not support reading or writing. +// +// +stateify savable +type NoReadWriteFile struct { + waiter.AlwaysReady `state:"nosave"` + FileGenericSeek `state:"nosave"` + FileNoIoctl `state:"nosave"` + FileNoMMap `state:"nosave"` + FileNoopFsync `state:"nosave"` + FileNoopFlush `state:"nosave"` + FileNoopRelease `state:"nosave"` + FileNoRead `state:"nosave"` + FileNoWrite `state:"nosave"` + FileNotDirReaddir `state:"nosave"` } -// Write implements FileOperations.Write. -func (*DirFileOperations) Write(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { - return 0, syserror.EISDIR +var _ fs.FileOperations = (*NoReadWriteFile)(nil) + +// FileStaticContentReader is a helper to implement fs.FileOperations.Read with +// static content. +// +// +stateify savable +type FileStaticContentReader struct { + // content is immutable. + content []byte +} + +// NewFileStaticContentReader initializes a FileStaticContentReader with the +// given content. +func NewFileStaticContentReader(b []byte) FileStaticContentReader { + return FileStaticContentReader{ + content: b, + } +} + +// Read implements fs.FileOperations.Read. +func (scr *FileStaticContentReader) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { + if offset < 0 { + return 0, syserror.EINVAL + } + if offset >= int64(len(scr.content)) { + return 0, nil + } + n, err := dst.CopyOut(ctx, scr.content[offset:]) + return int64(n), err +} + +// FileNoopWrite implements fs.FileOperations.Write as a noop. +type FileNoopWrite struct{} + +// Write implements fs.FileOperations.Write. +func (FileNoopWrite) Write(_ context.Context, _ *fs.File, src usermem.IOSequence, _ int64) (int64, error) { + return src.NumBytes(), nil +} + +// FileNoRead implements fs.FileOperations.Read to return EINVAL. +type FileNoRead struct{} + +// Read implements fs.FileOperations.Read. +func (FileNoRead) Read(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, syserror.EINVAL +} + +// FileNoWrite implements fs.FileOperations.Write to return EINVAL. +type FileNoWrite struct{} + +// Write implements fs.FileOperations.Write. +func (FileNoWrite) Write(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, syserror.EINVAL +} + +// FileNoopRead implement fs.FileOperations.Read as a noop. +type FileNoopRead struct{} + +// Read implements fs.FileOperations.Read. +func (FileNoopRead) Read(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, nil } diff --git a/pkg/sentry/fs/fsutil/fsutil.go b/pkg/sentry/fs/fsutil/fsutil.go index 3d7f3732d..319c4841b 100644 --- a/pkg/sentry/fs/fsutil/fsutil.go +++ b/pkg/sentry/fs/fsutil/fsutil.go @@ -20,7 +20,5 @@ // - For fs.Inodes that require a page cache to be memory mapped, see // inode_cache.go. // -// - For fs.Files that implement fs.HandleOps, see handle.go. -// // - For anon fs.Inodes, see anon.go. package fsutil diff --git a/pkg/sentry/fs/fsutil/handle.go b/pkg/sentry/fs/fsutil/handle.go deleted file mode 100644 index 8920b72ee..000000000 --- a/pkg/sentry/fs/fsutil/handle.go +++ /dev/null @@ -1,128 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package fsutil - -import ( - "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/waiter" -) - -// Handle implements FileOperations. -// -// FIXME: Remove Handle entirely in favor of individual fs.File -// implementations using simple generic utilities. -// -// +stateify savable -type Handle struct { - NoopRelease `state:"nosave"` - NoIoctl `state:"nosave"` - HandleOperations fs.HandleOperations - - // dirCursor is the directory cursor. - dirCursor string -} - -// NewHandle returns a File backed by the Dirent and FileFlags. -func NewHandle(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags, hops fs.HandleOperations) *fs.File { - if !fs.IsPipe(dirent.Inode.StableAttr) && !fs.IsSocket(dirent.Inode.StableAttr) { - // Allow reading/writing at an arbitrary offset for non-pipes - // and non-sockets. - flags.Pread = true - flags.Pwrite = true - } - - return fs.NewFile(ctx, dirent, flags, &Handle{HandleOperations: hops}) -} - -// Readiness implements waiter.Waitable.Readiness. -func (h *Handle) Readiness(mask waiter.EventMask) waiter.EventMask { - return h.HandleOperations.Readiness(mask) -} - -// EventRegister implements waiter.Waitable.EventRegister. -func (h *Handle) EventRegister(e *waiter.Entry, mask waiter.EventMask) { - h.HandleOperations.EventRegister(e, mask) -} - -// EventUnregister implements waiter.Waitable.EventUnregister. -func (h *Handle) EventUnregister(e *waiter.Entry) { - h.HandleOperations.EventUnregister(e) -} - -// Readdir implements FileOperations.Readdir. -func (h *Handle) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { - root := fs.RootFromContext(ctx) - defer root.DecRef() - dirCtx := &fs.DirCtx{ - Serializer: serializer, - DirCursor: &h.dirCursor, - } - n, err := fs.DirentReaddir(ctx, file.Dirent, h, root, dirCtx, file.Offset()) - return n, err -} - -// Seek implements FileOperations.Seek. -func (h *Handle) Seek(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64) (int64, error) { - return SeekWithDirCursor(ctx, file, whence, offset, &h.dirCursor) -} - -// IterateDir implements DirIterator.IterateDir. -func (h *Handle) IterateDir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { - return h.HandleOperations.DeprecatedReaddir(ctx, dirCtx, offset) -} - -// Read implements FileOperations.Read. -func (h *Handle) Read(ctx context.Context, file *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { - return h.HandleOperations.DeprecatedPreadv(ctx, dst, offset) -} - -// Write implements FileOperations.Write. -func (h *Handle) Write(ctx context.Context, file *fs.File, src usermem.IOSequence, offset int64) (int64, error) { - return h.HandleOperations.DeprecatedPwritev(ctx, src, offset) -} - -// Fsync implements FileOperations.Fsync. -func (h *Handle) Fsync(ctx context.Context, file *fs.File, start int64, end int64, syncType fs.SyncType) error { - switch syncType { - case fs.SyncAll, fs.SyncData: - // Write out metadata. - if err := file.Dirent.Inode.WriteOut(ctx); err != nil { - return err - } - fallthrough - case fs.SyncBackingStorage: - // Use DeprecatedFsync to sync disks. - return h.HandleOperations.DeprecatedFsync() - } - panic("invalid sync type") -} - -// Flush implements FileOperations.Flush. -func (h *Handle) Flush(context.Context, *fs.File) error { - return h.HandleOperations.DeprecatedFlush() -} - -// ConfigureMMap implements FileOperations.ConfigureMMap. -func (h *Handle) ConfigureMMap(ctx context.Context, file *fs.File, opts *memmap.MMapOpts) error { - mappable := file.Dirent.Inode.Mappable() - if mappable == nil { - return syserror.ENODEV - } - return GenericConfigureMMap(file, mappable, opts) -} diff --git a/pkg/sentry/fs/fsutil/handle_test.go b/pkg/sentry/fs/fsutil/handle_test.go deleted file mode 100644 index 43e1a3bdf..000000000 --- a/pkg/sentry/fs/fsutil/handle_test.go +++ /dev/null @@ -1,227 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package fsutil_test - -import ( - "io" - "syscall" - "testing" - - "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" - ramfstest "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" -) - -type testInodeOperations struct { - fs.InodeOperations - fs.InodeType - FileSize int64 - writes uint - reads uint -} - -func (t *testInodeOperations) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - return fs.UnstableAttr{Size: t.FileSize}, nil -} - -// Check implements InodeOperations.Check. -func (t *testInodeOperations) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { - return fs.ContextCanAccessFile(ctx, inode, p) -} - -func (t *testInodeOperations) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - t.reads++ - return t.InodeOperations.DeprecatedPreadv(ctx, dst, offset) -} - -func (t *testInodeOperations) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { - t.writes++ - return t.InodeOperations.DeprecatedPwritev(ctx, src, offset) -} - -// testHandle returns a handle for a test node. -// -// The size of the node is fixed at 20 bytes. -func testHandle(t *testing.T, flags fs.FileFlags, nt fs.InodeType) (*fs.File, *testInodeOperations) { - ctx := contexttest.Context(t) - m := fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}) - n := &testInodeOperations{ - InodeOperations: ramfstest.NewFile(ctx, fs.FilePermissions{User: fs.PermMask{Read: true, Write: true}}), - FileSize: 20, - } - d := fs.NewDirent(fs.NewInode(n, m, fs.StableAttr{Type: nt}), "test") - return fsutil.NewHandle(ctx, d, flags, d.Inode.HandleOps()), n -} - -func TestHandleOps(t *testing.T) { - h, n := testHandle(t, fs.FileFlags{Read: true, Write: true}, fs.RegularFile) - defer h.DecRef() - - // Make sure a write request works. - if n, err := h.Writev(contexttest.Context(t), usermem.BytesIOSequence([]byte("a"))); n != 1 || err != nil { - t.Fatalf("Writev: got (%d, %v), wanted (1, nil)", n, err) - } - if n.writes != 1 { - t.Errorf("found %d writes, expected 1", n.writes) - } - - // Make sure a read request works. - dst := make([]byte, 1) - if n, err := h.Preadv(contexttest.Context(t), usermem.BytesIOSequence(dst), 0); n != 1 || (err != nil && err != io.EOF) { - t.Errorf("Preadv: got (%d, %v), wanted (1, nil or EOF)", n, err) - } - if dst[0] != 'a' { - t.Errorf("Preadv: read %q, wanted 'a'", dst[0]) - } - if n.reads != 1 { - t.Errorf("found %d reads, expected 1", n.reads) - } -} - -type seekTest struct { - whence fs.SeekWhence - offset int64 - result int64 - err error -} - -type seekSuite struct { - nodeType fs.InodeType - cases []seekTest -} - -// FIXME: This is currently missing fs.SeekEnd tests due to the -// fact that NullInodeOperations returns an error on stat. -func TestHandleSeek(t *testing.T) { - ts := []seekSuite{ - { - nodeType: fs.RegularFile, - cases: []seekTest{ - {fs.SeekSet, 0, 0, nil}, - {fs.SeekSet, 10, 10, nil}, - {fs.SeekSet, -5, 10, syscall.EINVAL}, - {fs.SeekCurrent, -1, 9, nil}, - {fs.SeekCurrent, 2, 11, nil}, - {fs.SeekCurrent, -12, 11, syscall.EINVAL}, - {fs.SeekEnd, -1, 19, nil}, - {fs.SeekEnd, 0, 20, nil}, - {fs.SeekEnd, 2, 22, nil}, - }, - }, - { - nodeType: fs.Directory, - cases: []seekTest{ - {fs.SeekSet, 0, 0, nil}, - {fs.SeekSet, 10, 0, syscall.EINVAL}, - {fs.SeekSet, -5, 0, syscall.EINVAL}, - {fs.SeekCurrent, 0, 0, nil}, - {fs.SeekCurrent, 11, 0, syscall.EINVAL}, - {fs.SeekCurrent, -6, 0, syscall.EINVAL}, - {fs.SeekEnd, 0, 0, syscall.EINVAL}, - {fs.SeekEnd, -1, 0, syscall.EINVAL}, - {fs.SeekEnd, 2, 0, syscall.EINVAL}, - }, - }, - { - nodeType: fs.Symlink, - cases: []seekTest{ - {fs.SeekSet, 5, 0, syscall.EINVAL}, - {fs.SeekSet, -5, 0, syscall.EINVAL}, - {fs.SeekSet, 0, 0, syscall.EINVAL}, - {fs.SeekCurrent, 5, 0, syscall.EINVAL}, - {fs.SeekCurrent, -5, 0, syscall.EINVAL}, - {fs.SeekCurrent, 0, 0, syscall.EINVAL}, - {fs.SeekEnd, 5, 0, syscall.EINVAL}, - {fs.SeekEnd, -5, 0, syscall.EINVAL}, - {fs.SeekEnd, 0, 0, syscall.EINVAL}, - }, - }, - { - nodeType: fs.Pipe, - cases: []seekTest{ - {fs.SeekSet, 5, 0, syscall.ESPIPE}, - {fs.SeekSet, -5, 0, syscall.ESPIPE}, - {fs.SeekSet, 0, 0, syscall.ESPIPE}, - {fs.SeekCurrent, 5, 0, syscall.ESPIPE}, - {fs.SeekCurrent, -5, 0, syscall.ESPIPE}, - {fs.SeekCurrent, 0, 0, syscall.ESPIPE}, - {fs.SeekEnd, 5, 0, syscall.ESPIPE}, - {fs.SeekEnd, -5, 0, syscall.ESPIPE}, - {fs.SeekEnd, 0, 0, syscall.ESPIPE}, - }, - }, - { - nodeType: fs.Socket, - cases: []seekTest{ - {fs.SeekSet, 5, 0, syscall.ESPIPE}, - {fs.SeekSet, -5, 0, syscall.ESPIPE}, - {fs.SeekSet, 0, 0, syscall.ESPIPE}, - {fs.SeekCurrent, 5, 0, syscall.ESPIPE}, - {fs.SeekCurrent, -5, 0, syscall.ESPIPE}, - {fs.SeekCurrent, 0, 0, syscall.ESPIPE}, - {fs.SeekEnd, 5, 0, syscall.ESPIPE}, - {fs.SeekEnd, -5, 0, syscall.ESPIPE}, - {fs.SeekEnd, 0, 0, syscall.ESPIPE}, - }, - }, - { - nodeType: fs.CharacterDevice, - cases: []seekTest{ - {fs.SeekSet, 5, 0, nil}, - {fs.SeekSet, -5, 0, nil}, - {fs.SeekSet, 0, 0, nil}, - {fs.SeekCurrent, 5, 0, nil}, - {fs.SeekCurrent, -5, 0, nil}, - {fs.SeekCurrent, 0, 0, nil}, - {fs.SeekEnd, 5, 0, nil}, - {fs.SeekEnd, -5, 0, nil}, - {fs.SeekEnd, 0, 0, nil}, - }, - }, - { - nodeType: fs.BlockDevice, - cases: []seekTest{ - {fs.SeekSet, 0, 0, nil}, - {fs.SeekSet, 10, 10, nil}, - {fs.SeekSet, -5, 10, syscall.EINVAL}, - {fs.SeekCurrent, -1, 9, nil}, - {fs.SeekCurrent, 2, 11, nil}, - {fs.SeekCurrent, -12, 11, syscall.EINVAL}, - {fs.SeekEnd, -1, 19, nil}, - {fs.SeekEnd, 0, 20, nil}, - {fs.SeekEnd, 2, 22, nil}, - }, - }, - } - - for _, s := range ts { - h, _ := testHandle(t, fs.FileFlags{Read: true, Write: true}, s.nodeType) - defer h.DecRef() - - for _, c := range s.cases { - // Try the given seek. - offset, err := h.Seek(contexttest.Context(t), c.whence, c.offset) - if err != c.err { - t.Errorf("seek(%s, %d) on %s had unexpected error: expected %v, got %v", c.whence, c.offset, s.nodeType, c.err, err) - } - if err == nil && offset != c.result { - t.Errorf("seek(%s, %d) on %s had bad result: expected %v, got %v", c.whence, c.offset, s.nodeType, c.result, offset) - } - } - } -} diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index d4db1c2de..f1f5ec1de 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -15,213 +15,270 @@ package fsutil import ( + "sync" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// NewSimpleInodeOperations constructs fs.InodeOperations from InodeSimpleAttributes. -func NewSimpleInodeOperations(i InodeSimpleAttributes) fs.InodeOperations { - return &simpleInodeOperations{InodeSimpleAttributes: i} +// SimpleFileInode is a simple implementation of InodeOperations. +// +// +stateify savable +type SimpleFileInode struct { + InodeGenericChecker `state:"nosave"` + InodeNoExtendedAttributes `state:"nosave"` + InodeNoopRelease `state:"nosave"` + InodeNoopWriteOut `state:"nosave"` + InodeNotDirectory `state:"nosave"` + InodeNotMappable `state:"nosave"` + InodeNotOpenable `state:"nosave"` + InodeNotSocket `state:"nosave"` + InodeNotSymlink `state:"nosave"` + InodeNotTruncatable `state:"nosave"` + InodeNotVirtual `state:"nosave"` + + InodeSimpleAttributes +} + +// NewSimpleFileInode returns a new SimpleFileInode. +func NewSimpleFileInode(ctx context.Context, owner fs.FileOwner, perms fs.FilePermissions, typ uint64) *SimpleFileInode { + return &SimpleFileInode{ + InodeSimpleAttributes: NewInodeSimpleAttributes(ctx, owner, perms, typ), + } } -// simpleInodeOperations is a simple implementation of Inode. +// NoReadWriteFileInode is an implementation of InodeOperations that supports +// opening files that are not readable or writeable. // // +stateify savable -type simpleInodeOperations struct { - DeprecatedFileOperations `state:"nosave"` +type NoReadWriteFileInode struct { + InodeGenericChecker `state:"nosave"` + InodeNoExtendedAttributes `state:"nosave"` + InodeNoopRelease `state:"nosave"` + InodeNoopWriteOut `state:"nosave"` InodeNotDirectory `state:"nosave"` + InodeNotMappable `state:"nosave"` InodeNotSocket `state:"nosave"` - InodeNotRenameable `state:"nosave"` - InodeNotOpenable `state:"nosave"` - InodeNotVirtual `state:"nosave"` InodeNotSymlink `state:"nosave"` - InodeNoExtendedAttributes `state:"nosave"` - NoMappable `state:"nosave"` - NoopWriteOut `state:"nosave"` + InodeNotTruncatable `state:"nosave"` + InodeNotVirtual `state:"nosave"` InodeSimpleAttributes } -// InodeSimpleAttributes implements a subset of the Inode interface. It provides -// read-only access to attributes. +// NewNoReadWriteFileInode returns a new NoReadWriteFileInode. +func NewNoReadWriteFileInode(ctx context.Context, owner fs.FileOwner, perms fs.FilePermissions, typ uint64) *NoReadWriteFileInode { + return &NoReadWriteFileInode{ + InodeSimpleAttributes: NewInodeSimpleAttributes(ctx, owner, perms, typ), + } +} + +// GetFile implements fs.InodeOperations.GetFile. +func (*NoReadWriteFileInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &NoReadWriteFile{}), nil +} + +// InodeSimpleAttributes implements methods for updating in-memory unstable +// attributes. // // +stateify savable type InodeSimpleAttributes struct { - // FSType is the filesystem type reported by StatFS. + // FSType is the immutable filesystem type that will be returned by + // StatFS. FSType uint64 - // UAttr are the unstable attributes of the Inode. - UAttr fs.UnstableAttr + // mu protects unstable. + mu sync.RWMutex `state:"nosave"` + Unstable fs.UnstableAttr } -// Release implements fs.InodeOperations.Release. -func (i *InodeSimpleAttributes) Release(context.Context) {} - -// StatFS implements fs.InodeOperations.StatFS. -func (i *InodeSimpleAttributes) StatFS(context.Context) (fs.Info, error) { - return fs.Info{Type: i.FSType}, nil +// NewInodeSimpleAttributes returns a new InodeSimpleAttributes. +func NewInodeSimpleAttributes(ctx context.Context, owner fs.FileOwner, perms fs.FilePermissions, typ uint64) InodeSimpleAttributes { + return InodeSimpleAttributes{ + FSType: typ, + Unstable: fs.WithCurrentTime(ctx, fs.UnstableAttr{ + Owner: owner, + Perms: perms, + }), + } } // UnstableAttr implements fs.InodeOperations.UnstableAttr. -func (i *InodeSimpleAttributes) UnstableAttr(context.Context, *fs.Inode) (fs.UnstableAttr, error) { - return i.UAttr, nil +func (i *InodeSimpleAttributes) UnstableAttr(ctx context.Context, _ *fs.Inode) (fs.UnstableAttr, error) { + i.mu.RLock() + u := i.Unstable + i.mu.RUnlock() + return u, nil } -// Check implements fs.InodeOperations.Check. -func (i *InodeSimpleAttributes) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { - return fs.ContextCanAccessFile(ctx, inode, p) +// SetPermissions implements fs.InodeOperations.SetPermissions. +func (i *InodeSimpleAttributes) SetPermissions(ctx context.Context, _ *fs.Inode, p fs.FilePermissions) bool { + i.mu.Lock() + i.Unstable.SetPermissions(ctx, p) + i.mu.Unlock() + return true +} + +// SetOwner implements fs.InodeOperations.SetOwner. +func (i *InodeSimpleAttributes) SetOwner(ctx context.Context, _ *fs.Inode, owner fs.FileOwner) error { + i.mu.Lock() + i.Unstable.SetOwner(ctx, owner) + i.mu.Unlock() + return nil +} + +// SetTimestamps implements fs.InodeOperations.SetTimestamps. +func (i *InodeSimpleAttributes) SetTimestamps(ctx context.Context, _ *fs.Inode, ts fs.TimeSpec) error { + i.mu.Lock() + i.Unstable.SetTimestamps(ctx, ts) + i.mu.Unlock() + return nil } // AddLink implements fs.InodeOperations.AddLink. -func (*InodeSimpleAttributes) AddLink() {} +func (i *InodeSimpleAttributes) AddLink() { + i.mu.Lock() + i.Unstable.Links++ + i.mu.Unlock() +} // DropLink implements fs.InodeOperations.DropLink. -func (*InodeSimpleAttributes) DropLink() {} - -// NotifyStatusChange implements fs.fs.InodeOperations. -func (i *InodeSimpleAttributes) NotifyStatusChange(ctx context.Context) { - i.UAttr.StatusChangeTime = ktime.NowFromContext(ctx) +func (i *InodeSimpleAttributes) DropLink() { + i.mu.Lock() + i.Unstable.Links-- + i.mu.Unlock() } -// SetPermissions implements fs.InodeOperations.SetPermissions. -func (*InodeSimpleAttributes) SetPermissions(context.Context, *fs.Inode, fs.FilePermissions) bool { - return false +// StatFS implements fs.InodeOperations.StatFS. +func (i *InodeSimpleAttributes) StatFS(context.Context) (fs.Info, error) { + if i.FSType == 0 { + return fs.Info{}, syserror.ENOSYS + } + return fs.Info{Type: i.FSType}, nil } -// SetOwner implements fs.InodeOperations.SetOwner. -func (*InodeSimpleAttributes) SetOwner(context.Context, *fs.Inode, fs.FileOwner) error { - return syserror.EINVAL +// NotifyAccess updates the access time. +func (i *InodeSimpleAttributes) NotifyAccess(ctx context.Context) { + i.mu.Lock() + i.Unstable.AccessTime = ktime.NowFromContext(ctx) + i.mu.Unlock() } -// SetTimestamps implements fs.InodeOperations.SetTimestamps. -func (*InodeSimpleAttributes) SetTimestamps(context.Context, *fs.Inode, fs.TimeSpec) error { - return syserror.EINVAL +// NotifyModification updates the modification time. +func (i *InodeSimpleAttributes) NotifyModification(ctx context.Context) { + i.mu.Lock() + i.Unstable.ModificationTime = ktime.NowFromContext(ctx) + i.mu.Unlock() } -// Truncate implements fs.InodeOperations.Truncate. -func (*InodeSimpleAttributes) Truncate(context.Context, *fs.Inode, int64) error { - return syserror.EINVAL +// NotifyStatusChange updates the status change time. +func (i *InodeSimpleAttributes) NotifyStatusChange(ctx context.Context) { + i.mu.Lock() + i.Unstable.StatusChangeTime = ktime.NowFromContext(ctx) + i.mu.Unlock() } -// InMemoryAttributes implements utilities for updating in-memory unstable -// attributes and extended attributes. It is not thread-safe. -// -// Users need not initialize Xattrs to non-nil (it will be initialized -// when the first extended attribute is set. +// InodeSimpleExtendedAttributes implements +// fs.InodeOperations.{Get,Set,List}xattr. // // +stateify savable -type InMemoryAttributes struct { - Unstable fs.UnstableAttr - Xattrs map[string][]byte +type InodeSimpleExtendedAttributes struct { + // mu protects xattrs. + mu sync.RWMutex `state:"nosave"` + xattrs map[string][]byte } -// SetPermissions updates the permissions to p. -func (i *InMemoryAttributes) SetPermissions(ctx context.Context, p fs.FilePermissions) bool { - i.Unstable.Perms = p - i.Unstable.StatusChangeTime = ktime.NowFromContext(ctx) - return true +// Getxattr implements fs.InodeOperations.Getxattr. +func (i *InodeSimpleExtendedAttributes) Getxattr(_ *fs.Inode, name string) ([]byte, error) { + i.mu.RLock() + value, ok := i.xattrs[name] + i.mu.RUnlock() + if !ok { + return nil, syserror.ENOATTR + } + return value, nil } -// SetOwner updates the file owner to owner. -func (i *InMemoryAttributes) SetOwner(ctx context.Context, owner fs.FileOwner) error { - if owner.UID.Ok() { - i.Unstable.Owner.UID = owner.UID - } - if owner.GID.Ok() { - i.Unstable.Owner.GID = owner.GID +// Setxattr implements fs.InodeOperations.Setxattr. +func (i *InodeSimpleExtendedAttributes) Setxattr(_ *fs.Inode, name string, value []byte) error { + i.mu.Lock() + if i.xattrs == nil { + i.xattrs = make(map[string][]byte) } + i.xattrs[name] = value + i.mu.Unlock() return nil } -// SetTimestamps sets the timestamps to ts. -func (i *InMemoryAttributes) SetTimestamps(ctx context.Context, ts fs.TimeSpec) error { - if ts.ATimeOmit && ts.MTimeOmit { - return nil - } - - now := ktime.NowFromContext(ctx) - if !ts.ATimeOmit { - if ts.ATimeSetSystemTime { - i.Unstable.AccessTime = now - } else { - i.Unstable.AccessTime = ts.ATime - } - } - if !ts.MTimeOmit { - if ts.MTimeSetSystemTime { - i.Unstable.ModificationTime = now - } else { - i.Unstable.ModificationTime = ts.MTime - } +// Listxattr implements fs.InodeOperations.Listxattr. +func (i *InodeSimpleExtendedAttributes) Listxattr(_ *fs.Inode) (map[string]struct{}, error) { + i.mu.RLock() + names := make(map[string]struct{}, len(i.xattrs)) + for name := range i.xattrs { + names[name] = struct{}{} } - i.Unstable.StatusChangeTime = now - return nil + i.mu.RUnlock() + return names, nil } -// TouchAccessTime updates access time to the current time. -func (i *InMemoryAttributes) TouchAccessTime(ctx context.Context) { - i.Unstable.AccessTime = ktime.NowFromContext(ctx) -} +// staticFile is a file with static contents. It is returned by +// InodeStaticFileGetter.GetFile. +// +// +stateify savable +type staticFile struct { + waiter.AlwaysReady `state:"nosave"` + FileGenericSeek `state:"nosave"` + FileNoIoctl `state:"nosave"` + FileNoMMap `state:"nosave"` + FileNoopFsync `state:"nosave"` + FileNoopFlush `state:"nosave"` + FileNoopRelease `state:"nosave"` + FileNoopWrite `state:"nosave"` + FileNotDirReaddir `state:"nosave"` -// TouchModificationTime updates modification and status change -// time to the current time. -func (i *InMemoryAttributes) TouchModificationTime(ctx context.Context) { - now := ktime.NowFromContext(ctx) - i.Unstable.ModificationTime = now - i.Unstable.StatusChangeTime = now + FileStaticContentReader } -// TouchStatusChangeTime updates status change time to the current time. -func (i *InMemoryAttributes) TouchStatusChangeTime(ctx context.Context) { - i.Unstable.StatusChangeTime = ktime.NowFromContext(ctx) -} +// InodeNoStatFS implement StatFS by retuning ENOSYS. +type InodeNoStatFS struct{} -// Getxattr returns the extended attribute at name or ENOATTR if -// it isn't set. -func (i *InMemoryAttributes) Getxattr(name string) ([]byte, error) { - if value, ok := i.Xattrs[name]; ok { - return value, nil - } - return nil, syserror.ENOATTR +// StatFS implements fs.InodeOperations.StatFS. +func (InodeNoStatFS) StatFS(context.Context) (fs.Info, error) { + return fs.Info{}, syserror.ENOSYS } -// Setxattr sets the extended attribute at name to value. -func (i *InMemoryAttributes) Setxattr(name string, value []byte) error { - if i.Xattrs == nil { - i.Xattrs = make(map[string][]byte) - } - i.Xattrs[name] = value - return nil +// InodeStaticFileGetter implements GetFile for a file with static contents. +// +// +stateify savable +type InodeStaticFileGetter struct { + Contents []byte } -// Listxattr returns the set of all currently set extended attributes. -func (i *InMemoryAttributes) Listxattr() (map[string]struct{}, error) { - names := make(map[string]struct{}, len(i.Xattrs)) - for name := range i.Xattrs { - names[name] = struct{}{} - } - return names, nil +// GetFile implements fs.InodeOperations.GetFile. +func (i *InodeStaticFileGetter) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &staticFile{ + FileStaticContentReader: NewFileStaticContentReader(i.Contents), + }), nil } -// NoMappable returns a nil memmap.Mappable. -type NoMappable struct{} +// InodeNotMappable returns a nil memmap.Mappable. +type InodeNotMappable struct{} // Mappable implements fs.InodeOperations.Mappable. -func (NoMappable) Mappable(*fs.Inode) memmap.Mappable { +func (InodeNotMappable) Mappable(*fs.Inode) memmap.Mappable { return nil } -// NoopWriteOut is a no-op implementation of Inode.WriteOut. -type NoopWriteOut struct{} +// InodeNoopWriteOut is a no-op implementation of fs.InodeOperations.WriteOut. +type InodeNoopWriteOut struct{} // WriteOut is a no-op. -func (NoopWriteOut) WriteOut(context.Context, *fs.Inode) error { +func (InodeNoopWriteOut) WriteOut(context.Context, *fs.Inode) error { return nil } @@ -273,6 +330,11 @@ func (InodeNotDirectory) RemoveDirectory(context.Context, *fs.Inode, string) err return syserror.ENOTDIR } +// Rename implements fs.FileOperations.Rename. +func (InodeNotDirectory) Rename(context.Context, *fs.Inode, string, *fs.Inode, string) error { + return syserror.EINVAL +} + // InodeNotSocket can be used by Inodes that are not sockets. type InodeNotSocket struct{} @@ -281,7 +343,31 @@ func (InodeNotSocket) BoundEndpoint(*fs.Inode, string) transport.BoundEndpoint { return nil } -// InodeNotRenameable can be used by Inodes that cannot be renamed. +// InodeNotTruncatable can be used by Inodes that cannot be truncated. +type InodeNotTruncatable struct{} + +// Truncate implements fs.InodeOperations.Truncate. +func (InodeNotTruncatable) Truncate(context.Context, *fs.Inode, int64) error { + return syserror.EINVAL +} + +// InodeIsDirTruncate implements fs.InodeOperations.Truncate for directories. +type InodeIsDirTruncate struct{} + +// Truncate implements fs.InodeOperations.Truncate. +func (InodeIsDirTruncate) Truncate(context.Context, *fs.Inode, int64) error { + return syserror.EISDIR +} + +// InodeNoopTruncate implements fs.InodeOperations.Truncate as a noop. +type InodeNoopTruncate struct{} + +// Truncate implements fs.InodeOperations.Truncate. +func (InodeNoopTruncate) Truncate(context.Context, *fs.Inode, int64) error { + return nil +} + +// InodeNotRenameable can be used by Inodes that cannot be truncated. type InodeNotRenameable struct{} // Rename implements fs.InodeOperations.Rename. @@ -305,6 +391,14 @@ func (InodeNotVirtual) IsVirtual() bool { return false } +// InodeVirtual can be used by Inodes that are virtual. +type InodeVirtual struct{} + +// IsVirtual implements fs.InodeOperations.IsVirtual. +func (InodeVirtual) IsVirtual() bool { + return true +} + // InodeNotSymlink can be used by Inodes that are not symlinks. type InodeNotSymlink struct{} @@ -337,50 +431,17 @@ func (InodeNoExtendedAttributes) Listxattr(*fs.Inode) (map[string]struct{}, erro return nil, syserror.EOPNOTSUPP } -// DeprecatedFileOperations panics if any deprecated Inode method is called. -type DeprecatedFileOperations struct{} +// InodeNoopRelease implements fs.InodeOperations.Release as a noop. +type InodeNoopRelease struct{} -// Readiness implements fs.InodeOperations.Waitable.Readiness. -func (DeprecatedFileOperations) Readiness(waiter.EventMask) waiter.EventMask { - panic("not implemented") -} - -// EventRegister implements fs.InodeOperations.Waitable.EventRegister. -func (DeprecatedFileOperations) EventRegister(*waiter.Entry, waiter.EventMask) { - panic("not implemented") -} - -// EventUnregister implements fs.InodeOperations.Waitable.EventUnregister. -func (DeprecatedFileOperations) EventUnregister(*waiter.Entry) { - panic("not implemented") -} - -// DeprecatedPreadv implements fs.InodeOperations.DeprecatedPreadv. -func (DeprecatedFileOperations) DeprecatedPreadv(context.Context, usermem.IOSequence, int64) (int64, error) { - panic("not implemented") -} - -// DeprecatedPwritev implements fs.InodeOperations.DeprecatedPwritev. -func (DeprecatedFileOperations) DeprecatedPwritev(context.Context, usermem.IOSequence, int64) (int64, error) { - panic("not implemented") -} - -// DeprecatedReaddir implements fs.InodeOperations.DeprecatedReaddir. -func (DeprecatedFileOperations) DeprecatedReaddir(context.Context, *fs.DirCtx, int) (int, error) { - panic("not implemented") -} - -// DeprecatedFsync implements fs.InodeOperations.DeprecatedFsync. -func (DeprecatedFileOperations) DeprecatedFsync() error { - panic("not implemented") -} +// Release implements fs.InodeOperations.Release. +func (InodeNoopRelease) Release(context.Context) {} -// DeprecatedFlush implements fs.InodeOperations.DeprecatedFlush. -func (DeprecatedFileOperations) DeprecatedFlush() error { - panic("not implemented") -} +// InodeGenericChecker implements fs.InodeOperations.Check with a generic +// implementation. +type InodeGenericChecker struct{} -// DeprecatedMappable implements fs.InodeOperations.DeprecatedMappable. -func (DeprecatedFileOperations) DeprecatedMappable(context.Context, *fs.Inode) (memmap.Mappable, bool) { - panic("not implemented") +// Check implements fs.InodeOperations.Check. +func (InodeGenericChecker) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { + return fs.ContextCanAccessFile(ctx, inode, p) } diff --git a/pkg/sentry/fs/fsutil/inode_cached_test.go b/pkg/sentry/fs/fsutil/inode_cached_test.go index ce5201a40..9c9391511 100644 --- a/pkg/sentry/fs/fsutil/inode_cached_test.go +++ b/pkg/sentry/fs/fsutil/inode_cached_test.go @@ -261,15 +261,11 @@ func (noopMappingSpace) Invalidate(ar usermem.AddrRange, opts memmap.InvalidateO } func anonInode(ctx context.Context) *fs.Inode { - return fs.NewInode(NewSimpleInodeOperations(InodeSimpleAttributes{ - UAttr: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: fs.FileOwnerFromContext(ctx), - Perms: fs.FilePermissions{ - User: fs.PermMask{Read: true, Write: true}, - }, - Links: 1, - }), - }), fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}), fs.StableAttr{ + return fs.NewInode(&SimpleFileInode{ + InodeSimpleAttributes: NewInodeSimpleAttributes(ctx, fs.FileOwnerFromContext(ctx), fs.FilePermissions{ + User: fs.PermMask{Read: true, Write: true}, + }, 0), + }, fs.NewPseudoMountSource(), fs.StableAttr{ Type: fs.Anonymous, BlockSize: usermem.PageSize, }) diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index 6d961813d..3578b07a0 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -37,7 +37,7 @@ var openedWX = metric.MustCreateNewUint64Metric("/gofer/opened_write_execute_fil // // +stateify savable type fileOperations struct { - fsutil.NoIoctl `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` waiter.AlwaysReady `state:"nosave"` // inodeOperations is the inodeOperations backing the file. It is protected diff --git a/pkg/sentry/fs/gofer/fs.go b/pkg/sentry/fs/gofer/fs.go index ed30cb1f1..2dc000c6f 100644 --- a/pkg/sentry/fs/gofer/fs.go +++ b/pkg/sentry/fs/gofer/fs.go @@ -75,11 +75,11 @@ var ( // ErrNoTransport is returned when there is no 'trans' option. ErrNoTransport = errors.New("missing required option: 'trans='") - // ErrNoReadFD is returned when there is no 'rfdno' option. - ErrNoReadFD = errors.New("missing required option: 'rfdno='") + // ErrFileNoReadFD is returned when there is no 'rfdno' option. + ErrFileNoReadFD = errors.New("missing required option: 'rfdno='") - // ErrNoWriteFD is returned when there is no 'wfdno' option. - ErrNoWriteFD = errors.New("missing required option: 'wfdno='") + // ErrFileNoWriteFD is returned when there is no 'wfdno' option. + ErrFileNoWriteFD = errors.New("missing required option: 'wfdno='") ) // filesystem is a 9p client. @@ -87,6 +87,8 @@ var ( // +stateify savable type filesystem struct{} +var _ fs.Filesystem = (*filesystem)(nil) + func init() { fs.RegisterFilesystem(&filesystem{}) } @@ -160,14 +162,14 @@ func options(data string) (opts, error) { // Check for the required 'rfdno=' option. srfd, ok := options[readFDKey] if !ok { - return o, ErrNoReadFD + return o, ErrFileNoReadFD } delete(options, readFDKey) // Check for the required 'wfdno=' option. swfd, ok := options[writeFDKey] if !ok { - return o, ErrNoWriteFD + return o, ErrFileNoWriteFD } delete(options, writeFDKey) diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go index 7c6e5b025..f0dc99fd0 100644 --- a/pkg/sentry/fs/gofer/inode.go +++ b/pkg/sentry/fs/gofer/inode.go @@ -40,7 +40,6 @@ import ( type inodeOperations struct { fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` - fsutil.DeprecatedFileOperations `state:"nosave"` // fileState implements fs.CachedFileObject. It exists // to break a circular load dependency between inodeOperations diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index bc6ee7aa4..4e84d1d6c 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -36,8 +36,8 @@ import ( // // +stateify savable type fileOperations struct { - fsutil.NoIoctl `state:"nosave"` - fsutil.NoopRelease `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` // iops are the Inode operations for this file. iops *inodeOperations `state:"wait"` diff --git a/pkg/sentry/fs/host/fs.go b/pkg/sentry/fs/host/fs.go index 54cbb94f9..d2ba38449 100644 --- a/pkg/sentry/fs/host/fs.go +++ b/pkg/sentry/fs/host/fs.go @@ -58,6 +58,8 @@ type Filesystem struct { paths []string } +var _ fs.Filesystem = (*Filesystem)(nil) + // Name is the identifier of this file system. func (*Filesystem) Name() string { return FilesystemName diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index 08754bd6b..6ff6c3254 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -39,7 +39,6 @@ import ( type inodeOperations struct { fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` - fsutil.DeprecatedFileOperations `state:"nosave"` // fileState implements fs.CachedFileObject. It exists // to break a circular load dependency between inodeOperations diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index fa3beb111..d32f52d55 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -356,11 +356,10 @@ func (i *Inode) AddLink() { if i.overlay != nil { // FIXME: Remove this from InodeOperations altogether. // - // This interface (including DropLink and NotifyStatusChange) - // is only used by ramfs to update metadata of children. These - // filesystems should _never_ have overlay Inodes cached as - // children. So explicitly disallow this scenario and avoid plumbing - // Dirents through to do copy up. + // This interface is only used by ramfs to update metadata of + // children. These filesystems should _never_ have overlay + // Inodes cached as children. So explicitly disallow this + // scenario and avoid plumbing Dirents through to do copy up. panic("overlay Inodes cached in ramfs directories are not supported") } i.InodeOperations.AddLink() @@ -375,15 +374,6 @@ func (i *Inode) DropLink() { i.InodeOperations.DropLink() } -// NotifyStatusChange calls i.InodeOperations.NotifyStatusChange. -func (i *Inode) NotifyStatusChange(ctx context.Context) { - if i.overlay != nil { - // Same as AddLink. - panic("overlay Inodes cached in ramfs directories are not supported") - } - i.InodeOperations.NotifyStatusChange(ctx) -} - // IsVirtual calls i.InodeOperations.IsVirtual. func (i *Inode) IsVirtual() bool { if i.overlay != nil { @@ -401,17 +391,6 @@ func (i *Inode) StatFS(ctx context.Context) (Info, error) { return i.InodeOperations.StatFS(ctx) } -// HandleOps extracts HandleOperations from i. -func (i *Inode) HandleOps() HandleOperations { - if i.overlay != nil { - return overlayHandleOps(i.overlay) - } - if h, ok := i.InodeOperations.(HandleOperations); ok { - return h - } - return nil -} - // CheckOwnership checks whether `ctx` owns this Inode or may act as its owner. // Compare Linux's fs/inode.c:inode_owner_or_capable(). func (i *Inode) CheckOwnership(ctx context.Context) bool { diff --git a/pkg/sentry/fs/inode_operations.go b/pkg/sentry/fs/inode_operations.go index 77973ce79..db40b5256 100644 --- a/pkg/sentry/fs/inode_operations.go +++ b/pkg/sentry/fs/inode_operations.go @@ -21,8 +21,6 @@ import ( ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/waiter" ) var ( @@ -303,83 +301,5 @@ type InodeOperations interface { // StatFS returns a filesystem Info implementation or an error. If // the filesystem does not support this operation (maybe in the future // it will), then ENOSYS should be returned. - // - // Move to MountSourceOperations. StatFS(context.Context) (Info, error) - - HandleOperations -} - -// HandleOperations are extended InodeOperations that are only implemented -// for file systems that use fs/handle.go:Handle to generate open Files. -// -// Handle is deprecated; these methods are deprecated as well. -// -// Filesystems are encouraged to implement the File interface directly -// instead of using Handle. To indicate that the below methods should never -// be called, embed DeprecatedFileOperations to satisfy this interface. -type HandleOperations interface { - waiter.Waitable - - // DeprecatedPreadv is deprecated in favor of filesystems - // implementing File.Preadv directly. - // - // DeprecatedPreadv reads up to dst.NumBytes() bytes into dst, starting at - // the given offset, and returns the number of bytes read. - // - // Preadv may return a partial read result before EOF is reached. - // - // If a symlink, Preadv reads the target value of the symlink. - // - // Preadv should not check for readable permissions. - DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) - - // DeprecatedPwritev is deprecated in favor of filesystems - // implementing File.Pwritev directly. - // - // DeprecatedPwritev writes up to src.NumBytes() bytes from src to the - // Inode, starting at the given offset and returns the number of bytes - // written. - // - // Pwritev should not check that the Inode has writable permissions. - DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) - - // DeprecatedReaddir is deprecated in favor of filesystems - // implementing File.Readdir directly. - // - // DeprecatedReaddir emits directory entries by calling dirCtx.EmitDir, - // beginning with the entry at offset. - // - // Entries for "." and ".." must *not* be included. - // - // If the offset returned is the same as the argument offset, then - // nothing has been serialized. This is equivalent to reaching EOF. - // In this case serializer.Written() should return 0. - // - // The order of entries to emit must be consistent between Readdir - // calls, and must start with the given offset. - // - // The caller must ensure that this operation is permitted. - DeprecatedReaddir(ctx context.Context, dirCtx *DirCtx, offset int) (int, error) - - // DeprecatedFsync is deprecated in favor of filesystems implementing - // File.Fsync directly. - // - // DeprecatedFsync syncs a file. - DeprecatedFsync() error - - // DeprecatedMappable is deprecated in favor of filesystems implementing - // File.Mappable directly. - // - // DeprecatedMappable returns a Mappable if the Inode can be mapped. - DeprecatedMappable(ctx context.Context, inode *Inode) (memmap.Mappable, bool) - - // DeprecatedFlush is deprecated in favor of filesystems implementing - // File.Flush directly. - // - // DeprecatedFlush flushes a file. - // - // Implementations may choose to free up memory or complete pending I/O - // but also may implement Flush as a no-op. - DeprecatedFlush() error } diff --git a/pkg/sentry/fs/inode_overlay.go b/pkg/sentry/fs/inode_overlay.go index 512a0da28..77a2623ef 100644 --- a/pkg/sentry/fs/inode_overlay.go +++ b/pkg/sentry/fs/inode_overlay.go @@ -596,19 +596,6 @@ func overlayStatFS(ctx context.Context, o *overlayEntry) (Info, error) { return i, nil } -func overlayHandleOps(o *overlayEntry) HandleOperations { - // Hot path. Avoid defers. - var hops HandleOperations - o.copyMu.RLock() - if o.upper != nil { - hops = o.upper.HandleOps() - } else { - hops = o.lower.HandleOps() - } - o.copyMu.RUnlock() - return hops -} - // NewTestOverlayDir returns an overlay Inode for tests. // // If `revalidate` is true, then the upper filesystem will require diff --git a/pkg/sentry/fs/inode_overlay_test.go b/pkg/sentry/fs/inode_overlay_test.go index 9e922d008..bc91be226 100644 --- a/pkg/sentry/fs/inode_overlay_test.go +++ b/pkg/sentry/fs/inode_overlay_test.go @@ -19,7 +19,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - ramfstest "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -376,7 +377,8 @@ type dir struct { // List of negative child names. negative []string - // Whether DeprecatedReaddir has been called on this dir. + // ReaddirCalled records whether Readdir was called on a file + // corresponding to this inode. ReaddirCalled bool } @@ -390,10 +392,19 @@ func (d *dir) Getxattr(inode *fs.Inode, name string) ([]byte, error) { return nil, syserror.ENOATTR } -// DeprecatedReaddir implements InodeOperations.DeprecatedReaddir. -func (d *dir) DeprecatedReaddir(ctx context.Context, dirctx *fs.DirCtx, offset int) (int, error) { - d.ReaddirCalled = true - return d.InodeOperations.DeprecatedReaddir(ctx, dirctx, offset) +// GetFile implements InodeOperations.GetFile. +func (d *dir) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + file, err := d.InodeOperations.GetFile(ctx, dirent, flags) + if err != nil { + return nil, err + } + defer file.DecRef() + // Wrap the file's FileOperations in a dirFile. + fops := &dirFile{ + FileOperations: file.FileOperations, + inode: d, + } + return fs.NewFile(ctx, dirent, flags, fops), nil } type dirContent struct { @@ -401,12 +412,45 @@ type dirContent struct { dir bool } +type dirFile struct { + fs.FileOperations + inode *dir +} + +type inode struct { + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeNotTruncatable `state:"nosave"` + fsutil.InodeNotVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes + fsutil.InodeStaticFileGetter +} + +// Readdir implements fs.FileOperations.Readdir. It sets the ReaddirCalled +// field on the inode. +func (f *dirFile) Readdir(ctx context.Context, file *fs.File, ser fs.DentrySerializer) (int64, error) { + f.inode.ReaddirCalled = true + return f.FileOperations.Readdir(ctx, file, ser) +} + func newTestRamfsInode(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - return fs.NewInode(ramfstest.NewFile(ctx, fs.FilePermissions{}), msrc, fs.StableAttr{Type: fs.RegularFile}) + inode := fs.NewInode(&inode{ + InodeStaticFileGetter: fsutil.InodeStaticFileGetter{ + Contents: []byte("foobar"), + }, + }, msrc, fs.StableAttr{Type: fs.RegularFile}) + return inode } func newTestRamfsDir(ctx context.Context, contains []dirContent, negative []string) *fs.Inode { - msrc := fs.NewCachingMountSource(nil, fs.MountSourceFlags{}) + msrc := fs.NewPseudoMountSource() contents := make(map[string]*fs.Inode) for _, c := range contains { if c.dir { @@ -415,7 +459,7 @@ func newTestRamfsDir(ctx context.Context, contains []dirContent, negative []stri contents[c.name] = newTestRamfsInode(ctx, msrc) } } - dops := ramfstest.NewDir(ctx, contents, fs.FilePermissions{ + dops := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermissions{ User: fs.PermMask{Read: true, Execute: true}, }) return fs.NewInode(&dir{ diff --git a/pkg/sentry/fs/mock.go b/pkg/sentry/fs/mock.go index 6bfcda6bb..abfdc6a25 100644 --- a/pkg/sentry/fs/mock.go +++ b/pkg/sentry/fs/mock.go @@ -16,7 +16,6 @@ package fs import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -151,16 +150,6 @@ func (n *MockInodeOperations) Truncate(ctx context.Context, inode *Inode, size i return nil } -// DeprecatedPwritev implements fs.InodeOperations.DeprecatedPwritev. -func (n *MockInodeOperations) DeprecatedPwritev(context.Context, usermem.IOSequence, int64) (int64, error) { - return 0, nil -} - -// DeprecatedReaddir implements fs.InodeOperations.DeprecatedReaddir. -func (n *MockInodeOperations) DeprecatedReaddir(context.Context, *DirCtx, int) (int, error) { - return 0, nil -} - // Remove implements fs.InodeOperations.Remove. func (n *MockInodeOperations) Remove(context.Context, *Inode, string) error { return nil diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index 24e28ddb2..dd6e64b4c 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -43,8 +43,6 @@ type DirentOperations interface { // MountSourceOperations contains filesystem specific operations. type MountSourceOperations interface { // TODO: Add: - // - // StatFS() (Info, error) // BlockSize() int64 // FS() Filesystem @@ -249,7 +247,7 @@ func (msrc *MountSource) FlushDirentRefs() { } // NewCachingMountSource returns a generic mount that will cache dirents -// aggressively. Filesystem may be nil if there is no backing filesystem. +// aggressively. func NewCachingMountSource(filesystem Filesystem, flags MountSourceFlags) *MountSource { return NewMountSource(&SimpleMountSourceOperations{ keep: true, @@ -258,7 +256,6 @@ func NewCachingMountSource(filesystem Filesystem, flags MountSourceFlags) *Mount } // NewNonCachingMountSource returns a generic mount that will never cache dirents. -// Filesystem may be nil if there is no backing filesystem. func NewNonCachingMountSource(filesystem Filesystem, flags MountSourceFlags) *MountSource { return NewMountSource(&SimpleMountSourceOperations{ keep: false, @@ -275,6 +272,15 @@ func NewRevalidatingMountSource(filesystem Filesystem, flags MountSourceFlags) * }, filesystem, flags) } +// NewPseudoMountSource returns a "pseudo" mount source that is not backed by +// an actual filesystem. It is always non-caching. +func NewPseudoMountSource() *MountSource { + return NewMountSource(&SimpleMountSourceOperations{ + keep: false, + revalidate: false, + }, nil, MountSourceFlags{}) +} + // SimpleMountSourceOperations implements MountSourceOperations. // // +stateify savable diff --git a/pkg/sentry/fs/mounts_test.go b/pkg/sentry/fs/mounts_test.go index 7d682d99b..54000614f 100644 --- a/pkg/sentry/fs/mounts_test.go +++ b/pkg/sentry/fs/mounts_test.go @@ -19,7 +19,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - ramfstest "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/contexttest" ) @@ -29,15 +30,15 @@ import ( // |-bar (file) func createMountNamespace(ctx context.Context) (*fs.MountNamespace, error) { perms := fs.FilePermsFromMode(0777) - m := fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}) + m := fs.NewPseudoMountSource() - barFile := ramfstest.NewFile(ctx, perms) - fooDir := ramfstest.NewDir(ctx, map[string]*fs.Inode{ + barFile := fsutil.NewSimpleFileInode(ctx, fs.RootOwner, perms, 0) + fooDir := ramfs.NewDir(ctx, map[string]*fs.Inode{ "bar": fs.NewInode(barFile, m, fs.StableAttr{Type: fs.RegularFile}), - }, perms) - rootDir := ramfstest.NewDir(ctx, map[string]*fs.Inode{ + }, fs.RootOwner, perms) + rootDir := ramfs.NewDir(ctx, map[string]*fs.Inode{ "foo": fs.NewInode(fooDir, m, fs.StableAttr{Type: fs.Directory}), - }, perms) + }, fs.RootOwner, perms) return fs.NewMountNamespace(ctx, fs.NewInode(rootDir, m, fs.StableAttr{Type: fs.Directory})) } diff --git a/pkg/sentry/fs/proc/BUILD b/pkg/sentry/fs/proc/BUILD index aff3c3c01..74954f213 100644 --- a/pkg/sentry/fs/proc/BUILD +++ b/pkg/sentry/fs/proc/BUILD @@ -8,9 +8,9 @@ go_library( "cpuinfo.go", "exec_args.go", "fds.go", - "file.go", "filesystems.go", "fs.go", + "inode.go", "loadavg.go", "meminfo.go", "mounts.go", @@ -32,6 +32,7 @@ go_library( "//pkg/abi/linux", "//pkg/sentry/context", "//pkg/sentry/fs", + "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/proc/device", "//pkg/sentry/fs/proc/seqfile", "//pkg/sentry/fs/ramfs", @@ -45,6 +46,7 @@ go_library( "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/syserror", + "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/proc/cpuinfo.go b/pkg/sentry/fs/proc/cpuinfo.go index f8be06dc3..f756c45bf 100644 --- a/pkg/sentry/fs/proc/cpuinfo.go +++ b/pkg/sentry/fs/proc/cpuinfo.go @@ -15,52 +15,21 @@ package proc import ( - "io" - "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// cpuinfo is a file describing the CPU capabilities. -// -// Presently cpuinfo never changes, so it doesn't need to be a SeqFile. -// -// +stateify savable -type cpuinfo struct { - ramfs.Entry - - // k is the system kernel. - k *kernel.Kernel -} - -// DeprecatedPreadv implements fs.InodeOperations.DeprecatedPreadv. -func (c *cpuinfo) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - features := c.k.FeatureSet() +func newCPUInfo(ctx context.Context, msrc *fs.MountSource) *fs.Inode { + k := kernel.KernelFromContext(ctx) + features := k.FeatureSet() if features == nil { // Kernel is always initialized with a FeatureSet. panic("cpuinfo read with nil FeatureSet") } - contents := make([]byte, 0, 1024) - for i, max := uint(0), c.k.ApplicationCores(); i < max; i++ { + for i, max := uint(0), k.ApplicationCores(); i < max; i++ { contents = append(contents, []byte(features.CPUInfo(i))...) } - if offset >= int64(len(contents)) { - return 0, io.EOF - } - - n, err := dst.CopyOut(ctx, contents[offset:]) - return int64(n), err -} - -func (p *proc) newCPUInfo(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - f := &cpuinfo{ - k: p.k, - } - f.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0444)) - - return newFile(f, msrc, fs.SpecialFile, nil) + return newStaticProcInode(ctx, msrc, contents) } diff --git a/pkg/sentry/fs/proc/exec_args.go b/pkg/sentry/fs/proc/exec_args.go index b4896053f..ddda67f54 100644 --- a/pkg/sentry/fs/proc/exec_args.go +++ b/pkg/sentry/fs/proc/exec_args.go @@ -18,12 +18,14 @@ import ( "fmt" "io" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // execArgType enumerates the types of exec arguments that are exposed through @@ -35,12 +37,12 @@ const ( environExecArg ) -// execArgFile is a file containing the exec args (either cmdline or environ) +// execArgInode is a inode containing the exec args (either cmdline or environ) // for a given task. // // +stateify savable -type execArgFile struct { - ramfs.Entry +type execArgInode struct { + fsutil.SimpleFileInode // arg is the type of exec argument this file contains. arg execArgType @@ -49,21 +51,52 @@ type execArgFile struct { t *kernel.Task } +var _ fs.InodeOperations = (*execArgInode)(nil) + // newExecArgFile creates a file containing the exec args of the given type. -func newExecArgFile(t *kernel.Task, msrc *fs.MountSource, arg execArgType) *fs.Inode { +func newExecArgInode(t *kernel.Task, msrc *fs.MountSource, arg execArgType) *fs.Inode { if arg != cmdlineExecArg && arg != environExecArg { panic(fmt.Sprintf("unknown exec arg type %v", arg)) } - f := &execArgFile{ - arg: arg, - t: t, + f := &execArgInode{ + SimpleFileInode: *fsutil.NewSimpleFileInode(t, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + arg: arg, + t: t, } - f.InitEntry(t, fs.RootOwner, fs.FilePermsFromMode(0444)) - return newFile(f, msrc, fs.SpecialFile, t) + return newProcInode(f, msrc, fs.SpecialFile, t) +} + +// GetFile implements fs.InodeOperations.GetFile. +func (i *execArgInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &execArgFile{ + arg: i.arg, + t: i.t, + }), nil } -// DeprecatedPreadv reads the exec arg from the process's address space.. -func (f *execArgFile) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { +// +stateify savable +type execArgFile struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` + + // arg is the type of exec argument this file contains. + arg execArgType + + // t is the Task to read the exec arg line from. + t *kernel.Task +} + +var _ fs.FileOperations = (*execArgFile)(nil) + +// Read reads the exec arg from the process's address space.. +func (f *execArgFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { if offset < 0 { return 0, syserror.EINVAL } diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index 5acbce75e..b8a0a5eff 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -21,11 +21,11 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/device" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -55,7 +55,7 @@ func walkDescriptors(t *kernel.Task, p string, toInode func(*fs.File, kernel.FDF // readDescriptors reads fds in the task starting at offset, and calls the // toDentAttr callback for each to get a DentAttr, which it then emits. This is // a helper for implementing fs.InodeOperations.Readdir. -func readDescriptors(t *kernel.Task, c *fs.DirCtx, offset int, toDentAttr func(int) fs.DentAttr) (int, error) { +func readDescriptors(t *kernel.Task, c *fs.DirCtx, offset int64, toDentAttr func(int) fs.DentAttr) (int64, error) { var fds kernel.FDs t.WithMuLocked(func(t *kernel.Task) { if fdm := t.FDMap(); fdm != nil { @@ -69,7 +69,7 @@ func readDescriptors(t *kernel.Task, c *fs.DirCtx, offset int, toDentAttr func(i } // Find the fd to start at. - idx := sort.SearchInts(fdInts, offset) + idx := sort.SearchInts(fdInts, int(offset)) if idx == len(fdInts) { return offset, nil } @@ -80,28 +80,32 @@ func readDescriptors(t *kernel.Task, c *fs.DirCtx, offset int, toDentAttr func(i name := strconv.FormatUint(uint64(fd), 10) if err := c.DirEmit(name, toDentAttr(fd)); err != nil { // Returned offset is the next fd to serialize. - return fd, err + return int64(fd), err } } // We serialized them all. Next offset should be higher than last // serialized fd. - return fd + 1, nil + return int64(fd + 1), nil } -// fd is a single file in /proc/TID/fd/. +// fd implements fs.InodeOperations for a file in /proc/TID/fd/. type fd struct { ramfs.Symlink *fs.File } +var _ fs.InodeOperations = (*fd)(nil) + // newFd returns a new fd based on an existing file. // // This inherits one reference to the file. func newFd(t *kernel.Task, f *fs.File, msrc *fs.MountSource) *fs.Inode { - fd := &fd{File: f} - // RootOwner by default, is overridden in UnstableAttr() - fd.InitSymlink(t, fs.RootOwner, "") - return newFile(fd, msrc, fs.Symlink, t) + fd := &fd{ + // RootOwner overridden by taskOwnedInodeOps.UnstableAttrs(). + Symlink: *ramfs.NewSymlink(t, fs.RootOwner, ""), + File: f, + } + return newProcInode(fd, msrc, fs.Symlink, t) } // GetFile returns the fs.File backing this fd. The dirent and flags @@ -142,7 +146,7 @@ func (f *fd) Close() error { return nil } -// fdDir implements /proc/TID/fd. +// fdDir is an InodeOperations for /proc/TID/fd. // // +stateify savable type fdDir struct { @@ -154,11 +158,15 @@ type fdDir struct { t *kernel.Task } +var _ fs.InodeOperations = (*fdDir)(nil) + // newFdDir creates a new fdDir. func newFdDir(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - f := &fdDir{t: t} - f.InitDir(t, nil, fs.RootOwner, fs.FilePermissions{User: fs.PermMask{Read: true, Execute: true}}) - return newFile(f, msrc, fs.SpecialDirectory, t) + f := &fdDir{ + Dir: *ramfs.NewDir(t, nil, fs.RootOwner, fs.FilePermissions{User: fs.PermMask{Read: true, Execute: true}}), + t: t, + } + return newProcInode(f, msrc, fs.SpecialDirectory, t) } // Check implements InodeOperations.Check. @@ -191,49 +199,55 @@ func (f *fdDir) Lookup(ctx context.Context, dir *fs.Inode, p string) (*fs.Dirent return fs.NewDirent(n, p), nil } -// DeprecatedReaddir lists fds in /proc/TID/fd. -func (f *fdDir) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { - return readDescriptors(f.t, dirCtx, offset, func(fd int) fs.DentAttr { - return fs.GenericDentAttr(fs.Symlink, device.ProcDevice) +// GetFile implements fs.FileOperations.GetFile. +func (f *fdDir) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + fops := &fdDirFile{ + isInfoFile: false, + t: f.t, + } + return fs.NewFile(ctx, dirent, flags, fops), nil +} + +// +stateify savable +type fdDirFile struct { + fsutil.DirFileOperations `state:"nosave"` + + isInfoFile bool + + t *kernel.Task +} + +var _ fs.FileOperations = (*fdDirFile)(nil) + +// Readdir implements fs.FileOperations.Readdir. +func (f *fdDirFile) Readdir(ctx context.Context, file *fs.File, ser fs.DentrySerializer) (int64, error) { + dirCtx := &fs.DirCtx{ + Serializer: ser, + } + typ := fs.RegularFile + if f.isInfoFile { + typ = fs.Symlink + } + return readDescriptors(f.t, dirCtx, file.Offset(), func(fd int) fs.DentAttr { + return fs.GenericDentAttr(typ, device.ProcDevice) }) } -// fdInfo is a single file in /proc/TID/fdinfo/. +// fdInfoInode is a single file in /proc/TID/fdinfo/. // // +stateify savable -type fdInfo struct { - ramfs.File +type fdInfoInode struct { + staticFileInodeOps file *fs.File flags fs.FileFlags fdFlags kernel.FDFlags } -// newFdInfo returns a new fdInfo based on an existing file. -func newFdInfo(t *kernel.Task, file *fs.File, fdFlags kernel.FDFlags, msrc *fs.MountSource) *fs.Inode { - fdi := &fdInfo{file: file, flags: file.Flags(), fdFlags: fdFlags} - fdi.InitFile(t, fs.RootOwner, fs.FilePermissions{User: fs.PermMask{Read: true}}) - // TODO: Get pos, locks, and other data. For now we only - // have flags. - // See https://www.kernel.org/doc/Documentation/filesystems/proc.txt - - flags := file.Flags().ToLinux() | fdFlags.ToLinuxFileFlags() - fdi.Append([]byte(fmt.Sprintf("flags:\t0%o\n", flags))) - return newFile(fdi, msrc, fs.SpecialFile, t) -} - -// DeprecatedPwritev implements fs.HandleOperations.DeprecatedPwritev. -func (*fdInfo) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { - return 0, ramfs.ErrInvalidOp -} - -// Truncate implements fs.InodeOperations.Truncate. -func (*fdInfo) Truncate(ctx context.Context, inode *fs.Inode, size int64) error { - return ramfs.ErrInvalidOp -} +var _ fs.InodeOperations = (*fdInfoInode)(nil) -func (f *fdInfo) Release(ctx context.Context) { - f.File.Release(ctx) +// Release implements fs.InodeOperations.Release. +func (f *fdInfoInode) Release(ctx context.Context) { f.file.DecRef() } @@ -249,25 +263,37 @@ type fdInfoDir struct { // newFdInfoDir creates a new fdInfoDir. func newFdInfoDir(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - fdid := &fdInfoDir{t: t} - fdid.InitDir(t, nil, fs.RootOwner, fs.FilePermsFromMode(0500)) - return newFile(fdid, msrc, fs.SpecialDirectory, t) + fdid := &fdInfoDir{ + Dir: *ramfs.NewDir(t, nil, fs.RootOwner, fs.FilePermsFromMode(0500)), + t: t, + } + return newProcInode(fdid, msrc, fs.SpecialDirectory, t) } // Lookup loads an fd in /proc/TID/fdinfo into a Dirent. func (fdid *fdInfoDir) Lookup(ctx context.Context, dir *fs.Inode, p string) (*fs.Dirent, error) { - n, err := walkDescriptors(fdid.t, p, func(file *fs.File, fdFlags kernel.FDFlags) *fs.Inode { - return newFdInfo(fdid.t, file, fdFlags, dir.MountSource) + inode, err := walkDescriptors(fdid.t, p, func(file *fs.File, fdFlags kernel.FDFlags) *fs.Inode { + // TODO: Using a static inode here means that the + // data can be out-of-date if, for instance, the flags on the + // FD change before we read this file. We should switch to + // generating the data on Read(). Also, we should include pos, + // locks, and other data. For now we only have flags. + // See https://www.kernel.org/doc/Documentation/filesystems/proc.txt + flags := file.Flags().ToLinux() | fdFlags.ToLinuxFileFlags() + contents := []byte(fmt.Sprintf("flags:\t0%o\n", flags)) + return newStaticProcInode(ctx, dir.MountSource, contents) }) if err != nil { return nil, err } - return fs.NewDirent(n, p), nil + return fs.NewDirent(inode, p), nil } -// DeprecatedReaddir lists fds in /proc/TID/fdinfo. -func (fdid *fdInfoDir) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { - return readDescriptors(fdid.t, dirCtx, offset, func(fd int) fs.DentAttr { - return fs.GenericDentAttr(fs.RegularFile, device.ProcDevice) - }) +// GetFile implements fs.FileOperations.GetFile. +func (fdid *fdInfoDir) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + fops := &fdDirFile{ + isInfoFile: true, + t: fdid.t, + } + return fs.NewFile(ctx, dirent, flags, fops), nil } diff --git a/pkg/sentry/fs/proc/file.go b/pkg/sentry/fs/proc/file.go deleted file mode 100644 index f659e590a..000000000 --- a/pkg/sentry/fs/proc/file.go +++ /dev/null @@ -1,58 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package proc - -import ( - "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/device" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" -) - -// +stateify savable -type file struct { - fs.InodeOperations - - // nodeType is the file type of this file. - nodeType fs.InodeType - - // t is the associated kernel task that owns this file. - t *kernel.Task -} - -func newFile(node fs.InodeOperations, msrc *fs.MountSource, nodeType fs.InodeType, t *kernel.Task) *fs.Inode { - iops := &file{node, nodeType, t} - sattr := fs.StableAttr{ - DeviceID: device.ProcDevice.DeviceID(), - InodeID: device.ProcDevice.NextIno(), - BlockSize: usermem.PageSize, - Type: nodeType, - } - return fs.NewInode(iops, msrc, sattr) -} - -// UnstableAttr returns all attributes of this file. -func (f *file) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - uattr, err := f.InodeOperations.UnstableAttr(ctx, inode) - if err != nil { - return fs.UnstableAttr{}, err - } - if f.t != nil { - creds := f.t.Credentials() - uattr.Owner = fs.FileOwner{creds.EffectiveKUID, creds.EffectiveKGID} - } - return uattr, nil -} diff --git a/pkg/sentry/fs/proc/inode.go b/pkg/sentry/fs/proc/inode.go new file mode 100644 index 000000000..3c36af5ea --- /dev/null +++ b/pkg/sentry/fs/proc/inode.go @@ -0,0 +1,96 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package proc + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/device" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" +) + +// taskOwnedInodeOps wraps an fs.InodeOperations and overrides the UnstableAttr +// method to return the task as the owner. +// +// +stateify savable +type taskOwnedInodeOps struct { + fs.InodeOperations + + // t is the task that owns this file. + t *kernel.Task +} + +// UnstableAttr implement fs.InodeOperations.UnstableAttr. +func (i *taskOwnedInodeOps) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { + uattr, err := i.InodeOperations.UnstableAttr(ctx, inode) + if err != nil { + return fs.UnstableAttr{}, err + } + // Set the task owner as the file owner. + creds := i.t.Credentials() + uattr.Owner = fs.FileOwner{creds.EffectiveKUID, creds.EffectiveKGID} + return uattr, nil +} + +// staticFileInodeOps is an InodeOperations implementation that can be used to +// return file contents which are constant. This file is not writable and will +// always have mode 0444. +// +// +stateify savable +type staticFileInodeOps struct { + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopTruncate `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes + fsutil.InodeStaticFileGetter +} + +var _ fs.InodeOperations = (*staticFileInodeOps)(nil) + +// newStaticFileInode returns a procfs InodeOperations with static contents. +func newStaticProcInode(ctx context.Context, msrc *fs.MountSource, contents []byte) *fs.Inode { + iops := &staticFileInodeOps{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + InodeStaticFileGetter: fsutil.InodeStaticFileGetter{ + Contents: contents, + }, + } + return newProcInode(iops, msrc, fs.SpecialFile, nil) +} + +// newProcInode creates a new inode from the given inode operations. +func newProcInode(iops fs.InodeOperations, msrc *fs.MountSource, typ fs.InodeType, t *kernel.Task) *fs.Inode { + sattr := fs.StableAttr{ + DeviceID: device.ProcDevice.DeviceID(), + InodeID: device.ProcDevice.NextIno(), + BlockSize: usermem.PageSize, + Type: typ, + } + if t != nil { + iops = &taskOwnedInodeOps{iops, t} + } + return fs.NewInode(iops, msrc, sattr) +} diff --git a/pkg/sentry/fs/proc/net.go b/pkg/sentry/fs/proc/net.go index 2806d6035..3ed85a538 100644 --- a/pkg/sentry/fs/proc/net.go +++ b/pkg/sentry/fs/proc/net.go @@ -28,35 +28,36 @@ import ( // newNet creates a new proc net entry. func (p *proc) newNetDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) + var contents map[string]*fs.Inode if s := p.k.NetworkStack(); s != nil && s.SupportsIPv6() { - d.AddChild(ctx, "dev", seqfile.NewSeqFileInode(ctx, &netDev{s: s}, msrc)) - d.AddChild(ctx, "if_inet6", seqfile.NewSeqFileInode(ctx, &ifinet6{s: s}, msrc)) - - // The following files are simple stubs until they are implemented in - // netstack, if the file contains a header the stub is just the header - // otherwise it is an empty file. - d.AddChild(ctx, "arp", p.newStubProcFSFile(ctx, msrc, []byte("IP address HW type Flags HW address Mask Device"))) - d.AddChild(ctx, "ipv6_route", p.newStubProcFSFile(ctx, msrc, []byte(""))) - d.AddChild(ctx, "netlink", p.newStubProcFSFile(ctx, msrc, []byte("sk Eth Pid Groups Rmem Wmem Dump Locks Drops Inode"))) - d.AddChild(ctx, "netstat", p.newStubProcFSFile(ctx, msrc, []byte("TcpExt: SyncookiesSent SyncookiesRecv SyncookiesFailed EmbryonicRsts PruneCalled RcvPruned OfoPruned OutOfWindowIcmps LockDroppedIcmps ArpFilter TW TWRecycled TWKilled PAWSPassive PAWSActive PAWSEstab DelayedACKs DelayedACKLocked DelayedACKLost ListenOverflows ListenDrops TCPPrequeued TCPDirectCopyFromBacklog TCPDirectCopyFromPrequeue TCPPrequeueDropped TCPHPHits TCPHPHitsToUser TCPPureAcks TCPHPAcks TCPRenoRecovery TCPSackRecovery TCPSACKReneging TCPFACKReorder TCPSACKReorder TCPRenoReorder TCPTSReorder TCPFullUndo TCPPartialUndo TCPDSACKUndo TCPLossUndo TCPLostRetransmit TCPRenoFailures TCPSackFailures TCPLossFailures TCPFastRetrans TCPForwardRetrans TCPSlowStartRetrans TCPTimeouts TCPLossProbes TCPLossProbeRecovery TCPRenoRecoveryFail TCPSackRecoveryFail TCPSchedulerFailed TCPRcvCollapsed TCPDSACKOldSent TCPDSACKOfoSent TCPDSACKRecv TCPDSACKOfoRecv TCPAbortOnData TCPAbortOnClose TCPAbortOnMemory TCPAbortOnTimeout TCPAbortOnLinger TCPAbortFailed TCPMemoryPressures TCPSACKDiscard TCPDSACKIgnoredOld TCPDSACKIgnoredNoUndo TCPSpuriousRTOs TCPMD5NotFound TCPMD5Unexpected TCPMD5Failure TCPSackShifted TCPSackMerged TCPSackShiftFallback TCPBacklogDrop TCPMinTTLDrop TCPDeferAcceptDrop IPReversePathFilter TCPTimeWaitOverflow TCPReqQFullDoCookies TCPReqQFullDrop TCPRetransFail TCPRcvCoalesce TCPOFOQueue TCPOFODrop TCPOFOMerge TCPChallengeACK TCPSYNChallenge TCPFastOpenActive TCPFastOpenActiveFail TCPFastOpenPassive TCPFastOpenPassiveFail TCPFastOpenListenOverflow TCPFastOpenCookieReqd TCPSpuriousRtxHostQueues BusyPollRxPackets TCPAutoCorking TCPFromZeroWindowAdv TCPToZeroWindowAdv TCPWantZeroWindowAdv TCPSynRetrans TCPOrigDataSent TCPHystartTrainDetect TCPHystartTrainCwnd TCPHystartDelayDetect TCPHystartDelayCwnd TCPACKSkippedSynRecv TCPACKSkippedPAWS TCPACKSkippedSeq TCPACKSkippedFinWait2 TCPACKSkippedTimeWait TCPACKSkippedChallenge TCPWinProbe TCPKeepAlive TCPMTUPFail TCPMTUPSuccess"))) - d.AddChild(ctx, "packet", p.newStubProcFSFile(ctx, msrc, []byte("sk RefCnt Type Proto Iface R Rmem User Inode"))) - d.AddChild(ctx, "protocols", p.newStubProcFSFile(ctx, msrc, []byte("protocol size sockets memory press maxhdr slab module cl co di ac io in de sh ss gs se re sp bi br ha uh gp em"))) - - // Linux sets these values to: nsec per usec, psched tick in ns, 1000000, - // high res timer ticks per sec (ClockGetres returns 1ns resolution). - psched := fmt.Sprintf("%08x %08x %08x %08x\n", uint64(time.Microsecond/time.Nanosecond), 64, 1000000, uint64(time.Second/time.Nanosecond)) - d.AddChild(ctx, "psched", p.newStubProcFSFile(ctx, msrc, []byte(psched))) - - d.AddChild(ctx, "ptype", p.newStubProcFSFile(ctx, msrc, []byte("Type Device Function"))) - d.AddChild(ctx, "route", p.newStubProcFSFile(ctx, msrc, []byte("Iface Destination Gateway Flags RefCnt Use Metric Mask MTU Window IRTT"))) - d.AddChild(ctx, "tcp", p.newStubProcFSFile(ctx, msrc, []byte(" sl local_address rem_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode"))) - d.AddChild(ctx, "tcp6", p.newStubProcFSFile(ctx, msrc, []byte(" sl local_address remote_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode"))) - d.AddChild(ctx, "udp", p.newStubProcFSFile(ctx, msrc, []byte(" sl local_address rem_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode ref pointer drops"))) - d.AddChild(ctx, "udp6", p.newStubProcFSFile(ctx, msrc, []byte(" sl local_address remote_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode"))) + contents = map[string]*fs.Inode{ + "dev": seqfile.NewSeqFileInode(ctx, &netDev{s: s}, msrc), + "if_inet6": seqfile.NewSeqFileInode(ctx, &ifinet6{s: s}, msrc), + + // The following files are simple stubs until they are + // implemented in netstack, if the file contains a + // header the stub is just the header otherwise it is + // an empty file. + "arp": newStaticProcInode(ctx, msrc, []byte("IP address HW type Flags HW address Mask Device")), + "ipv6_route": newStaticProcInode(ctx, msrc, []byte("")), + "netlink": newStaticProcInode(ctx, msrc, []byte("sk Eth Pid Groups Rmem Wmem Dump Locks Drops Inode")), + "netstat": newStaticProcInode(ctx, msrc, []byte("TcpExt: SyncookiesSent SyncookiesRecv SyncookiesFailed EmbryonicRsts PruneCalled RcvPruned OfoPruned OutOfWindowIcmps LockDroppedIcmps ArpFilter TW TWRecycled TWKilled PAWSPassive PAWSActive PAWSEstab DelayedACKs DelayedACKLocked DelayedACKLost ListenOverflows ListenDrops TCPPrequeued TCPDirectCopyFromBacklog TCPDirectCopyFromPrequeue TCPPrequeueDropped TCPHPHits TCPHPHitsToUser TCPPureAcks TCPHPAcks TCPRenoRecovery TCPSackRecovery TCPSACKReneging TCPFACKReorder TCPSACKReorder TCPRenoReorder TCPTSReorder TCPFullUndo TCPPartialUndo TCPDSACKUndo TCPLossUndo TCPLostRetransmit TCPRenoFailures TCPSackFailures TCPLossFailures TCPFastRetrans TCPForwardRetrans TCPSlowStartRetrans TCPTimeouts TCPLossProbes TCPLossProbeRecovery TCPRenoRecoveryFail TCPSackRecoveryFail TCPSchedulerFailed TCPRcvCollapsed TCPDSACKOldSent TCPDSACKOfoSent TCPDSACKRecv TCPDSACKOfoRecv TCPAbortOnData TCPAbortOnClose TCPAbortOnMemory TCPAbortOnTimeout TCPAbortOnLinger TCPAbortFailed TCPMemoryPressures TCPSACKDiscard TCPDSACKIgnoredOld TCPDSACKIgnoredNoUndo TCPSpuriousRTOs TCPMD5NotFound TCPMD5Unexpected TCPMD5Failure TCPSackShifted TCPSackMerged TCPSackShiftFallback TCPBacklogDrop TCPMinTTLDrop TCPDeferAcceptDrop IPReversePathFilter TCPTimeWaitOverflow TCPReqQFullDoCookies TCPReqQFullDrop TCPRetransFail TCPRcvCoalesce TCPOFOQueue TCPOFODrop TCPOFOMerge TCPChallengeACK TCPSYNChallenge TCPFastOpenActive TCPFastOpenActiveFail TCPFastOpenPassive TCPFastOpenPassiveFail TCPFastOpenListenOverflow TCPFastOpenCookieReqd TCPSpuriousRtxHostQueues BusyPollRxPackets TCPAutoCorking TCPFromZeroWindowAdv TCPToZeroWindowAdv TCPWantZeroWindowAdv TCPSynRetrans TCPOrigDataSent TCPHystartTrainDetect TCPHystartTrainCwnd TCPHystartDelayDetect TCPHystartDelayCwnd TCPACKSkippedSynRecv TCPACKSkippedPAWS TCPACKSkippedSeq TCPACKSkippedFinWait2 TCPACKSkippedTimeWait TCPACKSkippedChallenge TCPWinProbe TCPKeepAlive TCPMTUPFail TCPMTUPSuccess")), + "packet": newStaticProcInode(ctx, msrc, []byte("sk RefCnt Type Proto Iface R Rmem User Inode")), + "protocols": newStaticProcInode(ctx, msrc, []byte("protocol size sockets memory press maxhdr slab module cl co di ac io in de sh ss gs se re sp bi br ha uh gp em")), + // Linux sets psched values to: nsec per usec, psched + // tick in ns, 1000000, high res timer ticks per sec + // (ClockGetres returns 1ns resolution). + "psched": newStaticProcInode(ctx, msrc, []byte(fmt.Sprintf("%08x %08x %08x %08x\n", uint64(time.Microsecond/time.Nanosecond), 64, 1000000, uint64(time.Second/time.Nanosecond)))), + "ptype": newStaticProcInode(ctx, msrc, []byte("Type Device Function")), + "route": newStaticProcInode(ctx, msrc, []byte("Iface Destination Gateway Flags RefCnt Use Metric Mask MTU Window IRTT")), + "tcp": newStaticProcInode(ctx, msrc, []byte(" sl local_address rem_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode")), + "tcp6": newStaticProcInode(ctx, msrc, []byte(" sl local_address remote_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode")), + "udp": newStaticProcInode(ctx, msrc, []byte(" sl local_address rem_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode ref pointer drops")), + "udp6": newStaticProcInode(ctx, msrc, []byte(" sl local_address remote_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode")), + } } - return newFile(d, msrc, fs.SpecialDirectory, nil) + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } // ifinet6 implements seqfile.SeqSource for /proc/net/if_inet6. diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index 70e549c31..d1c699418 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -17,18 +17,17 @@ package proc import ( "fmt" - "io" "sort" "strconv" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/device" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/rpcinet" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -46,32 +45,6 @@ type proc struct { pidns *kernel.PIDNamespace } -// stubProcFSFile is a file type that can be used to return file contents -// which are constant. This file is not writable and will always have mode -// 0444. -// -// +stateify savable -type stubProcFSFile struct { - ramfs.Entry - - // contents are the immutable file contents that will always be returned. - contents []byte -} - -// DeprecatedPreadv implements fs.InodeOperations.DeprecatedPreadv. -func (s *stubProcFSFile) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - if offset < 0 { - return 0, syserror.EINVAL - } - - if offset >= int64(len(s.contents)) { - return 0, io.EOF - } - - n, err := dst.CopyOut(ctx, s.contents[offset:]) - return int64(n), err -} - // New returns the root node of a partial simple procfs. func New(ctx context.Context, msrc *fs.MountSource) (*fs.Inode, error) { k := kernel.KernelFromContext(ctx) @@ -83,29 +56,39 @@ func New(ctx context.Context, msrc *fs.MountSource) (*fs.Inode, error) { return nil, fmt.Errorf("procfs requires a PID namespace") } - p := &proc{k: k, pidns: pidns} - p.InitDir(ctx, map[string]*fs.Inode{ + // Note that these are just the static members. There are dynamic + // members populated in Readdir and Lookup below. + contents := map[string]*fs.Inode{ + "cpuinfo": newCPUInfo(ctx, msrc), "filesystems": seqfile.NewSeqFileInode(ctx, &filesystemsData{}, msrc), "loadavg": seqfile.NewSeqFileInode(ctx, &loadavgData{}, msrc), "meminfo": seqfile.NewSeqFileInode(ctx, &meminfoData{k}, msrc), - "mounts": newMountsSymlink(ctx, msrc), + "mounts": newProcInode(ramfs.NewSymlink(ctx, fs.RootOwner, "self/mounts"), msrc, fs.Symlink, nil), + "self": newSelf(ctx, pidns, msrc), "stat": seqfile.NewSeqFileInode(ctx, &statData{k}, msrc), + "thread-self": newThreadSelf(ctx, pidns, msrc), + "uptime": newUptime(ctx, msrc), "version": seqfile.NewSeqFileInode(ctx, &versionData{k}, msrc), - }, fs.RootOwner, fs.FilePermsFromMode(0555)) + } + + // Construct the proc InodeOperations. + p := &proc{ + Dir: *ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)), + k: k, + pidns: pidns, + } + + // Add more contents that need proc to be initialized. + p.AddChild(ctx, "sys", p.newSysDir(ctx, msrc)) - p.AddChild(ctx, "cpuinfo", p.newCPUInfo(ctx, msrc)) // If we're using rpcinet we will let it manage /proc/net. if _, ok := p.k.NetworkStack().(*rpcinet.Stack); ok { - p.AddChild(ctx, "net", newRPCInetProcNet(ctx, msrc)) + contents["net"] = newRPCInetProcNet(ctx, msrc) } else { - p.AddChild(ctx, "net", p.newNetDir(ctx, msrc)) + contents["net"] = p.newNetDir(ctx, msrc) } - p.AddChild(ctx, "self", p.newSelf(ctx, msrc)) - p.AddChild(ctx, "sys", p.newSysDir(ctx, msrc)) - p.AddChild(ctx, "thread-self", p.newThreadSelf(ctx, msrc)) - p.AddChild(ctx, "uptime", p.newUptime(ctx, msrc)) - return newFile(p, msrc, fs.SpecialDirectory, nil), nil + return newProcInode(p, msrc, fs.SpecialDirectory, nil), nil } // self is a magical link. @@ -118,26 +101,21 @@ type self struct { } // newSelf returns a new "self" node. -func (p *proc) newSelf(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - s := &self{pidns: p.pidns} - s.InitSymlink(ctx, fs.RootOwner, "") - return newFile(s, msrc, fs.Symlink, nil) +func newSelf(ctx context.Context, pidns *kernel.PIDNamespace, msrc *fs.MountSource) *fs.Inode { + s := &self{ + Symlink: *ramfs.NewSymlink(ctx, fs.RootOwner, ""), + pidns: pidns, + } + return newProcInode(s, msrc, fs.Symlink, nil) } // newThreadSelf returns a new "threadSelf" node. -func (p *proc) newThreadSelf(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - s := &threadSelf{pidns: p.pidns} - s.InitSymlink(ctx, fs.RootOwner, "") - return newFile(s, msrc, fs.Symlink, nil) -} - -// newStubProcFsFile returns a procfs file with constant contents. -func (p *proc) newStubProcFSFile(ctx context.Context, msrc *fs.MountSource, c []byte) *fs.Inode { - u := &stubProcFSFile{ - contents: c, +func newThreadSelf(ctx context.Context, pidns *kernel.PIDNamespace, msrc *fs.MountSource) *fs.Inode { + s := &threadSelf{ + Symlink: *ramfs.NewSymlink(ctx, fs.RootOwner, ""), + pidns: pidns, } - u.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0444)) - return newFile(u, msrc, fs.SpecialFile, nil) + return newProcInode(s, msrc, fs.Symlink, nil) } // Readlink implements fs.InodeOperations.Readlink. @@ -145,13 +123,13 @@ func (s *self) Readlink(ctx context.Context, inode *fs.Inode) (string, error) { if t := kernel.TaskFromContext(ctx); t != nil { tgid := s.pidns.IDOfThreadGroup(t.ThreadGroup()) if tgid == 0 { - return "", ramfs.ErrNotFound + return "", syserror.ENOENT } return strconv.FormatUint(uint64(tgid), 10), nil } // Who is reading this link? - return "", ramfs.ErrInvalidOp + return "", syserror.EINVAL } // threadSelf is more magical than "self" link. @@ -169,13 +147,13 @@ func (s *threadSelf) Readlink(ctx context.Context, inode *fs.Inode) (string, err tgid := s.pidns.IDOfThreadGroup(t.ThreadGroup()) tid := s.pidns.IDOfTask(t) if tid == 0 || tgid == 0 { - return "", ramfs.ErrNotFound + return "", syserror.ENOENT } return fmt.Sprintf("%d/task/%d", tgid, tid), nil } // Who is reading this link? - return "", ramfs.ErrInvalidOp + return "", syserror.EINVAL } // Lookup loads an Inode at name into a Dirent. @@ -204,25 +182,44 @@ func (p *proc) Lookup(ctx context.Context, dir *fs.Inode, name string) (*fs.Dire return fs.NewDirent(td, name), nil } -// Readdir synthesizes proc contents. -func (p *proc) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { - // Serialize normal contents. - _, err := p.Dir.DeprecatedReaddir(ctx, dirCtx, offset) - if err != nil { - return offset, err +// GetFile implements fs.InodeOperations. +func (p *proc) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &rootProcFile{iops: p}), nil +} + +// rootProcFile implements fs.FileOperations for the proc directory. +// +// +stateify savable +type rootProcFile struct { + fsutil.DirFileOperations `state:"nosave"` + + iops *proc +} + +var _ fs.FileOperations = (*rootProcFile)(nil) + +// Readdir implements fs.FileOperations.Readdir. +func (rpf *rootProcFile) Readdir(ctx context.Context, file *fs.File, ser fs.DentrySerializer) (int64, error) { + offset := file.Offset() + dirCtx := &fs.DirCtx{ + Serializer: ser, } - m := make(map[string]fs.DentAttr) - var names []string + // Get normal directory contents from ramfs dir. + names, m := rpf.iops.Dir.Children() - // Add special files. - m["sys"] = fs.GenericDentAttr(fs.SpecialFile, device.ProcDevice) - names = append(names, "sys") + // Add dot and dotdot. + root := fs.RootFromContext(ctx) + defer root.DecRef() + dot, dotdot := file.Dirent.GetDotAttrs(root) + names = append(names, ".", "..") + m["."] = dot + m[".."] = dotdot // Collect tasks. // Per linux we only include it in directory listings if it's the leader. // But for whatever crazy reason, you can still walk to the given node. - for _, tg := range p.pidns.ThreadGroups() { + for _, tg := range rpf.iops.pidns.ThreadGroups() { if leader := tg.Leader(); leader != nil { name := strconv.FormatUint(uint64(tg.ID()), 10) m[name] = fs.GenericDentAttr(fs.SpecialDirectory, device.ProcDevice) @@ -230,7 +227,7 @@ func (p *proc) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset } } - if offset >= len(m) { + if offset >= int64(len(m)) { return offset, nil } sort.Strings(names) @@ -241,12 +238,5 @@ func (p *proc) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset } offset++ } - return offset, err -} - -// newMountsSymlink returns a symlink to "self/mounts". -func newMountsSymlink(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - s := &ramfs.Symlink{} - s.InitSymlink(ctx, fs.RootOwner, "self/mounts") - return newFile(s, msrc, fs.Symlink, nil) + return offset, nil } diff --git a/pkg/sentry/fs/proc/rpcinet_proc.go b/pkg/sentry/fs/proc/rpcinet_proc.go index d025069df..65faa21f2 100644 --- a/pkg/sentry/fs/proc/rpcinet_proc.go +++ b/pkg/sentry/fs/proc/rpcinet_proc.go @@ -20,32 +20,72 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/rpcinet" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// rpcinetFile implments fs.InodeOperations as RPCs. -type rpcinetFile struct { - ramfs.Entry +// rpcInetInode implments fs.InodeOperations. +type rpcInetInode struct { + fsutil.SimpleFileInode - // filepath is the full path of this rpcinetFile. + // filepath is the full path of this rpcInetInode. filepath string k *kernel.Kernel } -// DeprecatedPreadv implements fs.InodeOperations.DeprecatedPreadv. -// This method can panic if an rpcinetFile was created without an rpcinet +func newRPCInetInode(ctx context.Context, msrc *fs.MountSource, filepath string, mode linux.FileMode) *fs.Inode { + f := &rpcInetInode{ + SimpleFileInode: *fsutil.NewSimpleFileInode(ctx, fs.RootOwner, fs.FilePermsFromMode(mode), linux.PROC_SUPER_MAGIC), + filepath: filepath, + k: kernel.KernelFromContext(ctx), + } + return newProcInode(f, msrc, fs.SpecialFile, nil) +} + +// GetFile implements fs.InodeOperations.GetFile. +func (i *rpcInetInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + flags.Pread = true + flags.Pwrite = true + fops := &rpcInetFile{ + inode: i, + } + return fs.NewFile(ctx, dirent, flags, fops), nil +} + +// rpcInetFile implements fs.FileOperations as RPCs. +type rpcInetFile struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + + inode *rpcInetInode +} + +// Read implements fs.FileOperations.Read. +// +// This method can panic if an rpcInetInode was created without an rpcinet // stack. -func (r rpcinetFile) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - s, ok := r.k.NetworkStack().(*rpcinet.Stack) +func (f *rpcInetFile) Read(ctx context.Context, file *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { + if offset < 0 { + return 0, syserror.EINVAL + } + s, ok := f.inode.k.NetworkStack().(*rpcinet.Stack) if !ok { panic("Network stack is not a rpcinet.") } - contents, se := s.RPCReadFile(r.filepath) + contents, se := s.RPCReadFile(f.inode.filepath) if se != nil || offset >= int64(len(contents)) { return 0, io.EOF } @@ -54,16 +94,12 @@ func (r rpcinetFile) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequenc return int64(n), err } -// Truncate implements fs.InodeOperations.Truncate. -func (r rpcinetFile) Truncate(context.Context, *fs.Inode, int64) error { - return nil -} - -// DeprecatedPwritev implements fs.InodeOperations.DeprecatedPwritev. -// This method can panic if an rpcinetFile was created without an rpcinet +// Write implements fs.FileOperations.Write. +// +// This method can panic if an rpcInetInode was created without an rpcInet // stack. -func (r rpcinetFile) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { - s, ok := r.k.NetworkStack().(*rpcinet.Stack) +func (f *rpcInetFile) Write(ctx context.Context, file *fs.File, src usermem.IOSequence, offset int64) (int64, error) { + s, ok := f.inode.k.NetworkStack().(*rpcinet.Stack) if !ok { panic("Network stack is not a rpcinet.") } @@ -78,116 +114,102 @@ func (r rpcinetFile) DeprecatedPwritev(ctx context.Context, src usermem.IOSequen return int64(n), err } - written, se := s.RPCWriteFile(r.filepath, b) + written, se := s.RPCWriteFile(f.inode.filepath, b) return int64(written), se.ToError() } -func newRPCProcFSFile(ctx context.Context, msrc *fs.MountSource, filepath string, mode linux.FileMode) *fs.Inode { - f := &rpcinetFile{ - filepath: filepath, - k: kernel.KernelFromContext(ctx), - } - f.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(mode)) - - fi := newFile(f, msrc, fs.SpecialFile, nil) - return fi -} - // newRPCInetProcNet will build an inode for /proc/net. func newRPCInetProcNet(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - - // Add all the files we want to forward for /proc/net. - d.AddChild(ctx, "arp", newRPCProcFSFile(ctx, msrc, "/proc/net/arp", 0444)) - d.AddChild(ctx, "dev", newRPCProcFSFile(ctx, msrc, "/proc/net/dev", 0444)) - d.AddChild(ctx, "if_inet6", newRPCProcFSFile(ctx, msrc, "/proc/net/if_inet6", 0444)) - d.AddChild(ctx, "ipv6_route", newRPCProcFSFile(ctx, msrc, "/proc/net/ipv6_route", 0444)) - d.AddChild(ctx, "netlink", newRPCProcFSFile(ctx, msrc, "/proc/net/netlink", 0444)) - d.AddChild(ctx, "netstat", newRPCProcFSFile(ctx, msrc, "/proc/net/netstat", 0444)) - d.AddChild(ctx, "packet", newRPCProcFSFile(ctx, msrc, "/proc/net/packet", 0444)) - d.AddChild(ctx, "protocols", newRPCProcFSFile(ctx, msrc, "/proc/net/protocols", 0444)) - d.AddChild(ctx, "psched", newRPCProcFSFile(ctx, msrc, "/proc/net/psched", 0444)) - d.AddChild(ctx, "ptype", newRPCProcFSFile(ctx, msrc, "/proc/net/ptype", 0444)) - d.AddChild(ctx, "route", newRPCProcFSFile(ctx, msrc, "/proc/net/route", 0444)) - d.AddChild(ctx, "tcp", newRPCProcFSFile(ctx, msrc, "/proc/net/tcp", 0444)) - d.AddChild(ctx, "tcp6", newRPCProcFSFile(ctx, msrc, "/proc/net/tcp6", 0444)) - d.AddChild(ctx, "udp", newRPCProcFSFile(ctx, msrc, "/proc/net/udp", 0444)) - d.AddChild(ctx, "udp6", newRPCProcFSFile(ctx, msrc, "/proc/net/udp6", 0444)) - - return newFile(d, msrc, fs.SpecialDirectory, nil) + contents := map[string]*fs.Inode{ + "arp": newRPCInetInode(ctx, msrc, "/proc/net/arp", 0444), + "dev": newRPCInetInode(ctx, msrc, "/proc/net/dev", 0444), + "if_inet6": newRPCInetInode(ctx, msrc, "/proc/net/if_inet6", 0444), + "ipv6_route": newRPCInetInode(ctx, msrc, "/proc/net/ipv6_route", 0444), + "netlink": newRPCInetInode(ctx, msrc, "/proc/net/netlink", 0444), + "netstat": newRPCInetInode(ctx, msrc, "/proc/net/netstat", 0444), + "packet": newRPCInetInode(ctx, msrc, "/proc/net/packet", 0444), + "protocols": newRPCInetInode(ctx, msrc, "/proc/net/protocols", 0444), + "psched": newRPCInetInode(ctx, msrc, "/proc/net/psched", 0444), + "ptype": newRPCInetInode(ctx, msrc, "/proc/net/ptype", 0444), + "route": newRPCInetInode(ctx, msrc, "/proc/net/route", 0444), + "tcp": newRPCInetInode(ctx, msrc, "/proc/net/tcp", 0444), + "tcp6": newRPCInetInode(ctx, msrc, "/proc/net/tcp6", 0444), + "udp": newRPCInetInode(ctx, msrc, "/proc/net/udp", 0444), + "udp6": newRPCInetInode(ctx, msrc, "/proc/net/udp6", 0444), + } + + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } // newRPCInetProcSysNet will build an inode for /proc/sys/net. func newRPCInetProcSysNet(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - d.AddChild(ctx, "ipv4", newRPCInetSysNetIPv4Dir(ctx, msrc)) - d.AddChild(ctx, "core", newRPCInetSysNetCore(ctx, msrc)) + contents := map[string]*fs.Inode{ + "ipv4": newRPCInetSysNetIPv4Dir(ctx, msrc), + "core": newRPCInetSysNetCore(ctx, msrc), + } - return newFile(d, msrc, fs.SpecialDirectory, nil) + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } // newRPCInetSysNetCore builds the /proc/sys/net/core directory. func newRPCInetSysNetCore(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - - // Add all the files we want to forward over RPC for /proc/sys/net/core - d.AddChild(ctx, "default_qdisc", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/default_qdisc", 0444)) - d.AddChild(ctx, "message_burst", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/message_burst", 0444)) - d.AddChild(ctx, "message_cost", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/message_cost", 0444)) - d.AddChild(ctx, "optmem_max", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/optmem_max", 0444)) - d.AddChild(ctx, "rmem_default", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/rmem_default", 0444)) - d.AddChild(ctx, "rmem_max", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/rmem_max", 0444)) - d.AddChild(ctx, "somaxconn", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/somaxconn", 0444)) - d.AddChild(ctx, "wmem_default", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/wmem_default", 0444)) - d.AddChild(ctx, "wmem_max", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/core/wmem_max", 0444)) - - return newFile(d, msrc, fs.SpecialDirectory, nil) + contents := map[string]*fs.Inode{ + "default_qdisc": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/default_qdisc", 0444), + "message_burst": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/message_burst", 0444), + "message_cost": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/message_cost", 0444), + "optmem_max": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/optmem_max", 0444), + "rmem_default": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/rmem_default", 0444), + "rmem_max": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/rmem_max", 0444), + "somaxconn": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/somaxconn", 0444), + "wmem_default": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/wmem_default", 0444), + "wmem_max": newRPCInetInode(ctx, msrc, "/proc/sys/net/core/wmem_max", 0444), + } + + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } // newRPCInetSysNetIPv4Dir builds the /proc/sys/net/ipv4 directory. func newRPCInetSysNetIPv4Dir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - - // Add all the files we want to forward over RPC for /proc/sys/net/ipv4. - d.AddChild(ctx, "ip_local_port_range", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/ip_local_port_range", 0444)) - d.AddChild(ctx, "ip_local_reserved_ports", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/ip_local_reserved_ports", 0444)) - d.AddChild(ctx, "ipfrag_time", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/ipfrag_time", 0444)) - d.AddChild(ctx, "ip_nonlocal_bind", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/ip_nonlocal_bind", 0444)) - d.AddChild(ctx, "ip_no_pmtu_disc", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/ip_no_pmtu_disc", 0444)) - - d.AddChild(ctx, "tcp_allowed_congestion_control", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_allowed_congestion_control", 0444)) - d.AddChild(ctx, "tcp_available_congestion_control", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_available_congestion_control", 0444)) - d.AddChild(ctx, "tcp_base_mss", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_base_mss", 0444)) - d.AddChild(ctx, "tcp_congestion_control", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_congestion_control", 0644)) - d.AddChild(ctx, "tcp_dsack", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_dsack", 0644)) - d.AddChild(ctx, "tcp_early_retrans", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_early_retrans", 0644)) - d.AddChild(ctx, "tcp_fack", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_fack", 0644)) - d.AddChild(ctx, "tcp_fastopen", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_fastopen", 0644)) - d.AddChild(ctx, "tcp_fastopen_key", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_fastopen_key", 0444)) - d.AddChild(ctx, "tcp_fin_timeout", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_fin_timeout", 0644)) - d.AddChild(ctx, "tcp_invalid_ratelimit", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_invalid_ratelimit", 0444)) - d.AddChild(ctx, "tcp_keepalive_intvl", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_keepalive_intvl", 0644)) - d.AddChild(ctx, "tcp_keepalive_probes", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_keepalive_probes", 0644)) - d.AddChild(ctx, "tcp_keepalive_time", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_keepalive_time", 0644)) - d.AddChild(ctx, "tcp_mem", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_mem", 0444)) - d.AddChild(ctx, "tcp_mtu_probing", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_mtu_probing", 0644)) - d.AddChild(ctx, "tcp_no_metrics_save", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_no_metrics_save", 0444)) - d.AddChild(ctx, "tcp_probe_interval", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_probe_interval", 0444)) - d.AddChild(ctx, "tcp_probe_threshold", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_probe_threshold", 0444)) - d.AddChild(ctx, "tcp_retries1", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_retries1", 0644)) - d.AddChild(ctx, "tcp_retries2", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_retries2", 0644)) - d.AddChild(ctx, "tcp_rfc1337", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_rfc1337", 0444)) - d.AddChild(ctx, "tcp_rmem", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_rmem", 0444)) - d.AddChild(ctx, "tcp_sack", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_sack", 0644)) - d.AddChild(ctx, "tcp_slow_start_after_idle", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_slow_start_after_idle", 0644)) - d.AddChild(ctx, "tcp_synack_retries", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_synack_retries", 0644)) - d.AddChild(ctx, "tcp_syn_retries", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_syn_retries", 0644)) - d.AddChild(ctx, "tcp_timestamps", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_timestamps", 0644)) - d.AddChild(ctx, "tcp_wmem", newRPCProcFSFile(ctx, msrc, "/proc/sys/net/ipv4/tcp_wmem", 0444)) - - return newFile(d, msrc, fs.SpecialDirectory, nil) + contents := map[string]*fs.Inode{ + "ip_local_port_range": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/ip_local_port_range", 0444), + "ip_local_reserved_ports": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/ip_local_reserved_ports", 0444), + "ipfrag_time": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/ipfrag_time", 0444), + "ip_nonlocal_bind": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/ip_nonlocal_bind", 0444), + "ip_no_pmtu_disc": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/ip_no_pmtu_disc", 0444), + "tcp_allowed_congestion_control": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_allowed_congestion_control", 0444), + "tcp_available_congestion_control": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_available_congestion_control", 0444), + "tcp_base_mss": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_base_mss", 0444), + "tcp_congestion_control": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_congestion_control", 0644), + "tcp_dsack": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_dsack", 0644), + "tcp_early_retrans": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_early_retrans", 0644), + "tcp_fack": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_fack", 0644), + "tcp_fastopen": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_fastopen", 0644), + "tcp_fastopen_key": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_fastopen_key", 0444), + "tcp_fin_timeout": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_fin_timeout", 0644), + "tcp_invalid_ratelimit": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_invalid_ratelimit", 0444), + "tcp_keepalive_intvl": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_keepalive_intvl", 0644), + "tcp_keepalive_probes": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_keepalive_probes", 0644), + "tcp_keepalive_time": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_keepalive_time", 0644), + "tcp_mem": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_mem", 0444), + "tcp_mtu_probing": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_mtu_probing", 0644), + "tcp_no_metrics_save": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_no_metrics_save", 0444), + "tcp_probe_interval": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_probe_interval", 0444), + "tcp_probe_threshold": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_probe_threshold", 0444), + "tcp_retries1": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_retries1", 0644), + "tcp_retries2": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_retries2", 0644), + "tcp_rfc1337": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_rfc1337", 0444), + "tcp_rmem": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_rmem", 0444), + "tcp_sack": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_sack", 0644), + "tcp_slow_start_after_idle": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_slow_start_after_idle", 0644), + "tcp_synack_retries": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_synack_retries", 0644), + "tcp_syn_retries": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_syn_retries", 0644), + "tcp_timestamps": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_timestamps", 0644), + "tcp_wmem": newRPCInetInode(ctx, msrc, "/proc/sys/net/ipv4/tcp_wmem", 0444), + } + + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } diff --git a/pkg/sentry/fs/proc/seqfile/BUILD b/pkg/sentry/fs/proc/seqfile/BUILD index 53c475652..b4ba64e10 100644 --- a/pkg/sentry/fs/proc/seqfile/BUILD +++ b/pkg/sentry/fs/proc/seqfile/BUILD @@ -8,12 +8,15 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile", visibility = ["//pkg/sentry:internal"], deps = [ + "//pkg/abi/linux", "//pkg/sentry/context", "//pkg/sentry/fs", + "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/proc/device", - "//pkg/sentry/fs/ramfs", "//pkg/sentry/kernel/time", "//pkg/sentry/usermem", + "//pkg/syserror", + "//pkg/waiter", ], ) @@ -26,7 +29,7 @@ go_test( "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/fs", - "//pkg/sentry/fs/ramfs/test", + "//pkg/sentry/fs/ramfs", "//pkg/sentry/usermem", ], ) diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index 0499ba65b..16fc6789e 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -18,12 +18,15 @@ import ( "io" "sync" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/device" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // SeqHandle is a helper handle to seek in the file. @@ -87,7 +90,18 @@ func (s *SeqGenerationCounter) IsCurrent(generation int64) bool { // // +stateify savable type SeqFile struct { - ramfs.Entry + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeNotTruncatable `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + fsutil.InodeSimpleExtendedAttributes + fsutil.InodeSimpleAttributes // mu protects the fields below. mu sync.Mutex `state:"nosave"` @@ -99,11 +113,14 @@ type SeqFile struct { lastRead int64 } +var _ fs.InodeOperations = (*SeqFile)(nil) + // NewSeqFile returns a seqfile suitable for use by external consumers. func NewSeqFile(ctx context.Context, source SeqSource) *SeqFile { - s := &SeqFile{SeqSource: source} - s.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0444)) - return s + return &SeqFile{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + SeqSource: source, + } } // NewSeqFileInode returns an Inode with SeqFile InodeOperations. @@ -120,11 +137,19 @@ func NewSeqFileInode(ctx context.Context, source SeqSource, msrc *fs.MountSource // UnstableAttr returns unstable attributes of the SeqFile. func (s *SeqFile) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - uattr, _ := s.Entry.UnstableAttr(ctx, inode) + uattr, err := s.InodeSimpleAttributes.UnstableAttr(ctx, inode) + if err != nil { + return fs.UnstableAttr{}, err + } uattr.ModificationTime = ktime.NowFromContext(ctx) return uattr, nil } +// GetFile implements fs.InodeOperations.GetFile. +func (s *SeqFile) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &seqFileOperations{seqFile: s}), nil +} + // findIndexAndOffset finds the unit that corresponds to a certain offset. // Returns the unit and the offset within the unit. If there are not enough // units len(data) and leftover offset is returned. @@ -139,36 +164,74 @@ func findIndexAndOffset(data []SeqData, offset int64) (int, int64) { return len(data), offset } -// DeprecatedPreadv reads from the file at the given offset. -func (s *SeqFile) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - s.mu.Lock() - defer s.mu.Unlock() +// updateSourceLocked requires that s.mu is held. +func (s *SeqFile) updateSourceLocked(ctx context.Context, record int) { + var h SeqHandle + if record == 0 { + h = nil + } else { + h = s.source[record-1].Handle + } + // Save what we have previously read. + s.source = s.source[:record] + var newSource []SeqData + newSource, s.generation = s.SeqSource.ReadSeqFileData(ctx, h) + s.source = append(s.source, newSource...) +} + +// seqFileOperations implements fs.FileOperations. +// +// +stateify savable +type seqFileOperations struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + + seqFile *SeqFile +} + +var _ fs.FileOperations = (*seqFileOperations)(nil) + +// Write implements fs.FileOperations.Write. +func (*seqFileOperations) Write(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { + return 0, syserror.EACCES +} + +// Read implements fs.FileOperations.Read. +func (sfo *seqFileOperations) Read(ctx context.Context, file *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { + sfo.seqFile.mu.Lock() + defer sfo.seqFile.mu.Unlock() - s.Entry.NotifyAccess(ctx) - defer func() { s.lastRead = offset }() + sfo.seqFile.NotifyAccess(ctx) + defer func() { sfo.seqFile.lastRead = offset }() updated := false // Try to find where we should start reading this file. - i, recordOffset := findIndexAndOffset(s.source, offset) - if i == len(s.source) { + i, recordOffset := findIndexAndOffset(sfo.seqFile.source, offset) + if i == len(sfo.seqFile.source) { // Ok, we're at EOF. Let's first check to see if there might be // more data available to us. If there is more data, add it to // the end and try reading again. - if !s.SeqSource.NeedsUpdate(s.generation) { + if !sfo.seqFile.SeqSource.NeedsUpdate(sfo.seqFile.generation) { return 0, io.EOF } - oldLen := len(s.source) - s.updateSourceLocked(ctx, len(s.source)) + oldLen := len(sfo.seqFile.source) + sfo.seqFile.updateSourceLocked(ctx, len(sfo.seqFile.source)) updated = true // We know that we had consumed everything up until this point // so we search in the new slice instead of starting over. - i, recordOffset = findIndexAndOffset(s.source[oldLen:], recordOffset) + i, recordOffset = findIndexAndOffset(sfo.seqFile.source[oldLen:], recordOffset) i += oldLen // i is at most the length of the slice which is - // len(s.source) - oldLen. So at most i will be equal to - // len(s.source). - if i == len(s.source) { + // len(sfo.seqFile.source) - oldLen. So at most i will be equal to + // len(sfo.seqFile.source). + if i == len(sfo.seqFile.source) { return 0, io.EOF } } @@ -178,7 +241,7 @@ func (s *SeqFile) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, // before continuing on to the next. We don't refresh our data source // before this record is completed. if recordOffset != 0 { - n, err := dst.CopyOut(ctx, s.source[i].Buf[recordOffset:]) + n, err := dst.CopyOut(ctx, sfo.seqFile.source[i].Buf[recordOffset:]) done += int64(n) dst = dst.DropFirst(n) if dst.NumBytes() == 0 || err != nil { @@ -190,15 +253,15 @@ func (s *SeqFile) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, // Next/New unit, update the source file if necessary. Make an extra // check to see if we've seeked backwards and if so always update our // data source. - if !updated && (s.SeqSource.NeedsUpdate(s.generation) || s.lastRead > offset) { - s.updateSourceLocked(ctx, i) + if !updated && (sfo.seqFile.SeqSource.NeedsUpdate(sfo.seqFile.generation) || sfo.seqFile.lastRead > offset) { + sfo.seqFile.updateSourceLocked(ctx, i) // recordOffset is 0 here and we won't update records behind the // current one so recordOffset is still 0 even though source // just got updated. Just read the next record. } // Finish by reading all the available data. - for _, buf := range s.source[i:] { + for _, buf := range sfo.seqFile.source[i:] { n, err := dst.CopyOut(ctx, buf.Buf) done += int64(n) dst = dst.DropFirst(n) @@ -214,23 +277,3 @@ func (s *SeqFile) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, } return done, nil } - -// updateSourceLocked requires that s.mu is held. -func (s *SeqFile) updateSourceLocked(ctx context.Context, record int) { - var h SeqHandle - if record == 0 { - h = nil - } else { - h = s.source[record-1].Handle - } - // Save what we have previously read. - s.source = s.source[:record] - var newSource []SeqData - newSource, s.generation = s.SeqSource.ReadSeqFileData(ctx, h) - s.source = append(s.source, newSource...) -} - -// DeprecatedPwritev is always denied. -func (*SeqFile) DeprecatedPwritev(context.Context, usermem.IOSequence, int64) (int64, error) { - return 0, ramfs.ErrDenied -} diff --git a/pkg/sentry/fs/proc/seqfile/seqfile_test.go b/pkg/sentry/fs/proc/seqfile/seqfile_test.go index f9a2ca38e..35403ab7f 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile_test.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile_test.go @@ -23,7 +23,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - ramfstest "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) @@ -91,10 +91,15 @@ type testTable struct { expectedError error } -func runTableTests(ctx context.Context, table []testTable, n fs.InodeOperations) error { +func runTableTests(ctx context.Context, table []testTable, dirent *fs.Dirent) error { for _, tt := range table { + file, err := dirent.Inode.InodeOperations.GetFile(ctx, dirent, fs.FileFlags{Read: true}) + if err != nil { + return fmt.Errorf("GetFile returned error: %v", err) + } + data := make([]byte, tt.readBufferSize) - resultLen, err := n.DeprecatedPreadv(ctx, usermem.BytesIOSequence(data), tt.offset) + resultLen, err := file.Preadv(ctx, usermem.BytesIOSequence(data), tt.offset) if err != tt.expectedError { return fmt.Errorf("t.Preadv(len: %v, offset: %v) (error) => %v expected %v", tt.readBufferSize, tt.offset, err, tt.expectedError) } @@ -115,12 +120,12 @@ func TestSeqFile(t *testing.T) { testSource.Init() // Create a file that can be R/W. - m := fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}) + m := fs.NewPseudoMountSource() ctx := contexttest.Context(t) contents := map[string]*fs.Inode{ "foo": NewSeqFileInode(ctx, testSource, m), } - root := ramfstest.NewDir(ctx, contents, fs.FilePermsFromMode(0777)) + root := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0777)) // How about opening it? inode := fs.NewInode(root, m, fs.StableAttr{Type: fs.Directory}) @@ -129,9 +134,13 @@ func TestSeqFile(t *testing.T) { t.Fatalf("failed to walk to foo for n2: %v", err) } n2 := dirent2.Inode.InodeOperations + file2, err := n2.GetFile(ctx, dirent2, fs.FileFlags{Read: true, Write: true}) + if err != nil { + t.Fatalf("GetFile returned error: %v", err) + } // Writing? - if _, err := n2.DeprecatedPwritev(nil, usermem.BytesIOSequence([]byte("test")), 0); err == nil { + if _, err := file2.Writev(ctx, usermem.BytesIOSequence([]byte("test"))); err == nil { t.Fatalf("managed to write to n2: %v", err) } @@ -141,7 +150,6 @@ func TestSeqFile(t *testing.T) { t.Fatalf("failed to walk to foo: %v", err) } n3 := dirent3.Inode.InodeOperations - if n2 != n3 { t.Error("got n2 != n3, want same") } @@ -170,13 +178,13 @@ func TestSeqFile(t *testing.T) { // Read the last 3 bytes. {97, 10, testSource.actual[9].Buf[7:], nil}, } - if err := runTableTests(ctx, table, n2); err != nil { + if err := runTableTests(ctx, table, dirent2); err != nil { t.Errorf("runTableTest failed with testSource.update = %v : %v", testSource.update, err) } // Disable updates and do it again. testSource.update = false - if err := runTableTests(ctx, table, n2); err != nil { + if err := runTableTests(ctx, table, dirent2); err != nil { t.Errorf("runTableTest failed with testSource.update = %v: %v", testSource.update, err) } } @@ -188,25 +196,24 @@ func TestSeqFileFileUpdated(t *testing.T) { testSource.update = true // Create a file that can be R/W. - m := fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}) + m := fs.NewPseudoMountSource() ctx := contexttest.Context(t) contents := map[string]*fs.Inode{ "foo": NewSeqFileInode(ctx, testSource, m), } - root := ramfstest.NewDir(ctx, contents, fs.FilePermsFromMode(0777)) + root := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0777)) // How about opening it? inode := fs.NewInode(root, m, fs.StableAttr{Type: fs.Directory}) dirent2, err := root.Lookup(ctx, inode, "foo") if err != nil { - t.Fatalf("failed to walk to foo for n2: %v", err) + t.Fatalf("failed to walk to foo for dirent2: %v", err) } - n2 := dirent2.Inode.InodeOperations table := []testTable{ {0, 16, flatten(testSource.actual[0].Buf, testSource.actual[1].Buf[:6]), nil}, } - if err := runTableTests(ctx, table, n2); err != nil { + if err := runTableTests(ctx, table, dirent2); err != nil { t.Errorf("runTableTest failed: %v", err) } // Delete the first entry. @@ -224,7 +231,7 @@ func TestSeqFileFileUpdated(t *testing.T) { // Read the following two lines. {30, 20, flatten(testSource.actual[3].Buf, testSource.actual[4].Buf), nil}, } - if err := runTableTests(ctx, table, n2); err != nil { + if err := runTableTests(ctx, table, dirent2); err != nil { t.Errorf("runTableTest failed after removing first entry: %v", err) } @@ -238,7 +245,7 @@ func TestSeqFileFileUpdated(t *testing.T) { table = []testTable{ {50, 20, flatten(testSource.actual[4].Buf, testSource.actual[5].Buf), nil}, } - if err := runTableTests(ctx, table, n2); err != nil { + if err := runTableTests(ctx, table, dirent2); err != nil { t.Errorf("runTableTest failed after adding middle entry: %v", err) } // This will be used in a later test. @@ -249,7 +256,7 @@ func TestSeqFileFileUpdated(t *testing.T) { table = []testTable{ {20, 20, []byte{}, io.EOF}, } - if err := runTableTests(ctx, table, n2); err != nil { + if err := runTableTests(ctx, table, dirent2); err != nil { t.Errorf("runTableTest failed after removing all entries: %v", err) } // Restore some of the data. @@ -257,7 +264,7 @@ func TestSeqFileFileUpdated(t *testing.T) { table = []testTable{ {6, 20, testSource.actual[0].Buf[6:], nil}, } - if err := runTableTests(ctx, table, n2); err != nil { + if err := runTableTests(ctx, table, dirent2); err != nil { t.Errorf("runTableTest failed after adding first entry back: %v", err) } @@ -266,7 +273,7 @@ func TestSeqFileFileUpdated(t *testing.T) { table = []testTable{ {30, 20, flatten(testSource.actual[3].Buf, testSource.actual[4].Buf), nil}, } - if err := runTableTests(ctx, table, n2); err != nil { + if err := runTableTests(ctx, table, dirent2); err != nil { t.Errorf("runTableTest failed after extending testSource: %v", err) } } diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index 54562508d..ee6b9f262 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -22,39 +22,15 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/rpcinet" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// hostname is a file containing the system hostname. -// -// +stateify savable -type hostname struct { - ramfs.Entry -} - -// DeprecatedPreadv implements fs.InodeOperations.DeprecatedPreadv. -func (hostname) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - utsns := kernel.UTSNamespaceFromContext(ctx) - contents := []byte(utsns.HostName() + "\n") - - if offset >= int64(len(contents)) { - return 0, io.EOF - } - - n, err := dst.CopyOut(ctx, contents[offset:]) - return int64(n), err -} - -func (p *proc) newHostname(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - h := &hostname{} - h.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0444)) - return newFile(h, msrc, fs.SpecialFile, nil) -} - // mmapMinAddrData backs /proc/sys/vm/mmap_min_addr. // // +stateify savable @@ -101,36 +77,84 @@ func (*overcommitMemory) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandl } func (p *proc) newKernelDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - d.AddChild(ctx, "hostname", p.newHostname(ctx, msrc)) - - d.AddChild(ctx, "shmmax", p.newStubProcFSFile(ctx, msrc, []byte(strconv.FormatUint(linux.SHMMAX, 10)))) - d.AddChild(ctx, "shmall", p.newStubProcFSFile(ctx, msrc, []byte(strconv.FormatUint(linux.SHMALL, 10)))) - d.AddChild(ctx, "shmmni", p.newStubProcFSFile(ctx, msrc, []byte(strconv.FormatUint(linux.SHMMNI, 10)))) - return newFile(d, msrc, fs.SpecialDirectory, nil) + h := hostname{ + SimpleFileInode: *fsutil.NewSimpleFileInode(ctx, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + } + + children := map[string]*fs.Inode{ + "hostname": newProcInode(&h, msrc, fs.SpecialFile, nil), + "shmall": newStaticProcInode(ctx, msrc, []byte(strconv.FormatUint(linux.SHMALL, 10))), + "shmmax": newStaticProcInode(ctx, msrc, []byte(strconv.FormatUint(linux.SHMMAX, 10))), + "shmmni": newStaticProcInode(ctx, msrc, []byte(strconv.FormatUint(linux.SHMMNI, 10))), + } + + d := ramfs.NewDir(ctx, children, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } func (p *proc) newVMDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - d.AddChild(ctx, "mmap_min_addr", seqfile.NewSeqFileInode(ctx, &mmapMinAddrData{p.k}, msrc)) - d.AddChild(ctx, "overcommit_memory", seqfile.NewSeqFileInode(ctx, &overcommitMemory{}, msrc)) - return newFile(d, msrc, fs.SpecialDirectory, nil) + children := map[string]*fs.Inode{ + "mmap_min_addr": seqfile.NewSeqFileInode(ctx, &mmapMinAddrData{p.k}, msrc), + "overcommit_memory": seqfile.NewSeqFileInode(ctx, &overcommitMemory{}, msrc), + } + d := ramfs.NewDir(ctx, children, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } func (p *proc) newSysDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - d.AddChild(ctx, "kernel", p.newKernelDir(ctx, msrc)) - d.AddChild(ctx, "vm", p.newVMDir(ctx, msrc)) + children := map[string]*fs.Inode{ + "kernel": p.newKernelDir(ctx, msrc), + "vm": p.newVMDir(ctx, msrc), + } // If we're using rpcinet we will let it manage /proc/sys/net. if _, ok := p.k.NetworkStack().(*rpcinet.Stack); ok { - d.AddChild(ctx, "net", newRPCInetProcSysNet(ctx, msrc)) + children["net"] = newRPCInetProcSysNet(ctx, msrc) } else { - d.AddChild(ctx, "net", p.newSysNetDir(ctx, msrc)) + children["net"] = p.newSysNetDir(ctx, msrc) } - return newFile(d, msrc, fs.SpecialDirectory, nil) + d := ramfs.NewDir(ctx, children, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) +} + +// hostname is the inode for a file containing the system hostname. +// +// +stateify savable +type hostname struct { + fsutil.SimpleFileInode +} + +// GetFile implements fs.InodeOperations.GetFile. +func (h *hostname) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, d, flags, &hostnameFile{}), nil +} + +var _ fs.InodeOperations = (*hostname)(nil) + +// +stateify savable +type hostnameFile struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSeek `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` } + +// Read implements fs.FileOperations.Read. +func (hf *hostnameFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { + utsns := kernel.UTSNamespaceFromContext(ctx) + contents := []byte(utsns.HostName() + "\n") + if offset >= int64(len(contents)) { + return 0, io.EOF + } + n, err := dst.CopyOut(ctx, contents[offset:]) + return int64(n), err + +} + +var _ fs.FileOperations = (*hostnameFile)(nil) diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index b50d43d70..42e9bc47f 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -17,13 +17,17 @@ package proc import ( "fmt" "io" + "sync" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/device" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/inet" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) type tcpMemDir int @@ -33,21 +37,37 @@ const ( tcpWMem ) +// tcpMemInode is used to read/write the size of netstack tcp buffers. +// +// TODO: If we have multiple proc mounts, concurrent writes can +// leave netstack and the proc files in an inconsistent state. Since we set the +// buffer size from these proc files on restore, we may also race and end up in +// an inconsistent state on restore. +// // +stateify savable -type tcpMem struct { - ramfs.Entry - s inet.Stack `state:"wait"` +type tcpMemInode struct { + fsutil.SimpleFileInode + dir tcpMemDir + s inet.Stack `state:"wait"` + + // size stores the tcp buffer size during save, and sets the buffer + // size in netstack in restore. We must save/restore this here, since + // netstack itself is stateless. size inet.TCPBufferSize - dir tcpMemDir -} -func newTCPMem(s inet.Stack, size inet.TCPBufferSize, dir tcpMemDir) *tcpMem { - return &tcpMem{s: s, size: size, dir: dir} + // mu protects against concurrent reads/writes to files based on this + // inode. + mu sync.Mutex `state:"nosave"` } -func newTCPMemInode(ctx context.Context, msrc *fs.MountSource, s inet.Stack, size inet.TCPBufferSize, dir tcpMemDir) *fs.Inode { - tm := newTCPMem(s, size, dir) - tm.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0644)) +var _ fs.InodeOperations = (*tcpMemInode)(nil) + +func newTCPMemInode(ctx context.Context, msrc *fs.MountSource, s inet.Stack, dir tcpMemDir) *fs.Inode { + tm := &tcpMemInode{ + SimpleFileInode: *fsutil.NewSimpleFileInode(ctx, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + s: s, + dir: dir, + } sattr := fs.StableAttr{ DeviceID: device.ProcDevice.DeviceID(), InodeID: device.ProcDevice.NextIno(), @@ -57,62 +77,105 @@ func newTCPMemInode(ctx context.Context, msrc *fs.MountSource, s inet.Stack, siz return fs.NewInode(tm, msrc, sattr) } -// DeprecatedPreadv implements fs.InodeOperations.DeprecatedPreadv. -func (m *tcpMem) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { +// GetFile implements fs.InodeOperations.GetFile. +func (m *tcpMemInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + flags.Pread = true + return fs.NewFile(ctx, dirent, flags, &tcpMemFile{tcpMemInode: m}), nil +} + +// +stateify savable +type tcpMemFile struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + + tcpMemInode *tcpMemInode +} + +var _ fs.FileOperations = (*tcpMemFile)(nil) + +// Read implements fs.FileOperations.Read. +func (f *tcpMemFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { if offset != 0 { return 0, io.EOF } - s := fmt.Sprintf("%d\t%d\t%d\n", m.size.Min, m.size.Default, m.size.Max) + f.tcpMemInode.mu.Lock() + defer f.tcpMemInode.mu.Unlock() + + size, err := readSize(f.tcpMemInode.dir, f.tcpMemInode.s) + if err != nil { + return 0, err + } + s := fmt.Sprintf("%d\t%d\t%d\n", size.Min, size.Default, size.Max) n, err := dst.CopyOut(ctx, []byte(s)) return int64(n), err } -// Truncate implements fs.InodeOperations.Truncate. -func (*tcpMem) Truncate(context.Context, *fs.Inode, int64) error { - return nil -} - -// DeprecatedPwritev implements fs.InodeOperations.DeprecatedPwritev. -func (m *tcpMem) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { +// Write implements fs.FileOperations.Write. +func (f *tcpMemFile) Write(ctx context.Context, _ *fs.File, src usermem.IOSequence, offset int64) (int64, error) { if src.NumBytes() == 0 { return 0, nil } - src = src.TakeFirst(usermem.PageSize - 1) + f.tcpMemInode.mu.Lock() + defer f.tcpMemInode.mu.Unlock() - buf := []int32{int32(m.size.Min), int32(m.size.Default), int32(m.size.Max)} + src = src.TakeFirst(usermem.PageSize - 1) + size, err := readSize(f.tcpMemInode.dir, f.tcpMemInode.s) + if err != nil { + return 0, err + } + buf := []int32{int32(size.Min), int32(size.Default), int32(size.Max)} n, cperr := usermem.CopyInt32StringsInVec(ctx, src.IO, src.Addrs, buf, src.Opts) - m.size = inet.TCPBufferSize{ + newSize := inet.TCPBufferSize{ Min: int(buf[0]), Default: int(buf[1]), Max: int(buf[2]), } - if err := m.writeSize(); err != nil { + if err := writeSize(f.tcpMemInode.dir, f.tcpMemInode.s, newSize); err != nil { return n, err } return n, cperr } -func (m *tcpMem) writeSize() error { - switch m.dir { +func readSize(dirType tcpMemDir, s inet.Stack) (inet.TCPBufferSize, error) { + switch dirType { + case tcpRMem: + return s.TCPReceiveBufferSize() + case tcpWMem: + return s.TCPSendBufferSize() + default: + panic(fmt.Sprintf("unknown tcpMemFile type: %v", dirType)) + } +} + +func writeSize(dirType tcpMemDir, s inet.Stack, size inet.TCPBufferSize) error { + switch dirType { case tcpRMem: - return m.s.SetTCPReceiveBufferSize(m.size) + return s.SetTCPReceiveBufferSize(size) case tcpWMem: - return m.s.SetTCPSendBufferSize(m.size) + return s.SetTCPSendBufferSize(size) default: - panic(fmt.Sprintf("unknown tcpMem.dir: %v", m.dir)) + panic(fmt.Sprintf("unknown tcpMemFile type: %v", dirType)) } } // +stateify savable type tcpSack struct { - ramfs.Entry - s inet.Stack `state:"wait"` + stack inet.Stack `state:"wait"` enabled *bool + fsutil.SimpleFileInode } func newTCPSackInode(ctx context.Context, msrc *fs.MountSource, s inet.Stack) *fs.Inode { - ts := &tcpSack{s: s} - ts.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0644)) + ts := &tcpSack{ + SimpleFileInode: *fsutil.NewSimpleFileInode(ctx, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + stack: s, + } sattr := fs.StableAttr{ DeviceID: device.ProcDevice.DeviceID(), InodeID: device.ProcDevice.NextIno(), @@ -122,21 +185,48 @@ func newTCPSackInode(ctx context.Context, msrc *fs.MountSource, s inet.Stack) *f return fs.NewInode(ts, msrc, sattr) } -func (s *tcpSack) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { +// GetFile implements fs.InodeOperations.GetFile. +func (s *tcpSack) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + flags.Pread = true + flags.Pwrite = true + return fs.NewFile(ctx, dirent, flags, &tcpSackFile{ + tcpSack: s, + stack: s.stack, + }), nil +} + +// +stateify savable +type tcpSackFile struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + + tcpSack *tcpSack + + stack inet.Stack `state:"wait"` +} + +// Read implements fs.FileOperations.Read. +func (f *tcpSackFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { if offset != 0 { return 0, io.EOF } - if s.enabled == nil { - sack, err := s.s.TCPSACKEnabled() + if f.tcpSack.enabled == nil { + sack, err := f.stack.TCPSACKEnabled() if err != nil { return 0, err } - s.enabled = &sack + f.tcpSack.enabled = &sack } val := "0\n" - if *s.enabled { + if *f.tcpSack.enabled { // Technically, this is not quite compatible with Linux. Linux // stores these as an integer, so if you write "2" into // tcp_sack, you should get 2 back. Tough luck. @@ -146,13 +236,8 @@ func (s *tcpSack) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, return int64(n), err } -// Truncate implements fs.InodeOperations.Truncate. -func (*tcpSack) Truncate(context.Context, *fs.Inode, int64) error { - return nil -} - -// DeprecatedPwritev implements fs.InodeOperations.DeprecatedPwritev. -func (s *tcpSack) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { +// Write implements fs.FileOperations.Write. +func (f *tcpSackFile) Write(ctx context.Context, _ *fs.File, src usermem.IOSequence, offset int64) (int64, error) { if src.NumBytes() == 0 { return 0, nil } @@ -163,100 +248,104 @@ func (s *tcpSack) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, if err != nil { return n, err } - if s.enabled == nil { - s.enabled = new(bool) + if f.tcpSack.enabled == nil { + f.tcpSack.enabled = new(bool) } - *s.enabled = v != 0 - return n, s.s.SetTCPSACKEnabled(*s.enabled) + *f.tcpSack.enabled = v != 0 + return n, f.tcpSack.stack.SetTCPSACKEnabled(*f.tcpSack.enabled) } func (p *proc) newSysNetCore(ctx context.Context, msrc *fs.MountSource, s inet.Stack) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - // The following files are simple stubs until they are implemented in // netstack, most of these files are configuration related. We use the // value closest to the actual netstack behavior or any empty file, // all of these files will have mode 0444 (read-only for all users). - d.AddChild(ctx, "default_qdisc", p.newStubProcFSFile(ctx, msrc, []byte("pfifo_fast"))) - d.AddChild(ctx, "message_burst", p.newStubProcFSFile(ctx, msrc, []byte("10"))) - d.AddChild(ctx, "message_cost", p.newStubProcFSFile(ctx, msrc, []byte("5"))) - d.AddChild(ctx, "optmem_max", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "rmem_default", p.newStubProcFSFile(ctx, msrc, []byte("212992"))) - d.AddChild(ctx, "rmem_max", p.newStubProcFSFile(ctx, msrc, []byte("212992"))) - d.AddChild(ctx, "somaxconn", p.newStubProcFSFile(ctx, msrc, []byte("128"))) - d.AddChild(ctx, "wmem_default", p.newStubProcFSFile(ctx, msrc, []byte("212992"))) - d.AddChild(ctx, "wmem_max", p.newStubProcFSFile(ctx, msrc, []byte("212992"))) - - return newFile(d, msrc, fs.SpecialDirectory, nil) + contents := map[string]*fs.Inode{ + "default_qdisc": newStaticProcInode(ctx, msrc, []byte("pfifo_fast")), + "message_burst": newStaticProcInode(ctx, msrc, []byte("10")), + "message_cost": newStaticProcInode(ctx, msrc, []byte("5")), + "optmem_max": newStaticProcInode(ctx, msrc, []byte("0")), + "rmem_default": newStaticProcInode(ctx, msrc, []byte("212992")), + "rmem_max": newStaticProcInode(ctx, msrc, []byte("212992")), + "somaxconn": newStaticProcInode(ctx, msrc, []byte("128")), + "wmem_default": newStaticProcInode(ctx, msrc, []byte("212992")), + "wmem_max": newStaticProcInode(ctx, msrc, []byte("212992")), + } + + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } func (p *proc) newSysNetIPv4Dir(ctx context.Context, msrc *fs.MountSource, s inet.Stack) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) + contents := map[string]*fs.Inode{ + // Add tcp_sack. + "tcp_sack": newTCPSackInode(ctx, msrc, s), + + // The following files are simple stubs until they are + // implemented in netstack, most of these files are + // configuration related. We use the value closest to the + // actual netstack behavior or any empty file, all of these + // files will have mode 0444 (read-only for all users). + "ip_local_port_range": newStaticProcInode(ctx, msrc, []byte("16000 65535")), + "ip_local_reserved_ports": newStaticProcInode(ctx, msrc, []byte("")), + "ipfrag_time": newStaticProcInode(ctx, msrc, []byte("30")), + "ip_nonlocal_bind": newStaticProcInode(ctx, msrc, []byte("0")), + "ip_no_pmtu_disc": newStaticProcInode(ctx, msrc, []byte("1")), + + // tcp_allowed_congestion_control tell the user what they are + // able to do as an unprivledged process so we leave it empty. + "tcp_allowed_congestion_control": newStaticProcInode(ctx, msrc, []byte("")), + "tcp_available_congestion_control": newStaticProcInode(ctx, msrc, []byte("reno")), + "tcp_congestion_control": newStaticProcInode(ctx, msrc, []byte("reno")), + + // Many of the following stub files are features netstack + // doesn't support. The unsupported features return "0" to + // indicate they are disabled. + "tcp_base_mss": newStaticProcInode(ctx, msrc, []byte("1280")), + "tcp_dsack": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_early_retrans": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_fack": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_fastopen": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_fastopen_key": newStaticProcInode(ctx, msrc, []byte("")), + "tcp_invalid_ratelimit": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_keepalive_intvl": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_keepalive_probes": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_keepalive_time": newStaticProcInode(ctx, msrc, []byte("7200")), + "tcp_mtu_probing": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_no_metrics_save": newStaticProcInode(ctx, msrc, []byte("1")), + "tcp_probe_interval": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_probe_threshold": newStaticProcInode(ctx, msrc, []byte("0")), + "tcp_retries1": newStaticProcInode(ctx, msrc, []byte("3")), + "tcp_retries2": newStaticProcInode(ctx, msrc, []byte("15")), + "tcp_rfc1337": newStaticProcInode(ctx, msrc, []byte("1")), + "tcp_slow_start_after_idle": newStaticProcInode(ctx, msrc, []byte("1")), + "tcp_synack_retries": newStaticProcInode(ctx, msrc, []byte("5")), + "tcp_syn_retries": newStaticProcInode(ctx, msrc, []byte("3")), + "tcp_timestamps": newStaticProcInode(ctx, msrc, []byte("1")), + } // Add tcp_rmem. - if rs, err := s.TCPReceiveBufferSize(); err == nil { - d.AddChild(ctx, "tcp_rmem", newTCPMemInode(ctx, msrc, s, rs, tcpRMem)) + if _, err := s.TCPReceiveBufferSize(); err == nil { + contents["tcp_rmem"] = newTCPMemInode(ctx, msrc, s, tcpRMem) } // Add tcp_wmem. - if ss, err := s.TCPSendBufferSize(); err == nil { - d.AddChild(ctx, "tcp_wmem", newTCPMemInode(ctx, msrc, s, ss, tcpWMem)) + if _, err := s.TCPSendBufferSize(); err == nil { + contents["tcp_wmem"] = newTCPMemInode(ctx, msrc, s, tcpWMem) } - // Add tcp_sack. - d.AddChild(ctx, "tcp_sack", newTCPSackInode(ctx, msrc, s)) - - // The following files are simple stubs until they are implemented in - // netstack, most of these files are configuration related. We use the - // value closest to the actual netstack behavior or any empty file, - // all of these files will have mode 0444 (read-only for all users). - d.AddChild(ctx, "ip_local_port_range", p.newStubProcFSFile(ctx, msrc, []byte("16000 65535"))) - d.AddChild(ctx, "ip_local_reserved_ports", p.newStubProcFSFile(ctx, msrc, []byte(""))) - d.AddChild(ctx, "ipfrag_time", p.newStubProcFSFile(ctx, msrc, []byte("30"))) - d.AddChild(ctx, "ip_nonlocal_bind", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "ip_no_pmtu_disc", p.newStubProcFSFile(ctx, msrc, []byte("1"))) - - // tcp_allowed_congestion_control tell the user what they are able to do as an - // unprivledged process so we leave it empty. - d.AddChild(ctx, "tcp_allowed_congestion_control", p.newStubProcFSFile(ctx, msrc, []byte(""))) - d.AddChild(ctx, "tcp_available_congestion_control", p.newStubProcFSFile(ctx, msrc, []byte("reno"))) - d.AddChild(ctx, "tcp_congestion_control", p.newStubProcFSFile(ctx, msrc, []byte("reno"))) - - // Many of the following stub files are features netstack doesn't support - // and are therefore "0" for disabled. - d.AddChild(ctx, "tcp_base_mss", p.newStubProcFSFile(ctx, msrc, []byte("1280"))) - d.AddChild(ctx, "tcp_dsack", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_early_retrans", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_fack", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_fastopen", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_fastopen_key", p.newStubProcFSFile(ctx, msrc, []byte(""))) - d.AddChild(ctx, "tcp_invalid_ratelimit", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_keepalive_intvl", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_keepalive_probes", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_keepalive_time", p.newStubProcFSFile(ctx, msrc, []byte("7200"))) - d.AddChild(ctx, "tcp_mtu_probing", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_no_metrics_save", p.newStubProcFSFile(ctx, msrc, []byte("1"))) - d.AddChild(ctx, "tcp_probe_interval", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_probe_threshold", p.newStubProcFSFile(ctx, msrc, []byte("0"))) - d.AddChild(ctx, "tcp_retries1", p.newStubProcFSFile(ctx, msrc, []byte("3"))) - d.AddChild(ctx, "tcp_retries2", p.newStubProcFSFile(ctx, msrc, []byte("15"))) - d.AddChild(ctx, "tcp_rfc1337", p.newStubProcFSFile(ctx, msrc, []byte("1"))) - d.AddChild(ctx, "tcp_slow_start_after_idle", p.newStubProcFSFile(ctx, msrc, []byte("1"))) - d.AddChild(ctx, "tcp_synack_retries", p.newStubProcFSFile(ctx, msrc, []byte("5"))) - d.AddChild(ctx, "tcp_syn_retries", p.newStubProcFSFile(ctx, msrc, []byte("3"))) - d.AddChild(ctx, "tcp_timestamps", p.newStubProcFSFile(ctx, msrc, []byte("1"))) - - return newFile(d, msrc, fs.SpecialDirectory, nil) + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } func (p *proc) newSysNetDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(ctx, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) + var contents map[string]*fs.Inode if s := p.k.NetworkStack(); s != nil { - d.AddChild(ctx, "ipv4", p.newSysNetIPv4Dir(ctx, msrc, s)) - d.AddChild(ctx, "core", p.newSysNetCore(ctx, msrc, s)) + contents = map[string]*fs.Inode{ + "ipv4": p.newSysNetIPv4Dir(ctx, msrc, s), + "core": p.newSysNetCore(ctx, msrc, s), + } } - return newFile(d, msrc, fs.SpecialDirectory, nil) + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + return newProcInode(d, msrc, fs.SpecialDirectory, nil) } diff --git a/pkg/sentry/fs/proc/sys_net_state.go b/pkg/sentry/fs/proc/sys_net_state.go index 7f46776c0..5f481a1cf 100644 --- a/pkg/sentry/fs/proc/sys_net_state.go +++ b/pkg/sentry/fs/proc/sys_net_state.go @@ -16,17 +16,26 @@ package proc import "fmt" +// beforeSave is invoked by stateify. +func (t *tcpMemInode) beforeSave() { + size, err := readSize(t.dir, t.s) + if err != nil { + panic(fmt.Sprintf("failed to read TCP send / receive buffer sizes: %v", err)) + } + t.size = size +} + // afterLoad is invoked by stateify. -func (m *tcpMem) afterLoad() { - if err := m.writeSize(); err != nil { - panic(fmt.Sprintf("failed to write previous TCP send / receive buffer sizes [%v]: %v", m.size, err)) +func (t *tcpMemInode) afterLoad() { + if err := writeSize(t.dir, t.s, t.size); err != nil { + panic(fmt.Sprintf("failed to write previous TCP send / receive buffer sizes [%v]: %v", t.size, err)) } } // afterLoad is invoked by stateify. func (s *tcpSack) afterLoad() { if s.enabled != nil { - if err := s.s.SetTCPSACKEnabled(*s.enabled); err != nil { + if err := s.stack.SetTCPSACKEnabled(*s.enabled); err != nil { panic(fmt.Sprintf("failed to set previous TCP sack configuration [%v]: %v", *s.enabled, err)) } } diff --git a/pkg/sentry/fs/proc/sys_net_test.go b/pkg/sentry/fs/proc/sys_net_test.go index 0ce9d30f1..ea0d94fce 100644 --- a/pkg/sentry/fs/proc/sys_net_test.go +++ b/pkg/sentry/fs/proc/sys_net_test.go @@ -26,13 +26,14 @@ func TestQuerySendBufferSize(t *testing.T) { ctx := context.Background() s := inet.NewTestStack() s.TCPSendBufSize = inet.TCPBufferSize{100, 200, 300} - tm := newTCPMem(s, s.TCPSendBufSize, tcpWMem) + tmi := &tcpMemInode{s: s, dir: tcpWMem} + tmf := &tcpMemFile{tcpMemInode: tmi} buf := make([]byte, 100) dst := usermem.BytesIOSequence(buf) - n, err := tm.DeprecatedPreadv(ctx, dst, 0) + n, err := tmf.Read(ctx, nil, dst, 0) if err != nil { - t.Fatalf("DeprecatedPreadv failed: %v", err) + t.Fatalf("Read failed: %v", err) } if got, want := string(buf[:n]), "100\t200\t300\n"; got != want { @@ -44,13 +45,14 @@ func TestQueryRecvBufferSize(t *testing.T) { ctx := context.Background() s := inet.NewTestStack() s.TCPRecvBufSize = inet.TCPBufferSize{100, 200, 300} - tm := newTCPMem(s, s.TCPRecvBufSize, tcpRMem) + tmi := &tcpMemInode{s: s, dir: tcpRMem} + tmf := &tcpMemFile{tcpMemInode: tmi} buf := make([]byte, 100) dst := usermem.BytesIOSequence(buf) - n, err := tm.DeprecatedPreadv(ctx, dst, 0) + n, err := tmf.Read(ctx, nil, dst, 0) if err != nil { - t.Fatalf("DeprecatedPreadv failed: %v", err) + t.Fatalf("Read failed: %v", err) } if got, want := string(buf[:n]), "100\t200\t300\n"; got != want { @@ -85,12 +87,13 @@ func TestConfigureSendBufferSize(t *testing.T) { s := inet.NewTestStack() for _, c := range cases { s.TCPSendBufSize = c.initial - tm := newTCPMem(s, c.initial, tcpWMem) + tmi := &tcpMemInode{s: s, dir: tcpWMem} + tmf := &tcpMemFile{tcpMemInode: tmi} // Write the values. src := usermem.BytesIOSequence([]byte(c.str)) - if n, err := tm.DeprecatedPwritev(ctx, src, 0); n != int64(len(c.str)) || err != nil { - t.Errorf("DeprecatedPwritev, case = %q: got (%d, %v), wanted (%d, nil)", c.str, n, err, len(c.str)) + if n, err := tmf.Write(ctx, nil, src, 0); n != int64(len(c.str)) || err != nil { + t.Errorf("Write, case = %q: got (%d, %v), wanted (%d, nil)", c.str, n, err, len(c.str)) } // Read the values from the stack and check them. @@ -105,12 +108,13 @@ func TestConfigureRecvBufferSize(t *testing.T) { s := inet.NewTestStack() for _, c := range cases { s.TCPRecvBufSize = c.initial - tm := newTCPMem(s, c.initial, tcpRMem) + tmi := &tcpMemInode{s: s, dir: tcpRMem} + tmf := &tcpMemFile{tcpMemInode: tmi} // Write the values. src := usermem.BytesIOSequence([]byte(c.str)) - if n, err := tm.DeprecatedPwritev(ctx, src, 0); n != int64(len(c.str)) || err != nil { - t.Errorf("DeprecatedPwritev, case = %q: got (%d, %v), wanted (%d, nil)", c.str, n, err, len(c.str)) + if n, err := tmf.Write(ctx, nil, src, 0); n != int64(len(c.str)) || err != nil { + t.Errorf("Write, case = %q: got (%d, %v), wanted (%d, nil)", c.str, n, err, len(c.str)) } // Read the values from the stack and check them. diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 91bda8a95..41981a973 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -24,6 +24,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/device" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" @@ -32,6 +33,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // getTaskMM returns t's MemoryManager. If getTaskMM succeeds, the MemoryManager's @@ -57,19 +59,19 @@ func getTaskMM(t *kernel.Task) (*mm.MemoryManager, error) { type taskDir struct { ramfs.Dir - // t is the associated kernel task that owns this file. - t *kernel.Task + t *kernel.Task + pidns *kernel.PIDNamespace } +var _ fs.InodeOperations = (*taskDir)(nil) + // newTaskDir creates a new proc task entry. func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace, showSubtasks bool) *fs.Inode { - d := &taskDir{t: t} - // TODO: Set EUID/EGID based on dumpability. - d.InitDir(t, map[string]*fs.Inode{ + contents := map[string]*fs.Inode{ "auxv": newAuxvec(t, msrc), - "cmdline": newExecArgFile(t, msrc, cmdlineExecArg), + "cmdline": newExecArgInode(t, msrc, cmdlineExecArg), "comm": newComm(t, msrc), - "environ": newExecArgFile(t, msrc, environExecArg), + "environ": newExecArgInode(t, msrc, environExecArg), "exe": newExe(t, msrc), "fd": newFdDir(t, msrc), "fdinfo": newFdInfoDir(t, msrc), @@ -87,11 +89,18 @@ func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace "statm": newStatm(t, msrc), "status": newStatus(t, msrc, pidns), "uid_map": newUIDMap(t, msrc), - }, fs.RootOwner, fs.FilePermsFromMode(0555)) + } if showSubtasks { - d.AddChild(t, "task", newSubtasks(t, msrc, pidns)) + contents["task"] = newSubtasks(t, msrc, pidns) } - return newFile(d, msrc, fs.SpecialDirectory, t) + + // TODO: Set EUID/EGID based on dumpability. + d := &taskDir{ + Dir: *ramfs.NewDir(t, contents, fs.RootOwner, fs.FilePermsFromMode(0555)), + t: t, + pidns: pidns, + } + return newProcInode(d, msrc, fs.SpecialDirectory, t) } // subtasks represents a /proc/TID/task directory. @@ -100,15 +109,19 @@ func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace type subtasks struct { ramfs.Dir - t *kernel.Task - + t *kernel.Task pidns *kernel.PIDNamespace } +var _ fs.InodeOperations = (*subtasks)(nil) + func newSubtasks(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace) *fs.Inode { - s := &subtasks{t: t, pidns: pidns} - s.InitDir(t, nil, fs.RootOwner, fs.FilePermsFromMode(0555)) - return newFile(s, msrc, fs.SpecialDirectory, t) + s := &subtasks{ + Dir: *ramfs.NewDir(t, nil, fs.RootOwner, fs.FilePermsFromMode(0555)), + t: t, + pidns: pidns, + } + return newProcInode(s, msrc, fs.SpecialDirectory, t) } // UnstableAttr returns unstable attributes of the subtasks. @@ -123,35 +136,52 @@ func (s *subtasks) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.Unstab return uattr, nil } -// Lookup loads an Inode in a task's subtask directory into a Dirent. -func (s *subtasks) Lookup(ctx context.Context, dir *fs.Inode, p string) (*fs.Dirent, error) { - tid, err := strconv.ParseUint(p, 10, 32) - if err != nil { - return nil, syserror.ENOENT - } +// GetFile implements fs.InodeOperations.GetFile. +func (s *subtasks) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &subtasksFile{t: s.t, pidns: s.pidns}), nil +} - task := s.pidns.TaskWithID(kernel.ThreadID(tid)) - if task == nil { - return nil, syserror.ENOENT - } - if task.ThreadGroup() != s.t.ThreadGroup() { - return nil, syserror.ENOENT - } +// +stateify savable +type subtasksFile struct { + fsutil.DirFileOperations `state:"nosave"` - td := newTaskDir(task, dir.MountSource, s.pidns, false) - return fs.NewDirent(td, p), nil + t *kernel.Task + pidns *kernel.PIDNamespace } -// DeprecatedReaddir lists a task's subtask directory. -func (s *subtasks) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { - tasks := s.t.ThreadGroup().MemberIDs(s.pidns) +// Readdir implements fs.FileOperations.Readdir. +func (f *subtasksFile) Readdir(ctx context.Context, file *fs.File, ser fs.DentrySerializer) (int64, error) { + dirCtx := fs.DirCtx{ + Serializer: ser, + } + + // Note that unlike most Readdir implementations, the offset here is + // not an index into the subtasks, but rather the TID of the next + // subtask to emit. + offset := file.Offset() + + if offset == 0 { + // Serialize "." and "..". + root := fs.RootFromContext(ctx) + defer root.DecRef() + dot, dotdot := file.Dirent.GetDotAttrs(root) + if err := dirCtx.DirEmit(".", dot); err != nil { + return offset, err + } + if err := dirCtx.DirEmit("..", dotdot); err != nil { + return offset, err + } + } + + // Serialize tasks. + tasks := f.t.ThreadGroup().MemberIDs(f.pidns) taskInts := make([]int, 0, len(tasks)) for _, tid := range tasks { taskInts = append(taskInts, int(tid)) } // Find the task to start at. - idx := sort.SearchInts(taskInts, offset) + idx := sort.SearchInts(taskInts, int(offset)) if idx == len(taskInts) { return offset, nil } @@ -163,12 +193,33 @@ func (s *subtasks) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, off attr := fs.GenericDentAttr(fs.SpecialDirectory, device.ProcDevice) if err := dirCtx.DirEmit(name, attr); err != nil { // Returned offset is next tid to serialize. - return tid, err + return int64(tid), err } } // We serialized them all. Next offset should be higher than last // serialized tid. - return tid + 1, nil + return int64(tid) + 1, nil +} + +var _ fs.FileOperations = (*subtasksFile)(nil) + +// Lookup loads an Inode in a task's subtask directory into a Dirent. +func (s *subtasks) Lookup(ctx context.Context, dir *fs.Inode, p string) (*fs.Dirent, error) { + tid, err := strconv.ParseUint(p, 10, 32) + if err != nil { + return nil, syserror.ENOENT + } + + task := s.pidns.TaskWithID(kernel.ThreadID(tid)) + if task == nil { + return nil, syserror.ENOENT + } + if task.ThreadGroup() != s.t.ThreadGroup() { + return nil, syserror.ENOENT + } + + td := newTaskDir(task, dir.MountSource, s.pidns, false) + return fs.NewDirent(td, p), nil } // exe is an fs.InodeOperations symlink for the /proc/PID/exe file. @@ -181,9 +232,11 @@ type exe struct { } func newExe(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - exeSymlink := &exe{t: t} - exeSymlink.InitSymlink(t, fs.RootOwner, "") - return newFile(exeSymlink, msrc, fs.Symlink, t) + exeSymlink := &exe{ + Symlink: *ramfs.NewSymlink(t, fs.RootOwner, ""), + t: t, + } + return newProcInode(exeSymlink, msrc, fs.Symlink, t) } func (e *exe) executable() (d *fs.Dirent, err error) { @@ -231,55 +284,48 @@ func (e *exe) Readlink(ctx context.Context, inode *fs.Inode) (string, error) { return n, nil } -// namespaceFile represents a file in the namespacefs, such as the files in -// /proc//ns. +// namespaceSymlink represents a symlink in the namespacefs, such as the files +// in /proc//ns. // // +stateify savable -type namespaceFile struct { +type namespaceSymlink struct { ramfs.Symlink t *kernel.Task } -func newNamespaceFile(t *kernel.Task, msrc *fs.MountSource, name string) *fs.Inode { - n := &namespaceFile{t: t} - n.InitSymlink(t, fs.RootOwner, "") - +func newNamespaceSymlink(t *kernel.Task, msrc *fs.MountSource, name string) *fs.Inode { // TODO: Namespace symlinks should contain the namespace name and the // inode number for the namespace instance, so for example user:[123456]. We // currently fake the inode number by sticking the symlink inode in its // place. - n.Target = fmt.Sprintf("%s:[%d]", name, device.ProcDevice.NextIno()) - - return newFile(n, msrc, fs.Symlink, t) + target := fmt.Sprintf("%s:[%d]", name, device.ProcDevice.NextIno()) + n := &namespaceSymlink{ + Symlink: *ramfs.NewSymlink(t, fs.RootOwner, target), + t: t, + } + return newProcInode(n, msrc, fs.Symlink, t) } // Getlink implements fs.InodeOperations.Getlink. -func (n *namespaceFile) Getlink(ctx context.Context, inode *fs.Inode) (*fs.Dirent, error) { +func (n *namespaceSymlink) Getlink(ctx context.Context, inode *fs.Inode) (*fs.Dirent, error) { if !kernel.ContextCanTrace(ctx, n.t, false) { return nil, syserror.EACCES } // Create a new regular file to fake the namespace file. - node := &ramfs.Entry{} - node.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0777)) - sattr := fs.StableAttr{ - DeviceID: device.ProcDevice.DeviceID(), - InodeID: device.ProcDevice.NextIno(), - BlockSize: usermem.PageSize, - Type: fs.RegularFile, - } - return fs.NewDirent(fs.NewInode(node, inode.MountSource, sattr), n.Symlink.Target), nil + iops := fsutil.NewNoReadWriteFileInode(ctx, fs.RootOwner, fs.FilePermsFromMode(0777), linux.PROC_SUPER_MAGIC) + return fs.NewDirent(newProcInode(iops, inode.MountSource, fs.RegularFile, nil), n.Symlink.Target), nil } func newNamespaceDir(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - d := &ramfs.Dir{} - d.InitDir(t, map[string]*fs.Inode{ - "net": newNamespaceFile(t, msrc, "net"), - "pid": newNamespaceFile(t, msrc, "pid"), - "user": newNamespaceFile(t, msrc, "user"), - }, fs.RootOwner, fs.FilePermsFromMode(0511)) - return newFile(d, msrc, fs.SpecialDirectory, t) + contents := map[string]*fs.Inode{ + "net": newNamespaceSymlink(t, msrc, "net"), + "pid": newNamespaceSymlink(t, msrc, "pid"), + "user": newNamespaceSymlink(t, msrc, "user"), + } + d := ramfs.NewDir(t, contents, fs.RootOwner, fs.FilePermsFromMode(0511)) + return newProcInode(d, msrc, fs.SpecialDirectory, t) } // mapsData implements seqfile.SeqSource for /proc/[pid]/maps. @@ -290,7 +336,7 @@ type mapsData struct { } func newMaps(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - return newFile(seqfile.NewSeqFile(t, &mapsData{t}), msrc, fs.SpecialFile, t) + return newProcInode(seqfile.NewSeqFile(t, &mapsData{t}), msrc, fs.SpecialFile, t) } func (md *mapsData) mm() *mm.MemoryManager { @@ -330,7 +376,7 @@ type smapsData struct { } func newSmaps(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - return newFile(seqfile.NewSeqFile(t, &smapsData{t}), msrc, fs.SpecialFile, t) + return newProcInode(seqfile.NewSeqFile(t, &smapsData{t}), msrc, fs.SpecialFile, t) } func (sd *smapsData) mm() *mm.MemoryManager { @@ -376,7 +422,7 @@ type taskStatData struct { } func newTaskStat(t *kernel.Task, msrc *fs.MountSource, showSubtasks bool, pidns *kernel.PIDNamespace) *fs.Inode { - return newFile(seqfile.NewSeqFile(t, &taskStatData{t, showSubtasks /* tgstats */, pidns}), msrc, fs.SpecialFile, t) + return newProcInode(seqfile.NewSeqFile(t, &taskStatData{t, showSubtasks /* tgstats */, pidns}), msrc, fs.SpecialFile, t) } // NeedsUpdate returns whether the generation is old or not. @@ -450,7 +496,7 @@ type statmData struct { } func newStatm(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - return newFile(seqfile.NewSeqFile(t, &statmData{t}), msrc, fs.SpecialFile, t) + return newProcInode(seqfile.NewSeqFile(t, &statmData{t}), msrc, fs.SpecialFile, t) } // NeedsUpdate implements seqfile.SeqSource.NeedsUpdate. @@ -487,7 +533,7 @@ type statusData struct { } func newStatus(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace) *fs.Inode { - return newFile(seqfile.NewSeqFile(t, &statusData{t, pidns}), msrc, fs.SpecialFile, t) + return newProcInode(seqfile.NewSeqFile(t, &statusData{t, pidns}), msrc, fs.SpecialFile, t) } // NeedsUpdate implements seqfile.SeqSource.NeedsUpdate. @@ -552,7 +598,7 @@ type ioData struct { } func newIO(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - return newFile(seqfile.NewSeqFile(t, &ioData{t.ThreadGroup()}), msrc, fs.SpecialFile, t) + return newProcInode(seqfile.NewSeqFile(t, &ioData{t.ThreadGroup()}), msrc, fs.SpecialFile, t) } // NeedsUpdate returns whether the generation is old or not. @@ -590,25 +636,49 @@ func (i *ioData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([]se // // +stateify savable type comm struct { - ramfs.Entry + fsutil.SimpleFileInode t *kernel.Task } // newComm returns a new comm file. func newComm(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - c := &comm{t: t} - c.InitEntry(t, fs.RootOwner, fs.FilePermsFromMode(0444)) - return newFile(c, msrc, fs.SpecialFile, t) + c := &comm{ + SimpleFileInode: *fsutil.NewSimpleFileInode(t, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + t: t, + } + return newProcInode(c, msrc, fs.SpecialFile, t) +} + +// GetFile implements fs.InodeOperations.GetFile. +func (c *comm) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &commFile{t: c.t}), nil +} + +// +stateify savable +type commFile struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + + t *kernel.Task } -// DeprecatedPreadv reads the current command name. -func (c *comm) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { +var _ fs.FileOperations = (*commFile)(nil) + +// Read implements fs.FileOperations.Read. +func (f *commFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { if offset < 0 { return 0, syserror.EINVAL } - buf := []byte(c.t.Name() + "\n") + buf := []byte(f.t.Name() + "\n") if offset >= int64(len(buf)) { return 0, io.EOF } @@ -621,25 +691,47 @@ func (c *comm) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, off // // +stateify savable type auxvec struct { - ramfs.Entry + fsutil.SimpleFileInode t *kernel.Task } // newAuxvec returns a new auxvec file. func newAuxvec(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { - a := &auxvec{t: t} - a.InitEntry(t, fs.RootOwner, fs.FilePermsFromMode(0400)) - return newFile(a, msrc, fs.SpecialFile, t) + a := &auxvec{ + SimpleFileInode: *fsutil.NewSimpleFileInode(t, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + t: t, + } + return newProcInode(a, msrc, fs.SpecialFile, t) +} + +// GetFile implements fs.InodeOperations.GetFile. +func (a *auxvec) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &auxvecFile{t: a.t}), nil +} + +// +stateify savable +type auxvecFile struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + + t *kernel.Task } -// DeprecatedPreadv reads the current auxiliary vector. -func (a *auxvec) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { +// Read implements fs.FileOperations.Read. +func (f *auxvecFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { if offset < 0 { return 0, syserror.EINVAL } - m, err := getTaskMM(a.t) + m, err := getTaskMM(f.t) if err != nil { return 0, err } diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index f70399686..815c40b7f 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -90,12 +90,13 @@ func newGIDMap(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { } func newIDMap(t *kernel.Task, msrc *fs.MountSource, gids bool) *fs.Inode { - imsf := &idMapSeqFile{seqfile.SeqFile{SeqSource: &idMapSeqSource{ - t: t, - gids: gids, - }}} - imsf.InitEntry(t, fs.RootOwner, fs.FilePermsFromMode(0644)) - return newFile(imsf, msrc, fs.SpecialFile, t) + imsf := &idMapSeqFile{ + *seqfile.NewSeqFile(t, &idMapSeqSource{ + t: t, + gids: gids, + }), + } + return newProcInode(imsf, msrc, fs.SpecialFile, t) } func (imsf *idMapSeqFile) source() *idMapSeqSource { @@ -106,8 +107,8 @@ func (imsf *idMapSeqFile) source() *idMapSeqSource { // Linux 3.18, the limit is five lines." - user_namespaces(7) const maxIDMapLines = 5 -// DeprecatedPwritev implements fs.InodeOperations.DeprecatedPwritev. -func (imsf *idMapSeqFile) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { +// Write implements fs.FileOperations.Write. +func (imsf *idMapSeqFile) Write(ctx context.Context, _ *fs.File, src usermem.IOSequence, offset int64) (int64, error) { // "In addition, the number of bytes written to the file must be less than // the system page size, and the write must be performed at the start of // the file ..." - user_namespaces(7) diff --git a/pkg/sentry/fs/proc/uptime.go b/pkg/sentry/fs/proc/uptime.go index 80c7ce0b4..40d0fd1fd 100644 --- a/pkg/sentry/fs/proc/uptime.go +++ b/pkg/sentry/fs/proc/uptime.go @@ -18,42 +18,64 @@ import ( "fmt" "io" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // uptime is a file containing the system uptime. // // +stateify savable type uptime struct { - ramfs.Entry + fsutil.SimpleFileInode // The "start time" of the sandbox. startTime ktime.Time } // newUptime returns a new uptime file. -func (p *proc) newUptime(ctx context.Context, msrc *fs.MountSource) *fs.Inode { +func newUptime(ctx context.Context, msrc *fs.MountSource) *fs.Inode { u := &uptime{ - startTime: ktime.NowFromContext(ctx), + SimpleFileInode: *fsutil.NewSimpleFileInode(ctx, fs.RootOwner, fs.FilePermsFromMode(0444), linux.PROC_SUPER_MAGIC), + startTime: ktime.NowFromContext(ctx), } - u.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0444)) - return newFile(u, msrc, fs.SpecialFile, nil) + return newProcInode(u, msrc, fs.SpecialFile, nil) } -// DeprecatedPreadv reads the current uptime. -func (u *uptime) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { +// GetFile implements fs.InodeOperations.GetFile. +func (u *uptime) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &uptimeFile{startTime: u.startTime}), nil +} + +// +stateify savable +type uptimeFile struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + + startTime ktime.Time +} + +// Read implements fs.FileOperations.Read. +func (f *uptimeFile) Read(ctx context.Context, _ *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { if offset < 0 { return 0, syserror.EINVAL } now := ktime.NowFromContext(ctx) // Pretend that we've spent zero time sleeping (second number). - s := []byte(fmt.Sprintf("%.2f 0.00\n", now.Sub(u.startTime).Seconds())) + s := []byte(fmt.Sprintf("%.2f 0.00\n", now.Sub(f.startTime).Seconds())) if offset >= int64(len(s)) { return 0, io.EOF } diff --git a/pkg/sentry/fs/ramfs/BUILD b/pkg/sentry/fs/ramfs/BUILD index a93ad6240..a476c9cce 100644 --- a/pkg/sentry/fs/ramfs/BUILD +++ b/pkg/sentry/fs/ramfs/BUILD @@ -6,8 +6,6 @@ go_library( name = "ramfs", srcs = [ "dir.go", - "file.go", - "ramfs.go", "socket.go", "symlink.go", "tree.go", @@ -15,14 +13,12 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs", visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/secio", + "//pkg/abi/linux", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/anon", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/time", - "//pkg/sentry/memmap", - "//pkg/sentry/safemem", "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usermem", "//pkg/syserror", diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index 0a911b155..729f37694 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -18,10 +18,12 @@ import ( "sync" "syscall" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -47,7 +49,17 @@ type CreateOps struct { // // +stateify savable type Dir struct { - Entry + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeIsDirTruncate `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes + fsutil.InodeSimpleExtendedAttributes // CreateOps may be provided. // @@ -64,17 +76,23 @@ type Dir struct { children map[string]*fs.Inode // dentryMap is a sortedDentryMap containing entries for all children. - // Its entries ar kept up-to-date with d.children. + // Its entries are kept up-to-date with d.children. dentryMap *fs.SortedDentryMap } -// InitDir initializes a directory. -func (d *Dir) InitDir(ctx context.Context, contents map[string]*fs.Inode, owner fs.FileOwner, perms fs.FilePermissions) { - d.InitEntry(ctx, owner, perms) +var _ fs.InodeOperations = (*Dir)(nil) + +// NewDir returns a new Dir with the given contents and attributes. +func NewDir(ctx context.Context, contents map[string]*fs.Inode, owner fs.FileOwner, perms fs.FilePermissions) *Dir { + d := &Dir{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, perms, linux.RAMFS_MAGIC), + } + if contents == nil { contents = make(map[string]*fs.Inode) } d.children = contents + // Build the entries map ourselves, rather than calling addChildLocked, // because it will be faster. entries := make(map[string]fs.DentAttr, len(contents)) @@ -88,6 +106,8 @@ func (d *Dir) InitDir(ctx context.Context, contents map[string]*fs.Inode, owner // Directories have an extra link, corresponding to '.'. d.AddLink() + + return d } // addChildLocked add the child inode, inheriting its reference. @@ -124,17 +144,24 @@ func (d *Dir) FindChild(name string) (*fs.Inode, bool) { return child, ok } +// Children returns the names and DentAttrs of all children. It can be used to +// implement Readdir for types that embed ramfs.Dir. +func (d *Dir) Children() ([]string, map[string]fs.DentAttr) { + d.mu.Lock() + defer d.mu.Unlock() + return d.dentryMap.GetAll() +} + // removeChildLocked attempts to remove an entry from this directory. -// This Entry's mutex must be held. It returns the removed Inode. func (d *Dir) removeChildLocked(ctx context.Context, name string) (*fs.Inode, error) { inode, ok := d.children[name] if !ok { - return nil, ErrNotFound + return nil, syserror.EACCES } delete(d.children, name) d.dentryMap.Remove(name) - d.Entry.NotifyModification(ctx) + d.NotifyModification(ctx) // If the child was a subdirectory, then we must decrement this dir's // link count which was the child's ".." directory entry. @@ -143,7 +170,7 @@ func (d *Dir) removeChildLocked(ctx context.Context, name string) (*fs.Inode, er } // Update ctime. - inode.NotifyStatusChange(ctx) + inode.InodeOperations.NotifyStatusChange(ctx) // Given we're now removing this inode to the directory we must also // decrease its link count. Similarly it is increased in addChildLocked. @@ -152,8 +179,8 @@ func (d *Dir) removeChildLocked(ctx context.Context, name string) (*fs.Inode, er return inode, nil } -// RemoveEntry attempts to remove an entry from this directory. -func (d *Dir) RemoveEntry(ctx context.Context, name string) error { +// Remove removes the named non-directory. +func (d *Dir) Remove(ctx context.Context, _ *fs.Inode, name string) error { d.mu.Lock() defer d.mu.Unlock() inode, err := d.removeChildLocked(ctx, name) @@ -166,27 +193,23 @@ func (d *Dir) RemoveEntry(ctx context.Context, name string) error { return nil } -// Remove removes the named non-directory. -func (d *Dir) Remove(ctx context.Context, dir *fs.Inode, name string) error { - return d.RemoveEntry(ctx, name) -} - // RemoveDirectory removes the named directory. -func (d *Dir) RemoveDirectory(ctx context.Context, dir *fs.Inode, name string) error { +func (d *Dir) RemoveDirectory(ctx context.Context, _ *fs.Inode, name string) error { d.mu.Lock() defer d.mu.Unlock() - n, err := d.walkLocked(ctx, name) + // Get the child and make sure it is not empty. + childInode, err := d.walkLocked(ctx, name) if err != nil { return err } - dirCtx := &fs.DirCtx{} - if _, err := n.HandleOps().DeprecatedReaddir(ctx, dirCtx, 0); err != nil { + if ok, err := hasChildren(ctx, childInode); err != nil { return err + } else if ok { + return syserror.ENOTEMPTY } - if len(dirCtx.DentAttrs()) > 0 { - return ErrNotEmpty - } + + // Child was empty. Proceed with removal. inode, err := d.removeChildLocked(ctx, name) if err != nil { return err @@ -195,11 +218,11 @@ func (d *Dir) RemoveDirectory(ctx context.Context, dir *fs.Inode, name string) e // Remove our reference on the inode. inode.DecRef() - return err + return nil } // Lookup loads an inode at p into a Dirent. -func (d *Dir) Lookup(ctx context.Context, dir *fs.Inode, p string) (*fs.Dirent, error) { +func (d *Dir) Lookup(ctx context.Context, _ *fs.Inode, p string) (*fs.Dirent, error) { d.mu.Lock() defer d.mu.Unlock() @@ -214,9 +237,9 @@ func (d *Dir) Lookup(ctx context.Context, dir *fs.Inode, p string) (*fs.Dirent, return fs.NewDirent(inode, p), nil } -// walkLocked must be called with this Entry's mutex held. +// walkLocked must be called with d.mu held. func (d *Dir) walkLocked(ctx context.Context, p string) (*fs.Inode, error) { - d.Entry.NotifyAccess(ctx) + d.NotifyAccess(ctx) // Lookup a child node. if inode, ok := d.children[p]; ok { @@ -244,7 +267,7 @@ func (d *Dir) createInodeOperationsCommon(ctx context.Context, name string, make } d.addChildLocked(name, inode) - d.Entry.NotifyModification(ctx) + d.NotifyModification(ctx) return inode, nil } @@ -252,7 +275,7 @@ func (d *Dir) createInodeOperationsCommon(ctx context.Context, name string, make // Create creates a new Inode with the given name and returns its File. func (d *Dir) Create(ctx context.Context, dir *fs.Inode, name string, flags fs.FileFlags, perms fs.FilePermissions) (*fs.File, error) { if d.CreateOps == nil || d.CreateOps.NewFile == nil { - return nil, ErrDenied + return nil, syserror.EACCES } inode, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) { @@ -274,7 +297,7 @@ func (d *Dir) Create(ctx context.Context, dir *fs.Inode, name string, flags fs.F // CreateLink returns a new link. func (d *Dir) CreateLink(ctx context.Context, dir *fs.Inode, oldname, newname string) error { if d.CreateOps == nil || d.CreateOps.NewSymlink == nil { - return ErrDenied + return syserror.EACCES } _, err := d.createInodeOperationsCommon(ctx, newname, func() (*fs.Inode, error) { return d.NewSymlink(ctx, dir, oldname) @@ -292,10 +315,10 @@ func (d *Dir) CreateHardLink(ctx context.Context, dir *fs.Inode, target *fs.Inod // The link count will be incremented in addChildLocked. d.addChildLocked(name, target) - d.Entry.NotifyModification(ctx) + d.NotifyModification(ctx) // Update ctime. - target.NotifyStatusChange(ctx) + target.InodeOperations.NotifyStatusChange(ctx) return nil } @@ -303,7 +326,7 @@ func (d *Dir) CreateHardLink(ctx context.Context, dir *fs.Inode, target *fs.Inod // CreateDirectory returns a new subdirectory. func (d *Dir) CreateDirectory(ctx context.Context, dir *fs.Inode, name string, perms fs.FilePermissions) error { if d.CreateOps == nil || d.CreateOps.NewDir == nil { - return ErrDenied + return syserror.EACCES } _, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) { return d.NewDir(ctx, dir, perms) @@ -316,7 +339,7 @@ func (d *Dir) CreateDirectory(ctx context.Context, dir *fs.Inode, name string, p // Bind implements fs.InodeOperations.Bind. func (d *Dir) Bind(ctx context.Context, dir *fs.Inode, name string, ep transport.BoundEndpoint, perms fs.FilePermissions) (*fs.Dirent, error) { if d.CreateOps == nil || d.CreateOps.NewBoundEndpoint == nil { - return nil, ErrDenied + return nil, syserror.EACCES } inode, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) { return d.NewBoundEndpoint(ctx, dir, ep, perms) @@ -335,7 +358,7 @@ func (d *Dir) Bind(ctx context.Context, dir *fs.Inode, name string, ep transport // CreateFifo implements fs.InodeOperations.CreateFifo. func (d *Dir) CreateFifo(ctx context.Context, dir *fs.Inode, name string, perms fs.FilePermissions) error { if d.CreateOps == nil || d.CreateOps.NewFifo == nil { - return ErrDenied + return syserror.EACCES } _, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) { return d.NewFifo(ctx, dir, perms) @@ -343,29 +366,125 @@ func (d *Dir) CreateFifo(ctx context.Context, dir *fs.Inode, name string, perms return err } -func (d *Dir) readdirLocked(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { - // Serialize the entries in dentryMap. - n, err := fs.GenericReaddir(dirCtx, d.dentryMap) +// GetFile implements fs.InodeOperations.GetFile. +func (d *Dir) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + flags.Pread = true + return fs.NewFile(ctx, dirent, flags, &dirFileOperations{dir: d}), nil +} - // Touch the access time. - d.Entry.NotifyAccess(ctx) +// Rename implements fs.InodeOperations.Rename. +func (*Dir) Rename(ctx context.Context, oldParent *fs.Inode, oldName string, newParent *fs.Inode, newName string) error { + return Rename(ctx, oldParent.InodeOperations, oldName, newParent.InodeOperations, newName) +} +// dirFileOperations implements fs.FileOperations for a ramfs directory. +// +// +stateify savable +type dirFileOperations struct { + fsutil.DirFileOperations `state:"nosave"` + + // dirCursor contains the name of the last directory entry that was + // serialized. + dirCursor string + + // dir is the ramfs dir that this file corresponds to. + dir *Dir +} + +var _ fs.FileOperations = (*dirFileOperations)(nil) + +// Seek implements fs.FileOperations.Seek. +func (dfo *dirFileOperations) Seek(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64) (int64, error) { + return fsutil.SeekWithDirCursor(ctx, file, whence, offset, &dfo.dirCursor) +} + +// IterateDir implements DirIterator.IterateDir. +func (dfo *dirFileOperations) IterateDir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { + dfo.dir.mu.Lock() + defer dfo.dir.mu.Unlock() + + n, err := fs.GenericReaddir(dirCtx, dfo.dir.dentryMap) return offset + n, err } -// DeprecatedReaddir emits the entries contained in this directory. -func (d *Dir) DeprecatedReaddir(ctx context.Context, dirCtx *fs.DirCtx, offset int) (int, error) { - d.mu.Lock() - defer d.mu.Unlock() - return d.readdirLocked(ctx, dirCtx, offset) +// Readdir implements FileOperations.Readdir. +func (dfo *dirFileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { + root := fs.RootFromContext(ctx) + defer root.DecRef() + dirCtx := &fs.DirCtx{ + Serializer: serializer, + DirCursor: &dfo.dirCursor, + } + dfo.dir.mu.Lock() + dfo.dir.InodeSimpleAttributes.Unstable.AccessTime = ktime.NowFromContext(ctx) + dfo.dir.mu.Unlock() + return fs.DirentReaddir(ctx, file.Dirent, dfo, root, dirCtx, file.Offset()) } -// DeprecatedPreadv always returns ErrIsDirectory -func (*Dir) DeprecatedPreadv(context.Context, usermem.IOSequence, int64) (int64, error) { - return 0, ErrIsDirectory +// hasChildren is a helper method that determines whether an arbitrary inode +// (not necessarily ramfs) has any children. +func hasChildren(ctx context.Context, inode *fs.Inode) (bool, error) { + // Take an extra ref on inode which will be given to the dirent and + // dropped when that dirent is destroyed. + inode.IncRef() + d := fs.NewTransientDirent(inode) + defer d.DecRef() + + file, err := inode.GetFile(ctx, d, fs.FileFlags{Read: true}) + if err != nil { + return false, err + } + defer file.DecRef() + + ser := &fs.CollectEntriesSerializer{} + if err := file.Readdir(ctx, ser); err != nil { + return false, err + } + // We will always write "." and "..", so ignore those two. + if ser.Written() > 2 { + return true, nil + } + return false, nil } -// DeprecatedPwritev always returns ErrIsDirectory -func (*Dir) DeprecatedPwritev(context.Context, usermem.IOSequence, int64) (int64, error) { - return 0, ErrIsDirectory +// Rename renames from a *ramfs.Dir to another *ramfs.Dir. +func Rename(ctx context.Context, oldParent fs.InodeOperations, oldName string, newParent fs.InodeOperations, newName string) error { + op, ok := oldParent.(*Dir) + if !ok { + return syserror.EXDEV + } + np, ok := newParent.(*Dir) + if !ok { + return syserror.EXDEV + } + + np.mu.Lock() + defer np.mu.Unlock() + + // Check whether the ramfs entry to be replaced is a non-empty directory. + if replaced, ok := np.children[newName]; ok { + if fs.IsDir(replaced.StableAttr) { + if ok, err := hasChildren(ctx, replaced); err != nil { + return err + } else if ok { + return syserror.ENOTEMPTY + } + } + } + + // Be careful, we may have already grabbed this mutex above. + if op != np { + op.mu.Lock() + defer op.mu.Unlock() + } + + // Do the swap. + n := op.children[oldName] + op.removeChildLocked(ctx, oldName) + np.addChildLocked(newName, n) + + // Update ctime. + n.InodeOperations.NotifyStatusChange(ctx) + + return nil } diff --git a/pkg/sentry/fs/ramfs/file.go b/pkg/sentry/fs/ramfs/file.go deleted file mode 100644 index b7fc98ffc..000000000 --- a/pkg/sentry/fs/ramfs/file.go +++ /dev/null @@ -1,150 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package ramfs - -import ( - "io" - "sync" - - "gvisor.googlesource.com/gvisor/pkg/secio" - "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" -) - -// File represents a unique file. It uses a simple byte slice as storage, and -// thus should only be used for small files. -// -// A File is not mappable. -// -// +stateify savable -type File struct { - Entry - - // mu protects the fields below. - mu sync.Mutex `state:"nosave"` - - // data tracks backing data for the file. - data []byte -} - -// InitFile initializes a file. -func (f *File) InitFile(ctx context.Context, owner fs.FileOwner, perms fs.FilePermissions) { - f.InitEntry(ctx, owner, perms) -} - -// UnstableAttr returns unstable attributes of this ramfs file. -func (f *File) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - f.mu.Lock() - defer f.mu.Unlock() - - uattr, _ := f.Entry.UnstableAttr(ctx, inode) - uattr.Size = int64(len(f.data)) - uattr.Usage = f.usageLocked() - - return uattr, nil -} - -// usageLocked returns the disk usage. Caller must hold f.mu. -func (f *File) usageLocked() int64 { - return int64(len(f.data)) -} - -// Append appends the given data. This is for internal use. -func (f *File) Append(data []byte) { - f.mu.Lock() - defer f.mu.Unlock() - f.data = append(f.data, data...) -} - -// Truncate truncates this node. -func (f *File) Truncate(ctx context.Context, inode *fs.Inode, l int64) error { - f.mu.Lock() - defer f.mu.Unlock() - if l < int64(len(f.data)) { - // Remove excess bytes. - f.data = f.data[:l] - return nil - } else if l > int64(len(f.data)) { - // Create a new slice with size l, and copy f.data into it. - d := make([]byte, l) - copy(d, f.data) - f.data = d - } - f.Entry.NotifyModification(ctx) - return nil -} - -// ReadAt implements io.ReaderAt. -func (f *File) ReadAt(data []byte, offset int64) (int, error) { - if offset < 0 { - return 0, ErrInvalidOp - } - if offset >= int64(len(f.data)) { - return 0, io.EOF - } - n := copy(data, f.data[offset:]) - // Did we read past the end? - if offset+int64(len(data)) >= int64(len(f.data)) { - return n, io.EOF - } - return n, nil -} - -// DeprecatedPreadv reads into a collection of slices from a given offset. -func (f *File) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - f.mu.Lock() - defer f.mu.Unlock() - if offset >= int64(len(f.data)) { - return 0, io.EOF - } - n, err := dst.CopyOut(ctx, f.data[offset:]) - if n > 0 { - f.Entry.NotifyAccess(ctx) - } - return int64(n), err -} - -// WriteAt implements io.WriterAt. -func (f *File) WriteAt(data []byte, offset int64) (int, error) { - if offset < 0 { - return 0, ErrInvalidOp - } - newLen := offset + int64(len(data)) - if newLen < 0 { - // Overflow. - return 0, syserror.EINVAL - } - if newLen > int64(len(f.data)) { - // Copy f.data into new slice with expanded length. - d := make([]byte, newLen) - copy(d, f.data) - f.data = d - } - return copy(f.data[offset:], data), nil -} - -// DeprecatedPwritev writes from a collection of slices at a given offset. -func (f *File) DeprecatedPwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { - f.mu.Lock() - defer f.mu.Unlock() - n, err := src.CopyInTo(ctx, safemem.FromIOWriter{secio.NewOffsetWriter(f, offset)}) - if n > 0 { - f.Entry.NotifyModification(ctx) - } - return n, err -} diff --git a/pkg/sentry/fs/ramfs/ramfs.go b/pkg/sentry/fs/ramfs/ramfs.go deleted file mode 100644 index d77688a34..000000000 --- a/pkg/sentry/fs/ramfs/ramfs.go +++ /dev/null @@ -1,441 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package ramfs implements an in-memory file system that can be associated with -// any device. -package ramfs - -import ( - "errors" - "sync" - "syscall" - - "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" - "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" - "gvisor.googlesource.com/gvisor/pkg/waiter" -) - -var ( - // ErrInvalidOp indicates the operation is not valid. - ErrInvalidOp = errors.New("invalid operation") - - // ErrDenied indicates the operation was denied. - ErrDenied = errors.New("operation denied") - - // ErrNotFound indicates that a node was not found on a walk. - ErrNotFound = errors.New("node not found") - - // ErrCrossDevice indicates a cross-device link or rename. - ErrCrossDevice = errors.New("can't link across filesystems") - - // ErrIsDirectory indicates that the operation failed because - // the node is a directory. - ErrIsDirectory = errors.New("is a directory") - - // ErrNotDirectory indicates that the operation failed because - // the node is a not directory. - ErrNotDirectory = errors.New("not a directory") - - // ErrNotEmpty indicates that the operation failed because the - // directory is not empty. - ErrNotEmpty = errors.New("directory not empty") -) - -// Entry represents common internal state for file and directory nodes. -// This may be used by other packages to easily create ramfs files. -// -// +stateify savable -type Entry struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.NoMappable `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` - fsutil.InodeNotSocket `state:"nosave"` - - // mu protects the fields below. - mu sync.Mutex `state:"nosave"` - - // unstable is unstable attributes. - unstable fs.UnstableAttr - - // xattrs are the extended attributes of the Entry. - xattrs map[string][]byte -} - -// InitEntry initializes an entry. -func (e *Entry) InitEntry(ctx context.Context, owner fs.FileOwner, p fs.FilePermissions) { - e.InitEntryWithAttr(ctx, fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: owner, - Perms: p, - // Always start unlinked. - Links: 0, - })) -} - -// InitEntryWithAttr initializes an entry with a complete set of attributes. -func (e *Entry) InitEntryWithAttr(ctx context.Context, uattr fs.UnstableAttr) { - e.unstable = uattr - e.xattrs = make(map[string][]byte) -} - -// UnstableAttr implements fs.InodeOperations.UnstableAttr. -func (e *Entry) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - e.mu.Lock() - attr := e.unstable - e.mu.Unlock() - return attr, nil -} - -// Check implements fs.InodeOperations.Check. -func (*Entry) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { - return fs.ContextCanAccessFile(ctx, inode, p) -} - -// Getxattr implements fs.InodeOperations.Getxattr. -func (e *Entry) Getxattr(inode *fs.Inode, name string) ([]byte, error) { - // Hot path. Avoid defers. - e.mu.Lock() - value, ok := e.xattrs[name] - e.mu.Unlock() - if ok { - return value, nil - } - return nil, syserror.ENOATTR -} - -// Setxattr implements fs.InodeOperations.Setxattr. -func (e *Entry) Setxattr(inode *fs.Inode, name string, value []byte) error { - e.mu.Lock() - e.xattrs[name] = value - e.mu.Unlock() - return nil -} - -// Listxattr implements fs.InodeOperations.Listxattr. -func (e *Entry) Listxattr(inode *fs.Inode) (map[string]struct{}, error) { - e.mu.Lock() - names := make(map[string]struct{}, len(e.xattrs)) - for name := range e.xattrs { - names[name] = struct{}{} - } - e.mu.Unlock() - return names, nil -} - -// GetFile returns a fs.File backed by the dirent argument and flags. -func (*Entry) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { - return fsutil.NewHandle(ctx, d, flags, d.Inode.HandleOps()), nil -} - -// SetPermissions always sets the permissions. -func (e *Entry) SetPermissions(ctx context.Context, inode *fs.Inode, p fs.FilePermissions) bool { - e.mu.Lock() - e.unstable.Perms = p - e.unstable.StatusChangeTime = ktime.NowFromContext(ctx) - e.mu.Unlock() - return true -} - -// SetOwner always sets ownership. -func (e *Entry) SetOwner(ctx context.Context, inode *fs.Inode, owner fs.FileOwner) error { - e.mu.Lock() - if owner.UID.Ok() { - e.unstable.Owner.UID = owner.UID - } - if owner.GID.Ok() { - e.unstable.Owner.GID = owner.GID - } - e.mu.Unlock() - return nil -} - -// SetTimestamps sets the timestamps. -func (e *Entry) SetTimestamps(ctx context.Context, inode *fs.Inode, ts fs.TimeSpec) error { - if ts.ATimeOmit && ts.MTimeOmit { - return nil - } - - e.mu.Lock() - now := ktime.NowFromContext(ctx) - if !ts.ATimeOmit { - if ts.ATimeSetSystemTime { - e.unstable.AccessTime = now - } else { - e.unstable.AccessTime = ts.ATime - } - } - if !ts.MTimeOmit { - if ts.MTimeSetSystemTime { - e.unstable.ModificationTime = now - } else { - e.unstable.ModificationTime = ts.MTime - } - } - e.unstable.StatusChangeTime = now - e.mu.Unlock() - return nil -} - -// NotifyStatusChange updates the status change time (ctime). -func (e *Entry) NotifyStatusChange(ctx context.Context) { - e.mu.Lock() - e.unstable.StatusChangeTime = ktime.NowFromContext(ctx) - e.mu.Unlock() -} - -// StatusChangeTime returns the last status change time for this node. -func (e *Entry) StatusChangeTime() ktime.Time { - e.mu.Lock() - t := e.unstable.StatusChangeTime - e.mu.Unlock() - return t -} - -// NotifyModification updates the modification time and the status change time. -func (e *Entry) NotifyModification(ctx context.Context) { - e.mu.Lock() - now := ktime.NowFromContext(ctx) - e.unstable.ModificationTime = now - e.unstable.StatusChangeTime = now - e.mu.Unlock() -} - -// ModificationTime returns the last modification time for this node. -func (e *Entry) ModificationTime() ktime.Time { - e.mu.Lock() - t := e.unstable.ModificationTime - e.mu.Unlock() - return t -} - -// NotifyAccess updates the access time. -func (e *Entry) NotifyAccess(ctx context.Context) { - e.mu.Lock() - now := ktime.NowFromContext(ctx) - e.unstable.AccessTime = now - e.mu.Unlock() -} - -// AccessTime returns the last access time for this node. -func (e *Entry) AccessTime() ktime.Time { - e.mu.Lock() - t := e.unstable.AccessTime - e.mu.Unlock() - return t -} - -// Permissions returns permissions on this entry. -func (e *Entry) Permissions() fs.FilePermissions { - e.mu.Lock() - p := e.unstable.Perms - e.mu.Unlock() - return p -} - -// Lookup is not supported by default. -func (*Entry) Lookup(context.Context, *fs.Inode, string) (*fs.Dirent, error) { - return nil, ErrInvalidOp -} - -// Create is not supported by default. -func (*Entry) Create(context.Context, *fs.Inode, string, fs.FileFlags, fs.FilePermissions) (*fs.File, error) { - return nil, ErrInvalidOp -} - -// CreateLink is not supported by default. -func (*Entry) CreateLink(context.Context, *fs.Inode, string, string) error { - return ErrInvalidOp -} - -// CreateHardLink is not supported by default. -func (*Entry) CreateHardLink(context.Context, *fs.Inode, *fs.Inode, string) error { - return ErrInvalidOp -} - -// IsVirtual returns true. -func (*Entry) IsVirtual() bool { - return true -} - -// CreateDirectory is not supported by default. -func (*Entry) CreateDirectory(context.Context, *fs.Inode, string, fs.FilePermissions) error { - return ErrInvalidOp -} - -// Bind is not supported by default. -func (*Entry) Bind(context.Context, *fs.Inode, string, transport.BoundEndpoint, fs.FilePermissions) (*fs.Dirent, error) { - return nil, ErrInvalidOp -} - -// CreateFifo implements fs.InodeOperations.CreateFifo. CreateFifo is not supported by -// default. -func (*Entry) CreateFifo(context.Context, *fs.Inode, string, fs.FilePermissions) error { - return ErrInvalidOp -} - -// Remove is not supported by default. -func (*Entry) Remove(context.Context, *fs.Inode, string) error { - return ErrInvalidOp -} - -// RemoveDirectory is not supported by default. -func (*Entry) RemoveDirectory(context.Context, *fs.Inode, string) error { - return ErrInvalidOp -} - -// StatFS always returns ENOSYS. -func (*Entry) StatFS(context.Context) (fs.Info, error) { - return fs.Info{}, syscall.ENOSYS -} - -// Rename implements fs.InodeOperations.Rename. -func (e *Entry) Rename(ctx context.Context, oldParent *fs.Inode, oldName string, newParent *fs.Inode, newName string) error { - return Rename(ctx, oldParent.InodeOperations, oldName, newParent.InodeOperations, newName) -} - -// Rename renames from a *ramfs.Dir to another *ramfs.Dir. -func Rename(ctx context.Context, oldParent fs.InodeOperations, oldName string, newParent fs.InodeOperations, newName string) error { - op, ok := oldParent.(*Dir) - if !ok { - return ErrCrossDevice - } - np, ok := newParent.(*Dir) - if !ok { - return ErrCrossDevice - } - - np.mu.Lock() - defer np.mu.Unlock() - - // Check whether the ramfs entry to be replaced is a non-empty directory. - if replaced, ok := np.children[newName]; ok { - if fs.IsDir(replaced.StableAttr) { - // FIXME: simplify by pinning children of ramfs-backed directories - // in the Dirent tree: this allows us to generalize ramfs operations without - // relying on an implementation of Readdir (which may do anything, like require - // that the file be open ... which would be reasonable). - dirCtx := &fs.DirCtx{} - _, err := replaced.HandleOps().DeprecatedReaddir(ctx, dirCtx, 0) - if err != nil { - return err - } - attrs := dirCtx.DentAttrs() - - // ramfs-backed directories should not contain "." and "..", but we do this - // just in case. - delete(attrs, ".") - delete(attrs, "..") - - // If the directory to be replaced is not empty, reject the rename. - if len(attrs) != 0 { - return ErrNotEmpty - } - } - } - - // Be careful, we may have already grabbed this mutex above. - if op != np { - op.mu.Lock() - defer op.mu.Unlock() - } - - // Do the swap. - n := op.children[oldName] - op.removeChildLocked(ctx, oldName) - np.addChildLocked(newName, n) - - // Update ctime. - n.NotifyStatusChange(ctx) - - return nil -} - -// Truncate is not supported by default. -func (*Entry) Truncate(context.Context, *fs.Inode, int64) error { - return ErrInvalidOp -} - -// Readlink always returns ENOLINK. -func (*Entry) Readlink(context.Context, *fs.Inode) (string, error) { - return "", syscall.ENOLINK -} - -// Getlink always returns ENOLINK. -func (*Entry) Getlink(context.Context, *fs.Inode) (*fs.Dirent, error) { - return nil, syscall.ENOLINK -} - -// Release is a no-op. -func (e *Entry) Release(context.Context) {} - -// AddLink implements InodeOperationss.AddLink. -func (e *Entry) AddLink() { - e.mu.Lock() - e.unstable.Links++ - e.mu.Unlock() -} - -// DropLink implements InodeOperationss.DropLink. -func (e *Entry) DropLink() { - e.mu.Lock() - e.unstable.Links-- - e.mu.Unlock() -} - -// DeprecatedReaddir is not supported by default. -func (*Entry) DeprecatedReaddir(context.Context, *fs.DirCtx, int) (int, error) { - return 0, ErrNotDirectory -} - -// DeprecatedPreadv always returns ErrInvalidOp. -func (*Entry) DeprecatedPreadv(context.Context, usermem.IOSequence, int64) (int64, error) { - return 0, ErrInvalidOp -} - -// DeprecatedPwritev always returns ErrInvalidOp. -func (*Entry) DeprecatedPwritev(context.Context, usermem.IOSequence, int64) (int64, error) { - return 0, ErrInvalidOp -} - -// DeprecatedFsync is a noop. -func (*Entry) DeprecatedFsync() error { - // Ignore, this is in memory. - return nil -} - -// DeprecatedFlush always returns nil. -func (*Entry) DeprecatedFlush() error { - return nil -} - -// DeprecatedMappable implements fs.InodeOperations.DeprecatedMappable. -func (*Entry) DeprecatedMappable(context.Context, *fs.Inode) (memmap.Mappable, bool) { - return nil, false -} - -func init() { - // Register ramfs errors. - syserror.AddErrorTranslation(ErrInvalidOp, syscall.EINVAL) - syserror.AddErrorTranslation(ErrDenied, syscall.EACCES) - syserror.AddErrorTranslation(ErrNotFound, syscall.ENOENT) - syserror.AddErrorTranslation(ErrCrossDevice, syscall.EXDEV) - syserror.AddErrorTranslation(ErrIsDirectory, syscall.EISDIR) - syserror.AddErrorTranslation(ErrNotDirectory, syscall.ENOTDIR) - syserror.AddErrorTranslation(ErrNotEmpty, syscall.ENOTEMPTY) -} diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index 8c81478c8..2c1295897 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -15,25 +15,42 @@ package ramfs import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // Socket represents a socket. // // +stateify savable type Socket struct { - Entry + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeNotTruncatable `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes + fsutil.InodeSimpleExtendedAttributes // ep is the bound endpoint. ep transport.BoundEndpoint } -// InitSocket initializes a socket. -func (s *Socket) InitSocket(ctx context.Context, ep transport.BoundEndpoint, owner fs.FileOwner, perms fs.FilePermissions) { - s.InitEntry(ctx, owner, perms) - s.ep = ep +var _ fs.InodeOperations = (*Socket)(nil) + +// NewSocket returns a new Socket. +func NewSocket(ctx context.Context, ep transport.BoundEndpoint, owner fs.FileOwner, perms fs.FilePermissions) *Socket { + return &Socket{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, perms, linux.SOCKFS_MAGIC), + ep: ep, + } } // BoundEndpoint returns the socket data. @@ -42,3 +59,24 @@ func (s *Socket) BoundEndpoint(*fs.Inode, string) transport.BoundEndpoint { // care about the path argument. return s.ep } + +// GetFile implements fs.FileOperations.GetFile. +func (s *Socket) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &socketFileOperations{}), nil +} + +// +stateify savable +type socketFileOperations struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoRead `state:"nosave"` + fsutil.FileNoSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` +} + +var _ fs.FileOperations = (*socketFileOperations)(nil) diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index a21fac2c7..47dae380b 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -15,44 +15,55 @@ package ramfs import ( - "sync" - + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" + "gvisor.googlesource.com/gvisor/pkg/waiter" ) // Symlink represents a symlink. // // +stateify savable type Symlink struct { - Entry + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotTruncatable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` - mu sync.Mutex `state:"nosave"` + fsutil.InodeSimpleAttributes + fsutil.InodeSimpleExtendedAttributes // Target is the symlink target. Target string } -// InitSymlink initializes a symlink, pointing to the given target. -// A symlink is assumed to always have permissions 0777. -func (s *Symlink) InitSymlink(ctx context.Context, owner fs.FileOwner, target string) { - s.InitEntry(ctx, owner, fs.FilePermsFromMode(0777)) - s.Target = target +var _ fs.InodeOperations = (*Symlink)(nil) + +// NewSymlink returns a new Symlink. +func NewSymlink(ctx context.Context, owner fs.FileOwner, target string) *Symlink { + // A symlink is assumed to always have permissions 0777. + return &Symlink{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, owner, fs.FilePermsFromMode(0777), linux.RAMFS_MAGIC), + Target: target, + } } // UnstableAttr returns all attributes of this ramfs symlink. func (s *Symlink) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - uattr, _ := s.Entry.UnstableAttr(ctx, inode) + uattr, err := s.InodeSimpleAttributes.UnstableAttr(ctx, inode) + if err != nil { + return fs.UnstableAttr{}, err + } uattr.Size = int64(len(s.Target)) uattr.Usage = uattr.Size return uattr, nil } -// Check implements InodeOperations.Check. -func (s *Symlink) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { - return fs.ContextCanAccessFile(ctx, inode, p) -} - // SetPermissions on a symlink is always rejected. func (s *Symlink) SetPermissions(context.Context, *fs.Inode, fs.FilePermissions) bool { return false @@ -60,10 +71,7 @@ func (s *Symlink) SetPermissions(context.Context, *fs.Inode, fs.FilePermissions) // Readlink reads the symlink value. func (s *Symlink) Readlink(ctx context.Context, _ *fs.Inode) (string, error) { - s.mu.Lock() - defer s.mu.Unlock() - - s.Entry.NotifyAccess(ctx) + s.NotifyAccess(ctx) return s.Target, nil } @@ -72,3 +80,24 @@ func (s *Symlink) Readlink(ctx context.Context, _ *fs.Inode) (string, error) { func (*Symlink) Getlink(context.Context, *fs.Inode) (*fs.Dirent, error) { return nil, fs.ErrResolveViaReadlink } + +// GetFile implements fs.FileOperations.GetFile. +func (s *Symlink) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return fs.NewFile(ctx, dirent, flags, &symlinkFileOperations{}), nil +} + +// +stateify savable +type symlinkFileOperations struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoRead `state:"nosave"` + fsutil.FileNoSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` +} + +var _ fs.FileOperations = (*symlinkFileOperations)(nil) diff --git a/pkg/sentry/fs/ramfs/test/BUILD b/pkg/sentry/fs/ramfs/test/BUILD deleted file mode 100644 index 187eac49d..000000000 --- a/pkg/sentry/fs/ramfs/test/BUILD +++ /dev/null @@ -1,16 +0,0 @@ -package(licenses = ["notice"]) # Apache 2.0 - -load("//tools/go_stateify:defs.bzl", "go_library") - -go_library( - name = "test", - testonly = 1, - srcs = ["test.go"], - importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs/test", - visibility = ["//pkg/sentry:internal"], - deps = [ - "//pkg/sentry/context", - "//pkg/sentry/fs", - "//pkg/sentry/fs/ramfs", - ], -) diff --git a/pkg/sentry/fs/ramfs/test/test.go b/pkg/sentry/fs/ramfs/test/test.go deleted file mode 100644 index 11bff7729..000000000 --- a/pkg/sentry/fs/ramfs/test/test.go +++ /dev/null @@ -1,46 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package test provides a simple ramfs-based filesystem for use in testing. -package test - -import ( - "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" -) - -// Dir is a simple ramfs.Dir that supports save/restore as-is. -type Dir struct { - ramfs.Dir -} - -// NewDir returns a simple ramfs directory with the passed contents. -func NewDir(ctx context.Context, contents map[string]*fs.Inode, perms fs.FilePermissions) *Dir { - d := &Dir{} - d.InitDir(ctx, contents, fs.RootOwner, perms) - return d -} - -// File is a simple ramfs.File that supports save/restore as-is. -type File struct { - ramfs.File -} - -// NewFile returns a simple ramfs File. -func NewFile(ctx context.Context, perms fs.FilePermissions) *File { - f := &File{} - f.InitFile(ctx, fs.RootOwner, perms) - return f -} diff --git a/pkg/sentry/fs/ramfs/tree.go b/pkg/sentry/fs/ramfs/tree.go index 29a70f698..f6d5ffdec 100644 --- a/pkg/sentry/fs/ramfs/tree.go +++ b/pkg/sentry/fs/ramfs/tree.go @@ -60,8 +60,7 @@ func makeSubdir(ctx context.Context, msrc *fs.MountSource, root *Dir, subdir str // emptyDir returns an empty *ramfs.Dir that is traversable but not writable. func emptyDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - dir := &Dir{} - dir.InitDir(ctx, make(map[string]*fs.Inode), fs.RootOwner, fs.FilePermsFromMode(0555)) + dir := NewDir(ctx, make(map[string]*fs.Inode), fs.RootOwner, fs.FilePermsFromMode(0555)) return fs.NewInode(dir, msrc, fs.StableAttr{ DeviceID: anon.PseudoDevice.DeviceID(), InodeID: anon.PseudoDevice.NextIno(), diff --git a/pkg/sentry/fs/ramfs/tree_test.go b/pkg/sentry/fs/ramfs/tree_test.go index 54df2143c..8bee9cfc1 100644 --- a/pkg/sentry/fs/ramfs/tree_test.go +++ b/pkg/sentry/fs/ramfs/tree_test.go @@ -22,7 +22,7 @@ import ( ) func TestMakeDirectoryTree(t *testing.T) { - mount := fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}) + mount := fs.NewPseudoMountSource() for _, test := range []struct { name string diff --git a/pkg/sentry/fs/sys/BUILD b/pkg/sentry/fs/sys/BUILD index 5ba23d5da..7de928e16 100644 --- a/pkg/sentry/fs/sys/BUILD +++ b/pkg/sentry/fs/sys/BUILD @@ -13,12 +13,13 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs/sys", visibility = ["//pkg/sentry:internal"], deps = [ + "//pkg/abi/linux", "//pkg/sentry/context", "//pkg/sentry/device", "//pkg/sentry/fs", + "//pkg/sentry/fs/fsutil", "//pkg/sentry/fs/ramfs", "//pkg/sentry/kernel", "//pkg/sentry/usermem", - "//pkg/syserror", ], ) diff --git a/pkg/sentry/fs/sys/devices.go b/pkg/sentry/fs/sys/devices.go index e64aa0edc..8b728a4e4 100644 --- a/pkg/sentry/fs/sys/devices.go +++ b/pkg/sentry/fs/sys/devices.go @@ -16,43 +16,50 @@ package sys import ( "fmt" - "io" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" ) // +stateify savable type cpunum struct { - ramfs.Entry + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeNotVirtual `state:"nosave"` + fsutil.InodeNotTruncatable `state:"nosave"` + + fsutil.InodeSimpleAttributes + fsutil.InodeStaticFileGetter + + // k is the system kernel. + k *kernel.Kernel } -func (c *cpunum) DeprecatedPreadv(ctx context.Context, dst usermem.IOSequence, offset int64) (int64, error) { - if offset < 0 { - return 0, syserror.EINVAL - } +var _ fs.InodeOperations = (*cpunum)(nil) +func newPossible(ctx context.Context, msrc *fs.MountSource) *fs.Inode { + var maxCore uint k := kernel.KernelFromContext(ctx) - if k == nil { - return 0, io.EOF + if k != nil { + maxCore = k.ApplicationCores() - 1 } + contents := []byte(fmt.Sprintf("0-%d\n", maxCore)) - str := []byte(fmt.Sprintf("0-%d\n", k.ApplicationCores()-1)) - if offset >= int64(len(str)) { - return 0, io.EOF + c := &cpunum{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, fs.RootOwner, fs.FilePermsFromMode(0444), linux.SYSFS_MAGIC), + InodeStaticFileGetter: fsutil.InodeStaticFileGetter{ + Contents: contents, + }, } - - n, err := dst.CopyOut(ctx, str[offset:]) - return int64(n), err -} - -func newPossible(ctx context.Context, msrc *fs.MountSource) *fs.Inode { - c := &cpunum{} - c.InitEntry(ctx, fs.RootOwner, fs.FilePermsFromMode(0444)) return newFile(c, msrc) } diff --git a/pkg/sentry/fs/sys/fs.go b/pkg/sentry/fs/sys/fs.go index 5ce33f87f..301fef038 100644 --- a/pkg/sentry/fs/sys/fs.go +++ b/pkg/sentry/fs/sys/fs.go @@ -24,6 +24,8 @@ import ( // +stateify savable type filesystem struct{} +var _ fs.Filesystem = (*filesystem)(nil) + func init() { fs.RegisterFilesystem(&filesystem{}) } diff --git a/pkg/sentry/fs/sys/sys.go b/pkg/sentry/fs/sys/sys.go index 7cc1942c7..c5b56fe69 100644 --- a/pkg/sentry/fs/sys/sys.go +++ b/pkg/sentry/fs/sys/sys.go @@ -22,13 +22,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// sys is a root sys node. -// -// +stateify savable -type sys struct { - ramfs.Dir -} - func newFile(node fs.InodeOperations, msrc *fs.MountSource) *fs.Inode { sattr := fs.StableAttr{ DeviceID: sysfsDevice.DeviceID(), @@ -40,8 +33,7 @@ func newFile(node fs.InodeOperations, msrc *fs.MountSource) *fs.Inode { } func newDir(ctx context.Context, msrc *fs.MountSource, contents map[string]*fs.Inode) *fs.Inode { - d := &sys{} - d.InitDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) + d := ramfs.NewDir(ctx, contents, fs.RootOwner, fs.FilePermsFromMode(0555)) return fs.NewInode(d, msrc, fs.StableAttr{ DeviceID: sysfsDevice.DeviceID(), InodeID: sysfsDevice.NextIno(), diff --git a/pkg/sentry/fs/timerfd/timerfd.go b/pkg/sentry/fs/timerfd/timerfd.go index 7423e816c..b26466b9d 100644 --- a/pkg/sentry/fs/timerfd/timerfd.go +++ b/pkg/sentry/fs/timerfd/timerfd.go @@ -33,12 +33,12 @@ import ( // // +stateify savable type TimerOperations struct { - fsutil.ZeroSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` - fsutil.NoIoctl `state:"nosave"` + fsutil.FileZeroSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` events waiter.Queue `state:"zerovalue"` timer *ktime.Timer diff --git a/pkg/sentry/fs/tmpfs/BUILD b/pkg/sentry/fs/tmpfs/BUILD index 9065cdd5d..14c7a9e62 100644 --- a/pkg/sentry/fs/tmpfs/BUILD +++ b/pkg/sentry/fs/tmpfs/BUILD @@ -23,11 +23,13 @@ go_library( "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/pipe", + "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", "//pkg/sentry/safemem", "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usage", "//pkg/sentry/usermem", + "//pkg/syserror", "//pkg/waiter", ], ) diff --git a/pkg/sentry/fs/tmpfs/file_regular.go b/pkg/sentry/fs/tmpfs/file_regular.go index 1f9d69909..2c1eb0fd2 100644 --- a/pkg/sentry/fs/tmpfs/file_regular.go +++ b/pkg/sentry/fs/tmpfs/file_regular.go @@ -28,13 +28,13 @@ import ( // // +stateify savable type regularFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.NoopRelease `state:"nosave"` - fsutil.GenericSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoopFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoIoctl `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` // iops is the InodeOperations of a regular tmpfs file. It is // guaranteed to be the same as file.Dirent.Inode.InodeOperations, diff --git a/pkg/sentry/fs/tmpfs/file_test.go b/pkg/sentry/fs/tmpfs/file_test.go index 02da9af82..e7bbdc404 100644 --- a/pkg/sentry/fs/tmpfs/file_test.go +++ b/pkg/sentry/fs/tmpfs/file_test.go @@ -52,19 +52,19 @@ func TestGrow(t *testing.T) { abuf := bytes.Repeat([]byte{'a'}, 68) n, err := f.Pwritev(ctx, usermem.BytesIOSequence(abuf), 0) if n != int64(len(abuf)) || err != nil { - t.Fatalf("DeprecatedPwritev got (%d, %v) want (%d, nil)", n, err, len(abuf)) + t.Fatalf("Pwritev got (%d, %v) want (%d, nil)", n, err, len(abuf)) } bbuf := bytes.Repeat([]byte{'b'}, 856) n, err = f.Pwritev(ctx, usermem.BytesIOSequence(bbuf), 68) if n != int64(len(bbuf)) || err != nil { - t.Fatalf("DeprecatedPwritev got (%d, %v) want (%d, nil)", n, err, len(bbuf)) + t.Fatalf("Pwritev got (%d, %v) want (%d, nil)", n, err, len(bbuf)) } rbuf := make([]byte, len(abuf)+len(bbuf)) n, err = f.Preadv(ctx, usermem.BytesIOSequence(rbuf), 0) if n != int64(len(rbuf)) || err != nil { - t.Fatalf("DeprecatedPreadv got (%d, %v) want (%d, nil)", n, err, len(rbuf)) + t.Fatalf("Preadv got (%d, %v) want (%d, nil)", n, err, len(rbuf)) } if want := append(abuf, bbuf...); !bytes.Equal(rbuf, want) { diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go index 88f85b85a..caa3220ee 100644 --- a/pkg/sentry/fs/tmpfs/fs.go +++ b/pkg/sentry/fs/tmpfs/fs.go @@ -50,6 +50,8 @@ const ( // +stateify savable type Filesystem struct{} +var _ fs.Filesystem = (*Filesystem)(nil) + func init() { fs.RegisterFilesystem(&Filesystem{}) } diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index ca2b4aabb..42d4bc76f 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" @@ -46,11 +47,13 @@ import ( // // +stateify savable type fileInodeOperations struct { - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.InodeNotDirectory `state:"nosave"` - fsutil.InodeNotSocket `state:"nosave"` - fsutil.InodeNotSymlink `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + + fsutil.InodeSimpleExtendedAttributes // kernel is used to allocate platform memory that stores the file's contents. kernel *kernel.Kernel @@ -62,10 +65,10 @@ type fileInodeOperations struct { // attr contains the unstable metadata for the file. // - // attr is protected by attrMu. attr.Unstable.Size is protected by both - // attrMu and dataMu; reading it requires locking either mutex, while - // mutating it requires locking both. - attr fsutil.InMemoryAttributes + // attr is protected by attrMu. attr.Size is protected by both attrMu + // and dataMu; reading it requires locking either mutex, while mutating + // it requires locking both. + attr fs.UnstableAttr mapsMu sync.Mutex `state:"nosave"` @@ -83,12 +86,12 @@ type fileInodeOperations struct { data fsutil.FileRangeSet } +var _ fs.InodeOperations = (*fileInodeOperations)(nil) + // NewInMemoryFile returns a new file backed by p.Memory(). func NewInMemoryFile(ctx context.Context, usage usage.MemoryKind, uattr fs.UnstableAttr, k *kernel.Kernel) fs.InodeOperations { return &fileInodeOperations{ - attr: fsutil.InMemoryAttributes{ - Unstable: uattr, - }, + attr: uattr, kernel: k, memUsage: usage, } @@ -121,71 +124,56 @@ func (f *fileInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags f // UnstableAttr returns unstable attributes of this tmpfs file. func (f *fileInodeOperations) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { f.attrMu.Lock() - defer f.attrMu.Unlock() f.dataMu.RLock() - defer f.dataMu.RUnlock() - attr := f.attr.Unstable + attr := f.attr attr.Usage = int64(f.data.Span()) + f.dataMu.RUnlock() + f.attrMu.Unlock() return attr, nil } -// Getxattr implements fs.InodeOperations.Getxattr. -func (f *fileInodeOperations) Getxattr(inode *fs.Inode, name string) ([]byte, error) { - f.attrMu.Lock() - defer f.attrMu.Unlock() - return f.attr.Getxattr(name) -} - -// Setxattr implements fs.InodeOperations.Setxattr. -func (f *fileInodeOperations) Setxattr(inode *fs.Inode, name string, value []byte) error { - f.attrMu.Lock() - defer f.attrMu.Unlock() - return f.attr.Setxattr(name, value) -} - -// Listxattr implements fs.InodeOperations.Listxattr. -func (f *fileInodeOperations) Listxattr(inode *fs.Inode) (map[string]struct{}, error) { - f.attrMu.Lock() - defer f.attrMu.Unlock() - return f.attr.Listxattr() -} - // Check implements fs.InodeOperations.Check. func (f *fileInodeOperations) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { return fs.ContextCanAccessFile(ctx, inode, p) } // SetPermissions implements fs.InodeOperations.SetPermissions. -func (f *fileInodeOperations) SetPermissions(ctx context.Context, inode *fs.Inode, p fs.FilePermissions) bool { +func (f *fileInodeOperations) SetPermissions(ctx context.Context, _ *fs.Inode, p fs.FilePermissions) bool { f.attrMu.Lock() - defer f.attrMu.Unlock() - return f.attr.SetPermissions(ctx, p) + f.attr.SetPermissions(ctx, p) + f.attrMu.Unlock() + return true } // SetTimestamps implements fs.InodeOperations.SetTimestamps. -func (f *fileInodeOperations) SetTimestamps(ctx context.Context, inode *fs.Inode, ts fs.TimeSpec) error { +func (f *fileInodeOperations) SetTimestamps(ctx context.Context, _ *fs.Inode, ts fs.TimeSpec) error { f.attrMu.Lock() - defer f.attrMu.Unlock() - return f.attr.SetTimestamps(ctx, ts) + f.attr.SetTimestamps(ctx, ts) + f.attrMu.Unlock() + return nil } // SetOwner implements fs.InodeOperations.SetOwner. -func (f *fileInodeOperations) SetOwner(ctx context.Context, inode *fs.Inode, owner fs.FileOwner) error { +func (f *fileInodeOperations) SetOwner(ctx context.Context, _ *fs.Inode, owner fs.FileOwner) error { f.attrMu.Lock() - defer f.attrMu.Unlock() - return f.attr.SetOwner(ctx, owner) + f.attr.SetOwner(ctx, owner) + f.attrMu.Unlock() + return nil } // Truncate implements fs.InodeOperations.Truncate. -func (f *fileInodeOperations) Truncate(ctx context.Context, inode *fs.Inode, size int64) error { +func (f *fileInodeOperations) Truncate(ctx context.Context, _ *fs.Inode, size int64) error { f.attrMu.Lock() defer f.attrMu.Unlock() f.dataMu.Lock() - oldSize := f.attr.Unstable.Size + oldSize := f.attr.Size if oldSize != size { - f.attr.Unstable.Size = size - f.attr.TouchModificationTime(ctx) + f.attr.Size = size + // Update mtime and ctime. + now := ktime.NowFromContext(ctx) + f.attr.ModificationTime = now + f.attr.StatusChangeTime = now } f.dataMu.Unlock() @@ -220,21 +208,21 @@ func (f *fileInodeOperations) Truncate(ctx context.Context, inode *fs.Inode, siz // AddLink implements fs.InodeOperations.AddLink. func (f *fileInodeOperations) AddLink() { f.attrMu.Lock() - f.attr.Unstable.Links++ + f.attr.Links++ f.attrMu.Unlock() } // DropLink implements fs.InodeOperations.DropLink. func (f *fileInodeOperations) DropLink() { f.attrMu.Lock() - f.attr.Unstable.Links-- + f.attr.Links-- f.attrMu.Unlock() } // NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange. func (f *fileInodeOperations) NotifyStatusChange(ctx context.Context) { f.attrMu.Lock() - f.attr.TouchStatusChangeTime(ctx) + f.attr.StatusChangeTime = ktime.NowFromContext(ctx) f.attrMu.Unlock() } @@ -264,7 +252,7 @@ func (f *fileInodeOperations) read(ctx context.Context, dst usermem.IOSequence, // TODO: Separate out f.attr.Size and use atomics instead of // f.dataMu. f.dataMu.RLock() - size := f.attr.Unstable.Size + size := f.attr.Size f.dataMu.RUnlock() if offset >= size { return 0, io.EOF @@ -273,7 +261,7 @@ func (f *fileInodeOperations) read(ctx context.Context, dst usermem.IOSequence, n, err := dst.CopyOutFrom(ctx, &fileReadWriter{f, offset}) // Compare Linux's mm/filemap.c:do_generic_file_read() => file_accessed(). f.attrMu.Lock() - f.attr.TouchAccessTime(ctx) + f.attr.AccessTime = ktime.NowFromContext(ctx) f.attrMu.Unlock() return n, err } @@ -287,7 +275,9 @@ func (f *fileInodeOperations) write(ctx context.Context, src usermem.IOSequence, f.attrMu.Lock() defer f.attrMu.Unlock() // Compare Linux's mm/filemap.c:__generic_file_write_iter() => file_update_time(). - f.attr.TouchModificationTime(ctx) + now := ktime.NowFromContext(ctx) + f.attr.ModificationTime = now + f.attr.StatusChangeTime = now return src.CopyInTo(ctx, &fileReadWriter{f, offset}) } @@ -302,10 +292,10 @@ func (rw *fileReadWriter) ReadToBlocks(dsts safemem.BlockSeq) (uint64, error) { defer rw.f.dataMu.RUnlock() // Compute the range to read. - if rw.offset >= rw.f.attr.Unstable.Size { + if rw.offset >= rw.f.attr.Size { return 0, io.EOF } - end := fs.ReadEndOffset(rw.offset, int64(dsts.NumBytes()), rw.f.attr.Unstable.Size) + end := fs.ReadEndOffset(rw.offset, int64(dsts.NumBytes()), rw.f.attr.Size) if end == rw.offset { // dsts.NumBytes() == 0? return 0, nil } @@ -371,8 +361,8 @@ func (rw *fileReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error) defer func() { // If the write ends beyond the file's previous size, it causes the // file to grow. - if rw.offset > rw.f.attr.Unstable.Size { - rw.f.attr.Unstable.Size = rw.offset + if rw.offset > rw.f.attr.Size { + rw.f.attr.Size = rw.offset } }() @@ -450,9 +440,9 @@ func (f *fileInodeOperations) Translate(ctx context.Context, required, optional f.dataMu.Lock() defer f.dataMu.Unlock() - // Constrain translations to f.attr.Unstable.Size (rounded up) to prevent + // Constrain translations to f.attr.Size (rounded up) to prevent // translation to pages that may be concurrently truncated. - pgend := fs.OffsetPageEnd(f.attr.Unstable.Size) + pgend := fs.OffsetPageEnd(f.attr.Size) var beyondEOF bool if required.End > pgend { if required.Start >= pgend { diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 40a8c4b1e..a0277a132 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -19,12 +19,14 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/pipe" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" ) var fsInfo = fs.Info{ @@ -39,32 +41,54 @@ var fsInfo = fs.Info{ func rename(ctx context.Context, oldParent *fs.Inode, oldName string, newParent *fs.Inode, newName string) error { op, ok := oldParent.InodeOperations.(*Dir) if !ok { - return ramfs.ErrCrossDevice + return syserror.EXDEV } np, ok := newParent.InodeOperations.(*Dir) if !ok { - return ramfs.ErrCrossDevice + return syserror.EXDEV } - return ramfs.Rename(ctx, &op.Dir, oldName, &np.Dir, newName) + return ramfs.Rename(ctx, op.ramfsDir, oldName, np.ramfsDir, newName) } // Dir is a directory. // // +stateify savable type Dir struct { - ramfs.Dir + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeIsDirTruncate `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + // Ideally this would be embedded, so that we "inherit" all of the + // InodeOperations implemented by ramfs.Dir for free. + // + // However, ramfs.dirFileOperations stores a pointer to a ramfs.Dir, + // and our save/restore package does not allow saving a pointer to an + // embedded field elsewhere. + // + // Thus, we must make the ramfs.Dir is a field, and we delegate all the + // InodeOperation methods to it. + ramfsDir *ramfs.Dir // kernel is used to allocate platform memory as storage for tmpfs Files. kernel *kernel.Kernel } +var _ fs.InodeOperations = (*Dir)(nil) + // NewDir returns a new directory. func NewDir(ctx context.Context, contents map[string]*fs.Inode, owner fs.FileOwner, perms fs.FilePermissions, msrc *fs.MountSource, kernel *kernel.Kernel) *fs.Inode { - d := &Dir{kernel: kernel} - d.InitDir(ctx, contents, owner, perms) + d := &Dir{ + ramfsDir: ramfs.NewDir(ctx, contents, owner, perms), + kernel: kernel, + } // Manually set the CreateOps. - d.CreateOps = d.newCreateOps() + d.ramfsDir.CreateOps = d.newCreateOps() return fs.NewInode(d, msrc, fs.StableAttr{ DeviceID: tmpfsDevice.DeviceID(), @@ -77,7 +101,107 @@ func NewDir(ctx context.Context, contents map[string]*fs.Inode, owner fs.FileOwn // afterLoad is invoked by stateify. func (d *Dir) afterLoad() { // Per NewDir, manually set the CreateOps. - d.Dir.CreateOps = d.newCreateOps() + d.ramfsDir.CreateOps = d.newCreateOps() +} + +// GetFile implements fs.InodeOperations.GetFile. +func (d *Dir) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { + return d.ramfsDir.GetFile(ctx, dirent, flags) +} + +// AddLink implements fs.InodeOperations.AddLink. +func (d *Dir) AddLink() { + d.ramfsDir.AddLink() +} + +// DropLink implements fs.InodeOperations.DropLink. +func (d *Dir) DropLink() { + d.ramfsDir.DropLink() +} + +// Bind implements fs.InodeOperations.Bind. +func (d *Dir) Bind(ctx context.Context, dir *fs.Inode, name string, ep transport.BoundEndpoint, perms fs.FilePermissions) (*fs.Dirent, error) { + return d.ramfsDir.Bind(ctx, dir, name, ep, perms) +} + +// Create implements fs.InodeOperations.Create. +func (d *Dir) Create(ctx context.Context, dir *fs.Inode, name string, flags fs.FileFlags, perms fs.FilePermissions) (*fs.File, error) { + return d.ramfsDir.Create(ctx, dir, name, flags, perms) +} + +// CreateLink implements fs.InodeOperations.CreateLink. +func (d *Dir) CreateLink(ctx context.Context, dir *fs.Inode, oldname, newname string) error { + return d.ramfsDir.CreateLink(ctx, dir, oldname, newname) +} + +// CreateHardLink implements fs.InodeOperations.CreateHardLink. +func (d *Dir) CreateHardLink(ctx context.Context, dir *fs.Inode, target *fs.Inode, name string) error { + return d.ramfsDir.CreateHardLink(ctx, dir, target, name) +} + +// CreateDirectory implements fs.InodeOperations.CreateDirectory. +func (d *Dir) CreateDirectory(ctx context.Context, dir *fs.Inode, name string, perms fs.FilePermissions) error { + return d.ramfsDir.CreateDirectory(ctx, dir, name, perms) +} + +// CreateFifo implements fs.InodeOperations.CreateFifo. +func (d *Dir) CreateFifo(ctx context.Context, dir *fs.Inode, name string, perms fs.FilePermissions) error { + return d.ramfsDir.CreateFifo(ctx, dir, name, perms) +} + +// Getxattr implements fs.InodeOperations.Getxattr. +func (d *Dir) Getxattr(i *fs.Inode, name string) ([]byte, error) { + return d.ramfsDir.Getxattr(i, name) +} + +// Setxattr implements fs.InodeOperations.Setxattr. +func (d *Dir) Setxattr(i *fs.Inode, name string, value []byte) error { + return d.ramfsDir.Setxattr(i, name, value) +} + +// Listxattr implements fs.InodeOperations.Listxattr. +func (d *Dir) Listxattr(i *fs.Inode) (map[string]struct{}, error) { + return d.ramfsDir.Listxattr(i) +} + +// Lookup implements fs.InodeOperations.Lookup. +func (d *Dir) Lookup(ctx context.Context, i *fs.Inode, p string) (*fs.Dirent, error) { + return d.ramfsDir.Lookup(ctx, i, p) +} + +// NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange. +func (d *Dir) NotifyStatusChange(ctx context.Context) { + d.ramfsDir.NotifyStatusChange(ctx) +} + +// Remove implements fs.InodeOperations.Remove. +func (d *Dir) Remove(ctx context.Context, i *fs.Inode, name string) error { + return d.ramfsDir.Remove(ctx, i, name) +} + +// RemoveDirectory implements fs.InodeOperations.RemoveDirectory. +func (d *Dir) RemoveDirectory(ctx context.Context, i *fs.Inode, name string) error { + return d.ramfsDir.RemoveDirectory(ctx, i, name) +} + +// UnstableAttr implements fs.InodeOperations.UnstableAttr. +func (d *Dir) UnstableAttr(ctx context.Context, i *fs.Inode) (fs.UnstableAttr, error) { + return d.ramfsDir.UnstableAttr(ctx, i) +} + +// SetPermissions implements fs.InodeOperations.SetPermissions. +func (d *Dir) SetPermissions(ctx context.Context, i *fs.Inode, p fs.FilePermissions) bool { + return d.ramfsDir.SetPermissions(ctx, i, p) +} + +// SetOwner implements fs.InodeOperations.SetOwner. +func (d *Dir) SetOwner(ctx context.Context, i *fs.Inode, owner fs.FileOwner) error { + return d.ramfsDir.SetOwner(ctx, i, owner) +} + +// SetTimestamps implements fs.InodeOperations.SetTimestamps. +func (d *Dir) SetTimestamps(ctx context.Context, i *fs.Inode, ts fs.TimeSpec) error { + return d.ramfsDir.SetTimestamps(ctx, i, ts) } // newCreateOps builds the custom CreateOps for this Dir. @@ -132,8 +256,7 @@ type Symlink struct { // NewSymlink returns a new symlink with the provided permissions. func NewSymlink(ctx context.Context, target string, owner fs.FileOwner, msrc *fs.MountSource) *fs.Inode { - s := &Symlink{} - s.InitSymlink(ctx, owner, target) + s := &Symlink{Symlink: *ramfs.NewSymlink(ctx, owner, target)} return fs.NewInode(s, msrc, fs.StableAttr{ DeviceID: tmpfsDevice.DeviceID(), InodeID: tmpfsDevice.NextIno(), @@ -157,12 +280,12 @@ func (s *Symlink) StatFS(context.Context) (fs.Info, error) { // +stateify savable type Socket struct { ramfs.Socket + fsutil.InodeNotTruncatable `state:"nosave"` } // NewSocket returns a new socket with the provided permissions. func NewSocket(ctx context.Context, socket transport.BoundEndpoint, owner fs.FileOwner, perms fs.FilePermissions, msrc *fs.MountSource) *fs.Inode { - s := &Socket{} - s.InitSocket(ctx, socket, owner, perms) + s := &Socket{Socket: *ramfs.NewSocket(ctx, socket, owner, perms)} return fs.NewInode(s, msrc, fs.StableAttr{ DeviceID: tmpfsDevice.DeviceID(), InodeID: tmpfsDevice.NextIno(), @@ -185,15 +308,22 @@ func (s *Socket) StatFS(context.Context) (fs.Info, error) { // // +stateify savable type Fifo struct { - ramfs.Entry + fs.InodeOperations } // NewFifo creates a new named pipe. func NewFifo(ctx context.Context, owner fs.FileOwner, perms fs.FilePermissions, msrc *fs.MountSource) *fs.Inode { - f := &Fifo{} - f.InitEntry(ctx, owner, perms) - iops := pipe.NewInodeOperations(f, pipe.NewPipe(ctx, true /* isNamed */, pipe.DefaultPipeSize, usermem.PageSize)) - return fs.NewInode(iops, msrc, fs.StableAttr{ + // First create a pipe. + p := pipe.NewPipe(ctx, true /* isNamed */, pipe.DefaultPipeSize, usermem.PageSize) + + // Build pipe InodeOperations. + iops := pipe.NewInodeOperations(ctx, perms, p) + + // Wrap the iops with our Fifo. + fifoIops := &Fifo{iops} + + // Build a new Inode. + return fs.NewInode(fifoIops, msrc, fs.StableAttr{ DeviceID: tmpfsDevice.DeviceID(), InodeID: tmpfsDevice.NextIno(), BlockSize: usermem.PageSize, diff --git a/pkg/sentry/fs/tty/BUILD b/pkg/sentry/fs/tty/BUILD index 2b45069a6..011cb6955 100644 --- a/pkg/sentry/fs/tty/BUILD +++ b/pkg/sentry/fs/tty/BUILD @@ -7,7 +7,6 @@ go_library( srcs = [ "dir.go", "fs.go", - "inode.go", "line_discipline.go", "master.go", "queue.go", @@ -25,7 +24,6 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel/auth", - "//pkg/sentry/kernel/time", "//pkg/sentry/socket/unix/transport", "//pkg/sentry/unimpl", "//pkg/sentry/usermem", diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index e32b05c1d..485cdb456 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -52,13 +52,17 @@ import ( // // +stateify savable type dirInodeOperations struct { - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotRenameable `state:"nosave"` fsutil.InodeNotSymlink `state:"nosave"` - fsutil.InodeNoExtendedAttributes `state:"nosave"` - fsutil.NoMappable `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotTruncatable `state:"nosave"` + fsutil.InodeVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes // msrc is the super block this directory is on. // @@ -68,9 +72,6 @@ type dirInodeOperations struct { // mu protects the fields below. mu sync.Mutex `state:"nosave"` - // attr contains the UnstableAttrs. - attr fsutil.InMemoryAttributes - // master is the master PTY inode. master *fs.Inode @@ -97,15 +98,10 @@ var _ fs.InodeOperations = (*dirInodeOperations)(nil) // newDir creates a new dir with a ptmx file and no terminals. func newDir(ctx context.Context, m *fs.MountSource) *fs.Inode { d := &dirInodeOperations{ - attr: fsutil.InMemoryAttributes{ - Unstable: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: fs.RootOwner, - Perms: fs.FilePermsFromMode(0555), - }), - }, - msrc: m, - slaves: make(map[uint32]*fs.Inode), - dentryMap: fs.NewSortedDentryMap(nil), + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, fs.RootOwner, fs.FilePermsFromMode(0555), linux.DEVPTS_SUPER_MAGIC), + msrc: m, + slaves: make(map[uint32]*fs.Inode), + dentryMap: fs.NewSortedDentryMap(nil), } // Linux devpts uses a default mode of 0000 for ptmx which can be // changed with the ptmxmode mount option. However, that default is not @@ -224,70 +220,6 @@ func (d *dirInodeOperations) GetFile(ctx context.Context, dirent *fs.Dirent, fla return fs.NewFile(ctx, dirent, flags, &dirFileOperations{di: d}), nil } -// UnstableAttr implements fs.InodeOperations.UnstableAttr. -func (d *dirInodeOperations) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - d.mu.Lock() - defer d.mu.Unlock() - return d.attr.Unstable, nil -} - -// Check implements fs.InodeOperations.Check. -func (d *dirInodeOperations) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { - return fs.ContextCanAccessFile(ctx, inode, p) -} - -// SetPermissions implements fs.InodeOperations.SetPermissions. -func (d *dirInodeOperations) SetPermissions(ctx context.Context, inode *fs.Inode, p fs.FilePermissions) bool { - d.mu.Lock() - defer d.mu.Unlock() - return d.attr.SetPermissions(ctx, p) -} - -// SetOwner implements fs.InodeOperations.SetOwner. -func (d *dirInodeOperations) SetOwner(ctx context.Context, inode *fs.Inode, owner fs.FileOwner) error { - d.mu.Lock() - defer d.mu.Unlock() - return d.attr.SetOwner(ctx, owner) -} - -// SetTimestamps implements fs.InodeOperations.SetTimestamps. -func (d *dirInodeOperations) SetTimestamps(ctx context.Context, inode *fs.Inode, ts fs.TimeSpec) error { - d.mu.Lock() - defer d.mu.Unlock() - return d.attr.SetTimestamps(ctx, ts) -} - -// Truncate implements fs.InodeOperations.Truncate. -func (d *dirInodeOperations) Truncate(ctx context.Context, inode *fs.Inode, size int64) error { - return syserror.EINVAL -} - -// AddLink implements fs.InodeOperations.AddLink. -func (d *dirInodeOperations) AddLink() {} - -// DropLink implements fs.InodeOperations.DropLink. -func (d *dirInodeOperations) DropLink() {} - -// NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange. -func (d *dirInodeOperations) NotifyStatusChange(ctx context.Context) { - d.mu.Lock() - defer d.mu.Unlock() - - d.attr.TouchStatusChangeTime(ctx) -} - -// IsVirtual implements fs.InodeOperations.IsVirtual. -func (d *dirInodeOperations) IsVirtual() bool { - return true -} - -// StatFS implements fs.InodeOperations.StatFS. -func (d *dirInodeOperations) StatFS(ctx context.Context) (fs.Info, error) { - return fs.Info{ - Type: linux.DEVPTS_SUPER_MAGIC, - }, nil -} - // allocateTerminal creates a new Terminal and installs a pts node for it. // // The caller must call DecRef when done with the returned Terminal. @@ -353,13 +285,13 @@ func (d *dirInodeOperations) masterClose(t *Terminal) { // // +stateify savable type dirFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.NoopRelease `state:"nosave"` - fsutil.GenericSeek `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` - fsutil.NoIoctl `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` // di is the inode operations. di *dirInodeOperations diff --git a/pkg/sentry/fs/tty/inode.go b/pkg/sentry/fs/tty/inode.go deleted file mode 100644 index d5d1caafc..000000000 --- a/pkg/sentry/fs/tty/inode.go +++ /dev/null @@ -1,145 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package tty - -import ( - "sync" - - "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/context" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" - ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" - "gvisor.googlesource.com/gvisor/pkg/syserror" -) - -// inodeOperations are the base fs.InodeOperations for master and slave Inodes. -// -// inodeOperations does not implement: -// -// * fs.InodeOperations.Release -// * fs.InodeOperations.GetFile -// -// +stateify savable -type inodeOperations struct { - fsutil.DeprecatedFileOperations `state:"nosave"` - fsutil.InodeNoExtendedAttributes `state:"nosave"` - fsutil.InodeNotDirectory `state:"nosave"` - fsutil.InodeNotRenameable `state:"nosave"` - fsutil.InodeNotSocket `state:"nosave"` - fsutil.InodeNotSymlink `state:"nosave"` - fsutil.NoMappable `state:"nosave"` - fsutil.NoopWriteOut `state:"nosave"` - - // mu protects the fields below. - mu sync.Mutex `state:"nosave"` - - // uattr is the inode's UnstableAttr. - uattr fs.UnstableAttr -} - -// UnstableAttr implements fs.InodeOperations.UnstableAttr. -func (i *inodeOperations) UnstableAttr(ctx context.Context, inode *fs.Inode) (fs.UnstableAttr, error) { - i.mu.Lock() - defer i.mu.Unlock() - return i.uattr, nil -} - -// Check implements fs.InodeOperations.Check. -func (i *inodeOperations) Check(ctx context.Context, inode *fs.Inode, p fs.PermMask) bool { - return fs.ContextCanAccessFile(ctx, inode, p) -} - -// SetPermissions implements fs.InodeOperations.SetPermissions -func (i *inodeOperations) SetPermissions(ctx context.Context, inode *fs.Inode, p fs.FilePermissions) bool { - i.mu.Lock() - defer i.mu.Unlock() - i.uattr.Perms = p - i.uattr.StatusChangeTime = ktime.NowFromContext(ctx) - return true -} - -// SetOwner implements fs.InodeOperations.SetOwner. -func (i *inodeOperations) SetOwner(ctx context.Context, inode *fs.Inode, owner fs.FileOwner) error { - i.mu.Lock() - defer i.mu.Unlock() - if owner.UID.Ok() { - i.uattr.Owner.UID = owner.UID - } - if owner.GID.Ok() { - i.uattr.Owner.GID = owner.GID - } - return nil -} - -// SetTimestamps implements fs.InodeOperations.SetTimestamps. -func (i *inodeOperations) SetTimestamps(ctx context.Context, inode *fs.Inode, ts fs.TimeSpec) error { - if ts.ATimeOmit && ts.MTimeOmit { - return nil - } - - i.mu.Lock() - defer i.mu.Unlock() - - now := ktime.NowFromContext(ctx) - if !ts.ATimeOmit { - if ts.ATime.IsZero() { - i.uattr.AccessTime = now - } else { - i.uattr.AccessTime = ts.ATime - } - } - if !ts.MTimeOmit { - if ts.MTime.IsZero() { - i.uattr.ModificationTime = now - } else { - i.uattr.ModificationTime = ts.MTime - } - } - i.uattr.StatusChangeTime = now - return nil -} - -// Truncate implements fs.InodeOperations.Truncate. -func (i *inodeOperations) Truncate(ctx context.Context, inode *fs.Inode, size int64) error { - return syserror.EINVAL -} - -// AddLink implements fs.InodeOperations.AddLink. -func (i *inodeOperations) AddLink() { -} - -// DropLink implements fs.InodeOperations.DropLink. -func (i *inodeOperations) DropLink() { -} - -// NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange. -func (i *inodeOperations) NotifyStatusChange(ctx context.Context) { - i.mu.Lock() - defer i.mu.Unlock() - i.uattr.StatusChangeTime = ktime.NowFromContext(ctx) -} - -// IsVirtual implements fs.InodeOperations.IsVirtual. -func (i *inodeOperations) IsVirtual() bool { - return true -} - -// StatFS implements fs.InodeOperations.StatFS. -func (i *inodeOperations) StatFS(ctx context.Context) (fs.Info, error) { - return fs.Info{ - Type: linux.DEVPTS_SUPER_MAGIC, - }, nil -} diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index 00bec4c2c..b5e13ab36 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -31,7 +31,7 @@ import ( // // +stateify savable type masterInodeOperations struct { - inodeOperations + fsutil.SimpleFileInode // d is the containing dir. d *dirInodeOperations @@ -42,15 +42,8 @@ var _ fs.InodeOperations = (*masterInodeOperations)(nil) // newMasterInode creates an Inode for the master end of a terminal. func newMasterInode(ctx context.Context, d *dirInodeOperations, owner fs.FileOwner, p fs.FilePermissions) *fs.Inode { iops := &masterInodeOperations{ - inodeOperations: inodeOperations{ - uattr: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: owner, - Perms: p, - Links: 1, - // Size and Blocks are always 0. - }), - }, - d: d, + SimpleFileInode: *fsutil.NewSimpleFileInode(ctx, owner, p, linux.DEVPTS_SUPER_MAGIC), + d: d, } return fs.NewInode(iops, d.msrc, fs.StableAttr{ @@ -102,11 +95,11 @@ func (mi *masterInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flag // // +stateify savable type masterFileOperations struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` // d is the containing dir. d *dirInodeOperations diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index a696fbb51..6dbce90b4 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -30,7 +30,7 @@ import ( // // +stateify savable type slaveInodeOperations struct { - inodeOperations + fsutil.SimpleFileInode // d is the containing dir. d *dirInodeOperations @@ -46,16 +46,9 @@ var _ fs.InodeOperations = (*slaveInodeOperations)(nil) // newSlaveInode takes ownership of t. func newSlaveInode(ctx context.Context, d *dirInodeOperations, t *Terminal, owner fs.FileOwner, p fs.FilePermissions) *fs.Inode { iops := &slaveInodeOperations{ - inodeOperations: inodeOperations{ - uattr: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: owner, - Perms: p, - Links: 1, - // Size and Blocks are always 0. - }), - }, - d: d, - t: t, + SimpleFileInode: *fsutil.NewSimpleFileInode(ctx, owner, p, linux.DEVPTS_SUPER_MAGIC), + d: d, + t: t, } return fs.NewInode(iops, d.msrc, fs.StableAttr{ @@ -91,11 +84,11 @@ func (si *slaveInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags // // +stateify savable type slaveFileOperations struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` // si is the inode operations. si *slaveInodeOperations diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index 9c13ecfcc..502395f18 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -99,12 +99,12 @@ func (p *pollEntry) WeakRefGone() { // // +stateify savable type EventPoll struct { - fsutil.PipeSeek `state:"zerovalue"` - fsutil.NotDirReaddir `state:"zerovalue"` - fsutil.NoFsync `state:"zerovalue"` - fsutil.NoopFlush `state:"zerovalue"` - fsutil.NoMMap `state:"zerovalue"` - fsutil.NoIoctl `state:"zerovalue"` + fsutil.FilePipeSeek `state:"zerovalue"` + fsutil.FileNotDirReaddir `state:"zerovalue"` + fsutil.FileNoFsync `state:"zerovalue"` + fsutil.FileNoopFlush `state:"zerovalue"` + fsutil.FileNoMMap `state:"zerovalue"` + fsutil.FileNoIoctl `state:"zerovalue"` // Wait queue is used to notify interested parties when the event poll // object itself becomes readable or writable. diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index 063a1d5f5..2d43c986d 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -38,13 +38,13 @@ import ( // // +stateify savable type EventOperations struct { - fsutil.NoopRelease `state:"nosave"` - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` - fsutil.NoIoctl `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` // Mutex that protects accesses to the fields of this event. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index 4b0e00b85..1336b6293 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -17,17 +17,30 @@ package pipe import ( "sync" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/amutex" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/syserror" ) -// inodeOperations wraps fs.InodeOperations operations with common pipe opening semantics. +// inodeOperations implements fs.InodeOperations for pipes. // // +stateify savable type inodeOperations struct { - fs.InodeOperations + fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopRelease `state:"nosave"` + fsutil.InodeNoopTruncate `state:"nosave"` + fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotDirectory `state:"nosave"` + fsutil.InodeNotMappable `state:"nosave"` + fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` + fsutil.InodeNotVirtual `state:"nosave"` + + fsutil.InodeSimpleAttributes // mu protects the fields below. mu sync.Mutex `state:"nosave"` @@ -46,12 +59,15 @@ type inodeOperations struct { wWakeup chan struct{} `state:"nosave"` } -// NewInodeOperations creates a new pipe fs.InodeOperations. -func NewInodeOperations(base fs.InodeOperations, p *Pipe) fs.InodeOperations { +var _ fs.InodeOperations = (*inodeOperations)(nil) + +// NewInodeOperations returns a new fs.InodeOperations for a given pipe. +func NewInodeOperations(ctx context.Context, perms fs.FilePermissions, p *Pipe) *inodeOperations { return &inodeOperations{ - InodeOperations: base, - p: p, + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, fs.FileOwnerFromContext(ctx), perms, linux.PIPEFS_MAGIC), + p: p, } + } // GetFile implements fs.InodeOperations.GetFile. Named pipes have special blocking @@ -164,18 +180,6 @@ func (i *inodeOperations) waitFor(wakeupChan *chan struct{}, sleeper amutex.Slee } } -// Truncate implements fs.InodeOperations.Truncate -// -// This method is required to override the default i.InodeOperations.Truncate -// which may return ErrInvalidOperation, this allows open related -// syscalls to set the O_TRUNC flag without returning an error by -// calling Truncate directly during openat. The ftruncate and truncate -// system calls will check that the file is an actual file and return -// EINVAL because it's a PIPE, making this behavior consistent with linux. -func (i *inodeOperations) Truncate(context.Context, *fs.Inode, int64) error { - return nil -} - // newHandleLocked signals a new pipe reader or writer depending on where // 'wakeupChan' points. This unblocks any corresponding reader or writer // waiting for the other end of the channel to be opened, see Fifo.waitFor. diff --git a/pkg/sentry/kernel/pipe/node_test.go b/pkg/sentry/kernel/pipe/node_test.go index eda551594..ad103b195 100644 --- a/pkg/sentry/kernel/pipe/node_test.go +++ b/pkg/sentry/kernel/pipe/node_test.go @@ -53,6 +53,10 @@ type openResult struct { error } +var perms fs.FilePermissions = fs.FilePermissions{ + User: fs.PermMask{Read: true, Write: true}, +} + func testOpenOrDie(ctx context.Context, t *testing.T, n fs.InodeOperations, flags fs.FileFlags, doneChan chan<- struct{}) (*fs.File, error) { file, err := n.GetFile(ctx, nil, flags) if err != nil { @@ -93,8 +97,8 @@ func assertRecvBlocks(t *testing.T, c <-chan struct{}, blockDuration time.Durati } func TestReadOpenBlocksForWriteOpen(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) rDone := make(chan struct{}) go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone) @@ -111,8 +115,8 @@ func TestReadOpenBlocksForWriteOpen(t *testing.T) { } func TestWriteOpenBlocksForReadOpen(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) wDone := make(chan struct{}) go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) @@ -129,8 +133,8 @@ func TestWriteOpenBlocksForReadOpen(t *testing.T) { } func TestMultipleWriteOpenDoesntCountAsReadOpen(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) rDone1 := make(chan struct{}) rDone2 := make(chan struct{}) @@ -151,8 +155,8 @@ func TestMultipleWriteOpenDoesntCountAsReadOpen(t *testing.T) { } func TestClosedReaderBlocksWriteOpen(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) rFile, _ := testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true, NonBlocking: true}, nil) rFile.DecRef() @@ -172,8 +176,8 @@ func TestClosedReaderBlocksWriteOpen(t *testing.T) { } func TestReadWriteOpenNeverBlocks(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) rwDone := make(chan struct{}) // Open for read-write never wait for a reader or writer, even if the @@ -183,8 +187,8 @@ func TestReadWriteOpenNeverBlocks(t *testing.T) { } func TestReadWriteOpenUnblocksReadOpen(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) rDone := make(chan struct{}) go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone) @@ -197,8 +201,8 @@ func TestReadWriteOpenUnblocksReadOpen(t *testing.T) { } func TestReadWriteOpenUnblocksWriteOpen(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) wDone := make(chan struct{}) go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) @@ -211,8 +215,8 @@ func TestReadWriteOpenUnblocksWriteOpen(t *testing.T) { } func TestBlockedOpenIsCancellable(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) done := make(chan openResult) go testOpen(ctx, t, f, fs.FileFlags{Read: true}, done) @@ -233,18 +237,18 @@ func TestBlockedOpenIsCancellable(t *testing.T) { } } -func TestNonblockingReadOpenNoWriters(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) +func TestNonblockingReadOpenFileNoWriters(t *testing.T) { ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) if _, err := testOpen(ctx, t, f, fs.FileFlags{Read: true, NonBlocking: true}, nil); err != nil { t.Fatalf("Nonblocking open for read failed with error %v.", err) } } -func TestNonblockingWriteOpenNoReaders(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) +func TestNonblockingWriteOpenFileNoReaders(t *testing.T) { ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) if _, err := testOpen(ctx, t, f, fs.FileFlags{Write: true, NonBlocking: true}, nil); err != syserror.ENXIO { t.Fatalf("Nonblocking open for write failed unexpected error %v.", err) @@ -252,8 +256,8 @@ func TestNonblockingWriteOpenNoReaders(t *testing.T) { } func TestNonBlockingReadOpenWithWriter(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) wDone := make(chan struct{}) go testOpenOrDie(ctx, t, f, fs.FileFlags{Write: true}, wDone) @@ -271,8 +275,8 @@ func TestNonBlockingReadOpenWithWriter(t *testing.T) { } func TestNonBlockingWriteOpenWithReader(t *testing.T) { - f := NewInodeOperations(nil, newNamedPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newNamedPipe(t)) rDone := make(chan struct{}) go testOpenOrDie(ctx, t, f, fs.FileFlags{Read: true}, rDone) @@ -290,8 +294,8 @@ func TestNonBlockingWriteOpenWithReader(t *testing.T) { } func TestAnonReadOpen(t *testing.T) { - f := NewInodeOperations(nil, newAnonPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newAnonPipe(t)) if _, err := testOpen(ctx, t, f, fs.FileFlags{Read: true}, nil); err != nil { t.Fatalf("open anon pipe for read failed: %v", err) @@ -299,8 +303,8 @@ func TestAnonReadOpen(t *testing.T) { } func TestAnonWriteOpen(t *testing.T) { - f := NewInodeOperations(nil, newAnonPipe(t)) ctx := newSleeperContext(t) + f := NewInodeOperations(ctx, perms, newAnonPipe(t)) if _, err := testOpen(ctx, t, f, fs.FileFlags{Write: true}, nil); err != nil { t.Fatalf("open anon pipe for write failed: %v", err) diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index 126054826..fad077d2d 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -25,11 +25,9 @@ import ( "sync/atomic" "syscall" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/ilist" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" - "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/pkg/waiter" @@ -50,7 +48,7 @@ type Pipe struct { isNamed bool // The dirent backing this pipe. Shared by all readers and writers. - dirent *fs.Dirent + Dirent *fs.Dirent // The buffered byte queue. data ilist.List @@ -97,28 +95,19 @@ func NewPipe(ctx context.Context, isNamed bool, sizeBytes, atomicIOBytes int) *P // Build the fs.Dirent of this pipe, shared by all fs.Files associated // with this pipe. + perms := fs.FilePermissions{ + User: fs.PermMask{Read: true, Write: true}, + } + iops := NewInodeOperations(ctx, perms, p) ino := pipeDevice.NextIno() - base := fsutil.NewSimpleInodeOperations(fsutil.InodeSimpleAttributes{ - FSType: linux.PIPEFS_MAGIC, - UAttr: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: fs.FileOwnerFromContext(ctx), - Perms: fs.FilePermissions{ - User: fs.PermMask{Read: true, Write: true}, - }, - Links: 1, - }), - }) sattr := fs.StableAttr{ Type: fs.Pipe, DeviceID: pipeDevice.DeviceID(), InodeID: ino, BlockSize: int64(atomicIOBytes), } - // There is no real filesystem backing this pipe, so we pass in a nil - // Filesystem. - sb := fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}) - p.dirent = fs.NewDirent(fs.NewInode(NewInodeOperations(base, p), sb, sattr), fmt.Sprintf("pipe:[%d]", ino)) - + ms := fs.NewPseudoMountSource() + p.Dirent = fs.NewDirent(fs.NewInode(iops, ms, sattr), fmt.Sprintf("pipe:[%d]", ino)) return p } @@ -135,7 +124,7 @@ func NewConnectedPipe(ctx context.Context, sizeBytes int, atomicIOBytes int) (*f // ROpen opens the pipe for reading. func (p *Pipe) ROpen(ctx context.Context) *fs.File { p.rOpen() - return fs.NewFile(ctx, p.dirent, fs.FileFlags{Read: true}, &Reader{ + return fs.NewFile(ctx, p.Dirent, fs.FileFlags{Read: true}, &Reader{ ReaderWriter: ReaderWriter{Pipe: p}, }) } @@ -143,7 +132,7 @@ func (p *Pipe) ROpen(ctx context.Context) *fs.File { // WOpen opens the pipe for writing. func (p *Pipe) WOpen(ctx context.Context) *fs.File { p.wOpen() - return fs.NewFile(ctx, p.dirent, fs.FileFlags{Write: true}, &Writer{ + return fs.NewFile(ctx, p.Dirent, fs.FileFlags{Write: true}, &Writer{ ReaderWriter: ReaderWriter{Pipe: p}, }) } @@ -152,7 +141,7 @@ func (p *Pipe) WOpen(ctx context.Context) *fs.File { func (p *Pipe) RWOpen(ctx context.Context) *fs.File { p.rOpen() p.wOpen() - return fs.NewFile(ctx, p.dirent, fs.FileFlags{Read: true, Write: true}, &ReaderWriter{ + return fs.NewFile(ctx, p.Dirent, fs.FileFlags{Read: true, Write: true}, &ReaderWriter{ Pipe: p, }) } diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index 36be1efc3..028175530 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -35,11 +35,11 @@ import ( // // +stateify savable type ReaderWriter struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` *Pipe } diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index 437cc5da1..c070c7316 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -20,7 +20,6 @@ import ( "io" "gvisor.googlesource.com/gvisor/pkg/abi" - "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" @@ -38,20 +37,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// byteReaderFileOperations implements fs.FileOperations for reading -// from a []byte source. -type byteReader struct { - fsutil.NoopRelease - fsutil.PipeSeek - fsutil.NotDirReaddir - fsutil.NoFsync - fsutil.NoopFlush - fsutil.NoMMap - fsutil.NoIoctl - waiter.AlwaysReady - data []byte -} - type fileContext struct { context.Context } @@ -65,17 +50,34 @@ func (f *fileContext) Value(key interface{}) interface{} { } } +// byteReader implements fs.FileOperations for reading from a []byte source. +type byteReader struct { + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + + data []byte +} + +var _ fs.FileOperations = (*byteReader)(nil) + // newByteReaderFile creates a fake file to read data from. func newByteReaderFile(data []byte) *fs.File { // Create a fake inode. - inode := fs.NewInode(fsutil.NewSimpleInodeOperations(fsutil.InodeSimpleAttributes{ - FSType: linux.ANON_INODE_FS_MAGIC, - }), fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}), fs.StableAttr{ - Type: fs.Anonymous, - DeviceID: anon.PseudoDevice.DeviceID(), - InodeID: anon.PseudoDevice.NextIno(), - BlockSize: usermem.PageSize, - }) + inode := fs.NewInode( + &fsutil.SimpleFileInode{}, + fs.NewPseudoMountSource(), + fs.StableAttr{ + Type: fs.Anonymous, + DeviceID: anon.PseudoDevice.DeviceID(), + InodeID: anon.PseudoDevice.NextIno(), + BlockSize: usermem.PageSize, + }) // Use the fake inode to create a fake dirent. dirent := fs.NewTransientDirent(inode) diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index d65b5f49e..ca865b111 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -138,11 +138,11 @@ type commonEndpoint interface { // // +stateify savable type SocketOperations struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` socket.SendReceiveTimeout *waiter.Queue diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index f3ecb6dc3..2c54e8de2 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -46,11 +46,11 @@ const ( // socketOperations implements fs.FileOperations and socket.Socket for a socket // implemented using a host socket. type socketOperations struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` socket.SendReceiveTimeout fd int // must be O_NONBLOCK diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index 0a7d4772c..5b0c11c84 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -65,11 +65,11 @@ var netlinkSocketDevice = device.NewAnonDevice() // // +stateify savable type Socket struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` socket.SendReceiveTimeout // ports provides netlink port allocation. diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index 8c8ebadb7..13681100e 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -45,11 +45,11 @@ import ( // socketOperations implements fs.FileOperations and socket.Socket for a socket // implemented using a host socket. type socketOperations struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` socket.SendReceiveTimeout fd uint32 // must be O_NONBLOCK diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index 9d4aaeb9d..e28d2c4fa 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -178,18 +178,12 @@ func Pair(t *kernel.Task, family int, stype transport.SockType, protocol int) (* // NewDirent returns a sockfs fs.Dirent that resides on device d. func NewDirent(ctx context.Context, d *device.Device) *fs.Dirent { ino := d.NextIno() - // There is no real filesystem backing this pipe, so we pass in a nil - // Filesystem. - inode := fs.NewInode(fsutil.NewSimpleInodeOperations(fsutil.InodeSimpleAttributes{ - FSType: linux.SOCKFS_MAGIC, - UAttr: fs.WithCurrentTime(ctx, fs.UnstableAttr{ - Owner: fs.FileOwnerFromContext(ctx), - Perms: fs.FilePermissions{ - User: fs.PermMask{Read: true, Write: true}, - }, - Links: 1, - }), - }), fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}), fs.StableAttr{ + iops := &fsutil.SimpleFileInode{ + InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, fs.FileOwnerFromContext(ctx), fs.FilePermissions{ + User: fs.PermMask{Read: true, Write: true}, + }, linux.SOCKFS_MAGIC), + } + inode := fs.NewInode(iops, fs.NewPseudoMountSource(), fs.StableAttr{ Type: fs.Socket, DeviceID: d.DeviceID(), InodeID: ino, diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index da225eabb..19258e692 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -45,11 +45,11 @@ import ( // // +stateify savable type SocketOperations struct { - fsutil.PipeSeek `state:"nosave"` - fsutil.NotDirReaddir `state:"nosave"` - fsutil.NoFsync `state:"nosave"` - fsutil.NoopFlush `state:"nosave"` - fsutil.NoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` refs.AtomicRefCount socket.SendReceiveTimeout diff --git a/runsc/boot/fs.go b/runsc/boot/fs.go index 1e75b0efc..942315d6e 100644 --- a/runsc/boot/fs.go +++ b/runsc/boot/fs.go @@ -489,9 +489,7 @@ func mustFindFilesystem(name string) fs.Filesystem { // addSubmountOverlay overlays the inode over a ramfs tree containing the given // paths. func addSubmountOverlay(ctx context.Context, inode *fs.Inode, submounts []string) (*fs.Inode, error) { - // There is no real filesystem backing this ramfs tree, so we pass in - // "nil" here. - msrc := fs.NewNonCachingMountSource(nil, fs.MountSourceFlags{}) + msrc := fs.NewPseudoMountSource() mountTree, err := ramfs.MakeDirectoryTree(ctx, msrc, submounts) if err != nil { return nil, fmt.Errorf("error creating mount tree: %v", err) diff --git a/test/syscalls/linux/proc.cc b/test/syscalls/linux/proc.cc index e64df97b0..6ffe9aed6 100644 --- a/test/syscalls/linux/proc.cc +++ b/test/syscalls/linux/proc.cc @@ -334,6 +334,11 @@ int ReadlinkWhileExited(std::string const& basename, char* buf, size_t count) { return ret; } +TEST(ProcTest, NotFoundInRoot) { + struct stat s; + EXPECT_THAT(stat("/proc/foobar", &s), SyscallFailsWithErrno(ENOENT)); +} + TEST(ProcSelfTest, IsThreadGroupLeader) { ScopedThread([] { const pid_t tgid = getpid(); -- cgit v1.2.3 From 2a0c69b19f4b55c3f9777f0098a72af123ccff3c Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Thu, 31 Jan 2019 11:11:44 -0800 Subject: Remove license comments Nothing reads them and they can simply get stale. Generated with: $ sed -i "s/licenses(\(.*\)).*/licenses(\1)/" **/BUILD PiperOrigin-RevId: 231818945 Change-Id: Ibc3f9838546b7e94f13f217060d31f4ada9d4bf0 --- pkg/abi/BUILD | 2 +- pkg/abi/linux/BUILD | 2 +- pkg/amutex/BUILD | 2 +- pkg/atomicbitops/BUILD | 2 +- pkg/binary/BUILD | 2 +- pkg/bits/BUILD | 2 +- pkg/bpf/BUILD | 2 +- pkg/compressio/BUILD | 2 +- pkg/control/client/BUILD | 2 +- pkg/control/server/BUILD | 2 +- pkg/cpuid/BUILD | 2 +- pkg/dhcp/BUILD | 2 +- pkg/eventchannel/BUILD | 2 +- pkg/fd/BUILD | 2 +- pkg/fdnotifier/BUILD | 2 +- pkg/gate/BUILD | 2 +- pkg/ilist/BUILD | 2 +- pkg/linewriter/BUILD | 2 +- pkg/log/BUILD | 2 +- pkg/metric/BUILD | 2 +- pkg/p9/BUILD | 2 +- pkg/p9/local_server/BUILD | 2 +- pkg/p9/p9test/BUILD | 2 +- pkg/rand/BUILD | 2 +- pkg/refs/BUILD | 2 +- pkg/seccomp/BUILD | 2 +- pkg/secio/BUILD | 2 +- pkg/segment/BUILD | 2 +- pkg/segment/test/BUILD | 2 +- pkg/sentry/BUILD | 2 +- pkg/sentry/arch/BUILD | 2 +- pkg/sentry/context/BUILD | 2 +- pkg/sentry/context/contexttest/BUILD | 2 +- pkg/sentry/control/BUILD | 2 +- pkg/sentry/device/BUILD | 2 +- pkg/sentry/fs/BUILD | 2 +- pkg/sentry/fs/anon/BUILD | 2 +- pkg/sentry/fs/ashmem/BUILD | 2 +- pkg/sentry/fs/binder/BUILD | 2 +- pkg/sentry/fs/dev/BUILD | 2 +- pkg/sentry/fs/fdpipe/BUILD | 2 +- pkg/sentry/fs/filetest/BUILD | 2 +- pkg/sentry/fs/fsutil/BUILD | 2 +- pkg/sentry/fs/gofer/BUILD | 2 +- pkg/sentry/fs/host/BUILD | 2 +- pkg/sentry/fs/lock/BUILD | 2 +- pkg/sentry/fs/proc/BUILD | 2 +- pkg/sentry/fs/proc/device/BUILD | 2 +- pkg/sentry/fs/proc/seqfile/BUILD | 2 +- pkg/sentry/fs/ramfs/BUILD | 2 +- pkg/sentry/fs/sys/BUILD | 2 +- pkg/sentry/fs/timerfd/BUILD | 2 +- pkg/sentry/fs/tmpfs/BUILD | 2 +- pkg/sentry/fs/tty/BUILD | 2 +- pkg/sentry/hostcpu/BUILD | 2 +- pkg/sentry/inet/BUILD | 2 +- pkg/sentry/kernel/BUILD | 2 +- pkg/sentry/kernel/auth/BUILD | 2 +- pkg/sentry/kernel/contexttest/BUILD | 2 +- pkg/sentry/kernel/epoll/BUILD | 2 +- pkg/sentry/kernel/eventfd/BUILD | 2 +- pkg/sentry/kernel/fasync/BUILD | 2 +- pkg/sentry/kernel/futex/BUILD | 2 +- pkg/sentry/kernel/kdefs/BUILD | 2 +- pkg/sentry/kernel/memevent/BUILD | 2 +- pkg/sentry/kernel/pipe/BUILD | 2 +- pkg/sentry/kernel/sched/BUILD | 2 +- pkg/sentry/kernel/semaphore/BUILD | 2 +- pkg/sentry/kernel/shm/BUILD | 2 +- pkg/sentry/kernel/time/BUILD | 2 +- pkg/sentry/limits/BUILD | 2 +- pkg/sentry/loader/BUILD | 2 +- pkg/sentry/memmap/BUILD | 2 +- pkg/sentry/memutil/BUILD | 2 +- pkg/sentry/mm/BUILD | 2 +- pkg/sentry/platform/BUILD | 2 +- pkg/sentry/platform/filemem/BUILD | 2 +- pkg/sentry/platform/interrupt/BUILD | 2 +- pkg/sentry/platform/kvm/BUILD | 2 +- pkg/sentry/platform/kvm/testutil/BUILD | 2 +- pkg/sentry/platform/procid/BUILD | 2 +- pkg/sentry/platform/ptrace/BUILD | 2 +- pkg/sentry/platform/ring0/BUILD | 2 +- pkg/sentry/platform/ring0/gen_offsets/BUILD | 2 +- pkg/sentry/platform/ring0/pagetables/BUILD | 2 +- pkg/sentry/platform/safecopy/BUILD | 2 +- pkg/sentry/safemem/BUILD | 2 +- pkg/sentry/sighandling/BUILD | 2 +- pkg/sentry/socket/BUILD | 2 +- pkg/sentry/socket/control/BUILD | 2 +- pkg/sentry/socket/epsocket/BUILD | 2 +- pkg/sentry/socket/hostinet/BUILD | 2 +- pkg/sentry/socket/netlink/BUILD | 2 +- pkg/sentry/socket/netlink/port/BUILD | 2 +- pkg/sentry/socket/netlink/route/BUILD | 2 +- pkg/sentry/socket/rpcinet/BUILD | 2 +- pkg/sentry/socket/rpcinet/conn/BUILD | 2 +- pkg/sentry/socket/rpcinet/notifier/BUILD | 2 +- pkg/sentry/socket/unix/BUILD | 2 +- pkg/sentry/socket/unix/transport/BUILD | 2 +- pkg/sentry/state/BUILD | 2 +- pkg/sentry/strace/BUILD | 2 +- pkg/sentry/syscalls/BUILD | 2 +- pkg/sentry/syscalls/linux/BUILD | 2 +- pkg/sentry/time/BUILD | 2 +- pkg/sentry/unimpl/BUILD | 2 +- pkg/sentry/uniqueid/BUILD | 2 +- pkg/sentry/usage/BUILD | 2 +- pkg/sentry/usermem/BUILD | 2 +- pkg/sentry/watchdog/BUILD | 2 +- pkg/sleep/BUILD | 2 +- pkg/state/BUILD | 2 +- pkg/state/statefile/BUILD | 2 +- pkg/sync/BUILD | 2 +- pkg/sync/atomicptrtest/BUILD | 2 +- pkg/sync/seqatomictest/BUILD | 2 +- pkg/syserr/BUILD | 2 +- pkg/syserror/BUILD | 2 +- pkg/tcpip/BUILD | 2 +- pkg/tcpip/adapters/gonet/BUILD | 2 +- pkg/tcpip/buffer/BUILD | 2 +- pkg/tcpip/checker/BUILD | 2 +- pkg/tcpip/hash/jenkins/BUILD | 2 +- pkg/tcpip/header/BUILD | 2 +- pkg/tcpip/link/channel/BUILD | 2 +- pkg/tcpip/link/fdbased/BUILD | 2 +- pkg/tcpip/link/loopback/BUILD | 2 +- pkg/tcpip/link/rawfile/BUILD | 2 +- pkg/tcpip/link/sharedmem/BUILD | 2 +- pkg/tcpip/link/sharedmem/pipe/BUILD | 2 +- pkg/tcpip/link/sharedmem/queue/BUILD | 2 +- pkg/tcpip/link/sniffer/BUILD | 2 +- pkg/tcpip/link/tun/BUILD | 2 +- pkg/tcpip/link/waitable/BUILD | 2 +- pkg/tcpip/network/BUILD | 2 +- pkg/tcpip/network/arp/BUILD | 2 +- pkg/tcpip/network/fragmentation/BUILD | 2 +- pkg/tcpip/network/hash/BUILD | 2 +- pkg/tcpip/network/ipv4/BUILD | 2 +- pkg/tcpip/network/ipv6/BUILD | 2 +- pkg/tcpip/ports/BUILD | 2 +- pkg/tcpip/sample/tun_tcp_connect/BUILD | 2 +- pkg/tcpip/sample/tun_tcp_echo/BUILD | 2 +- pkg/tcpip/seqnum/BUILD | 2 +- pkg/tcpip/stack/BUILD | 2 +- pkg/tcpip/transport/ping/BUILD | 2 +- pkg/tcpip/transport/tcp/BUILD | 2 +- pkg/tcpip/transport/tcp/testing/context/BUILD | 2 +- pkg/tcpip/transport/tcpconntrack/BUILD | 2 +- pkg/tcpip/transport/udp/BUILD | 2 +- pkg/tmutex/BUILD | 2 +- pkg/unet/BUILD | 2 +- pkg/urpc/BUILD | 2 +- pkg/waiter/BUILD | 2 +- runsc/boot/BUILD | 2 +- runsc/boot/filter/BUILD | 2 +- runsc/cgroup/BUILD | 2 +- runsc/cmd/BUILD | 2 +- runsc/console/BUILD | 2 +- runsc/container/BUILD | 2 +- runsc/fsgofer/BUILD | 2 +- runsc/fsgofer/filter/BUILD | 2 +- runsc/sandbox/BUILD | 2 +- runsc/specutils/BUILD | 2 +- runsc/test/image/BUILD | 2 +- runsc/test/integration/BUILD | 2 +- runsc/test/root/BUILD | 2 +- runsc/test/root/testdata/BUILD | 2 +- runsc/test/testutil/BUILD | 2 +- runsc/tools/dockercfg/BUILD | 2 +- test/syscalls/BUILD | 2 +- test/syscalls/gtest/BUILD | 2 +- test/syscalls/linux/BUILD | 2 +- test/util/BUILD | 2 +- tools/go_generics/BUILD | 2 +- tools/go_generics/globals/BUILD | 2 +- tools/go_generics/go_merge/BUILD | 2 +- tools/go_generics/rules_tests/BUILD | 2 +- tools/go_stateify/BUILD | 2 +- vdso/BUILD | 2 +- 180 files changed, 180 insertions(+), 180 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/BUILD b/pkg/abi/BUILD index 1ba4f3a46..323263ebf 100644 --- a/pkg/abi/BUILD +++ b/pkg/abi/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index e6043abf4..7648c9469 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -2,7 +2,7 @@ # Linux kernel. It should be used instead of syscall or golang.org/x/sys/unix # when the host OS may not be Linux. -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/amutex/BUILD b/pkg/amutex/BUILD index 7cda07418..bdb6e8f2c 100644 --- a/pkg/amutex/BUILD +++ b/pkg/amutex/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "amutex", diff --git a/pkg/atomicbitops/BUILD b/pkg/atomicbitops/BUILD index 235188531..9555bf645 100644 --- a/pkg/atomicbitops/BUILD +++ b/pkg/atomicbitops/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "atomicbitops", diff --git a/pkg/binary/BUILD b/pkg/binary/BUILD index 571151f72..bd37376b0 100644 --- a/pkg/binary/BUILD +++ b/pkg/binary/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "binary", diff --git a/pkg/bits/BUILD b/pkg/bits/BUILD index 46794bdb8..5214b2c24 100644 --- a/pkg/bits/BUILD +++ b/pkg/bits/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") diff --git a/pkg/bpf/BUILD b/pkg/bpf/BUILD index 564df3af5..3c7ae3103 100644 --- a/pkg/bpf/BUILD +++ b/pkg/bpf/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/compressio/BUILD b/pkg/compressio/BUILD index 72952d735..3a0ac64e6 100644 --- a/pkg/compressio/BUILD +++ b/pkg/compressio/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "compressio", diff --git a/pkg/control/client/BUILD b/pkg/control/client/BUILD index 32853875d..22a4a4a5a 100644 --- a/pkg/control/client/BUILD +++ b/pkg/control/client/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "client", diff --git a/pkg/control/server/BUILD b/pkg/control/server/BUILD index ba2b1be9f..76b2e9787 100644 --- a/pkg/control/server/BUILD +++ b/pkg/control/server/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "server", diff --git a/pkg/cpuid/BUILD b/pkg/cpuid/BUILD index 46fc4703b..29cc38778 100644 --- a/pkg/cpuid/BUILD +++ b/pkg/cpuid/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/dhcp/BUILD b/pkg/dhcp/BUILD index c97dfc14b..003620b48 100644 --- a/pkg/dhcp/BUILD +++ b/pkg/dhcp/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "dhcp", diff --git a/pkg/eventchannel/BUILD b/pkg/eventchannel/BUILD index 18348ef54..5c2a44aa1 100644 --- a/pkg/eventchannel/BUILD +++ b/pkg/eventchannel/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "eventchannel", diff --git a/pkg/fd/BUILD b/pkg/fd/BUILD index 06cfd445e..ab1109157 100644 --- a/pkg/fd/BUILD +++ b/pkg/fd/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "fd", diff --git a/pkg/fdnotifier/BUILD b/pkg/fdnotifier/BUILD index 27d378d5b..8c8d193cc 100644 --- a/pkg/fdnotifier/BUILD +++ b/pkg/fdnotifier/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "fdnotifier", diff --git a/pkg/gate/BUILD b/pkg/gate/BUILD index 9a87a3a31..83679f2da 100644 --- a/pkg/gate/BUILD +++ b/pkg/gate/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "gate", diff --git a/pkg/ilist/BUILD b/pkg/ilist/BUILD index a67aa2cff..dbd65ab12 100644 --- a/pkg/ilist/BUILD +++ b/pkg/ilist/BUILD @@ -1,7 +1,7 @@ load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "ilist", diff --git a/pkg/linewriter/BUILD b/pkg/linewriter/BUILD index 3f28ba867..d1aa2e7d6 100644 --- a/pkg/linewriter/BUILD +++ b/pkg/linewriter/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "linewriter", diff --git a/pkg/log/BUILD b/pkg/log/BUILD index 94ac66db3..b2d18eddb 100644 --- a/pkg/log/BUILD +++ b/pkg/log/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "log", diff --git a/pkg/metric/BUILD b/pkg/metric/BUILD index d96e5563b..4b2c7a00e 100644 --- a/pkg/metric/BUILD +++ b/pkg/metric/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "metric", diff --git a/pkg/p9/BUILD b/pkg/p9/BUILD index 2c224e65b..5d972309d 100644 --- a/pkg/p9/BUILD +++ b/pkg/p9/BUILD @@ -2,7 +2,7 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") package( default_visibility = ["//visibility:public"], - licenses = ["notice"], # Apache 2.0 + licenses = ["notice"], ) go_library( diff --git a/pkg/p9/local_server/BUILD b/pkg/p9/local_server/BUILD index b17ebb79d..aa6db186c 100644 --- a/pkg/p9/local_server/BUILD +++ b/pkg/p9/local_server/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_binary( name = "local_server", diff --git a/pkg/p9/p9test/BUILD b/pkg/p9/p9test/BUILD index 7c4b875ce..cf22edde8 100644 --- a/pkg/p9/p9test/BUILD +++ b/pkg/p9/p9test/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) alias( name = "mockgen", diff --git a/pkg/rand/BUILD b/pkg/rand/BUILD index 0c9efc709..4eec3a4dd 100644 --- a/pkg/rand/BUILD +++ b/pkg/rand/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "rand", diff --git a/pkg/refs/BUILD b/pkg/refs/BUILD index 98150ba8f..fc562f821 100644 --- a/pkg/refs/BUILD +++ b/pkg/refs/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/seccomp/BUILD b/pkg/seccomp/BUILD index 657f923ed..0e9c4692d 100644 --- a/pkg/seccomp/BUILD +++ b/pkg/seccomp/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//go:def.bzl", "go_binary", "go_embed_data") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_binary( name = "victim", diff --git a/pkg/secio/BUILD b/pkg/secio/BUILD index 29f751725..2b4b87c61 100644 --- a/pkg/secio/BUILD +++ b/pkg/secio/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "secio", diff --git a/pkg/segment/BUILD b/pkg/segment/BUILD index 964d73af8..700385907 100644 --- a/pkg/segment/BUILD +++ b/pkg/segment/BUILD @@ -1,6 +1,6 @@ package( default_visibility = ["//:sandbox"], - licenses = ["notice"], # Apache 2.0 + licenses = ["notice"], ) load("//tools/go_generics:defs.bzl", "go_template") diff --git a/pkg/segment/test/BUILD b/pkg/segment/test/BUILD index bdf53e24e..81e929b8c 100644 --- a/pkg/segment/test/BUILD +++ b/pkg/segment/test/BUILD @@ -2,7 +2,7 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") package( default_visibility = ["//visibility:private"], - licenses = ["notice"], # Apache 2.0 + licenses = ["notice"], ) load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/sentry/BUILD b/pkg/sentry/BUILD index d18cf3555..53989301f 100644 --- a/pkg/sentry/BUILD +++ b/pkg/sentry/BUILD @@ -1,7 +1,7 @@ # This BUILD file defines a package_group that allows for interdependencies for # sentry-internal packages. -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) package_group( name = "internal", diff --git a/pkg/sentry/arch/BUILD b/pkg/sentry/arch/BUILD index 9bf04360a..0c044bc33 100644 --- a/pkg/sentry/arch/BUILD +++ b/pkg/sentry/arch/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/context/BUILD b/pkg/sentry/context/BUILD index 02d24defd..a3c8d0177 100644 --- a/pkg/sentry/context/BUILD +++ b/pkg/sentry/context/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "context", diff --git a/pkg/sentry/context/contexttest/BUILD b/pkg/sentry/context/contexttest/BUILD index 01bb40b04..bed156b70 100644 --- a/pkg/sentry/context/contexttest/BUILD +++ b/pkg/sentry/context/contexttest/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/control/BUILD b/pkg/sentry/control/BUILD index c3b682d6f..f54e01ee8 100644 --- a/pkg/sentry/control/BUILD +++ b/pkg/sentry/control/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "control", diff --git a/pkg/sentry/device/BUILD b/pkg/sentry/device/BUILD index bebdb2939..01de708d3 100644 --- a/pkg/sentry/device/BUILD +++ b/pkg/sentry/device/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "device", diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 6f368b0da..e58333da3 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/anon/BUILD b/pkg/sentry/fs/anon/BUILD index 4bd912e95..2111df2e8 100644 --- a/pkg/sentry/fs/anon/BUILD +++ b/pkg/sentry/fs/anon/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "anon", diff --git a/pkg/sentry/fs/ashmem/BUILD b/pkg/sentry/fs/ashmem/BUILD index e5bb661b5..dcf620dca 100644 --- a/pkg/sentry/fs/ashmem/BUILD +++ b/pkg/sentry/fs/ashmem/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/sentry/fs/binder/BUILD b/pkg/sentry/fs/binder/BUILD index 27155819e..8a448175f 100644 --- a/pkg/sentry/fs/binder/BUILD +++ b/pkg/sentry/fs/binder/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/fs/dev/BUILD b/pkg/sentry/fs/dev/BUILD index 85371032a..e5b962c8c 100644 --- a/pkg/sentry/fs/dev/BUILD +++ b/pkg/sentry/fs/dev/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/fs/fdpipe/BUILD b/pkg/sentry/fs/fdpipe/BUILD index 8a0937cda..098463e97 100644 --- a/pkg/sentry/fs/fdpipe/BUILD +++ b/pkg/sentry/fs/fdpipe/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/filetest/BUILD b/pkg/sentry/fs/filetest/BUILD index d137fee4c..05ca72aa0 100644 --- a/pkg/sentry/fs/filetest/BUILD +++ b/pkg/sentry/fs/filetest/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index d4767642b..7dff970ea 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/gofer/BUILD b/pkg/sentry/fs/gofer/BUILD index 35ffadd13..f2c79b475 100644 --- a/pkg/sentry/fs/gofer/BUILD +++ b/pkg/sentry/fs/gofer/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/host/BUILD b/pkg/sentry/fs/host/BUILD index 6877eb161..ea2ca11bf 100644 --- a/pkg/sentry/fs/host/BUILD +++ b/pkg/sentry/fs/host/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/lock/BUILD b/pkg/sentry/fs/lock/BUILD index 3159ff1da..7164744b8 100644 --- a/pkg/sentry/fs/lock/BUILD +++ b/pkg/sentry/fs/lock/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/proc/BUILD b/pkg/sentry/fs/proc/BUILD index 74954f213..f6bc90634 100644 --- a/pkg/sentry/fs/proc/BUILD +++ b/pkg/sentry/fs/proc/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/proc/device/BUILD b/pkg/sentry/fs/proc/device/BUILD index ff7dacf07..64b0c5a3a 100644 --- a/pkg/sentry/fs/proc/device/BUILD +++ b/pkg/sentry/fs/proc/device/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "device", diff --git a/pkg/sentry/fs/proc/seqfile/BUILD b/pkg/sentry/fs/proc/seqfile/BUILD index b4ba64e10..6b44c0075 100644 --- a/pkg/sentry/fs/proc/seqfile/BUILD +++ b/pkg/sentry/fs/proc/seqfile/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/ramfs/BUILD b/pkg/sentry/fs/ramfs/BUILD index 4a629e38e..f36e4a5e8 100644 --- a/pkg/sentry/fs/ramfs/BUILD +++ b/pkg/sentry/fs/ramfs/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/sys/BUILD b/pkg/sentry/fs/sys/BUILD index 7de928e16..42e98230e 100644 --- a/pkg/sentry/fs/sys/BUILD +++ b/pkg/sentry/fs/sys/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/fs/timerfd/BUILD b/pkg/sentry/fs/timerfd/BUILD index ffdd7e0dc..0e06a5028 100644 --- a/pkg/sentry/fs/timerfd/BUILD +++ b/pkg/sentry/fs/timerfd/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/fs/tmpfs/BUILD b/pkg/sentry/fs/tmpfs/BUILD index c5ec85460..bf5b68869 100644 --- a/pkg/sentry/fs/tmpfs/BUILD +++ b/pkg/sentry/fs/tmpfs/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/fs/tty/BUILD b/pkg/sentry/fs/tty/BUILD index 011cb6955..bee2db3f3 100644 --- a/pkg/sentry/fs/tty/BUILD +++ b/pkg/sentry/fs/tty/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/hostcpu/BUILD b/pkg/sentry/hostcpu/BUILD index 33197cf14..b5067ae6d 100644 --- a/pkg/sentry/hostcpu/BUILD +++ b/pkg/sentry/hostcpu/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "hostcpu", diff --git a/pkg/sentry/inet/BUILD b/pkg/sentry/inet/BUILD index 159c50efb..e288d34e9 100644 --- a/pkg/sentry/inet/BUILD +++ b/pkg/sentry/inet/BUILD @@ -1,6 +1,6 @@ package( default_visibility = ["//:sandbox"], - licenses = ["notice"], # Apache 2.0 + licenses = ["notice"], ) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 7d41626dc..b230aff98 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/kernel/auth/BUILD b/pkg/sentry/kernel/auth/BUILD index a81085372..abd4f2dae 100644 --- a/pkg/sentry/kernel/auth/BUILD +++ b/pkg/sentry/kernel/auth/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/kernel/contexttest/BUILD b/pkg/sentry/kernel/contexttest/BUILD index 391986291..5769a3b28 100644 --- a/pkg/sentry/kernel/contexttest/BUILD +++ b/pkg/sentry/kernel/contexttest/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/kernel/epoll/BUILD b/pkg/sentry/kernel/epoll/BUILD index 5e8b36ed6..1567d5050 100644 --- a/pkg/sentry/kernel/epoll/BUILD +++ b/pkg/sentry/kernel/epoll/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/kernel/eventfd/BUILD b/pkg/sentry/kernel/eventfd/BUILD index d96803fc9..f2f1a1223 100644 --- a/pkg/sentry/kernel/eventfd/BUILD +++ b/pkg/sentry/kernel/eventfd/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/kernel/fasync/BUILD b/pkg/sentry/kernel/fasync/BUILD index 17749c0de..5faf95909 100644 --- a/pkg/sentry/kernel/fasync/BUILD +++ b/pkg/sentry/kernel/fasync/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index afd35985f..da24c36c1 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/kernel/kdefs/BUILD b/pkg/sentry/kernel/kdefs/BUILD index 3f8fa206c..38aaca134 100644 --- a/pkg/sentry/kernel/kdefs/BUILD +++ b/pkg/sentry/kernel/kdefs/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "kdefs", diff --git a/pkg/sentry/kernel/memevent/BUILD b/pkg/sentry/kernel/memevent/BUILD index dfd8dd062..347a69062 100644 --- a/pkg/sentry/kernel/memevent/BUILD +++ b/pkg/sentry/kernel/memevent/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "memevent", diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD index 19b23c6d2..011a3f349 100644 --- a/pkg/sentry/kernel/pipe/BUILD +++ b/pkg/sentry/kernel/pipe/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/kernel/sched/BUILD b/pkg/sentry/kernel/sched/BUILD index 52e226a39..184e8a35b 100644 --- a/pkg/sentry/kernel/sched/BUILD +++ b/pkg/sentry/kernel/sched/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "sched", diff --git a/pkg/sentry/kernel/semaphore/BUILD b/pkg/sentry/kernel/semaphore/BUILD index bdcf4ce5c..840943ca8 100644 --- a/pkg/sentry/kernel/semaphore/BUILD +++ b/pkg/sentry/kernel/semaphore/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/kernel/shm/BUILD b/pkg/sentry/kernel/shm/BUILD index 40e641355..f45770eef 100644 --- a/pkg/sentry/kernel/shm/BUILD +++ b/pkg/sentry/kernel/shm/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/kernel/time/BUILD b/pkg/sentry/kernel/time/BUILD index 5d8db2273..584f7c7cc 100644 --- a/pkg/sentry/kernel/time/BUILD +++ b/pkg/sentry/kernel/time/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/limits/BUILD b/pkg/sentry/limits/BUILD index 90f4395d4..800166675 100644 --- a/pkg/sentry/limits/BUILD +++ b/pkg/sentry/limits/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/loader/BUILD b/pkg/sentry/loader/BUILD index 24e734b49..1ea260a4e 100644 --- a/pkg/sentry/loader/BUILD +++ b/pkg/sentry/loader/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_embed_data") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/memmap/BUILD b/pkg/sentry/memmap/BUILD index c9e0b95a0..9c2cbd18b 100644 --- a/pkg/sentry/memmap/BUILD +++ b/pkg/sentry/memmap/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/memutil/BUILD b/pkg/sentry/memutil/BUILD index 88738d65d..68b03d4cc 100644 --- a/pkg/sentry/memutil/BUILD +++ b/pkg/sentry/memutil/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "memutil", diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index 0997ec0a7..f679262d0 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/platform/BUILD b/pkg/sentry/platform/BUILD index af9ba5394..ac8a6cb7f 100644 --- a/pkg/sentry/platform/BUILD +++ b/pkg/sentry/platform/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/platform/filemem/BUILD b/pkg/sentry/platform/filemem/BUILD index 2a5982763..1a61cfaa5 100644 --- a/pkg/sentry/platform/filemem/BUILD +++ b/pkg/sentry/platform/filemem/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/platform/interrupt/BUILD b/pkg/sentry/platform/interrupt/BUILD index dbafa3204..eeccd4d0e 100644 --- a/pkg/sentry/platform/interrupt/BUILD +++ b/pkg/sentry/platform/interrupt/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "interrupt", diff --git a/pkg/sentry/platform/kvm/BUILD b/pkg/sentry/platform/kvm/BUILD index 1b71e629f..6e40b3177 100644 --- a/pkg/sentry/platform/kvm/BUILD +++ b/pkg/sentry/platform/kvm/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/sentry/platform/kvm/testutil/BUILD b/pkg/sentry/platform/kvm/testutil/BUILD index 1dffe94a4..e10087e8e 100644 --- a/pkg/sentry/platform/kvm/testutil/BUILD +++ b/pkg/sentry/platform/kvm/testutil/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "testutil", diff --git a/pkg/sentry/platform/procid/BUILD b/pkg/sentry/platform/procid/BUILD index 20c8bc02c..277509624 100644 --- a/pkg/sentry/platform/procid/BUILD +++ b/pkg/sentry/platform/procid/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "procid", diff --git a/pkg/sentry/platform/ptrace/BUILD b/pkg/sentry/platform/ptrace/BUILD index 2eb354ad4..f86790942 100644 --- a/pkg/sentry/platform/ptrace/BUILD +++ b/pkg/sentry/platform/ptrace/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "ptrace", diff --git a/pkg/sentry/platform/ring0/BUILD b/pkg/sentry/platform/ring0/BUILD index c35d49f2d..ecb3e9a9c 100644 --- a/pkg/sentry/platform/ring0/BUILD +++ b/pkg/sentry/platform/ring0/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") diff --git a/pkg/sentry/platform/ring0/gen_offsets/BUILD b/pkg/sentry/platform/ring0/gen_offsets/BUILD index b76d7974e..d7029d5a9 100644 --- a/pkg/sentry/platform/ring0/gen_offsets/BUILD +++ b/pkg/sentry/platform/ring0/gen_offsets/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/sentry/platform/ring0/pagetables/BUILD b/pkg/sentry/platform/ring0/pagetables/BUILD index de1b920af..fe93d3030 100644 --- a/pkg/sentry/platform/ring0/pagetables/BUILD +++ b/pkg/sentry/platform/ring0/pagetables/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") diff --git a/pkg/sentry/platform/safecopy/BUILD b/pkg/sentry/platform/safecopy/BUILD index cb8347dd8..05a6a61ae 100644 --- a/pkg/sentry/platform/safecopy/BUILD +++ b/pkg/sentry/platform/safecopy/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0, portions BSD, MIT +package(licenses = ["notice"]) go_library( name = "safecopy", diff --git a/pkg/sentry/safemem/BUILD b/pkg/sentry/safemem/BUILD index 87a9bff12..3ab453718 100644 --- a/pkg/sentry/safemem/BUILD +++ b/pkg/sentry/safemem/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "safemem", diff --git a/pkg/sentry/sighandling/BUILD b/pkg/sentry/sighandling/BUILD index 41313d334..cec3af92e 100644 --- a/pkg/sentry/sighandling/BUILD +++ b/pkg/sentry/sighandling/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "sighandling", diff --git a/pkg/sentry/socket/BUILD b/pkg/sentry/socket/BUILD index 3a8044b5f..076f953e7 100644 --- a/pkg/sentry/socket/BUILD +++ b/pkg/sentry/socket/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/socket/control/BUILD b/pkg/sentry/socket/control/BUILD index d3a63f15f..9f4763906 100644 --- a/pkg/sentry/socket/control/BUILD +++ b/pkg/sentry/socket/control/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/socket/epsocket/BUILD b/pkg/sentry/socket/epsocket/BUILD index da4aaf510..45e418db3 100644 --- a/pkg/sentry/socket/epsocket/BUILD +++ b/pkg/sentry/socket/epsocket/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD index b8dceb102..a469af7ac 100644 --- a/pkg/sentry/socket/hostinet/BUILD +++ b/pkg/sentry/socket/hostinet/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/socket/netlink/BUILD b/pkg/sentry/socket/netlink/BUILD index cff922cb8..148306329 100644 --- a/pkg/sentry/socket/netlink/BUILD +++ b/pkg/sentry/socket/netlink/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/socket/netlink/port/BUILD b/pkg/sentry/socket/netlink/port/BUILD index 3a7dbc5ed..a7370a4ec 100644 --- a/pkg/sentry/socket/netlink/port/BUILD +++ b/pkg/sentry/socket/netlink/port/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/socket/netlink/route/BUILD b/pkg/sentry/socket/netlink/route/BUILD index e1bcfe252..be0419679 100644 --- a/pkg/sentry/socket/netlink/route/BUILD +++ b/pkg/sentry/socket/netlink/route/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/socket/rpcinet/BUILD b/pkg/sentry/socket/rpcinet/BUILD index 06e121946..4da14a1e0 100644 --- a/pkg/sentry/socket/rpcinet/BUILD +++ b/pkg/sentry/socket/rpcinet/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "rpcinet", diff --git a/pkg/sentry/socket/rpcinet/conn/BUILD b/pkg/sentry/socket/rpcinet/conn/BUILD index a16977f29..4336ae9b4 100644 --- a/pkg/sentry/socket/rpcinet/conn/BUILD +++ b/pkg/sentry/socket/rpcinet/conn/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # BSD +package(licenses = ["notice"]) go_library( name = "conn", diff --git a/pkg/sentry/socket/rpcinet/notifier/BUILD b/pkg/sentry/socket/rpcinet/notifier/BUILD index 2bab01774..b0b107ddb 100644 --- a/pkg/sentry/socket/rpcinet/notifier/BUILD +++ b/pkg/sentry/socket/rpcinet/notifier/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # BSD +package(licenses = ["notice"]) go_library( name = "notifier", diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index a12fa93db..fe6871cc6 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/socket/unix/transport/BUILD b/pkg/sentry/socket/unix/transport/BUILD index 5a90837bc..5a2de0c4c 100644 --- a/pkg/sentry/socket/unix/transport/BUILD +++ b/pkg/sentry/socket/unix/transport/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/sentry/state/BUILD b/pkg/sentry/state/BUILD index f1f6fdb7d..42c459acc 100644 --- a/pkg/sentry/state/BUILD +++ b/pkg/sentry/state/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "state", diff --git a/pkg/sentry/strace/BUILD b/pkg/sentry/strace/BUILD index 8517db1ac..552e79686 100644 --- a/pkg/sentry/strace/BUILD +++ b/pkg/sentry/strace/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "strace", diff --git a/pkg/sentry/syscalls/BUILD b/pkg/sentry/syscalls/BUILD index 35192ff49..6b5469e45 100644 --- a/pkg/sentry/syscalls/BUILD +++ b/pkg/sentry/syscalls/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "syscalls", diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index 7621bfdbd..846601881 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/time/BUILD b/pkg/sentry/time/BUILD index 1191010e6..c4b6dcc63 100644 --- a/pkg/sentry/time/BUILD +++ b/pkg/sentry/time/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0, portions BSD +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/sentry/unimpl/BUILD b/pkg/sentry/unimpl/BUILD index 42e24ace5..b608867a9 100644 --- a/pkg/sentry/unimpl/BUILD +++ b/pkg/sentry/unimpl/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) proto_library( name = "unimplemented_syscall_proto", diff --git a/pkg/sentry/uniqueid/BUILD b/pkg/sentry/uniqueid/BUILD index 0929497c3..ccc5a28d3 100644 --- a/pkg/sentry/uniqueid/BUILD +++ b/pkg/sentry/uniqueid/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "uniqueid", diff --git a/pkg/sentry/usage/BUILD b/pkg/sentry/usage/BUILD index 868dfd400..09198496b 100644 --- a/pkg/sentry/usage/BUILD +++ b/pkg/sentry/usage/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/sentry/usermem/BUILD b/pkg/sentry/usermem/BUILD index dae41ed0e..1a560b6f3 100644 --- a/pkg/sentry/usermem/BUILD +++ b/pkg/sentry/usermem/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/sentry/watchdog/BUILD b/pkg/sentry/watchdog/BUILD index b2c687b20..0bbf3705c 100644 --- a/pkg/sentry/watchdog/BUILD +++ b/pkg/sentry/watchdog/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "watchdog", diff --git a/pkg/sleep/BUILD b/pkg/sleep/BUILD index 338fd9336..2b005bf66 100644 --- a/pkg/sleep/BUILD +++ b/pkg/sleep/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "sleep", diff --git a/pkg/state/BUILD b/pkg/state/BUILD index dd0f250fa..0a975e162 100644 --- a/pkg/state/BUILD +++ b/pkg/state/BUILD @@ -1,7 +1,7 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") load("@io_bazel_rules_go//proto:def.bzl", "go_proto_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/state/statefile/BUILD b/pkg/state/statefile/BUILD index 66c8f3807..5967781e8 100644 --- a/pkg/state/statefile/BUILD +++ b/pkg/state/statefile/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "statefile", diff --git a/pkg/sync/BUILD b/pkg/sync/BUILD index 6ddc6e812..1624e681c 100644 --- a/pkg/sync/BUILD +++ b/pkg/sync/BUILD @@ -2,7 +2,7 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") package( default_visibility = ["//:sandbox"], - licenses = ["notice"], # Apache 2.0, portions BSD + licenses = ["notice"], ) load("//tools/go_generics:defs.bzl", "go_template") diff --git a/pkg/sync/atomicptrtest/BUILD b/pkg/sync/atomicptrtest/BUILD index 9cb7f66fe..198fbb895 100644 --- a/pkg/sync/atomicptrtest/BUILD +++ b/pkg/sync/atomicptrtest/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/sync/seqatomictest/BUILD b/pkg/sync/seqatomictest/BUILD index 54f8e59b1..23132650a 100644 --- a/pkg/sync/seqatomictest/BUILD +++ b/pkg/sync/seqatomictest/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") diff --git a/pkg/syserr/BUILD b/pkg/syserr/BUILD index 30ae20772..0d65115ef 100644 --- a/pkg/syserr/BUILD +++ b/pkg/syserr/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "syserr", diff --git a/pkg/syserror/BUILD b/pkg/syserror/BUILD index d4c6da97a..ac478d0ff 100644 --- a/pkg/syserror/BUILD +++ b/pkg/syserror/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "syserror", diff --git a/pkg/tcpip/BUILD b/pkg/tcpip/BUILD index daff9a0a0..83524cc8a 100644 --- a/pkg/tcpip/BUILD +++ b/pkg/tcpip/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/tcpip/adapters/gonet/BUILD b/pkg/tcpip/adapters/gonet/BUILD index 723ad668f..ee2417238 100644 --- a/pkg/tcpip/adapters/gonet/BUILD +++ b/pkg/tcpip/adapters/gonet/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "gonet", diff --git a/pkg/tcpip/buffer/BUILD b/pkg/tcpip/buffer/BUILD index 11a725423..648d12cdf 100644 --- a/pkg/tcpip/buffer/BUILD +++ b/pkg/tcpip/buffer/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/tcpip/checker/BUILD b/pkg/tcpip/checker/BUILD index a1de808b9..f597d0b24 100644 --- a/pkg/tcpip/checker/BUILD +++ b/pkg/tcpip/checker/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "checker", diff --git a/pkg/tcpip/hash/jenkins/BUILD b/pkg/tcpip/hash/jenkins/BUILD index bbb764db8..ce2194a4d 100644 --- a/pkg/tcpip/hash/jenkins/BUILD +++ b/pkg/tcpip/hash/jenkins/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "jenkins", diff --git a/pkg/tcpip/header/BUILD b/pkg/tcpip/header/BUILD index 8e455fe1e..a5c7290ee 100644 --- a/pkg/tcpip/header/BUILD +++ b/pkg/tcpip/header/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/tcpip/link/channel/BUILD b/pkg/tcpip/link/channel/BUILD index 25f6c1457..ae285e495 100644 --- a/pkg/tcpip/link/channel/BUILD +++ b/pkg/tcpip/link/channel/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "channel", diff --git a/pkg/tcpip/link/fdbased/BUILD b/pkg/tcpip/link/fdbased/BUILD index a4aa3feec..0d78c9b15 100644 --- a/pkg/tcpip/link/fdbased/BUILD +++ b/pkg/tcpip/link/fdbased/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "fdbased", diff --git a/pkg/tcpip/link/loopback/BUILD b/pkg/tcpip/link/loopback/BUILD index a46ba7f11..710a05ede 100644 --- a/pkg/tcpip/link/loopback/BUILD +++ b/pkg/tcpip/link/loopback/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "loopback", diff --git a/pkg/tcpip/link/rawfile/BUILD b/pkg/tcpip/link/rawfile/BUILD index 2746d4ced..f01bb2c07 100644 --- a/pkg/tcpip/link/rawfile/BUILD +++ b/pkg/tcpip/link/rawfile/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "rawfile", diff --git a/pkg/tcpip/link/sharedmem/BUILD b/pkg/tcpip/link/sharedmem/BUILD index d7f1e66ef..dc8f1543e 100644 --- a/pkg/tcpip/link/sharedmem/BUILD +++ b/pkg/tcpip/link/sharedmem/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "sharedmem", diff --git a/pkg/tcpip/link/sharedmem/pipe/BUILD b/pkg/tcpip/link/sharedmem/pipe/BUILD index 12e813509..85deafa38 100644 --- a/pkg/tcpip/link/sharedmem/pipe/BUILD +++ b/pkg/tcpip/link/sharedmem/pipe/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "pipe", diff --git a/pkg/tcpip/link/sharedmem/queue/BUILD b/pkg/tcpip/link/sharedmem/queue/BUILD index 661037bb2..d7dc631eb 100644 --- a/pkg/tcpip/link/sharedmem/queue/BUILD +++ b/pkg/tcpip/link/sharedmem/queue/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "queue", diff --git a/pkg/tcpip/link/sniffer/BUILD b/pkg/tcpip/link/sniffer/BUILD index 52e237c25..7d0d1781e 100644 --- a/pkg/tcpip/link/sniffer/BUILD +++ b/pkg/tcpip/link/sniffer/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "sniffer", diff --git a/pkg/tcpip/link/tun/BUILD b/pkg/tcpip/link/tun/BUILD index 5ec01cec9..e54852d3f 100644 --- a/pkg/tcpip/link/tun/BUILD +++ b/pkg/tcpip/link/tun/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "tun", diff --git a/pkg/tcpip/link/waitable/BUILD b/pkg/tcpip/link/waitable/BUILD index ba495c437..89a9eee23 100644 --- a/pkg/tcpip/link/waitable/BUILD +++ b/pkg/tcpip/link/waitable/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "waitable", diff --git a/pkg/tcpip/network/BUILD b/pkg/tcpip/network/BUILD index a2a07f533..f36f49453 100644 --- a/pkg/tcpip/network/BUILD +++ b/pkg/tcpip/network/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_test( name = "ip_test", diff --git a/pkg/tcpip/network/arp/BUILD b/pkg/tcpip/network/arp/BUILD index f6fb7daf7..ef18bb93d 100644 --- a/pkg/tcpip/network/arp/BUILD +++ b/pkg/tcpip/network/arp/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "arp", diff --git a/pkg/tcpip/network/fragmentation/BUILD b/pkg/tcpip/network/fragmentation/BUILD index aaabfcb9a..bf0a7b99c 100644 --- a/pkg/tcpip/network/fragmentation/BUILD +++ b/pkg/tcpip/network/fragmentation/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/tcpip/network/hash/BUILD b/pkg/tcpip/network/hash/BUILD index 401dce646..ea520c6ed 100644 --- a/pkg/tcpip/network/hash/BUILD +++ b/pkg/tcpip/network/hash/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "hash", diff --git a/pkg/tcpip/network/ipv4/BUILD b/pkg/tcpip/network/ipv4/BUILD index e72317e9f..7a5341def 100644 --- a/pkg/tcpip/network/ipv4/BUILD +++ b/pkg/tcpip/network/ipv4/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "ipv4", diff --git a/pkg/tcpip/network/ipv6/BUILD b/pkg/tcpip/network/ipv6/BUILD index 808c37df3..000e00dba 100644 --- a/pkg/tcpip/network/ipv6/BUILD +++ b/pkg/tcpip/network/ipv6/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "ipv6", diff --git a/pkg/tcpip/ports/BUILD b/pkg/tcpip/ports/BUILD index a2fa9b84a..3ee80c62b 100644 --- a/pkg/tcpip/ports/BUILD +++ b/pkg/tcpip/ports/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "ports", diff --git a/pkg/tcpip/sample/tun_tcp_connect/BUILD b/pkg/tcpip/sample/tun_tcp_connect/BUILD index 32baf2115..996939581 100644 --- a/pkg/tcpip/sample/tun_tcp_connect/BUILD +++ b/pkg/tcpip/sample/tun_tcp_connect/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_binary( name = "tun_tcp_connect", diff --git a/pkg/tcpip/sample/tun_tcp_echo/BUILD b/pkg/tcpip/sample/tun_tcp_echo/BUILD index 760445843..dad8ef399 100644 --- a/pkg/tcpip/sample/tun_tcp_echo/BUILD +++ b/pkg/tcpip/sample/tun_tcp_echo/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_binary( name = "tun_tcp_echo", diff --git a/pkg/tcpip/seqnum/BUILD b/pkg/tcpip/seqnum/BUILD index c5c889239..a63665efc 100644 --- a/pkg/tcpip/seqnum/BUILD +++ b/pkg/tcpip/seqnum/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/tcpip/stack/BUILD b/pkg/tcpip/stack/BUILD index 8a598c57d..551c3c73e 100644 --- a/pkg/tcpip/stack/BUILD +++ b/pkg/tcpip/stack/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/tcpip/transport/ping/BUILD b/pkg/tcpip/transport/ping/BUILD index 982b6795c..4d4241d4b 100644 --- a/pkg/tcpip/transport/ping/BUILD +++ b/pkg/tcpip/transport/ping/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library") diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index 726107739..e5c05f8c0 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/tcpip/transport/tcp/testing/context/BUILD b/pkg/tcpip/transport/tcp/testing/context/BUILD index 814e5c1ea..1584e4095 100644 --- a/pkg/tcpip/transport/tcp/testing/context/BUILD +++ b/pkg/tcpip/transport/tcp/testing/context/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "context", diff --git a/pkg/tcpip/transport/tcpconntrack/BUILD b/pkg/tcpip/transport/tcpconntrack/BUILD index ac1a94d4d..31a845dee 100644 --- a/pkg/tcpip/transport/tcpconntrack/BUILD +++ b/pkg/tcpip/transport/tcpconntrack/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "tcpconntrack", diff --git a/pkg/tcpip/transport/udp/BUILD b/pkg/tcpip/transport/udp/BUILD index 4225e28dc..8ccb79c48 100644 --- a/pkg/tcpip/transport/udp/BUILD +++ b/pkg/tcpip/transport/udp/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/pkg/tmutex/BUILD b/pkg/tmutex/BUILD index c20df7005..69035044d 100644 --- a/pkg/tmutex/BUILD +++ b/pkg/tmutex/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "tmutex", diff --git a/pkg/unet/BUILD b/pkg/unet/BUILD index f90e43c89..5e177e78e 100644 --- a/pkg/unet/BUILD +++ b/pkg/unet/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "unet", diff --git a/pkg/urpc/BUILD b/pkg/urpc/BUILD index 21008cf6c..36cae67e1 100644 --- a/pkg/urpc/BUILD +++ b/pkg/urpc/BUILD @@ -1,6 +1,6 @@ load("//tools/go_stateify:defs.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "urpc", diff --git a/pkg/waiter/BUILD b/pkg/waiter/BUILD index 5e611c54f..b748246da 100644 --- a/pkg/waiter/BUILD +++ b/pkg/waiter/BUILD @@ -1,4 +1,4 @@ -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_stateify:defs.bzl", "go_library", "go_test") diff --git a/runsc/boot/BUILD b/runsc/boot/BUILD index 15a7cdae1..540e99151 100644 --- a/runsc/boot/BUILD +++ b/runsc/boot/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "boot", diff --git a/runsc/boot/filter/BUILD b/runsc/boot/filter/BUILD index 004222242..3b6020cf3 100644 --- a/runsc/boot/filter/BUILD +++ b/runsc/boot/filter/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "filter", diff --git a/runsc/cgroup/BUILD b/runsc/cgroup/BUILD index 4f9a25a25..620d33a19 100644 --- a/runsc/cgroup/BUILD +++ b/runsc/cgroup/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "cgroup", diff --git a/runsc/cmd/BUILD b/runsc/cmd/BUILD index a908172af..9e2be0d37 100644 --- a/runsc/cmd/BUILD +++ b/runsc/cmd/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "cmd", diff --git a/runsc/console/BUILD b/runsc/console/BUILD index ff4ccff69..3ff9eba27 100644 --- a/runsc/console/BUILD +++ b/runsc/console/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "console", diff --git a/runsc/container/BUILD b/runsc/container/BUILD index 354ce2661..3b25ff79a 100644 --- a/runsc/container/BUILD +++ b/runsc/container/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "container", diff --git a/runsc/fsgofer/BUILD b/runsc/fsgofer/BUILD index 756c20ad7..4adc9c1bc 100644 --- a/runsc/fsgofer/BUILD +++ b/runsc/fsgofer/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "fsgofer", diff --git a/runsc/fsgofer/filter/BUILD b/runsc/fsgofer/filter/BUILD index c7848d10c..78c5b526c 100644 --- a/runsc/fsgofer/filter/BUILD +++ b/runsc/fsgofer/filter/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "filter", diff --git a/runsc/sandbox/BUILD b/runsc/sandbox/BUILD index 899fd99de..2ed793333 100644 --- a/runsc/sandbox/BUILD +++ b/runsc/sandbox/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "sandbox", diff --git a/runsc/specutils/BUILD b/runsc/specutils/BUILD index 77a10e2b6..372799850 100644 --- a/runsc/specutils/BUILD +++ b/runsc/specutils/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "specutils", diff --git a/runsc/test/image/BUILD b/runsc/test/image/BUILD index 22b3ebd2a..e8b629c6a 100644 --- a/runsc/test/image/BUILD +++ b/runsc/test/image/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_test( name = "image_test", diff --git a/runsc/test/integration/BUILD b/runsc/test/integration/BUILD index e7204dc66..779d30ec9 100644 --- a/runsc/test/integration/BUILD +++ b/runsc/test/integration/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_test( name = "integration_test", diff --git a/runsc/test/root/BUILD b/runsc/test/root/BUILD index 77dcbd79e..75826a521 100644 --- a/runsc/test/root/BUILD +++ b/runsc/test/root/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "root", diff --git a/runsc/test/root/testdata/BUILD b/runsc/test/root/testdata/BUILD index 6c9fe0aea..7f272dcd3 100644 --- a/runsc/test/root/testdata/BUILD +++ b/runsc/test/root/testdata/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "testdata", diff --git a/runsc/test/testutil/BUILD b/runsc/test/testutil/BUILD index 8c3919320..ddec81444 100644 --- a/runsc/test/testutil/BUILD +++ b/runsc/test/testutil/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "testutil", diff --git a/runsc/tools/dockercfg/BUILD b/runsc/tools/dockercfg/BUILD index a80b3abab..fd406ab93 100644 --- a/runsc/tools/dockercfg/BUILD +++ b/runsc/tools/dockercfg/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_binary( name = "dockercfg", diff --git a/test/syscalls/BUILD b/test/syscalls/BUILD index 8c391c8a6..148d9c366 100644 --- a/test/syscalls/BUILD +++ b/test/syscalls/BUILD @@ -1,7 +1,7 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") load("//test/syscalls:build_defs.bzl", "syscall_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) syscall_test(test = "//test/syscalls/linux:32bit_test") diff --git a/test/syscalls/gtest/BUILD b/test/syscalls/gtest/BUILD index d078fd3d5..22e061652 100644 --- a/test/syscalls/gtest/BUILD +++ b/test/syscalls/gtest/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "gtest", diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index e70742875..a311ca12c 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -1,6 +1,6 @@ package( default_visibility = ["//:sandbox"], - licenses = ["notice"], # Apache 2.0 + licenses = ["notice"], ) cc_binary( diff --git a/test/util/BUILD b/test/util/BUILD index f2e563507..fac0730b4 100644 --- a/test/util/BUILD +++ b/test/util/BUILD @@ -1,6 +1,6 @@ package( default_visibility = ["//:sandbox"], - licenses = ["notice"], # Apache 2.0 + licenses = ["notice"], ) cc_library( diff --git a/tools/go_generics/BUILD b/tools/go_generics/BUILD index 2d97d99dc..39318b877 100644 --- a/tools/go_generics/BUILD +++ b/tools/go_generics/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_binary( name = "go_generics", diff --git a/tools/go_generics/globals/BUILD b/tools/go_generics/globals/BUILD index c26ac56d2..6628132f5 100644 --- a/tools/go_generics/globals/BUILD +++ b/tools/go_generics/globals/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_library( name = "globals", diff --git a/tools/go_generics/go_merge/BUILD b/tools/go_generics/go_merge/BUILD index a60437962..02b09120e 100644 --- a/tools/go_generics/go_merge/BUILD +++ b/tools/go_generics/go_merge/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_binary( name = "go_merge", diff --git a/tools/go_generics/rules_tests/BUILD b/tools/go_generics/rules_tests/BUILD index 23b2d656d..a6f8cdd3c 100644 --- a/tools/go_generics/rules_tests/BUILD +++ b/tools/go_generics/rules_tests/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_test") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template", "go_template_instance") diff --git a/tools/go_stateify/BUILD b/tools/go_stateify/BUILD index 68d37f5d7..bb53f8ae9 100644 --- a/tools/go_stateify/BUILD +++ b/tools/go_stateify/BUILD @@ -1,6 +1,6 @@ load("@io_bazel_rules_go//go:def.bzl", "go_binary") -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) go_binary( name = "stateify", diff --git a/vdso/BUILD b/vdso/BUILD index fd395511c..c43d24070 100644 --- a/vdso/BUILD +++ b/vdso/BUILD @@ -3,7 +3,7 @@ # normal system VDSO (time, gettimeofday, clock_gettimeofday) but which uses # timekeeping parameters managed by the sandbox kernel. -package(licenses = ["notice"]) # Apache 2.0 +package(licenses = ["notice"]) genrule( name = "vdso", -- cgit v1.2.3 From fe1369ac98a4f1d8af5e8be6da71165339e52034 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Thu, 31 Jan 2019 17:47:24 -0800 Subject: Move package sync to third_party PiperOrigin-RevId: 231889261 Change-Id: I482f1df055bcedf4edb9fe3fe9b8e9c80085f1a0 --- pkg/sentry/kernel/BUILD | 4 +- pkg/sentry/kernel/futex/BUILD | 2 +- pkg/sentry/kernel/task.go | 4 +- pkg/sentry/mm/BUILD | 2 +- pkg/sentry/mm/mm.go | 6 +- pkg/sentry/time/BUILD | 4 +- pkg/sync/BUILD | 50 ------- pkg/sync/LICENSE | 27 ---- pkg/sync/atomicptr_unsafe.go | 46 ------ pkg/sync/atomicptrtest/BUILD | 29 ---- pkg/sync/atomicptrtest/atomicptr_test.go | 40 ----- pkg/sync/downgradable_rwmutex_test.go | 149 ------------------- pkg/sync/downgradable_rwmutex_unsafe.go | 143 ------------------ pkg/sync/memmove_unsafe.go | 32 ---- pkg/sync/norace_unsafe.go | 44 ------ pkg/sync/race_unsafe.go | 50 ------- pkg/sync/seqatomic_unsafe.go | 81 ----------- pkg/sync/seqatomictest/BUILD | 35 ----- pkg/sync/seqatomictest/seqatomic_test.go | 132 ----------------- pkg/sync/seqcount.go | 158 -------------------- pkg/sync/seqcount_test.go | 162 --------------------- pkg/sync/sync.go | 16 -- third_party/gvsync/BUILD | 50 +++++++ third_party/gvsync/LICENSE | 27 ++++ third_party/gvsync/README.md | 3 + third_party/gvsync/atomicptr_unsafe.go | 37 +++++ third_party/gvsync/atomicptrtest/BUILD | 29 ++++ third_party/gvsync/atomicptrtest/atomicptr_test.go | 31 ++++ third_party/gvsync/downgradable_rwmutex_test.go | 150 +++++++++++++++++++ third_party/gvsync/downgradable_rwmutex_unsafe.go | 144 ++++++++++++++++++ third_party/gvsync/gvsync.go | 7 + third_party/gvsync/memmove_unsafe.go | 23 +++ third_party/gvsync/norace_unsafe.go | 35 +++++ third_party/gvsync/race_unsafe.go | 41 ++++++ third_party/gvsync/seqatomic_unsafe.go | 72 +++++++++ third_party/gvsync/seqatomictest/BUILD | 35 +++++ third_party/gvsync/seqatomictest/seqatomic_test.go | 132 +++++++++++++++++ third_party/gvsync/seqcount.go | 149 +++++++++++++++++++ third_party/gvsync/seqcount_test.go | 153 +++++++++++++++++++ 39 files changed, 1129 insertions(+), 1205 deletions(-) delete mode 100644 pkg/sync/BUILD delete mode 100644 pkg/sync/LICENSE delete mode 100644 pkg/sync/atomicptr_unsafe.go delete mode 100644 pkg/sync/atomicptrtest/BUILD delete mode 100644 pkg/sync/atomicptrtest/atomicptr_test.go delete mode 100644 pkg/sync/downgradable_rwmutex_test.go delete mode 100644 pkg/sync/downgradable_rwmutex_unsafe.go delete mode 100644 pkg/sync/memmove_unsafe.go delete mode 100644 pkg/sync/norace_unsafe.go delete mode 100644 pkg/sync/race_unsafe.go delete mode 100644 pkg/sync/seqatomic_unsafe.go delete mode 100644 pkg/sync/seqatomictest/BUILD delete mode 100644 pkg/sync/seqatomictest/seqatomic_test.go delete mode 100644 pkg/sync/seqcount.go delete mode 100644 pkg/sync/seqcount_test.go delete mode 100644 pkg/sync/sync.go create mode 100644 third_party/gvsync/BUILD create mode 100644 third_party/gvsync/LICENSE create mode 100644 third_party/gvsync/README.md create mode 100644 third_party/gvsync/atomicptr_unsafe.go create mode 100644 third_party/gvsync/atomicptrtest/BUILD create mode 100644 third_party/gvsync/atomicptrtest/atomicptr_test.go create mode 100644 third_party/gvsync/downgradable_rwmutex_test.go create mode 100644 third_party/gvsync/downgradable_rwmutex_unsafe.go create mode 100644 third_party/gvsync/gvsync.go create mode 100644 third_party/gvsync/memmove_unsafe.go create mode 100644 third_party/gvsync/norace_unsafe.go create mode 100644 third_party/gvsync/race_unsafe.go create mode 100644 third_party/gvsync/seqatomic_unsafe.go create mode 100644 third_party/gvsync/seqatomictest/BUILD create mode 100644 third_party/gvsync/seqatomictest/seqatomic_test.go create mode 100644 third_party/gvsync/seqcount.go create mode 100644 third_party/gvsync/seqcount_test.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index b230aff98..773cb8c91 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -34,7 +34,7 @@ go_template_instance( out = "seqatomic_taskgoroutineschedinfo.go", package = "kernel", suffix = "TaskGoroutineSchedInfo", - template = "//pkg/sync:generic_seqatomic", + template = "//third_party/gvsync:generic_seqatomic", types = { "Value": "TaskGoroutineSchedInfo", }, @@ -183,12 +183,12 @@ go_library( "//pkg/sentry/usermem", "//pkg/state", "//pkg/state/statefile", - "//pkg/sync", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", "//pkg/tcpip/stack", "//pkg/waiter", + "//third_party/gvsync", ], ) diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index da24c36c1..91feeb5ed 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -8,7 +8,7 @@ go_template_instance( out = "atomicptr_bucket.go", package = "futex", suffix = "Bucket", - template = "//pkg/sync:generic_atomicptr", + template = "//third_party/gvsync:generic_atomicptr", types = { "Value": "bucket", }, diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 2982bc5d1..702e40cce 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -34,7 +34,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - ssync "gvisor.googlesource.com/gvisor/pkg/sync" + "gvisor.googlesource.com/gvisor/third_party/gvsync" ) // Task represents a thread of execution in the untrusted app. It @@ -81,7 +81,7 @@ type Task struct { // // gosched is protected by goschedSeq. gosched is owned by the task // goroutine. - goschedSeq ssync.SeqCount `state:"nosave"` + goschedSeq gvsync.SeqCount `state:"nosave"` gosched TaskGoroutineSchedInfo // yieldCount is the number of times the task goroutine has called diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index f679262d0..a85ffdef8 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -116,9 +116,9 @@ go_library( "//pkg/sentry/safemem", "//pkg/sentry/usage", "//pkg/sentry/usermem", - "//pkg/sync", "//pkg/syserror", "//pkg/tcpip/buffer", + "//third_party/gvsync", ], ) diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index 2154e7918..a69b8c7be 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -42,7 +42,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - ssync "gvisor.googlesource.com/gvisor/pkg/sync" + "gvisor.googlesource.com/gvisor/third_party/gvsync" ) // MemoryManager implements a virtual address space. @@ -84,7 +84,7 @@ type MemoryManager struct { users int32 // mappingMu is analogous to Linux's struct mm_struct::mmap_sem. - mappingMu ssync.DowngradableRWMutex `state:"nosave"` + mappingMu gvsync.DowngradableRWMutex `state:"nosave"` // vmas stores virtual memory areas. Since vmas are stored by value, // clients should usually use vmaIterator.ValuePtr() instead of @@ -121,7 +121,7 @@ type MemoryManager struct { // activeMu is loosely analogous to Linux's struct // mm_struct::page_table_lock. - activeMu ssync.DowngradableRWMutex `state:"nosave"` + activeMu gvsync.DowngradableRWMutex `state:"nosave"` // pmas stores platform mapping areas used to implement vmas. Since pmas // are stored by value, clients should usually use pmaIterator.ValuePtr() diff --git a/pkg/sentry/time/BUILD b/pkg/sentry/time/BUILD index c4b6dcc63..b2f8f6832 100644 --- a/pkg/sentry/time/BUILD +++ b/pkg/sentry/time/BUILD @@ -9,7 +9,7 @@ go_template_instance( out = "seqatomic_parameters.go", package = "time", suffix = "Parameters", - template = "//pkg/sync:generic_seqatomic", + template = "//third_party/gvsync:generic_seqatomic", types = { "Value": "Parameters", }, @@ -36,8 +36,8 @@ go_library( deps = [ "//pkg/log", "//pkg/metric", - "//pkg/sync", "//pkg/syserror", + "//third_party/gvsync", ], ) diff --git a/pkg/sync/BUILD b/pkg/sync/BUILD deleted file mode 100644 index 1624e681c..000000000 --- a/pkg/sync/BUILD +++ /dev/null @@ -1,50 +0,0 @@ -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") - -package( - default_visibility = ["//:sandbox"], - licenses = ["notice"], -) - -load("//tools/go_generics:defs.bzl", "go_template") - -go_template( - name = "generic_atomicptr", - srcs = ["atomicptr_unsafe.go"], - types = [ - "Value", - ], -) - -go_template( - name = "generic_seqatomic", - srcs = ["seqatomic_unsafe.go"], - types = [ - "Value", - ], - deps = [ - ":sync", - ], -) - -go_library( - name = "sync", - srcs = [ - "downgradable_rwmutex_unsafe.go", - "memmove_unsafe.go", - "norace_unsafe.go", - "race_unsafe.go", - "seqcount.go", - "sync.go", - ], - importpath = "gvisor.googlesource.com/gvisor/pkg/sync", -) - -go_test( - name = "sync_test", - size = "small", - srcs = [ - "downgradable_rwmutex_test.go", - "seqcount_test.go", - ], - embed = [":sync"], -) diff --git a/pkg/sync/LICENSE b/pkg/sync/LICENSE deleted file mode 100644 index 6a66aea5e..000000000 --- a/pkg/sync/LICENSE +++ /dev/null @@ -1,27 +0,0 @@ -Copyright (c) 2009 The Go Authors. All rights reserved. - -Redistribution and use in source and binary forms, with or without -modification, are permitted provided that the following conditions are -met: - - * Redistributions of source code must retain the above copyright -notice, this list of conditions and the following disclaimer. - * Redistributions in binary form must reproduce the above -copyright notice, this list of conditions and the following disclaimer -in the documentation and/or other materials provided with the -distribution. - * Neither the name of Google Inc. nor the names of its -contributors may be used to endorse or promote products derived from -this software without specific prior written permission. - -THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS -"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT -LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR -A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT -OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, -SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT -LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, -DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY -THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT -(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE -OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/pkg/sync/atomicptr_unsafe.go b/pkg/sync/atomicptr_unsafe.go deleted file mode 100644 index d943b7ff4..000000000 --- a/pkg/sync/atomicptr_unsafe.go +++ /dev/null @@ -1,46 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package template doesn't exist. This file must be instantiated using the -// go_template_instance rule in tools/go_generics/defs.bzl. -package template - -import ( - "sync/atomic" - "unsafe" -) - -// Value is a required type parameter. -type Value struct{} - -// An AtomicPtr is a pointer to a value of type Value that can be atomically -// loaded and stored. The zero value of an AtomicPtr represents nil. -// -// Note that copying AtomicPtr by value performs a non-atomic read of the -// stored pointer, which is unsafe if Store() can be called concurrently; in -// this case, do `dst.Store(src.Load())` instead. -type AtomicPtr struct { - ptr unsafe.Pointer -} - -// Load returns the value set by the most recent Store. It returns nil if there -// has been no previous call to Store. -func (p *AtomicPtr) Load() *Value { - return (*Value)(atomic.LoadPointer(&p.ptr)) -} - -// Store sets the value returned by Load to x. -func (p *AtomicPtr) Store(x *Value) { - atomic.StorePointer(&p.ptr, (unsafe.Pointer)(x)) -} diff --git a/pkg/sync/atomicptrtest/BUILD b/pkg/sync/atomicptrtest/BUILD deleted file mode 100644 index 198fbb895..000000000 --- a/pkg/sync/atomicptrtest/BUILD +++ /dev/null @@ -1,29 +0,0 @@ -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") - -package(licenses = ["notice"]) - -load("//tools/go_generics:defs.bzl", "go_template_instance") - -go_template_instance( - name = "atomicptr_int", - out = "atomicptr_int.go", - package = "atomicptr", - suffix = "Int", - template = "//pkg/sync:generic_atomicptr", - types = { - "Value": "int", - }, -) - -go_library( - name = "atomicptr", - srcs = ["atomicptr_int.go"], - importpath = "gvisor.googlesource.com/gvisor/pkg/sync/atomicptr", -) - -go_test( - name = "atomicptr_test", - size = "small", - srcs = ["atomicptr_test.go"], - embed = [":atomicptr"], -) diff --git a/pkg/sync/atomicptrtest/atomicptr_test.go b/pkg/sync/atomicptrtest/atomicptr_test.go deleted file mode 100644 index 3262785ce..000000000 --- a/pkg/sync/atomicptrtest/atomicptr_test.go +++ /dev/null @@ -1,40 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package atomicptr - -import ( - "testing" -) - -func newInt(val int) *int { - return &val -} - -func TestAtomicPtr(t *testing.T) { - var p AtomicPtrInt - if got := p.Load(); got != nil { - t.Errorf("initial value is %p (%v), wanted nil", got, got) - } - want := newInt(42) - p.Store(want) - if got := p.Load(); got != want { - t.Errorf("wrong value: got %p (%v), wanted %p (%v)", got, got, want, want) - } - want = newInt(100) - p.Store(want) - if got := p.Load(); got != want { - t.Errorf("wrong value: got %p (%v), wanted %p (%v)", got, got, want, want) - } -} diff --git a/pkg/sync/downgradable_rwmutex_test.go b/pkg/sync/downgradable_rwmutex_test.go deleted file mode 100644 index 99c1c8be0..000000000 --- a/pkg/sync/downgradable_rwmutex_test.go +++ /dev/null @@ -1,149 +0,0 @@ -// Copyright 2009 The Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -// GOMAXPROCS=10 go test - -// Copy/pasted from the standard library's sync/rwmutex_test.go, except for the -// addition of downgradingWriter and the renaming of num_iterations to -// numIterations to shut up Golint. - -package sync - -import ( - "fmt" - "runtime" - "sync/atomic" - "testing" -) - -func parallelReader(m *DowngradableRWMutex, clocked, cunlock, cdone chan bool) { - m.RLock() - clocked <- true - <-cunlock - m.RUnlock() - cdone <- true -} - -func doTestParallelReaders(numReaders, gomaxprocs int) { - runtime.GOMAXPROCS(gomaxprocs) - var m DowngradableRWMutex - clocked := make(chan bool) - cunlock := make(chan bool) - cdone := make(chan bool) - for i := 0; i < numReaders; i++ { - go parallelReader(&m, clocked, cunlock, cdone) - } - // Wait for all parallel RLock()s to succeed. - for i := 0; i < numReaders; i++ { - <-clocked - } - for i := 0; i < numReaders; i++ { - cunlock <- true - } - // Wait for the goroutines to finish. - for i := 0; i < numReaders; i++ { - <-cdone - } -} - -func TestParallelReaders(t *testing.T) { - defer runtime.GOMAXPROCS(runtime.GOMAXPROCS(-1)) - doTestParallelReaders(1, 4) - doTestParallelReaders(3, 4) - doTestParallelReaders(4, 2) -} - -func reader(rwm *DowngradableRWMutex, numIterations int, activity *int32, cdone chan bool) { - for i := 0; i < numIterations; i++ { - rwm.RLock() - n := atomic.AddInt32(activity, 1) - if n < 1 || n >= 10000 { - panic(fmt.Sprintf("wlock(%d)\n", n)) - } - for i := 0; i < 100; i++ { - } - atomic.AddInt32(activity, -1) - rwm.RUnlock() - } - cdone <- true -} - -func writer(rwm *DowngradableRWMutex, numIterations int, activity *int32, cdone chan bool) { - for i := 0; i < numIterations; i++ { - rwm.Lock() - n := atomic.AddInt32(activity, 10000) - if n != 10000 { - panic(fmt.Sprintf("wlock(%d)\n", n)) - } - for i := 0; i < 100; i++ { - } - atomic.AddInt32(activity, -10000) - rwm.Unlock() - } - cdone <- true -} - -func downgradingWriter(rwm *DowngradableRWMutex, numIterations int, activity *int32, cdone chan bool) { - for i := 0; i < numIterations; i++ { - rwm.Lock() - n := atomic.AddInt32(activity, 10000) - if n != 10000 { - panic(fmt.Sprintf("wlock(%d)\n", n)) - } - for i := 0; i < 100; i++ { - } - atomic.AddInt32(activity, -10000) - rwm.DowngradeLock() - n = atomic.AddInt32(activity, 1) - if n < 1 || n >= 10000 { - panic(fmt.Sprintf("wlock(%d)\n", n)) - } - for i := 0; i < 100; i++ { - } - n = atomic.AddInt32(activity, -1) - rwm.RUnlock() - } - cdone <- true -} - -func HammerDowngradableRWMutex(gomaxprocs, numReaders, numIterations int) { - runtime.GOMAXPROCS(gomaxprocs) - // Number of active readers + 10000 * number of active writers. - var activity int32 - var rwm DowngradableRWMutex - cdone := make(chan bool) - go writer(&rwm, numIterations, &activity, cdone) - go downgradingWriter(&rwm, numIterations, &activity, cdone) - var i int - for i = 0; i < numReaders/2; i++ { - go reader(&rwm, numIterations, &activity, cdone) - } - go writer(&rwm, numIterations, &activity, cdone) - go downgradingWriter(&rwm, numIterations, &activity, cdone) - for ; i < numReaders; i++ { - go reader(&rwm, numIterations, &activity, cdone) - } - // Wait for the 4 writers and all readers to finish. - for i := 0; i < 4+numReaders; i++ { - <-cdone - } -} - -func TestDowngradableRWMutex(t *testing.T) { - defer runtime.GOMAXPROCS(runtime.GOMAXPROCS(-1)) - n := 1000 - if testing.Short() { - n = 5 - } - HammerDowngradableRWMutex(1, 1, n) - HammerDowngradableRWMutex(1, 3, n) - HammerDowngradableRWMutex(1, 10, n) - HammerDowngradableRWMutex(4, 1, n) - HammerDowngradableRWMutex(4, 3, n) - HammerDowngradableRWMutex(4, 10, n) - HammerDowngradableRWMutex(10, 1, n) - HammerDowngradableRWMutex(10, 3, n) - HammerDowngradableRWMutex(10, 10, n) - HammerDowngradableRWMutex(10, 5, n) -} diff --git a/pkg/sync/downgradable_rwmutex_unsafe.go b/pkg/sync/downgradable_rwmutex_unsafe.go deleted file mode 100644 index 9a96056fc..000000000 --- a/pkg/sync/downgradable_rwmutex_unsafe.go +++ /dev/null @@ -1,143 +0,0 @@ -// Copyright 2009 The Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -// This is mostly copied from the standard library's sync/rwmutex.go. -// -// Happens-before relationships indicated to the race detector: -// - Unlock -> Lock (via writerSem) -// - Unlock -> RLock (via readerSem) -// - RUnlock -> Lock (via writerSem) -// - DowngradeLock -> RLock (via readerSem) - -package sync - -import ( - "sync" - "sync/atomic" - "unsafe" -) - -// DowngradableRWMutex is identical to sync.RWMutex, but adds the DowngradeLock -// method. -type DowngradableRWMutex struct { - w sync.Mutex // held if there are pending writers - writerSem uint32 // semaphore for writers to wait for completing readers - readerSem uint32 // semaphore for readers to wait for completing writers - readerCount int32 // number of pending readers - readerWait int32 // number of departing readers -} - -const rwmutexMaxReaders = 1 << 30 - -// RLock locks rw for reading. -func (rw *DowngradableRWMutex) RLock() { - if RaceEnabled { - RaceDisable() - } - if atomic.AddInt32(&rw.readerCount, 1) < 0 { - // A writer is pending, wait for it. - runtimeSemacquire(&rw.readerSem) - } - if RaceEnabled { - RaceEnable() - RaceAcquire(unsafe.Pointer(&rw.readerSem)) - } -} - -// RUnlock undoes a single RLock call. -func (rw *DowngradableRWMutex) RUnlock() { - if RaceEnabled { - // TODO: Why does this need to be ReleaseMerge instead of - // Release? IIUC this establishes Unlock happens-before RUnlock, which - // seems unnecessary. - RaceReleaseMerge(unsafe.Pointer(&rw.writerSem)) - RaceDisable() - } - if r := atomic.AddInt32(&rw.readerCount, -1); r < 0 { - if r+1 == 0 || r+1 == -rwmutexMaxReaders { - panic("RUnlock of unlocked DowngradableRWMutex") - } - // A writer is pending. - if atomic.AddInt32(&rw.readerWait, -1) == 0 { - // The last reader unblocks the writer. - runtimeSemrelease(&rw.writerSem, false) - } - } - if RaceEnabled { - RaceEnable() - } -} - -// Lock locks rw for writing. -func (rw *DowngradableRWMutex) Lock() { - if RaceEnabled { - RaceDisable() - } - // First, resolve competition with other writers. - rw.w.Lock() - // Announce to readers there is a pending writer. - r := atomic.AddInt32(&rw.readerCount, -rwmutexMaxReaders) + rwmutexMaxReaders - // Wait for active readers. - if r != 0 && atomic.AddInt32(&rw.readerWait, r) != 0 { - runtimeSemacquire(&rw.writerSem) - } - if RaceEnabled { - RaceEnable() - RaceAcquire(unsafe.Pointer(&rw.writerSem)) - } -} - -// Unlock unlocks rw for writing. -func (rw *DowngradableRWMutex) Unlock() { - if RaceEnabled { - RaceRelease(unsafe.Pointer(&rw.writerSem)) - RaceRelease(unsafe.Pointer(&rw.readerSem)) - RaceDisable() - } - // Announce to readers there is no active writer. - r := atomic.AddInt32(&rw.readerCount, rwmutexMaxReaders) - if r >= rwmutexMaxReaders { - panic("Unlock of unlocked DowngradableRWMutex") - } - // Unblock blocked readers, if any. - for i := 0; i < int(r); i++ { - runtimeSemrelease(&rw.readerSem, false) - } - // Allow other writers to proceed. - rw.w.Unlock() - if RaceEnabled { - RaceEnable() - } -} - -// DowngradeLock atomically unlocks rw for writing and locks it for reading. -func (rw *DowngradableRWMutex) DowngradeLock() { - if RaceEnabled { - RaceRelease(unsafe.Pointer(&rw.readerSem)) - RaceDisable() - } - // Announce to readers there is no active writer and one additional reader. - r := atomic.AddInt32(&rw.readerCount, rwmutexMaxReaders+1) - if r >= rwmutexMaxReaders+1 { - panic("DowngradeLock of unlocked DowngradableRWMutex") - } - // Unblock blocked readers, if any. Note that this loop starts as 1 since r - // includes this goroutine. - for i := 1; i < int(r); i++ { - runtimeSemrelease(&rw.readerSem, false) - } - // Allow other writers to proceed to rw.w.Lock(). Note that they will still - // block on rw.writerSem since at least this reader exists, such that - // DowngradeLock() is atomic with the previous write lock. - rw.w.Unlock() - if RaceEnabled { - RaceEnable() - } -} - -//go:linkname runtimeSemacquire sync.runtime_Semacquire -func runtimeSemacquire(s *uint32) - -//go:linkname runtimeSemrelease sync.runtime_Semrelease -func runtimeSemrelease(s *uint32, handoff bool) diff --git a/pkg/sync/memmove_unsafe.go b/pkg/sync/memmove_unsafe.go deleted file mode 100644 index cd7a02dca..000000000 --- a/pkg/sync/memmove_unsafe.go +++ /dev/null @@ -1,32 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package sync - -import ( - "unsafe" -) - -// Memmove is exported for SeqAtomicLoad/SeqAtomicTryLoad, which can't -// define it because go_generics can't update the go:linkname annotation. -// Furthermore, go:linkname silently doesn't work if the local name is exported -// (this is of course undocumented), which is why this indirection is -// necessary. -func Memmove(to, from unsafe.Pointer, n uintptr) { - memmove(to, from, n) -} - -//go:linkname memmove runtime.memmove -//go:noescape -func memmove(to, from unsafe.Pointer, n uintptr) diff --git a/pkg/sync/norace_unsafe.go b/pkg/sync/norace_unsafe.go deleted file mode 100644 index 1593b9e5d..000000000 --- a/pkg/sync/norace_unsafe.go +++ /dev/null @@ -1,44 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// +build !race - -package sync - -import ( - "unsafe" -) - -// RaceEnabled is true if the Go data race detector is enabled. -const RaceEnabled = false - -// RaceDisable has the same semantics as runtime.RaceDisable. -func RaceDisable() { -} - -// RaceEnable has the same semantics as runtime.RaceEnable. -func RaceEnable() { -} - -// RaceAcquire has the same semantics as runtime.RaceAcquire. -func RaceAcquire(addr unsafe.Pointer) { -} - -// RaceRelease has the same semantics as runtime.RaceRelease. -func RaceRelease(addr unsafe.Pointer) { -} - -// RaceReleaseMerge has the same semantics as runtime.RaceReleaseMerge. -func RaceReleaseMerge(addr unsafe.Pointer) { -} diff --git a/pkg/sync/race_unsafe.go b/pkg/sync/race_unsafe.go deleted file mode 100644 index 473eaddc6..000000000 --- a/pkg/sync/race_unsafe.go +++ /dev/null @@ -1,50 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// +build race - -package sync - -import ( - "runtime" - "unsafe" -) - -// RaceEnabled is true if the Go data race detector is enabled. -const RaceEnabled = true - -// RaceDisable has the same semantics as runtime.RaceDisable. -func RaceDisable() { - runtime.RaceDisable() -} - -// RaceEnable has the same semantics as runtime.RaceEnable. -func RaceEnable() { - runtime.RaceEnable() -} - -// RaceAcquire has the same semantics as runtime.RaceAcquire. -func RaceAcquire(addr unsafe.Pointer) { - runtime.RaceAcquire(addr) -} - -// RaceRelease has the same semantics as runtime.RaceRelease. -func RaceRelease(addr unsafe.Pointer) { - runtime.RaceRelease(addr) -} - -// RaceReleaseMerge has the same semantics as runtime.RaceReleaseMerge. -func RaceReleaseMerge(addr unsafe.Pointer) { - runtime.RaceReleaseMerge(addr) -} diff --git a/pkg/sync/seqatomic_unsafe.go b/pkg/sync/seqatomic_unsafe.go deleted file mode 100644 index bea31adc5..000000000 --- a/pkg/sync/seqatomic_unsafe.go +++ /dev/null @@ -1,81 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package template doesn't exist. This file must be instantiated using the -// go_template_instance rule in tools/go_generics/defs.bzl. -package template - -import ( - "fmt" - "reflect" - "strings" - "unsafe" - - ssync "gvisor.googlesource.com/gvisor/pkg/sync" -) - -// Value is a required type parameter. -// -// Value must not contain any pointers, including interface objects, function -// objects, slices, maps, channels, unsafe.Pointer, and arrays or structs -// containing any of the above. An init() function will panic if this property -// does not hold. -type Value struct{} - -// SeqAtomicLoad returns a copy of *ptr, ensuring that the read does not race -// with any writer critical sections in sc. -func SeqAtomicLoad(sc *ssync.SeqCount, ptr *Value) Value { - // This function doesn't use SeqAtomicTryLoad because doing so is - // measurably, significantly (~20%) slower; Go is awful at inlining. - var val Value - for { - epoch := sc.BeginRead() - if ssync.RaceEnabled { - // runtime.RaceDisable() doesn't actually stop the race detector, - // so it can't help us here. Instead, call runtime.memmove - // directly, which is not instrumented by the race detector. - ssync.Memmove(unsafe.Pointer(&val), unsafe.Pointer(ptr), unsafe.Sizeof(val)) - } else { - // This is ~40% faster for short reads than going through memmove. - val = *ptr - } - if sc.ReadOk(epoch) { - break - } - } - return val -} - -// SeqAtomicTryLoad returns a copy of *ptr while in a reader critical section -// in sc initiated by a call to sc.BeginRead() that returned epoch. If the read -// would race with a writer critical section, SeqAtomicTryLoad returns -// (unspecified, false). -func SeqAtomicTryLoad(sc *ssync.SeqCount, epoch ssync.SeqCountEpoch, ptr *Value) (Value, bool) { - var val Value - if ssync.RaceEnabled { - ssync.Memmove(unsafe.Pointer(&val), unsafe.Pointer(ptr), unsafe.Sizeof(val)) - } else { - val = *ptr - } - return val, sc.ReadOk(epoch) -} - -func init() { - var val Value - typ := reflect.TypeOf(val) - name := typ.Name() - if ptrs := ssync.PointersInType(typ, name); len(ptrs) != 0 { - panic(fmt.Sprintf("SeqAtomicLoad<%s> is invalid since values %s of type %s contain pointers:\n%s", typ, name, typ, strings.Join(ptrs, "\n"))) - } -} diff --git a/pkg/sync/seqatomictest/BUILD b/pkg/sync/seqatomictest/BUILD deleted file mode 100644 index 23132650a..000000000 --- a/pkg/sync/seqatomictest/BUILD +++ /dev/null @@ -1,35 +0,0 @@ -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") - -package(licenses = ["notice"]) - -load("//tools/go_generics:defs.bzl", "go_template_instance") - -go_template_instance( - name = "seqatomic_int", - out = "seqatomic_int.go", - package = "seqatomic", - suffix = "Int", - template = "//pkg/sync:generic_seqatomic", - types = { - "Value": "int", - }, -) - -go_library( - name = "seqatomic", - srcs = ["seqatomic_int.go"], - importpath = "gvisor.googlesource.com/gvisor/pkg/sync/seqatomic", - deps = [ - "//pkg/sync", - ], -) - -go_test( - name = "seqatomic_test", - size = "small", - srcs = ["seqatomic_test.go"], - embed = [":seqatomic"], - deps = [ - "//pkg/sync", - ], -) diff --git a/pkg/sync/seqatomictest/seqatomic_test.go b/pkg/sync/seqatomictest/seqatomic_test.go deleted file mode 100644 index f5e1fbfff..000000000 --- a/pkg/sync/seqatomictest/seqatomic_test.go +++ /dev/null @@ -1,132 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package seqatomic - -import ( - "sync/atomic" - "testing" - "time" - - gvsync "gvisor.googlesource.com/gvisor/pkg/sync" -) - -func TestSeqAtomicLoadUncontended(t *testing.T) { - var seq gvsync.SeqCount - const want = 1 - data := want - if got := SeqAtomicLoadInt(&seq, &data); got != want { - t.Errorf("SeqAtomicLoadInt: got %v, wanted %v", got, want) - } -} - -func TestSeqAtomicLoadAfterWrite(t *testing.T) { - var seq gvsync.SeqCount - var data int - const want = 1 - seq.BeginWrite() - data = want - seq.EndWrite() - if got := SeqAtomicLoadInt(&seq, &data); got != want { - t.Errorf("SeqAtomicLoadInt: got %v, wanted %v", got, want) - } -} - -func TestSeqAtomicLoadDuringWrite(t *testing.T) { - var seq gvsync.SeqCount - var data int - const want = 1 - seq.BeginWrite() - go func() { - time.Sleep(time.Second) - data = want - seq.EndWrite() - }() - if got := SeqAtomicLoadInt(&seq, &data); got != want { - t.Errorf("SeqAtomicLoadInt: got %v, wanted %v", got, want) - } -} - -func TestSeqAtomicTryLoadUncontended(t *testing.T) { - var seq gvsync.SeqCount - const want = 1 - data := want - epoch := seq.BeginRead() - if got, ok := SeqAtomicTryLoadInt(&seq, epoch, &data); !ok || got != want { - t.Errorf("SeqAtomicTryLoadInt: got (%v, %v), wanted (%v, true)", got, ok, want) - } -} - -func TestSeqAtomicTryLoadDuringWrite(t *testing.T) { - var seq gvsync.SeqCount - var data int - epoch := seq.BeginRead() - seq.BeginWrite() - if got, ok := SeqAtomicTryLoadInt(&seq, epoch, &data); ok { - t.Errorf("SeqAtomicTryLoadInt: got (%v, true), wanted (_, false)", got) - } - seq.EndWrite() -} - -func TestSeqAtomicTryLoadAfterWrite(t *testing.T) { - var seq gvsync.SeqCount - var data int - epoch := seq.BeginRead() - seq.BeginWrite() - seq.EndWrite() - if got, ok := SeqAtomicTryLoadInt(&seq, epoch, &data); ok { - t.Errorf("SeqAtomicTryLoadInt: got (%v, true), wanted (_, false)", got) - } -} - -func BenchmarkSeqAtomicLoadIntUncontended(b *testing.B) { - var seq gvsync.SeqCount - const want = 42 - data := want - b.RunParallel(func(pb *testing.PB) { - for pb.Next() { - if got := SeqAtomicLoadInt(&seq, &data); got != want { - b.Fatalf("SeqAtomicLoadInt: got %v, wanted %v", got, want) - } - } - }) -} - -func BenchmarkSeqAtomicTryLoadIntUncontended(b *testing.B) { - var seq gvsync.SeqCount - const want = 42 - data := want - b.RunParallel(func(pb *testing.PB) { - epoch := seq.BeginRead() - for pb.Next() { - if got, ok := SeqAtomicTryLoadInt(&seq, epoch, &data); !ok || got != want { - b.Fatalf("SeqAtomicTryLoadInt: got (%v, %v), wanted (%v, true)", got, ok, want) - } - } - }) -} - -// For comparison: -func BenchmarkAtomicValueLoadIntUncontended(b *testing.B) { - var a atomic.Value - const want = 42 - a.Store(int(want)) - b.RunParallel(func(pb *testing.PB) { - for pb.Next() { - if got := a.Load().(int); got != want { - b.Fatalf("atomic.Value.Load: got %v, wanted %v", got, want) - } - } - }) -} diff --git a/pkg/sync/seqcount.go b/pkg/sync/seqcount.go deleted file mode 100644 index 732e856a4..000000000 --- a/pkg/sync/seqcount.go +++ /dev/null @@ -1,158 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package sync - -import ( - "fmt" - "reflect" - "runtime" - "sync/atomic" -) - -// SeqCount is a synchronization primitive for optimistic reader/writer -// synchronization in cases where readers can work with stale data and -// therefore do not need to block writers. -// -// Compared to sync/atomic.Value: -// -// - Mutation of SeqCount-protected data does not require memory allocation, -// whereas atomic.Value generally does. This is a significant advantage when -// writes are common. -// -// - Atomic reads of SeqCount-protected data require copying. This is a -// disadvantage when atomic reads are common. -// -// - SeqCount may be more flexible: correct use of SeqCount.ReadOk allows other -// operations to be made atomic with reads of SeqCount-protected data. -// -// - SeqCount may be less flexible: as of this writing, SeqCount-protected data -// cannot include pointers. -// -// - SeqCount is more cumbersome to use; atomic reads of SeqCount-protected -// data require instantiating function templates using go_generics (see -// seqatomic.go). -type SeqCount struct { - // epoch is incremented by BeginWrite and EndWrite, such that epoch is odd - // if a writer critical section is active, and a read from data protected - // by this SeqCount is atomic iff epoch is the same even value before and - // after the read. - epoch uint32 -} - -// SeqCountEpoch tracks writer critical sections in a SeqCount. -type SeqCountEpoch struct { - val uint32 -} - -// We assume that: -// -// - All functions in sync/atomic that perform a memory read are at least a -// read fence: memory reads before calls to such functions cannot be reordered -// after the call, and memory reads after calls to such functions cannot be -// reordered before the call, even if those reads do not use sync/atomic. -// -// - All functions in sync/atomic that perform a memory write are at least a -// write fence: memory writes before calls to such functions cannot be -// reordered after the call, and memory writes after calls to such functions -// cannot be reordered before the call, even if those writes do not use -// sync/atomic. -// -// As of this writing, the Go memory model completely fails to describe -// sync/atomic, but these properties are implied by -// https://groups.google.com/forum/#!topic/golang-nuts/7EnEhM3U7B8. - -// BeginRead indicates the beginning of a reader critical section. Reader -// critical sections DO NOT BLOCK writer critical sections, so operations in a -// reader critical section MAY RACE with writer critical sections. Races are -// detected by ReadOk at the end of the reader critical section. Thus, the -// low-level structure of readers is generally: -// -// for { -// epoch := seq.BeginRead() -// // do something idempotent with seq-protected data -// if seq.ReadOk(epoch) { -// break -// } -// } -// -// However, since reader critical sections may race with writer critical -// sections, the Go race detector will (accurately) flag data races in readers -// using this pattern. Most users of SeqCount will need to use the -// SeqAtomicLoad function template in seqatomic.go. -func (s *SeqCount) BeginRead() SeqCountEpoch { - epoch := atomic.LoadUint32(&s.epoch) - for epoch&1 != 0 { - runtime.Gosched() - epoch = atomic.LoadUint32(&s.epoch) - } - return SeqCountEpoch{epoch} -} - -// ReadOk returns true if the reader critical section initiated by a previous -// call to BeginRead() that returned epoch did not race with any writer critical -// sections. -// -// ReadOk may be called any number of times during a reader critical section. -// Reader critical sections do not need to be explicitly terminated; the last -// call to ReadOk is implicitly the end of the reader critical section. -func (s *SeqCount) ReadOk(epoch SeqCountEpoch) bool { - return atomic.LoadUint32(&s.epoch) == epoch.val -} - -// BeginWrite indicates the beginning of a writer critical section. -// -// SeqCount does not support concurrent writer critical sections; clients with -// concurrent writers must synchronize them using e.g. sync.Mutex. -func (s *SeqCount) BeginWrite() { - if epoch := atomic.AddUint32(&s.epoch, 1); epoch&1 == 0 { - panic("SeqCount.BeginWrite during writer critical section") - } -} - -// EndWrite ends the effect of a preceding BeginWrite. -func (s *SeqCount) EndWrite() { - if epoch := atomic.AddUint32(&s.epoch, 1); epoch&1 != 0 { - panic("SeqCount.EndWrite outside writer critical section") - } -} - -// PointersInType returns a list of pointers reachable from values named -// valName of the given type. -// -// PointersInType is not exhaustive, but it is guaranteed that if typ contains -// at least one pointer, then PointersInTypeOf returns a non-empty list. -func PointersInType(typ reflect.Type, valName string) []string { - switch kind := typ.Kind(); kind { - case reflect.Bool, reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64, reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr, reflect.Float32, reflect.Float64, reflect.Complex64, reflect.Complex128: - return nil - - case reflect.Chan, reflect.Func, reflect.Interface, reflect.Map, reflect.Ptr, reflect.Slice, reflect.String, reflect.UnsafePointer: - return []string{valName} - - case reflect.Array: - return PointersInType(typ.Elem(), valName+"[]") - - case reflect.Struct: - var ptrs []string - for i, n := 0, typ.NumField(); i < n; i++ { - field := typ.Field(i) - ptrs = append(ptrs, PointersInType(field.Type, fmt.Sprintf("%s.%s", valName, field.Name))...) - } - return ptrs - - default: - return []string{fmt.Sprintf("%s (of type %s with unknown kind %s)", valName, typ, kind)} - } -} diff --git a/pkg/sync/seqcount_test.go b/pkg/sync/seqcount_test.go deleted file mode 100644 index b14a8878e..000000000 --- a/pkg/sync/seqcount_test.go +++ /dev/null @@ -1,162 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package sync - -import ( - "reflect" - "testing" - "time" -) - -func TestSeqCountWriteUncontended(t *testing.T) { - var seq SeqCount - seq.BeginWrite() - seq.EndWrite() -} - -func TestSeqCountReadUncontended(t *testing.T) { - var seq SeqCount - epoch := seq.BeginRead() - if !seq.ReadOk(epoch) { - t.Errorf("ReadOk: got false, wanted true") - } -} - -func TestSeqCountBeginReadAfterWrite(t *testing.T) { - var seq SeqCount - var data int32 - const want = 1 - seq.BeginWrite() - data = want - seq.EndWrite() - epoch := seq.BeginRead() - if data != want { - t.Errorf("Reader: got %v, wanted %v", data, want) - } - if !seq.ReadOk(epoch) { - t.Errorf("ReadOk: got false, wanted true") - } -} - -func TestSeqCountBeginReadDuringWrite(t *testing.T) { - var seq SeqCount - var data int - const want = 1 - seq.BeginWrite() - go func() { - time.Sleep(time.Second) - data = want - seq.EndWrite() - }() - epoch := seq.BeginRead() - if data != want { - t.Errorf("Reader: got %v, wanted %v", data, want) - } - if !seq.ReadOk(epoch) { - t.Errorf("ReadOk: got false, wanted true") - } -} - -func TestSeqCountReadOkAfterWrite(t *testing.T) { - var seq SeqCount - epoch := seq.BeginRead() - seq.BeginWrite() - seq.EndWrite() - if seq.ReadOk(epoch) { - t.Errorf("ReadOk: got true, wanted false") - } -} - -func TestSeqCountReadOkDuringWrite(t *testing.T) { - var seq SeqCount - epoch := seq.BeginRead() - seq.BeginWrite() - if seq.ReadOk(epoch) { - t.Errorf("ReadOk: got true, wanted false") - } - seq.EndWrite() -} - -func BenchmarkSeqCountWriteUncontended(b *testing.B) { - var seq SeqCount - for i := 0; i < b.N; i++ { - seq.BeginWrite() - seq.EndWrite() - } -} - -func BenchmarkSeqCountReadUncontended(b *testing.B) { - var seq SeqCount - b.RunParallel(func(pb *testing.PB) { - for pb.Next() { - epoch := seq.BeginRead() - if !seq.ReadOk(epoch) { - b.Fatalf("ReadOk: got false, wanted true") - } - } - }) -} - -func TestPointersInType(t *testing.T) { - for _, test := range []struct { - name string // used for both test and value name - val interface{} - ptrs []string - }{ - { - name: "EmptyStruct", - val: struct{}{}, - }, - { - name: "Int", - val: int(0), - }, - { - name: "MixedStruct", - val: struct { - b bool - I int - ExportedPtr *struct{} - unexportedPtr *struct{} - arr [2]int - ptrArr [2]*int - nestedStruct struct { - nestedNonptr int - nestedPtr *int - } - structArr [1]struct { - nonptr int - ptr *int - } - }{}, - ptrs: []string{ - "MixedStruct.ExportedPtr", - "MixedStruct.unexportedPtr", - "MixedStruct.ptrArr[]", - "MixedStruct.nestedStruct.nestedPtr", - "MixedStruct.structArr[].ptr", - }, - }, - } { - t.Run(test.name, func(t *testing.T) { - typ := reflect.TypeOf(test.val) - ptrs := PointersInType(typ, test.name) - t.Logf("Found pointers: %v", ptrs) - if (len(ptrs) != 0 || len(test.ptrs) != 0) && !reflect.DeepEqual(ptrs, test.ptrs) { - t.Errorf("Got %v, wanted %v", ptrs, test.ptrs) - } - }) - } -} diff --git a/pkg/sync/sync.go b/pkg/sync/sync.go deleted file mode 100644 index 22c5348d7..000000000 --- a/pkg/sync/sync.go +++ /dev/null @@ -1,16 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package sync provides synchronization primitives. -package sync diff --git a/third_party/gvsync/BUILD b/third_party/gvsync/BUILD new file mode 100644 index 000000000..4764eaa83 --- /dev/null +++ b/third_party/gvsync/BUILD @@ -0,0 +1,50 @@ +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") + +package( + default_visibility = ["//:sandbox"], + licenses = ["notice"], +) + +load("//tools/go_generics:defs.bzl", "go_template") + +go_template( + name = "generic_atomicptr", + srcs = ["atomicptr_unsafe.go"], + types = [ + "Value", + ], +) + +go_template( + name = "generic_seqatomic", + srcs = ["seqatomic_unsafe.go"], + types = [ + "Value", + ], + deps = [ + ":sync", + ], +) + +go_library( + name = "gvsync", + srcs = [ + "downgradable_rwmutex_unsafe.go", + "gvsync.go", + "memmove_unsafe.go", + "norace_unsafe.go", + "race_unsafe.go", + "seqcount.go", + ], + importpath = "gvisor.googlesource.com/gvisor/third_party/gvsync", +) + +go_test( + name = "gvsync_test", + size = "small", + srcs = [ + "downgradable_rwmutex_test.go", + "seqcount_test.go", + ], + embed = [":gvsync"], +) diff --git a/third_party/gvsync/LICENSE b/third_party/gvsync/LICENSE new file mode 100644 index 000000000..6a66aea5e --- /dev/null +++ b/third_party/gvsync/LICENSE @@ -0,0 +1,27 @@ +Copyright (c) 2009 The Go Authors. All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + * Redistributions of source code must retain the above copyright +notice, this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above +copyright notice, this list of conditions and the following disclaimer +in the documentation and/or other materials provided with the +distribution. + * Neither the name of Google Inc. nor the names of its +contributors may be used to endorse or promote products derived from +this software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/third_party/gvsync/README.md b/third_party/gvsync/README.md new file mode 100644 index 000000000..fcc7e6f44 --- /dev/null +++ b/third_party/gvsync/README.md @@ -0,0 +1,3 @@ +This package provides additional synchronization primitives not provided by the +Go stdlib 'sync' package. It is partially derived from the upstream 'sync' +package. diff --git a/third_party/gvsync/atomicptr_unsafe.go b/third_party/gvsync/atomicptr_unsafe.go new file mode 100644 index 000000000..da9f16240 --- /dev/null +++ b/third_party/gvsync/atomicptr_unsafe.go @@ -0,0 +1,37 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package template doesn't exist. This file must be instantiated using the +// go_template_instance rule in tools/go_generics/defs.bzl. +package template + +import ( + "sync/atomic" + "unsafe" +) + +// Value is a required type parameter. +type Value struct{} + +// An AtomicPtr is a pointer to a value of type Value that can be atomically +// loaded and stored. The zero value of an AtomicPtr represents nil. +// +// Note that copying AtomicPtr by value performs a non-atomic read of the +// stored pointer, which is unsafe if Store() can be called concurrently; in +// this case, do `dst.Store(src.Load())` instead. +type AtomicPtr struct { + ptr unsafe.Pointer +} + +// Load returns the value set by the most recent Store. It returns nil if there +// has been no previous call to Store. +func (p *AtomicPtr) Load() *Value { + return (*Value)(atomic.LoadPointer(&p.ptr)) +} + +// Store sets the value returned by Load to x. +func (p *AtomicPtr) Store(x *Value) { + atomic.StorePointer(&p.ptr, (unsafe.Pointer)(x)) +} diff --git a/third_party/gvsync/atomicptrtest/BUILD b/third_party/gvsync/atomicptrtest/BUILD new file mode 100644 index 000000000..74c51fd18 --- /dev/null +++ b/third_party/gvsync/atomicptrtest/BUILD @@ -0,0 +1,29 @@ +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") + +package(licenses = ["notice"]) + +load("//tools/go_generics:defs.bzl", "go_template_instance") + +go_template_instance( + name = "atomicptr_int", + out = "atomicptr_int.go", + package = "atomicptr", + suffix = "Int", + template = "//third_party/gvsync:generic_atomicptr", + types = { + "Value": "int", + }, +) + +go_library( + name = "atomicptr", + srcs = ["atomicptr_int.go"], + importpath = "gvisor.googlesource.com/gvisor/third_party/gvsync/atomicptr", +) + +go_test( + name = "atomicptr_test", + size = "small", + srcs = ["atomicptr_test.go"], + embed = [":atomicptr"], +) diff --git a/third_party/gvsync/atomicptrtest/atomicptr_test.go b/third_party/gvsync/atomicptrtest/atomicptr_test.go new file mode 100644 index 000000000..15d0936d4 --- /dev/null +++ b/third_party/gvsync/atomicptrtest/atomicptr_test.go @@ -0,0 +1,31 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package atomicptr + +import ( + "testing" +) + +func newInt(val int) *int { + return &val +} + +func TestAtomicPtr(t *testing.T) { + var p AtomicPtrInt + if got := p.Load(); got != nil { + t.Errorf("initial value is %p (%v), wanted nil", got, got) + } + want := newInt(42) + p.Store(want) + if got := p.Load(); got != want { + t.Errorf("wrong value: got %p (%v), wanted %p (%v)", got, got, want, want) + } + want = newInt(100) + p.Store(want) + if got := p.Load(); got != want { + t.Errorf("wrong value: got %p (%v), wanted %p (%v)", got, got, want, want) + } +} diff --git a/third_party/gvsync/downgradable_rwmutex_test.go b/third_party/gvsync/downgradable_rwmutex_test.go new file mode 100644 index 000000000..6517dd5dc --- /dev/null +++ b/third_party/gvsync/downgradable_rwmutex_test.go @@ -0,0 +1,150 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Copyright 2019 Google LLC +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// GOMAXPROCS=10 go test + +// Copy/pasted from the standard library's sync/rwmutex_test.go, except for the +// addition of downgradingWriter and the renaming of num_iterations to +// numIterations to shut up Golint. + +package gvsync + +import ( + "fmt" + "runtime" + "sync/atomic" + "testing" +) + +func parallelReader(m *DowngradableRWMutex, clocked, cunlock, cdone chan bool) { + m.RLock() + clocked <- true + <-cunlock + m.RUnlock() + cdone <- true +} + +func doTestParallelReaders(numReaders, gomaxprocs int) { + runtime.GOMAXPROCS(gomaxprocs) + var m DowngradableRWMutex + clocked := make(chan bool) + cunlock := make(chan bool) + cdone := make(chan bool) + for i := 0; i < numReaders; i++ { + go parallelReader(&m, clocked, cunlock, cdone) + } + // Wait for all parallel RLock()s to succeed. + for i := 0; i < numReaders; i++ { + <-clocked + } + for i := 0; i < numReaders; i++ { + cunlock <- true + } + // Wait for the goroutines to finish. + for i := 0; i < numReaders; i++ { + <-cdone + } +} + +func TestParallelReaders(t *testing.T) { + defer runtime.GOMAXPROCS(runtime.GOMAXPROCS(-1)) + doTestParallelReaders(1, 4) + doTestParallelReaders(3, 4) + doTestParallelReaders(4, 2) +} + +func reader(rwm *DowngradableRWMutex, numIterations int, activity *int32, cdone chan bool) { + for i := 0; i < numIterations; i++ { + rwm.RLock() + n := atomic.AddInt32(activity, 1) + if n < 1 || n >= 10000 { + panic(fmt.Sprintf("wlock(%d)\n", n)) + } + for i := 0; i < 100; i++ { + } + atomic.AddInt32(activity, -1) + rwm.RUnlock() + } + cdone <- true +} + +func writer(rwm *DowngradableRWMutex, numIterations int, activity *int32, cdone chan bool) { + for i := 0; i < numIterations; i++ { + rwm.Lock() + n := atomic.AddInt32(activity, 10000) + if n != 10000 { + panic(fmt.Sprintf("wlock(%d)\n", n)) + } + for i := 0; i < 100; i++ { + } + atomic.AddInt32(activity, -10000) + rwm.Unlock() + } + cdone <- true +} + +func downgradingWriter(rwm *DowngradableRWMutex, numIterations int, activity *int32, cdone chan bool) { + for i := 0; i < numIterations; i++ { + rwm.Lock() + n := atomic.AddInt32(activity, 10000) + if n != 10000 { + panic(fmt.Sprintf("wlock(%d)\n", n)) + } + for i := 0; i < 100; i++ { + } + atomic.AddInt32(activity, -10000) + rwm.DowngradeLock() + n = atomic.AddInt32(activity, 1) + if n < 1 || n >= 10000 { + panic(fmt.Sprintf("wlock(%d)\n", n)) + } + for i := 0; i < 100; i++ { + } + n = atomic.AddInt32(activity, -1) + rwm.RUnlock() + } + cdone <- true +} + +func HammerDowngradableRWMutex(gomaxprocs, numReaders, numIterations int) { + runtime.GOMAXPROCS(gomaxprocs) + // Number of active readers + 10000 * number of active writers. + var activity int32 + var rwm DowngradableRWMutex + cdone := make(chan bool) + go writer(&rwm, numIterations, &activity, cdone) + go downgradingWriter(&rwm, numIterations, &activity, cdone) + var i int + for i = 0; i < numReaders/2; i++ { + go reader(&rwm, numIterations, &activity, cdone) + } + go writer(&rwm, numIterations, &activity, cdone) + go downgradingWriter(&rwm, numIterations, &activity, cdone) + for ; i < numReaders; i++ { + go reader(&rwm, numIterations, &activity, cdone) + } + // Wait for the 4 writers and all readers to finish. + for i := 0; i < 4+numReaders; i++ { + <-cdone + } +} + +func TestDowngradableRWMutex(t *testing.T) { + defer runtime.GOMAXPROCS(runtime.GOMAXPROCS(-1)) + n := 1000 + if testing.Short() { + n = 5 + } + HammerDowngradableRWMutex(1, 1, n) + HammerDowngradableRWMutex(1, 3, n) + HammerDowngradableRWMutex(1, 10, n) + HammerDowngradableRWMutex(4, 1, n) + HammerDowngradableRWMutex(4, 3, n) + HammerDowngradableRWMutex(4, 10, n) + HammerDowngradableRWMutex(10, 1, n) + HammerDowngradableRWMutex(10, 3, n) + HammerDowngradableRWMutex(10, 10, n) + HammerDowngradableRWMutex(10, 5, n) +} diff --git a/third_party/gvsync/downgradable_rwmutex_unsafe.go b/third_party/gvsync/downgradable_rwmutex_unsafe.go new file mode 100644 index 000000000..a63a0d084 --- /dev/null +++ b/third_party/gvsync/downgradable_rwmutex_unsafe.go @@ -0,0 +1,144 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Copyright 2019 Google LLC +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This is mostly copied from the standard library's sync/rwmutex.go. +// +// Happens-before relationships indicated to the race detector: +// - Unlock -> Lock (via writerSem) +// - Unlock -> RLock (via readerSem) +// - RUnlock -> Lock (via writerSem) +// - DowngradeLock -> RLock (via readerSem) + +package gvsync + +import ( + "sync" + "sync/atomic" + "unsafe" +) + +// DowngradableRWMutex is identical to sync.RWMutex, but adds the DowngradeLock +// method. +type DowngradableRWMutex struct { + w sync.Mutex // held if there are pending writers + writerSem uint32 // semaphore for writers to wait for completing readers + readerSem uint32 // semaphore for readers to wait for completing writers + readerCount int32 // number of pending readers + readerWait int32 // number of departing readers +} + +const rwmutexMaxReaders = 1 << 30 + +// RLock locks rw for reading. +func (rw *DowngradableRWMutex) RLock() { + if RaceEnabled { + RaceDisable() + } + if atomic.AddInt32(&rw.readerCount, 1) < 0 { + // A writer is pending, wait for it. + runtimeSemacquire(&rw.readerSem) + } + if RaceEnabled { + RaceEnable() + RaceAcquire(unsafe.Pointer(&rw.readerSem)) + } +} + +// RUnlock undoes a single RLock call. +func (rw *DowngradableRWMutex) RUnlock() { + if RaceEnabled { + // TODO: Why does this need to be ReleaseMerge instead of + // Release? IIUC this establishes Unlock happens-before RUnlock, which + // seems unnecessary. + RaceReleaseMerge(unsafe.Pointer(&rw.writerSem)) + RaceDisable() + } + if r := atomic.AddInt32(&rw.readerCount, -1); r < 0 { + if r+1 == 0 || r+1 == -rwmutexMaxReaders { + panic("RUnlock of unlocked DowngradableRWMutex") + } + // A writer is pending. + if atomic.AddInt32(&rw.readerWait, -1) == 0 { + // The last reader unblocks the writer. + runtimeSemrelease(&rw.writerSem, false) + } + } + if RaceEnabled { + RaceEnable() + } +} + +// Lock locks rw for writing. +func (rw *DowngradableRWMutex) Lock() { + if RaceEnabled { + RaceDisable() + } + // First, resolve competition with other writers. + rw.w.Lock() + // Announce to readers there is a pending writer. + r := atomic.AddInt32(&rw.readerCount, -rwmutexMaxReaders) + rwmutexMaxReaders + // Wait for active readers. + if r != 0 && atomic.AddInt32(&rw.readerWait, r) != 0 { + runtimeSemacquire(&rw.writerSem) + } + if RaceEnabled { + RaceEnable() + RaceAcquire(unsafe.Pointer(&rw.writerSem)) + } +} + +// Unlock unlocks rw for writing. +func (rw *DowngradableRWMutex) Unlock() { + if RaceEnabled { + RaceRelease(unsafe.Pointer(&rw.writerSem)) + RaceRelease(unsafe.Pointer(&rw.readerSem)) + RaceDisable() + } + // Announce to readers there is no active writer. + r := atomic.AddInt32(&rw.readerCount, rwmutexMaxReaders) + if r >= rwmutexMaxReaders { + panic("Unlock of unlocked DowngradableRWMutex") + } + // Unblock blocked readers, if any. + for i := 0; i < int(r); i++ { + runtimeSemrelease(&rw.readerSem, false) + } + // Allow other writers to proceed. + rw.w.Unlock() + if RaceEnabled { + RaceEnable() + } +} + +// DowngradeLock atomically unlocks rw for writing and locks it for reading. +func (rw *DowngradableRWMutex) DowngradeLock() { + if RaceEnabled { + RaceRelease(unsafe.Pointer(&rw.readerSem)) + RaceDisable() + } + // Announce to readers there is no active writer and one additional reader. + r := atomic.AddInt32(&rw.readerCount, rwmutexMaxReaders+1) + if r >= rwmutexMaxReaders+1 { + panic("DowngradeLock of unlocked DowngradableRWMutex") + } + // Unblock blocked readers, if any. Note that this loop starts as 1 since r + // includes this goroutine. + for i := 1; i < int(r); i++ { + runtimeSemrelease(&rw.readerSem, false) + } + // Allow other writers to proceed to rw.w.Lock(). Note that they will still + // block on rw.writerSem since at least this reader exists, such that + // DowngradeLock() is atomic with the previous write lock. + rw.w.Unlock() + if RaceEnabled { + RaceEnable() + } +} + +//go:linkname runtimeSemacquire sync.runtime_Semacquire +func runtimeSemacquire(s *uint32) + +//go:linkname runtimeSemrelease sync.runtime_Semrelease +func runtimeSemrelease(s *uint32, handoff bool) diff --git a/third_party/gvsync/gvsync.go b/third_party/gvsync/gvsync.go new file mode 100644 index 000000000..46a2565fd --- /dev/null +++ b/third_party/gvsync/gvsync.go @@ -0,0 +1,7 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package gvsync provides synchronization primitives. +package gvsync diff --git a/third_party/gvsync/memmove_unsafe.go b/third_party/gvsync/memmove_unsafe.go new file mode 100644 index 000000000..d483fc739 --- /dev/null +++ b/third_party/gvsync/memmove_unsafe.go @@ -0,0 +1,23 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package gvsync + +import ( + "unsafe" +) + +// Memmove is exported for SeqAtomicLoad/SeqAtomicTryLoad, which can't +// define it because go_generics can't update the go:linkname annotation. +// Furthermore, go:linkname silently doesn't work if the local name is exported +// (this is of course undocumented), which is why this indirection is +// necessary. +func Memmove(to, from unsafe.Pointer, n uintptr) { + memmove(to, from, n) +} + +//go:linkname memmove runtime.memmove +//go:noescape +func memmove(to, from unsafe.Pointer, n uintptr) diff --git a/third_party/gvsync/norace_unsafe.go b/third_party/gvsync/norace_unsafe.go new file mode 100644 index 000000000..f9c88d13f --- /dev/null +++ b/third_party/gvsync/norace_unsafe.go @@ -0,0 +1,35 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// +build !race + +package gvsync + +import ( + "unsafe" +) + +// RaceEnabled is true if the Go data race detector is enabled. +const RaceEnabled = false + +// RaceDisable has the same semantics as runtime.RaceDisable. +func RaceDisable() { +} + +// RaceEnable has the same semantics as runtime.RaceEnable. +func RaceEnable() { +} + +// RaceAcquire has the same semantics as runtime.RaceAcquire. +func RaceAcquire(addr unsafe.Pointer) { +} + +// RaceRelease has the same semantics as runtime.RaceRelease. +func RaceRelease(addr unsafe.Pointer) { +} + +// RaceReleaseMerge has the same semantics as runtime.RaceReleaseMerge. +func RaceReleaseMerge(addr unsafe.Pointer) { +} diff --git a/third_party/gvsync/race_unsafe.go b/third_party/gvsync/race_unsafe.go new file mode 100644 index 000000000..2cdcdf7f7 --- /dev/null +++ b/third_party/gvsync/race_unsafe.go @@ -0,0 +1,41 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// +build race + +package gvsync + +import ( + "runtime" + "unsafe" +) + +// RaceEnabled is true if the Go data race detector is enabled. +const RaceEnabled = true + +// RaceDisable has the same semantics as runtime.RaceDisable. +func RaceDisable() { + runtime.RaceDisable() +} + +// RaceEnable has the same semantics as runtime.RaceEnable. +func RaceEnable() { + runtime.RaceEnable() +} + +// RaceAcquire has the same semantics as runtime.RaceAcquire. +func RaceAcquire(addr unsafe.Pointer) { + runtime.RaceAcquire(addr) +} + +// RaceRelease has the same semantics as runtime.RaceRelease. +func RaceRelease(addr unsafe.Pointer) { + runtime.RaceRelease(addr) +} + +// RaceReleaseMerge has the same semantics as runtime.RaceReleaseMerge. +func RaceReleaseMerge(addr unsafe.Pointer) { + runtime.RaceReleaseMerge(addr) +} diff --git a/third_party/gvsync/seqatomic_unsafe.go b/third_party/gvsync/seqatomic_unsafe.go new file mode 100644 index 000000000..ef61503e2 --- /dev/null +++ b/third_party/gvsync/seqatomic_unsafe.go @@ -0,0 +1,72 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package template doesn't exist. This file must be instantiated using the +// go_template_instance rule in tools/go_generics/defs.bzl. +package template + +import ( + "fmt" + "reflect" + "strings" + "unsafe" + + "gvisor.googlesource.com/gvisor/third_party/gvsync" +) + +// Value is a required type parameter. +// +// Value must not contain any pointers, including interface objects, function +// objects, slices, maps, channels, unsafe.Pointer, and arrays or structs +// containing any of the above. An init() function will panic if this property +// does not hold. +type Value struct{} + +// SeqAtomicLoad returns a copy of *ptr, ensuring that the read does not race +// with any writer critical sections in sc. +func SeqAtomicLoad(sc *gvsync.SeqCount, ptr *Value) Value { + // This function doesn't use SeqAtomicTryLoad because doing so is + // measurably, significantly (~20%) slower; Go is awful at inlining. + var val Value + for { + epoch := sc.BeginRead() + if gvsync.RaceEnabled { + // runtime.RaceDisable() doesn't actually stop the race detector, + // so it can't help us here. Instead, call runtime.memmove + // directly, which is not instrumented by the race detector. + gvsync.Memmove(unsafe.Pointer(&val), unsafe.Pointer(ptr), unsafe.Sizeof(val)) + } else { + // This is ~40% faster for short reads than going through memmove. + val = *ptr + } + if sc.ReadOk(epoch) { + break + } + } + return val +} + +// SeqAtomicTryLoad returns a copy of *ptr while in a reader critical section +// in sc initiated by a call to sc.BeginRead() that returned epoch. If the read +// would race with a writer critical section, SeqAtomicTryLoad returns +// (unspecified, false). +func SeqAtomicTryLoad(sc *gvsync.SeqCount, epoch gvsync.SeqCountEpoch, ptr *Value) (Value, bool) { + var val Value + if gvsync.RaceEnabled { + gvsync.Memmove(unsafe.Pointer(&val), unsafe.Pointer(ptr), unsafe.Sizeof(val)) + } else { + val = *ptr + } + return val, sc.ReadOk(epoch) +} + +func init() { + var val Value + typ := reflect.TypeOf(val) + name := typ.Name() + if ptrs := gvsync.PointersInType(typ, name); len(ptrs) != 0 { + panic(fmt.Sprintf("SeqAtomicLoad<%s> is invalid since values %s of type %s contain pointers:\n%s", typ, name, typ, strings.Join(ptrs, "\n"))) + } +} diff --git a/third_party/gvsync/seqatomictest/BUILD b/third_party/gvsync/seqatomictest/BUILD new file mode 100644 index 000000000..d83149e81 --- /dev/null +++ b/third_party/gvsync/seqatomictest/BUILD @@ -0,0 +1,35 @@ +load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test") + +package(licenses = ["notice"]) + +load("//tools/go_generics:defs.bzl", "go_template_instance") + +go_template_instance( + name = "seqatomic_int", + out = "seqatomic_int.go", + package = "seqatomic", + suffix = "Int", + template = "//third_party/gvsync:generic_seqatomic", + types = { + "Value": "int", + }, +) + +go_library( + name = "seqatomic", + srcs = ["seqatomic_int.go"], + importpath = "gvisor.googlesource.com/gvisor/third_party/gvsync/seqatomic", + deps = [ + "//third_party/gvsync", + ], +) + +go_test( + name = "seqatomic_test", + size = "small", + srcs = ["seqatomic_test.go"], + embed = [":seqatomic"], + deps = [ + "//third_party/gvsync", + ], +) diff --git a/third_party/gvsync/seqatomictest/seqatomic_test.go b/third_party/gvsync/seqatomictest/seqatomic_test.go new file mode 100644 index 000000000..d0c373bae --- /dev/null +++ b/third_party/gvsync/seqatomictest/seqatomic_test.go @@ -0,0 +1,132 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package seqatomic + +import ( + "sync/atomic" + "testing" + "time" + + "gvisor.googlesource.com/gvisor/third_party/gvsync" +) + +func TestSeqAtomicLoadUncontended(t *testing.T) { + var seq gvsync.SeqCount + const want = 1 + data := want + if got := SeqAtomicLoadInt(&seq, &data); got != want { + t.Errorf("SeqAtomicLoadInt: got %v, wanted %v", got, want) + } +} + +func TestSeqAtomicLoadAfterWrite(t *testing.T) { + var seq gvsync.SeqCount + var data int + const want = 1 + seq.BeginWrite() + data = want + seq.EndWrite() + if got := SeqAtomicLoadInt(&seq, &data); got != want { + t.Errorf("SeqAtomicLoadInt: got %v, wanted %v", got, want) + } +} + +func TestSeqAtomicLoadDuringWrite(t *testing.T) { + var seq gvsync.SeqCount + var data int + const want = 1 + seq.BeginWrite() + go func() { + time.Sleep(time.Second) + data = want + seq.EndWrite() + }() + if got := SeqAtomicLoadInt(&seq, &data); got != want { + t.Errorf("SeqAtomicLoadInt: got %v, wanted %v", got, want) + } +} + +func TestSeqAtomicTryLoadUncontended(t *testing.T) { + var seq gvsync.SeqCount + const want = 1 + data := want + epoch := seq.BeginRead() + if got, ok := SeqAtomicTryLoadInt(&seq, epoch, &data); !ok || got != want { + t.Errorf("SeqAtomicTryLoadInt: got (%v, %v), wanted (%v, true)", got, ok, want) + } +} + +func TestSeqAtomicTryLoadDuringWrite(t *testing.T) { + var seq gvsync.SeqCount + var data int + epoch := seq.BeginRead() + seq.BeginWrite() + if got, ok := SeqAtomicTryLoadInt(&seq, epoch, &data); ok { + t.Errorf("SeqAtomicTryLoadInt: got (%v, true), wanted (_, false)", got) + } + seq.EndWrite() +} + +func TestSeqAtomicTryLoadAfterWrite(t *testing.T) { + var seq gvsync.SeqCount + var data int + epoch := seq.BeginRead() + seq.BeginWrite() + seq.EndWrite() + if got, ok := SeqAtomicTryLoadInt(&seq, epoch, &data); ok { + t.Errorf("SeqAtomicTryLoadInt: got (%v, true), wanted (_, false)", got) + } +} + +func BenchmarkSeqAtomicLoadIntUncontended(b *testing.B) { + var seq gvsync.SeqCount + const want = 42 + data := want + b.RunParallel(func(pb *testing.PB) { + for pb.Next() { + if got := SeqAtomicLoadInt(&seq, &data); got != want { + b.Fatalf("SeqAtomicLoadInt: got %v, wanted %v", got, want) + } + } + }) +} + +func BenchmarkSeqAtomicTryLoadIntUncontended(b *testing.B) { + var seq gvsync.SeqCount + const want = 42 + data := want + b.RunParallel(func(pb *testing.PB) { + epoch := seq.BeginRead() + for pb.Next() { + if got, ok := SeqAtomicTryLoadInt(&seq, epoch, &data); !ok || got != want { + b.Fatalf("SeqAtomicTryLoadInt: got (%v, %v), wanted (%v, true)", got, ok, want) + } + } + }) +} + +// For comparison: +func BenchmarkAtomicValueLoadIntUncontended(b *testing.B) { + var a atomic.Value + const want = 42 + a.Store(int(want)) + b.RunParallel(func(pb *testing.PB) { + for pb.Next() { + if got := a.Load().(int); got != want { + b.Fatalf("atomic.Value.Load: got %v, wanted %v", got, want) + } + } + }) +} diff --git a/third_party/gvsync/seqcount.go b/third_party/gvsync/seqcount.go new file mode 100644 index 000000000..c7ae91cfa --- /dev/null +++ b/third_party/gvsync/seqcount.go @@ -0,0 +1,149 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package gvsync + +import ( + "fmt" + "reflect" + "runtime" + "sync/atomic" +) + +// SeqCount is a synchronization primitive for optimistic reader/writer +// synchronization in cases where readers can work with stale data and +// therefore do not need to block writers. +// +// Compared to sync/atomic.Value: +// +// - Mutation of SeqCount-protected data does not require memory allocation, +// whereas atomic.Value generally does. This is a significant advantage when +// writes are common. +// +// - Atomic reads of SeqCount-protected data require copying. This is a +// disadvantage when atomic reads are common. +// +// - SeqCount may be more flexible: correct use of SeqCount.ReadOk allows other +// operations to be made atomic with reads of SeqCount-protected data. +// +// - SeqCount may be less flexible: as of this writing, SeqCount-protected data +// cannot include pointers. +// +// - SeqCount is more cumbersome to use; atomic reads of SeqCount-protected +// data require instantiating function templates using go_generics (see +// seqatomic.go). +type SeqCount struct { + // epoch is incremented by BeginWrite and EndWrite, such that epoch is odd + // if a writer critical section is active, and a read from data protected + // by this SeqCount is atomic iff epoch is the same even value before and + // after the read. + epoch uint32 +} + +// SeqCountEpoch tracks writer critical sections in a SeqCount. +type SeqCountEpoch struct { + val uint32 +} + +// We assume that: +// +// - All functions in sync/atomic that perform a memory read are at least a +// read fence: memory reads before calls to such functions cannot be reordered +// after the call, and memory reads after calls to such functions cannot be +// reordered before the call, even if those reads do not use sync/atomic. +// +// - All functions in sync/atomic that perform a memory write are at least a +// write fence: memory writes before calls to such functions cannot be +// reordered after the call, and memory writes after calls to such functions +// cannot be reordered before the call, even if those writes do not use +// sync/atomic. +// +// As of this writing, the Go memory model completely fails to describe +// sync/atomic, but these properties are implied by +// https://groups.google.com/forum/#!topic/golang-nuts/7EnEhM3U7B8. + +// BeginRead indicates the beginning of a reader critical section. Reader +// critical sections DO NOT BLOCK writer critical sections, so operations in a +// reader critical section MAY RACE with writer critical sections. Races are +// detected by ReadOk at the end of the reader critical section. Thus, the +// low-level structure of readers is generally: +// +// for { +// epoch := seq.BeginRead() +// // do something idempotent with seq-protected data +// if seq.ReadOk(epoch) { +// break +// } +// } +// +// However, since reader critical sections may race with writer critical +// sections, the Go race detector will (accurately) flag data races in readers +// using this pattern. Most users of SeqCount will need to use the +// SeqAtomicLoad function template in seqatomic.go. +func (s *SeqCount) BeginRead() SeqCountEpoch { + epoch := atomic.LoadUint32(&s.epoch) + for epoch&1 != 0 { + runtime.Gosched() + epoch = atomic.LoadUint32(&s.epoch) + } + return SeqCountEpoch{epoch} +} + +// ReadOk returns true if the reader critical section initiated by a previous +// call to BeginRead() that returned epoch did not race with any writer critical +// sections. +// +// ReadOk may be called any number of times during a reader critical section. +// Reader critical sections do not need to be explicitly terminated; the last +// call to ReadOk is implicitly the end of the reader critical section. +func (s *SeqCount) ReadOk(epoch SeqCountEpoch) bool { + return atomic.LoadUint32(&s.epoch) == epoch.val +} + +// BeginWrite indicates the beginning of a writer critical section. +// +// SeqCount does not support concurrent writer critical sections; clients with +// concurrent writers must synchronize them using e.g. sync.Mutex. +func (s *SeqCount) BeginWrite() { + if epoch := atomic.AddUint32(&s.epoch, 1); epoch&1 == 0 { + panic("SeqCount.BeginWrite during writer critical section") + } +} + +// EndWrite ends the effect of a preceding BeginWrite. +func (s *SeqCount) EndWrite() { + if epoch := atomic.AddUint32(&s.epoch, 1); epoch&1 != 0 { + panic("SeqCount.EndWrite outside writer critical section") + } +} + +// PointersInType returns a list of pointers reachable from values named +// valName of the given type. +// +// PointersInType is not exhaustive, but it is guaranteed that if typ contains +// at least one pointer, then PointersInTypeOf returns a non-empty list. +func PointersInType(typ reflect.Type, valName string) []string { + switch kind := typ.Kind(); kind { + case reflect.Bool, reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64, reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr, reflect.Float32, reflect.Float64, reflect.Complex64, reflect.Complex128: + return nil + + case reflect.Chan, reflect.Func, reflect.Interface, reflect.Map, reflect.Ptr, reflect.Slice, reflect.String, reflect.UnsafePointer: + return []string{valName} + + case reflect.Array: + return PointersInType(typ.Elem(), valName+"[]") + + case reflect.Struct: + var ptrs []string + for i, n := 0, typ.NumField(); i < n; i++ { + field := typ.Field(i) + ptrs = append(ptrs, PointersInType(field.Type, fmt.Sprintf("%s.%s", valName, field.Name))...) + } + return ptrs + + default: + return []string{fmt.Sprintf("%s (of type %s with unknown kind %s)", valName, typ, kind)} + } +} diff --git a/third_party/gvsync/seqcount_test.go b/third_party/gvsync/seqcount_test.go new file mode 100644 index 000000000..ee6579ed8 --- /dev/null +++ b/third_party/gvsync/seqcount_test.go @@ -0,0 +1,153 @@ +// Copyright 2019 Google LLC +// +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package gvsync + +import ( + "reflect" + "testing" + "time" +) + +func TestSeqCountWriteUncontended(t *testing.T) { + var seq SeqCount + seq.BeginWrite() + seq.EndWrite() +} + +func TestSeqCountReadUncontended(t *testing.T) { + var seq SeqCount + epoch := seq.BeginRead() + if !seq.ReadOk(epoch) { + t.Errorf("ReadOk: got false, wanted true") + } +} + +func TestSeqCountBeginReadAfterWrite(t *testing.T) { + var seq SeqCount + var data int32 + const want = 1 + seq.BeginWrite() + data = want + seq.EndWrite() + epoch := seq.BeginRead() + if data != want { + t.Errorf("Reader: got %v, wanted %v", data, want) + } + if !seq.ReadOk(epoch) { + t.Errorf("ReadOk: got false, wanted true") + } +} + +func TestSeqCountBeginReadDuringWrite(t *testing.T) { + var seq SeqCount + var data int + const want = 1 + seq.BeginWrite() + go func() { + time.Sleep(time.Second) + data = want + seq.EndWrite() + }() + epoch := seq.BeginRead() + if data != want { + t.Errorf("Reader: got %v, wanted %v", data, want) + } + if !seq.ReadOk(epoch) { + t.Errorf("ReadOk: got false, wanted true") + } +} + +func TestSeqCountReadOkAfterWrite(t *testing.T) { + var seq SeqCount + epoch := seq.BeginRead() + seq.BeginWrite() + seq.EndWrite() + if seq.ReadOk(epoch) { + t.Errorf("ReadOk: got true, wanted false") + } +} + +func TestSeqCountReadOkDuringWrite(t *testing.T) { + var seq SeqCount + epoch := seq.BeginRead() + seq.BeginWrite() + if seq.ReadOk(epoch) { + t.Errorf("ReadOk: got true, wanted false") + } + seq.EndWrite() +} + +func BenchmarkSeqCountWriteUncontended(b *testing.B) { + var seq SeqCount + for i := 0; i < b.N; i++ { + seq.BeginWrite() + seq.EndWrite() + } +} + +func BenchmarkSeqCountReadUncontended(b *testing.B) { + var seq SeqCount + b.RunParallel(func(pb *testing.PB) { + for pb.Next() { + epoch := seq.BeginRead() + if !seq.ReadOk(epoch) { + b.Fatalf("ReadOk: got false, wanted true") + } + } + }) +} + +func TestPointersInType(t *testing.T) { + for _, test := range []struct { + name string // used for both test and value name + val interface{} + ptrs []string + }{ + { + name: "EmptyStruct", + val: struct{}{}, + }, + { + name: "Int", + val: int(0), + }, + { + name: "MixedStruct", + val: struct { + b bool + I int + ExportedPtr *struct{} + unexportedPtr *struct{} + arr [2]int + ptrArr [2]*int + nestedStruct struct { + nestedNonptr int + nestedPtr *int + } + structArr [1]struct { + nonptr int + ptr *int + } + }{}, + ptrs: []string{ + "MixedStruct.ExportedPtr", + "MixedStruct.unexportedPtr", + "MixedStruct.ptrArr[]", + "MixedStruct.nestedStruct.nestedPtr", + "MixedStruct.structArr[].ptr", + }, + }, + } { + t.Run(test.name, func(t *testing.T) { + typ := reflect.TypeOf(test.val) + ptrs := PointersInType(typ, test.name) + t.Logf("Found pointers: %v", ptrs) + if (len(ptrs) != 0 || len(test.ptrs) != 0) && !reflect.DeepEqual(ptrs, test.ptrs) { + t.Errorf("Got %v, wanted %v", ptrs, test.ptrs) + } + }) + } +} -- cgit v1.2.3 From 9ef3427ac14a84002497f3c8bac346486cb36f2b Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Thu, 7 Feb 2019 11:40:45 -0800 Subject: Implement semctl(2) SETALL and GETALL PiperOrigin-RevId: 232914984 Change-Id: Id2643d7ad8e986ca9be76d860788a71db2674cda --- pkg/sentry/kernel/semaphore/semaphore.go | 65 +++++++++++++++++++++++++++++--- pkg/sentry/syscalls/linux/sys_sem.go | 38 +++++++++++++++++++ test/syscalls/linux/semaphore.cc | 25 +++++++++++- 3 files changed, 122 insertions(+), 6 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index 232a276dc..c134931cd 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -16,6 +16,7 @@ package semaphore import ( + "fmt" "sync" "gvisor.googlesource.com/gvisor/pkg/abi/linux" @@ -75,7 +76,10 @@ type Set struct { perms fs.FilePermissions opTime ktime.Time changeTime ktime.Time - sems []sem + + // sems holds all semaphores in the set. The slice itself is immutable after + // it's been set, however each 'sem' object in the slice requires 'mu' lock. + sems []sem // dead is set to true when the set is removed and can't be reached anymore. // All waiters must wake up and fail when set is dead. @@ -136,7 +140,7 @@ func (r *Registry) FindOrCreate(ctx context.Context, key, nsems int32, mode linu } // Validate parameters. - if nsems > int32(set.size()) { + if nsems > int32(set.Size()) { return nil, syserror.EINVAL } if create && exclusive { @@ -244,19 +248,20 @@ func (r *Registry) findByKey(key int32) *Set { func (r *Registry) totalSems() int { totalSems := 0 for _, v := range r.semaphores { - totalSems += v.size() + totalSems += v.Size() } return totalSems } func (s *Set) findSem(num int32) *sem { - if num < 0 || int(num) >= s.size() { + if num < 0 || int(num) >= s.Size() { return nil } return &s.sems[num] } -func (s *Set) size() int { +// Size returns the number of semaphores in the set. Size is immutable. +func (s *Set) Size() int { return len(s.sems) } @@ -303,6 +308,39 @@ func (s *Set) SetVal(ctx context.Context, num int32, val int16, creds *auth.Cred return nil } +// SetValAll overrides all semaphores values, waking up waiters as needed. +// +// 'len(vals)' must be equal to 's.Size()'. +func (s *Set) SetValAll(ctx context.Context, vals []uint16, creds *auth.Credentials) error { + if len(vals) != s.Size() { + panic(fmt.Sprintf("vals length (%d) different that Set.Size() (%d)", len(vals), s.Size())) + } + + for _, val := range vals { + if val < 0 || val > valueMax { + return syserror.ERANGE + } + } + + s.mu.Lock() + defer s.mu.Unlock() + + // "The calling process must have alter permission on the semaphore set." + if !s.checkPerms(creds, fs.PermMask{Write: true}) { + return syserror.EACCES + } + + for i, val := range vals { + sem := &s.sems[i] + + // TODO: Clear undo entries in all processes + sem.value = int16(val) + sem.wakeWaiters() + } + s.changeTime = ktime.NowFromContext(ctx) + return nil +} + // GetVal returns a semaphore value. func (s *Set) GetVal(num int32, creds *auth.Credentials) (int16, error) { s.mu.Lock() @@ -320,6 +358,23 @@ func (s *Set) GetVal(num int32, creds *auth.Credentials) (int16, error) { return sem.value, nil } +// GetValAll returns value for all semaphores. +func (s *Set) GetValAll(creds *auth.Credentials) ([]uint16, error) { + s.mu.Lock() + defer s.mu.Unlock() + + // "The calling process must have read permission on the semaphore set." + if !s.checkPerms(creds, fs.PermMask{Read: true}) { + return nil, syserror.EACCES + } + + vals := make([]uint16, s.Size()) + for i, sem := range s.sems { + vals[i] = uint16(sem.value) + } + return vals, nil +} + // ExecuteOps attempts to execute a list of operations to the set. It only // succeeds when all operations can be applied. No changes are made if it fails. // diff --git a/pkg/sentry/syscalls/linux/sys_sem.go b/pkg/sentry/syscalls/linux/sys_sem.go index 4ed52c4a7..6775725ca 100644 --- a/pkg/sentry/syscalls/linux/sys_sem.go +++ b/pkg/sentry/syscalls/linux/sys_sem.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -97,10 +98,18 @@ func Semctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal } return 0, nil, setVal(t, id, num, int16(val)) + case linux.SETALL: + array := args[3].Pointer() + return 0, nil, setValAll(t, id, array) + case linux.GETVAL: v, err := getVal(t, id, num) return uintptr(v), nil, err + case linux.GETALL: + array := args[3].Pointer() + return 0, nil, getValAll(t, id, array) + case linux.IPC_RMID: return 0, nil, remove(t, id) @@ -155,6 +164,20 @@ func setVal(t *kernel.Task, id int32, num int32, val int16) error { return set.SetVal(t, num, val, creds) } +func setValAll(t *kernel.Task, id int32, array usermem.Addr) error { + r := t.IPCNamespace().SemaphoreRegistry() + set := r.FindByID(id) + if set == nil { + return syserror.EINVAL + } + vals := make([]uint16, set.Size()) + if _, err := t.CopyIn(array, vals); err != nil { + return err + } + creds := auth.CredentialsFromContext(t) + return set.SetValAll(t, vals, creds) +} + func getVal(t *kernel.Task, id int32, num int32) (int16, error) { r := t.IPCNamespace().SemaphoreRegistry() set := r.FindByID(id) @@ -164,3 +187,18 @@ func getVal(t *kernel.Task, id int32, num int32) (int16, error) { creds := auth.CredentialsFromContext(t) return set.GetVal(num, creds) } + +func getValAll(t *kernel.Task, id int32, array usermem.Addr) error { + r := t.IPCNamespace().SemaphoreRegistry() + set := r.FindByID(id) + if set == nil { + return syserror.EINVAL + } + creds := auth.CredentialsFromContext(t) + vals, err := set.GetValAll(creds) + if err != nil { + return err + } + _, err = t.CopyOut(array, vals) + return err +} diff --git a/test/syscalls/linux/semaphore.cc b/test/syscalls/linux/semaphore.cc index 12e33732d..da3d2c6fe 100644 --- a/test/syscalls/linux/semaphore.cc +++ b/test/syscalls/linux/semaphore.cc @@ -223,7 +223,7 @@ TEST(SemaphoreTest, SemOpRemoveWithWaiter_NoRandomSave) { }); // This must happen before IPC_RMID runs above. Otherwise it fails with EINVAL - // instead because the semaphire has already been removed. + // instead because the semaphore has already been removed. struct sembuf buf = {}; buf.sem_op = -1; ASSERT_THAT(RetryEINTR(semop)(sem.get(), &buf, 1), @@ -408,6 +408,29 @@ TEST(SemaphoreTest, SemCtlVal) { thZero.Join(); } +TEST(SemaphoreTest, SemCtlValAll) { + AutoSem sem(semget(IPC_PRIVATE, 3, 0600 | IPC_CREAT)); + ASSERT_THAT(sem.get(), SyscallSucceeds()); + + // Semaphores must start with 0. + uint16_t get[3] = {10, 10, 10}; + EXPECT_THAT(semctl(sem.get(), 1, GETALL, get), SyscallSucceedsWithValue(0)); + for (auto v : get) { + EXPECT_EQ(v, 0); + } + + // SetAll and check that they were set. + uint16_t vals[3] = {0, 10, 20}; + EXPECT_THAT(semctl(sem.get(), 1, SETALL, vals), SyscallSucceedsWithValue(0)); + EXPECT_THAT(semctl(sem.get(), 1, GETALL, get), SyscallSucceedsWithValue(0)); + for (size_t i = 0; i < ABSL_ARRAYSIZE(vals); ++i) { + EXPECT_EQ(get[i], vals[i]); + } + + EXPECT_THAT(semctl(sem.get(), 1, SETALL, nullptr), + SyscallFailsWithErrno(EFAULT)); +} + TEST(SemaphoreTest, SemIpcSet) { // Drop CAP_IPC_OWNER which allows us to bypass semaphore permissions. ASSERT_NO_ERRNO(SetCapability(CAP_IPC_OWNER, false)); -- cgit v1.2.3 From 2ba74f84be8b9d3d588fb834414d151607799fd3 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Thu, 7 Feb 2019 14:43:18 -0800 Subject: Implement /proc/net/unix. PiperOrigin-RevId: 232948478 Change-Id: Ib830121e5e79afaf5d38d17aeef5a1ef97913d23 --- pkg/abi/linux/socket.go | 16 +++ pkg/sentry/fs/proc/BUILD | 3 + pkg/sentry/fs/proc/net.go | 126 +++++++++++++++++- pkg/sentry/fs/proc/proc.go | 2 +- pkg/sentry/kernel/kernel.go | 58 ++++++++- pkg/sentry/socket/socket.go | 10 +- pkg/sentry/socket/unix/unix.go | 2 + test/syscalls/BUILD | 7 + test/syscalls/linux/BUILD | 17 +++ test/syscalls/linux/proc_net_unix.cc | 246 +++++++++++++++++++++++++++++++++++ 10 files changed, 481 insertions(+), 6 deletions(-) create mode 100644 test/syscalls/linux/proc_net_unix.cc (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/socket.go b/pkg/abi/linux/socket.go index 929814752..a5f78506a 100644 --- a/pkg/abi/linux/socket.go +++ b/pkg/abi/linux/socket.go @@ -191,6 +191,15 @@ const ( SO_TXTIME = 61 ) +// enum socket_state, from uapi/linux/net.h. +const ( + SS_FREE = 0 // Not allocated. + SS_UNCONNECTED = 1 // Unconnected to any socket. + SS_CONNECTING = 2 // In process of connecting. + SS_CONNECTED = 3 // Connected to socket. + SS_DISCONNECTING = 4 // In process of disconnecting. +) + // SockAddrMax is the maximum size of a struct sockaddr, from // uapi/linux/socket.h. const SockAddrMax = 128 @@ -343,3 +352,10 @@ const SizeOfControlMessageRight = 4 // SCM_MAX_FD is the maximum number of FDs accepted in a single sendmsg call. // From net/scm.h. const SCM_MAX_FD = 253 + +// SO_ACCEPTCON is defined as __SO_ACCEPTCON in +// include/uapi/linux/net.h, which represents a listening socket +// state. Note that this is distinct from SO_ACCEPTCONN, which is a +// socket option for querying whether a socket is in a listening +// state. +const SO_ACCEPTCON = 1 << 16 diff --git a/pkg/sentry/fs/proc/BUILD b/pkg/sentry/fs/proc/BUILD index f6bc90634..666b0ab3a 100644 --- a/pkg/sentry/fs/proc/BUILD +++ b/pkg/sentry/fs/proc/BUILD @@ -30,6 +30,7 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/log", "//pkg/sentry/context", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", @@ -43,6 +44,8 @@ go_library( "//pkg/sentry/kernel/time", "//pkg/sentry/mm", "//pkg/sentry/socket/rpcinet", + "//pkg/sentry/socket/unix", + "//pkg/sentry/socket/unix/transport", "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/syserror", diff --git a/pkg/sentry/fs/proc/net.go b/pkg/sentry/fs/proc/net.go index 219eea7f8..55a958f9e 100644 --- a/pkg/sentry/fs/proc/net.go +++ b/pkg/sentry/fs/proc/net.go @@ -15,19 +15,24 @@ package proc import ( + "bytes" "fmt" "time" "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc/seqfile" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/sentry/inet" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix" + "gvisor.googlesource.com/gvisor/pkg/sentry/socket/unix/transport" ) // newNet creates a new proc net entry. -func (p *proc) newNetDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { +func (p *proc) newNetDir(ctx context.Context, k *kernel.Kernel, msrc *fs.MountSource) *fs.Inode { var contents map[string]*fs.Inode if s := p.k.NetworkStack(); s != nil { contents = map[string]*fs.Inode{ @@ -52,6 +57,8 @@ func (p *proc) newNetDir(ctx context.Context, msrc *fs.MountSource) *fs.Inode { "tcp": newStaticProcInode(ctx, msrc, []byte(" sl local_address rem_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode")), "udp": newStaticProcInode(ctx, msrc, []byte(" sl local_address rem_address st tx_queue rx_queue tr tm->when retrnsmt uid timeout inode ref pointer drops")), + + "unix": seqfile.NewSeqFileInode(ctx, &netUnix{k: k}, msrc), } if s.SupportsIPv6() { @@ -182,3 +189,120 @@ func (n *netDev) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([]se return data, 0 } + +// netUnix implements seqfile.SeqSource for /proc/net/unix. +// +// +stateify savable +type netUnix struct { + k *kernel.Kernel +} + +// NeedsUpdate implements seqfile.SeqSource.NeedsUpdate. +func (*netUnix) NeedsUpdate(generation int64) bool { + return true +} + +// ReadSeqFileData implements seqfile.SeqSource.ReadSeqFileData. +func (n *netUnix) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([]seqfile.SeqData, int64) { + if h != nil { + return []seqfile.SeqData{}, 0 + } + + var buf bytes.Buffer + // Header + fmt.Fprintf(&buf, "Num RefCount Protocol Flags Type St Inode Path\n") + + // Entries + for _, sref := range n.k.ListSockets(linux.AF_UNIX) { + s := sref.Get() + if s == nil { + log.Debugf("Couldn't resolve weakref %v in socket table, racing with destruction?", sref) + continue + } + sfile := s.(*fs.File) + sops, ok := sfile.FileOperations.(*unix.SocketOperations) + if !ok { + panic(fmt.Sprintf("Found non-unix socket file in unix socket table: %+v", sfile)) + } + + addr, err := sops.Endpoint().GetLocalAddress() + if err != nil { + log.Warningf("Failed to retrieve socket name from %+v: %v", sfile, err) + addr.Addr = "" + } + + sockFlags := 0 + if ce, ok := sops.Endpoint().(transport.ConnectingEndpoint); ok { + if ce.Listening() { + // For unix domain sockets, linux reports a single flag + // value if the socket is listening, of __SO_ACCEPTCON. + sockFlags = linux.SO_ACCEPTCON + } + } + + var sockState int + switch sops.Endpoint().Type() { + case linux.SOCK_DGRAM: + sockState = linux.SS_CONNECTING + // Unlike Linux, we don't have unbound connection-less sockets, + // so no SS_DISCONNECTING. + + case linux.SOCK_SEQPACKET: + fallthrough + case linux.SOCK_STREAM: + // Connectioned. + if sops.Endpoint().(transport.ConnectingEndpoint).Connected() { + sockState = linux.SS_CONNECTED + } else { + sockState = linux.SS_UNCONNECTED + } + } + + // In the socket entry below, the value for the 'Num' field requires + // some consideration. Linux prints the address to the struct + // unix_sock representing a socket in the kernel, but may redact the + // value for unprivileged users depending on the kptr_restrict + // sysctl. + // + // One use for this field is to allow a privileged user to + // introspect into the kernel memory to determine information about + // a socket not available through procfs, such as the socket's peer. + // + // On gvisor, returning a pointer to our internal structures would + // be pointless, as it wouldn't match the memory layout for struct + // unix_sock, making introspection difficult. We could populate a + // struct unix_sock with the appropriate data, but even that + // requires consideration for which kernel version to emulate, as + // the definition of this struct changes over time. + // + // For now, we always redact this pointer. + fmt.Fprintf(&buf, "%#016p: %08X %08X %08X %04X %02X %5d", + (*unix.SocketOperations)(nil), // Num, pointer to kernel socket struct. + sfile.ReadRefs()-1, // RefCount, don't count our own ref. + 0, // Protocol, always 0 for UDS. + sockFlags, // Flags. + sops.Endpoint().Type(), // Type. + sockState, // State. + sfile.InodeID(), // Inode. + ) + + // Path + if len(addr.Addr) != 0 { + if addr.Addr[0] == 0 { + // Abstract path. + fmt.Fprintf(&buf, " @%s", string(addr.Addr[1:])) + } else { + fmt.Fprintf(&buf, " %s", string(addr.Addr)) + } + } + fmt.Fprintf(&buf, "\n") + + sfile.DecRef() + } + + data := []seqfile.SeqData{{ + Buf: buf.Bytes(), + Handle: (*netUnix)(nil), + }} + return data, 0 +} diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index be04f94af..88018e707 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -85,7 +85,7 @@ func New(ctx context.Context, msrc *fs.MountSource) (*fs.Inode, error) { if _, ok := p.k.NetworkStack().(*rpcinet.Stack); ok { p.AddChild(ctx, "net", newRPCInetProcNet(ctx, msrc)) } else { - p.AddChild(ctx, "net", p.newNetDir(ctx, msrc)) + p.AddChild(ctx, "net", p.newNetDir(ctx, k, msrc)) } return newProcInode(p, msrc, fs.SpecialDirectory, nil), nil diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 43e9823cb..e7e5ff777 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -43,6 +43,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/cpuid" "gvisor.googlesource.com/gvisor/pkg/eventchannel" "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" @@ -164,7 +165,7 @@ type Kernel struct { // nextInotifyCookie is a monotonically increasing counter used for // generating unique inotify event cookies. // - // nextInotifyCookie is mutable, and is accesed using atomic memory + // nextInotifyCookie is mutable, and is accessed using atomic memory // operations. nextInotifyCookie uint32 @@ -177,6 +178,10 @@ type Kernel struct { // danglingEndpoints is used to save / restore tcpip.DanglingEndpoints. danglingEndpoints struct{} `state:".([]tcpip.Endpoint)"` + + // socketTable is used to track all sockets on the system. Protected by + // extMu. + socketTable map[int]map[*refs.WeakRef]struct{} } // InitKernelArgs holds arguments to Init. @@ -266,6 +271,7 @@ func (k *Kernel) Init(args InitKernelArgs) error { k.monotonicClock = &timekeeperClock{tk: args.Timekeeper, c: sentrytime.Monotonic} k.futexes = futex.NewManager() k.netlinkPorts = port.New() + k.socketTable = make(map[int]map[*refs.WeakRef]struct{}) return nil } @@ -1051,6 +1057,56 @@ func (k *Kernel) EmitUnimplementedEvent(ctx context.Context) { }) } +// socketEntry represents a socket recorded in Kernel.socketTable. It implements +// refs.WeakRefUser for sockets stored in the socket table. +// +// +stateify savable +type socketEntry struct { + k *Kernel + sock *refs.WeakRef + family int +} + +// WeakRefGone implements refs.WeakRefUser.WeakRefGone. +func (s *socketEntry) WeakRefGone() { + s.k.extMu.Lock() + // k.socketTable is guaranteed to point to a valid socket table for s.family + // at this point, since we made sure of the fact when we created this + // socketEntry, and we never delete socket tables. + delete(s.k.socketTable[s.family], s.sock) + s.k.extMu.Unlock() +} + +// RecordSocket adds a socket to the system-wide socket table for tracking. +// +// Precondition: Caller must hold a reference to sock. +func (k *Kernel) RecordSocket(sock *fs.File, family int) { + k.extMu.Lock() + table, ok := k.socketTable[family] + if !ok { + table = make(map[*refs.WeakRef]struct{}) + k.socketTable[family] = table + } + se := socketEntry{k: k, family: family} + se.sock = refs.NewWeakRef(sock, &se) + table[se.sock] = struct{}{} + k.extMu.Unlock() +} + +// ListSockets returns a snapshot of all sockets of a given family. +func (k *Kernel) ListSockets(family int) []*refs.WeakRef { + k.extMu.Lock() + socks := []*refs.WeakRef{} + if table, ok := k.socketTable[family]; ok { + socks = make([]*refs.WeakRef, 0, len(table)) + for s, _ := range table { + socks = append(socks, s) + } + } + k.extMu.Unlock() + return socks +} + type supervisorContext struct { context.NoopSleeper log.Logger diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index e28d2c4fa..5ab423f3c 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -147,6 +147,7 @@ func New(t *kernel.Task, family int, stype transport.SockType, protocol int) (*f return nil, err } if s != nil { + t.Kernel().RecordSocket(s, family) return s, nil } } @@ -163,12 +164,15 @@ func Pair(t *kernel.Task, family int, stype transport.SockType, protocol int) (* } for _, p := range providers { - s, t, err := p.Pair(t, stype, protocol) + s1, s2, err := p.Pair(t, stype, protocol) if err != nil { return nil, nil, err } - if s != nil && t != nil { - return s, t, nil + if s1 != nil && s2 != nil { + k := t.Kernel() + k.RecordSocket(s1, family) + k.RecordSocket(s2, family) + return s1, s2, nil } } diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 19258e692..c857a0f33 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -219,6 +219,8 @@ func (s *SocketOperations) Accept(t *kernel.Task, peerRequested bool, flags int, return 0, nil, 0, syserr.FromError(e) } + t.Kernel().RecordSocket(ns, linux.AF_UNIX) + return fd, addr, addrLen, nil } diff --git a/test/syscalls/BUILD b/test/syscalls/BUILD index 148d9c366..53da121ec 100644 --- a/test/syscalls/BUILD +++ b/test/syscalls/BUILD @@ -534,6 +534,13 @@ syscall_test( syscall_test(test = "//test/syscalls/linux:write_test") +syscall_test( + test = "//test/syscalls/linux:proc_net_unix_test", + # Unix domain socket creation isn't supported on all file systems. The + # sentry-internal tmpfs is known to support it. + use_tmpfs = True, +) + go_binary( name = "syscall_test_runner", srcs = ["syscall_test_runner.go"], diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index a311ca12c..590ee1659 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -3102,3 +3102,20 @@ cc_binary( "@com_google_googletest//:gtest", ], ) + +cc_binary( + name = "proc_net_unix_test", + testonly = 1, + srcs = ["proc_net_unix.cc"], + linkstatic = 1, + deps = [ + ":unix_domain_socket_test_util", + "//test/util:file_descriptor", + "//test/util:fs_util", + "//test/util:test_main", + "//test/util:test_util", + "@com_google_absl//absl/strings", + "@com_google_absl//absl/strings:str_format", + "@com_google_googletest//:gtest", + ], +) diff --git a/test/syscalls/linux/proc_net_unix.cc b/test/syscalls/linux/proc_net_unix.cc new file mode 100644 index 000000000..ea7c93012 --- /dev/null +++ b/test/syscalls/linux/proc_net_unix.cc @@ -0,0 +1,246 @@ +// Copyright 2019 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "gtest/gtest.h" +#include "gtest/gtest.h" +#include "absl/strings/numbers.h" +#include "absl/strings/str_format.h" +#include "absl/strings/str_join.h" +#include "absl/strings/str_split.h" +#include "test/syscalls/linux/unix_domain_socket_test_util.h" +#include "test/util/file_descriptor.h" +#include "test/util/fs_util.h" +#include "test/util/test_util.h" + +namespace gvisor { +namespace testing { +namespace { + +using absl::StrCat; +using absl::StreamFormat; +using absl::StrFormat; + +constexpr char kProcNetUnixHeader[] = + "Num RefCount Protocol Flags Type St Inode Path"; + +// UnixEntry represents a single entry from /proc/net/unix. +struct UnixEntry { + uintptr_t addr; + uint64_t refs; + uint64_t protocol; + uint64_t flags; + uint64_t type; + uint64_t state; + uint64_t inode; + std::string path; +}; + +std::string ExtractPath(const struct sockaddr* addr) { + const char* path = + reinterpret_cast(addr)->sun_path; + // Note: sockaddr_un.sun_path is an embedded character array of length + // UNIX_PATH_MAX, so we can always safely dereference the first 2 bytes below. + // + // The kernel also enforces that the path is always null terminated. + if (path[0] == 0) { + // Abstract socket paths are null padded to the end of the struct + // sockaddr. However, these null bytes may or may not show up in + // /proc/net/unix depending on the kernel version. Truncate after the first + // null byte (by treating path as a c-std::string). + return StrCat("@", &path[1]); + } + return std::string(path); +} + +// Returns a parsed representation of /proc/net/unix entries. +PosixErrorOr> ProcNetUnixEntries() { + std::string content; + RETURN_IF_ERRNO(GetContents("/proc/net/unix", &content)); + + bool skipped_header = false; + std::vector entries; + std::vector lines = absl::StrSplit(content, absl::ByAnyChar("\n")); + for (std::string line : lines) { + if (!skipped_header) { + EXPECT_EQ(line, kProcNetUnixHeader); + skipped_header = true; + continue; + } + if (line.empty()) { + continue; + } + + // Abstract socket paths can have trailing null bytes in them depending on + // the linux version. Strip off everything after a null byte, including the + // null byte. + std::size_t null_pos = line.find('\0'); + if (null_pos != std::string::npos) { + line.erase(null_pos); + } + + // Parse a single entry from /proc/net/unix. + // + // Sample file: + // + // clang-format off + // + // Num RefCount Protocol Flags Type St Inode Path" + // ffffa130e7041c00: 00000002 00000000 00010000 0001 01 1299413685 /tmp/control_server/13293772586877554487 + // ffffa14f547dc400: 00000002 00000000 00010000 0001 01 3793 @remote_coredump + // + // clang-format on + // + // Note that from the second entry, the inode number can be padded using + // spaces, so we need to handle it separately during parsing. See + // net/unix/af_unix.c:unix_seq_show() for how these entries are produced. In + // particular, only the inode field is padded with spaces. + UnixEntry entry; + + // Process the first 6 fields, up to but not including "Inode". + std::vector fields = absl::StrSplit(line, absl::MaxSplits(' ', 6)); + + if (fields.size() < 7) { + return PosixError(EINVAL, StrFormat("Invalid entry: '%s'\n", line)); + } + + // AtoiBase can't handle the ':' in the "Num" field, so strip it out. + std::vector addr = absl::StrSplit(fields[0], ':'); + ASSIGN_OR_RETURN_ERRNO(entry.addr, AtoiBase(addr[0], 16)); + + ASSIGN_OR_RETURN_ERRNO(entry.refs, AtoiBase(fields[1], 16)); + ASSIGN_OR_RETURN_ERRNO(entry.protocol, AtoiBase(fields[2], 16)); + ASSIGN_OR_RETURN_ERRNO(entry.flags, AtoiBase(fields[3], 16)); + ASSIGN_OR_RETURN_ERRNO(entry.type, AtoiBase(fields[4], 16)); + ASSIGN_OR_RETURN_ERRNO(entry.state, AtoiBase(fields[5], 16)); + + absl::string_view rest = absl::StripAsciiWhitespace(fields[6]); + fields = absl::StrSplit(rest, absl::MaxSplits(' ', 1)); + if (fields.empty()) { + return PosixError( + EINVAL, StrFormat("Invalid entry, missing 'Inode': '%s'\n", line)); + } + ASSIGN_OR_RETURN_ERRNO(entry.inode, AtoiBase(fields[0], 10)); + + entry.path = ""; + if (fields.size() > 1) { + entry.path = fields[1]; + } + + entries.push_back(entry); + } + + return entries; +} + +// Finds the first entry in 'entries' for which 'predicate' returns true. +// Returns true on match, and sets 'match' to point to the matching entry. +bool FindBy(std::vector entries, UnixEntry* match, + std::function predicate) { + for (int i = 0; i < entries.size(); ++i) { + if (predicate(entries[i])) { + *match = entries[i]; + return true; + } + } + return false; +} + +bool FindByPath(std::vector entries, UnixEntry* match, + const std::string& path) { + return FindBy(entries, match, [path](UnixEntry e) { return e.path == path; }); +} + +TEST(ProcNetUnix, Exists) { + const std::string content = + ASSERT_NO_ERRNO_AND_VALUE(GetContents("/proc/net/unix")); + const std::string header_line = StrCat(kProcNetUnixHeader, "\n"); + if (IsRunningOnGvisor()) { + // Should be just the header since we don't have any unix domain sockets + // yet. + EXPECT_EQ(content, header_line); + } else { + // However, on a general linux machine, we could have abitrary sockets on + // the system, so just check the header. + EXPECT_THAT(content, ::testing::StartsWith(header_line)); + } +} + +TEST(ProcNetUnix, FilesystemBindAcceptConnect) { + auto sockets = ASSERT_NO_ERRNO_AND_VALUE( + FilesystemBoundUnixDomainSocketPair(SOCK_STREAM).Create()); + + std::string path1 = ExtractPath(sockets->first_addr()); + std::string path2 = ExtractPath(sockets->second_addr()); + std::cout << StreamFormat("Server socket address: %s\n", path1); + std::cout << StreamFormat("Client socket address: %s\n", path2); + + std::vector entries = + ASSERT_NO_ERRNO_AND_VALUE(ProcNetUnixEntries()); + if (IsRunningOnGvisor()) { + EXPECT_EQ(entries.size(), 2); + } + + // The server-side socket's path is listed in the socket entry... + UnixEntry s1; + EXPECT_TRUE(FindByPath(entries, &s1, path1)); + + // ... but the client-side socket's path is not. + UnixEntry s2; + EXPECT_FALSE(FindByPath(entries, &s2, path2)); +} + +TEST(ProcNetUnix, AbstractBindAcceptConnect) { + auto sockets = ASSERT_NO_ERRNO_AND_VALUE( + AbstractBoundUnixDomainSocketPair(SOCK_STREAM).Create()); + + std::string path1 = ExtractPath(sockets->first_addr()); + std::string path2 = ExtractPath(sockets->second_addr()); + std::cout << StreamFormat("Server socket address: '%s'\n", path1); + std::cout << StreamFormat("Client socket address: '%s'\n", path2); + + std::vector entries = + ASSERT_NO_ERRNO_AND_VALUE(ProcNetUnixEntries()); + if (IsRunningOnGvisor()) { + EXPECT_EQ(entries.size(), 2); + } + + // The server-side socket's path is listed in the socket entry... + UnixEntry s1; + EXPECT_TRUE(FindByPath(entries, &s1, path1)); + + // ... but the client-side socket's path is not. + UnixEntry s2; + EXPECT_FALSE(FindByPath(entries, &s2, path2)); +} + +TEST(ProcNetUnix, SocketPair) { + // Under gvisor, ensure a socketpair() syscall creates exactly 2 new + // entries. We have no way to verify this under Linux, as we have no control + // over socket creation on a general Linux machine. + SKIP_IF(!IsRunningOnGvisor()); + + std::vector entries = + ASSERT_NO_ERRNO_AND_VALUE(ProcNetUnixEntries()); + ASSERT_EQ(entries.size(), 0); + + auto sockets = + ASSERT_NO_ERRNO_AND_VALUE(UnixDomainSocketPair(SOCK_STREAM).Create()); + + entries = ASSERT_NO_ERRNO_AND_VALUE(ProcNetUnixEntries()); + EXPECT_EQ(entries.size(), 2); +} + +} // namespace +} // namespace testing +} // namespace gvisor -- cgit v1.2.3 From 0a41ea72c1f70916bdbb68d9fdfa6c438e28b5b2 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Thu, 14 Feb 2019 15:46:25 -0800 Subject: Don't allow writing or reading to TTY unless process group is in foreground. If a background process tries to read from a TTY, linux sends it a SIGTTIN unless the signal is blocked or ignored, or the process group is an orphan, in which case the syscall returns EIO. See drivers/tty/n_tty.c:n_tty_read()=>job_control(). If a background process tries to write a TTY, set the termios, or set the foreground process group, linux then sends a SIGTTOU. If the signal is ignored or blocked, linux allows the write. If the process group is an orphan, the syscall returns EIO. See drivers/tty/tty_io.c:tty_check_change(). PiperOrigin-RevId: 234044367 Change-Id: I009461352ac4f3f11c5d42c43ac36bb0caa580f9 --- pkg/sentry/control/proc.go | 14 ++- pkg/sentry/fs/host/tty.go | 183 ++++++++++++++++++++++++++++++----- pkg/sentry/kernel/kernel.go | 44 +++------ pkg/sentry/kernel/sessions.go | 29 ++++++ pkg/sentry/kernel/signal_handlers.go | 8 ++ runsc/boot/loader.go | 14 ++- 6 files changed, 234 insertions(+), 58 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go index 923399fb2..e848def14 100644 --- a/pkg/sentry/control/proc.go +++ b/pkg/sentry/control/proc.go @@ -222,10 +222,18 @@ func (proc *Proc) execAsync(args *ExecArgs) (*kernel.ThreadGroup, kernel.ThreadI return nil, 0, nil, err } - if ttyFile == nil { - return tg, tid, nil, nil + var ttyFileOps *host.TTYFileOperations + if ttyFile != nil { + // Set the foreground process group on the TTY before starting + // the process. + ttyFileOps = ttyFile.FileOperations.(*host.TTYFileOperations) + ttyFileOps.InitForegroundProcessGroup(tg.ProcessGroup()) } - return tg, tid, ttyFile.FileOperations.(*host.TTYFileOperations), nil + + // Start the newly created process. + proc.Kernel.StartProcess(tg) + + return tg, tid, ttyFileOps, nil } // PsArgs is the set of arguments to ps. diff --git a/pkg/sentry/fs/host/tty.go b/pkg/sentry/fs/host/tty.go index ac6ad1b87..21db0086e 100644 --- a/pkg/sentry/fs/host/tty.go +++ b/pkg/sentry/fs/host/tty.go @@ -37,8 +37,11 @@ type TTYFileOperations struct { // mu protects the fields below. mu sync.Mutex `state:"nosave"` - // FGProcessGroup is the foreground process group this TTY. Will be - // nil if not set or if this file has been released. + // session is the session attached to this TTYFileOperations. + session *kernel.Session + + // fgProcessGroup is the foreground process group that is currently + // connected to this TTY. fgProcessGroup *kernel.ProcessGroup } @@ -49,15 +52,58 @@ func newTTYFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlags, iops }) } -// ForegroundProcessGroup returns the foreground process for the TTY. This will -// be nil if the foreground process has not been set or if the file has been -// released. +// InitForegroundProcessGroup sets the foreground process group and session for +// the TTY. This should only be called once, after the foreground process group +// has been created, but before it has started running. +func (t *TTYFileOperations) InitForegroundProcessGroup(pg *kernel.ProcessGroup) { + t.mu.Lock() + defer t.mu.Unlock() + if t.fgProcessGroup != nil { + panic("foreground process group is already set") + } + t.fgProcessGroup = pg + t.session = pg.Session() +} + +// ForegroundProcessGroup returns the foreground process for the TTY. func (t *TTYFileOperations) ForegroundProcessGroup() *kernel.ProcessGroup { t.mu.Lock() defer t.mu.Unlock() return t.fgProcessGroup } +// Read implements fs.FileOperations.Read. +// +// Reading from a TTY is only allowed for foreground process groups. Background +// process groups will either get EIO or a SIGTTIN. +// +// See drivers/tty/n_tty.c:n_tty_read()=>job_control(). +func (t *TTYFileOperations) Read(ctx context.Context, file *fs.File, dst usermem.IOSequence, offset int64) (int64, error) { + t.mu.Lock() + defer t.mu.Unlock() + + // Are we allowed to do the read? + // drivers/tty/n_tty.c:n_tty_read()=>job_control()=>tty_check_change(). + if err := t.checkChange(ctx, linux.SIGTTIN); err != nil { + return 0, err + } + + // Do the read. + return t.fileOperations.Read(ctx, file, dst, offset) +} + +// Write implements fs.FileOperations.Write. +func (t *TTYFileOperations) Write(ctx context.Context, file *fs.File, src usermem.IOSequence, offset int64) (int64, error) { + t.mu.Lock() + defer t.mu.Unlock() + + // Are we allowed to do the write? + if err := t.checkChange(ctx, linux.SIGTTOU); err != nil { + return 0, err + } + return t.fileOperations.Write(ctx, file, src, offset) +} + // Release implements fs.FileOperations.Release. func (t *TTYFileOperations) Release() { t.mu.Lock() @@ -84,6 +130,13 @@ func (t *TTYFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch. return 0, err case linux.TCSETS, linux.TCSETSW, linux.TCSETSF: + t.mu.Lock() + defer t.mu.Unlock() + + if err := t.checkChange(ctx, linux.SIGTTOU); err != nil { + return 0, err + } + var termios linux.Termios if _, err := usermem.CopyObjectIn(ctx, io, args[2].Pointer(), &termios, usermem.IOOpts{ AddressSpaceActive: true, @@ -99,20 +152,17 @@ func (t *TTYFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch. // Get the process group ID of the foreground process group on // this terminal. + pidns := kernel.PIDNamespaceFromContext(ctx) + if pidns == nil { + return 0, syserror.ENOTTY + } + t.mu.Lock() defer t.mu.Unlock() - if t.fgProcessGroup == nil { - // No process group has been set yet. Let's just lie - // and tell it the process group from the current task. - // The app is probably going to set it to something - // else very soon anyways. - t.fgProcessGroup = kernel.TaskFromContext(ctx).ThreadGroup().ProcessGroup() - } - // Map the ProcessGroup into a ProcessGroupID in the task's PID // namespace. - pgID := kernel.TaskFromContext(ctx).ThreadGroup().PIDNamespace().IDOfProcessGroup(t.fgProcessGroup) + pgID := pidns.IDOfProcessGroup(t.fgProcessGroup) _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), &pgID, usermem.IOOpts{ AddressSpaceActive: true, }) @@ -123,6 +173,30 @@ func (t *TTYFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch. // Equivalent to tcsetpgrp(fd, *argp). // Set the foreground process group ID of this terminal. + task := kernel.TaskFromContext(ctx) + if task == nil { + return 0, syserror.ENOTTY + } + + t.mu.Lock() + defer t.mu.Unlock() + + // Check that we are allowed to set the process group. + if err := t.checkChange(ctx, linux.SIGTTOU); err != nil { + // drivers/tty/tty_io.c:tiocspgrp() converts -EIO from + // tty_check_change() to -ENOTTY. + if err == syserror.EIO { + return 0, syserror.ENOTTY + } + return 0, err + } + + // Check that calling task's process group is in the TTY + // session. + if task.ThreadGroup().Session() != t.session { + return 0, syserror.ENOTTY + } + var pgID kernel.ProcessGroupID if _, err := usermem.CopyObjectIn(ctx, io, args[2].Pointer(), &pgID, usermem.IOOpts{ AddressSpaceActive: true, @@ -136,24 +210,18 @@ func (t *TTYFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch. } // Process group with pgID must exist in this PID namespace. - task := kernel.TaskFromContext(ctx) pidns := task.PIDNamespace() pg := pidns.ProcessGroupWithID(pgID) if pg == nil { return 0, syserror.ESRCH } - // Process group must be in same session as calling task's - // process group. - curSession := task.ThreadGroup().ProcessGroup().Session() - curSessionID := pidns.IDOfSession(curSession) - if pidns.IDOfSession(pg.Session()) != curSessionID { + // Check that new process group is in the TTY session. + if pg.Session() != t.session { return 0, syserror.EPERM } - t.mu.Lock() t.fgProcessGroup = pg - t.mu.Unlock() return 0, nil case linux.TIOCGWINSZ: @@ -171,6 +239,10 @@ func (t *TTYFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch. case linux.TIOCSWINSZ: // Args: const struct winsize *argp // Set window size. + + // Unlike setting the termios, any process group (even + // background ones) can set the winsize. + var winsize linux.Winsize if _, err := usermem.CopyObjectIn(ctx, io, args[2].Pointer(), &winsize, usermem.IOOpts{ AddressSpaceActive: true, @@ -213,3 +285,70 @@ func (t *TTYFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch. return 0, syserror.ENOTTY } } + +// checkChange checks that the process group is allowed to read, write, or +// change the state of the TTY. +// +// This corresponds to Linux drivers/tty/tty_io.c:tty_check_change(). The logic +// is a bit convoluted, but documented inline. +// +// Preconditions: t.mu must be held. +func (t *TTYFileOperations) checkChange(ctx context.Context, sig linux.Signal) error { + task := kernel.TaskFromContext(ctx) + if task == nil { + // No task? Linux does not have an analog for this case, but + // tty_check_change is more of a blacklist of cases than a + // whitelist, and is surprisingly permissive. Allowing the + // change seems most appropriate. + return nil + } + + tg := task.ThreadGroup() + pg := tg.ProcessGroup() + + // If the session for the task is different than the session for the + // controlling TTY, then the change is allowed. Seems like a bad idea, + // but that's exactly what linux does. + if tg.Session() != t.fgProcessGroup.Session() { + return nil + } + + // If we are the foreground process group, then the change is allowed. + if pg == t.fgProcessGroup { + return nil + } + + // We are not the foreground process group. + + // Is the provided signal blocked or ignored? + if (task.SignalMask()&linux.SignalSetOf(sig) != 0) || tg.SignalHandlers().IsIgnored(sig) { + // If the signal is SIGTTIN, then we are attempting to read + // from the TTY. Don't send the signal and return EIO. + if sig == linux.SIGTTIN { + return syserror.EIO + } + + // Otherwise, we are writing or changing terminal state. This is allowed. + return nil + } + + // If the process group is an orphan, return EIO. + if pg.IsOrphan() { + return syserror.EIO + } + + // Otherwise, send the signal to the process group and return ERESTARTSYS. + // + // Note that Linux also unconditionally sets TIF_SIGPENDING on current, + // but this isn't necessary in gVisor because the rationale given in + // 040b6362d58f "tty: fix leakage of -ERESTARTSYS to userland" doesn't + // apply: the sentry will handle -ERESTARTSYS in + // kernel.runApp.execute() even if the kernel.Task isn't interrupted. + si := arch.SignalInfo{ + Code: arch.SignalInfoKernel, + Signo: int32(sig), + } + // Linux ignores the result of kill_pgrp(). + _ = pg.SendSignal(&si) + return kernel.ERESTARTSYS +} diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index e7e5ff777..c6afae2e6 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -615,8 +615,11 @@ func (ctx *createProcessContext) Value(key interface{}) interface{} { // CreateProcess creates a new task in a new thread group with the given // options. The new task has no parent and is in the root PID namespace. // -// If k.Start() has already been called, the created task will begin running -// immediately. Otherwise, it will be started when k.Start() is called. +// If k.Start() has already been called, then the created process must be +// started by calling kernel.StartProcess(tg). +// +// If k.Start() has not yet been called, then the created task will begin +// running when k.Start() is called. // // CreateProcess has no analogue in Linux; it is used to create the initial // application task, as well as processes started by the control server. @@ -688,22 +691,25 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, ThreadID, AbstractSocketNamespace: args.AbstractSocketNamespace, ContainerID: args.ContainerID, } - t, err := k.tasks.NewTask(config) - if err != nil { + if _, err := k.tasks.NewTask(config); err != nil { return nil, 0, err } // Success. tgid := k.tasks.Root.IDOfThreadGroup(tg) - if k.started { - tid := k.tasks.Root.IDOfTask(t) - t.Start(tid) - } else if k.globalInit == nil { + if k.globalInit == nil { k.globalInit = tg } return tg, tgid, nil } +// StartProcess starts running a process that was created with CreateProcess. +func (k *Kernel) StartProcess(tg *ThreadGroup) { + t := tg.Leader() + tid := k.tasks.Root.IDOfTask(t) + t.Start(tid) +} + // Start starts execution of all tasks in k. // // Preconditions: Start may be called exactly once. @@ -866,28 +872,6 @@ func (k *Kernel) SendContainerSignal(cid string, info *arch.SignalInfo) error { return lastErr } -// SendProcessGroupSignal sends a signal to all processes inside the process -// group. It is analagous to kernel/signal.c:kill_pgrp. -func (k *Kernel) SendProcessGroupSignal(pg *ProcessGroup, info *arch.SignalInfo) error { - k.extMu.Lock() - defer k.extMu.Unlock() - k.tasks.mu.RLock() - defer k.tasks.mu.RUnlock() - - var lastErr error - for t := range k.tasks.Root.tids { - if t == t.tg.leader && t.tg.ProcessGroup() == pg { - t.tg.signalHandlers.mu.Lock() - defer t.tg.signalHandlers.mu.Unlock() - infoCopy := *info - if err := t.sendSignalLocked(&infoCopy, true /*group*/); err != nil { - lastErr = err - } - } - } - return lastErr -} - // FeatureSet returns the FeatureSet. func (k *Kernel) FeatureSet() *cpuid.FeatureSet { return k.featureSet diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index 78a5b4063..6fd65f2b0 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -17,6 +17,7 @@ package kernel import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/refs" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -119,6 +120,13 @@ func (pg *ProcessGroup) Originator() *ThreadGroup { return pg.originator } +// IsOrphan returns true if this process group is an orphan. +func (pg *ProcessGroup) IsOrphan() bool { + pg.originator.TaskSet().mu.RLock() + defer pg.originator.TaskSet().mu.RUnlock() + return pg.ancestors == 0 +} + // incRefWithParent grabs a reference. // // This function is called when this ProcessGroup is being associated with some @@ -224,6 +232,27 @@ func (pg *ProcessGroup) Session() *Session { return pg.session } +// SendSignal sends a signal to all processes inside the process group. It is +// analagous to kernel/signal.c:kill_pgrp. +func (pg *ProcessGroup) SendSignal(info *arch.SignalInfo) error { + tasks := pg.originator.TaskSet() + tasks.mu.RLock() + defer tasks.mu.RUnlock() + + var lastErr error + for t := range tasks.Root.tids { + if t == t.tg.leader && t.tg.ProcessGroup() == pg { + t.tg.signalHandlers.mu.Lock() + defer t.tg.signalHandlers.mu.Unlock() + infoCopy := *info + if err := t.sendSignalLocked(&infoCopy, true /*group*/); err != nil { + lastErr = err + } + } + } + return lastErr +} + // CreateSession creates a new Session, with the ThreadGroup as the leader. // // EPERM may be returned if either the given ThreadGroup is already a Session diff --git a/pkg/sentry/kernel/signal_handlers.go b/pkg/sentry/kernel/signal_handlers.go index 3f1ac9898..60cbe85b8 100644 --- a/pkg/sentry/kernel/signal_handlers.go +++ b/pkg/sentry/kernel/signal_handlers.go @@ -69,6 +69,14 @@ func (sh *SignalHandlers) CopyForExec() *SignalHandlers { return sh2 } +// IsIgnored returns true if the signal is ignored. +func (sh *SignalHandlers) IsIgnored(sig linux.Signal) bool { + sh.mu.Lock() + defer sh.mu.Unlock() + sa, ok := sh.actions[sig] + return ok && sa.Handler == arch.SignalActIgnore +} + // dequeueActionLocked returns the SignalAct that should be used to handle sig. // // Preconditions: sh.mu must be locked. diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 973578484..41f456af7 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -477,9 +477,9 @@ func (l *Loader) run() error { return err } - // Create the root container init task. - _, _, err := l.k.CreateProcess(l.rootProcArgs) - if err != nil { + // Create the root container init task. It will begin running + // when the kernel is started. + if _, _, err := l.k.CreateProcess(l.rootProcArgs); err != nil { return fmt.Errorf("creating init process: %v", err) } @@ -492,6 +492,11 @@ func (l *Loader) run() error { ttyFile := l.rootProcArgs.FDMap.GetFile(0) defer ttyFile.DecRef() ep.tty = ttyFile.FileOperations.(*host.TTYFileOperations) + + // Set the foreground process group on the TTY to the global + // init process group, since that is what we are about to + // start running. + ep.tty.InitForegroundProcessGroup(ep.tg.ProcessGroup()) } // Start signal forwarding only after an init process is created. @@ -595,10 +600,13 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config return fmt.Errorf("setting executable path for %+v: %v", procArgs, err) } + // Create and start the new process. tg, _, err := l.k.CreateProcess(procArgs) if err != nil { return fmt.Errorf("creating process: %v", err) } + l.k.StartProcess(tg) + // CreateProcess takes a reference on FDMap if successful. procArgs.FDMap.DecRef() -- cgit v1.2.3 From bb47d8a545f82849f637c480459109e16be336cf Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Tue, 19 Feb 2019 14:19:07 -0800 Subject: Fix clone(CLONE_NEWUSER). - Use new user namespace for namespace creation checks. - Ensure userns is never nil since it's used by other namespaces. PiperOrigin-RevId: 234673175 Change-Id: I4b9d9d1e63ce4e24362089793961a996f7540cd9 --- pkg/sentry/kernel/task_clone.go | 7 +++---- test/syscalls/linux/BUILD | 2 ++ test/syscalls/linux/fork.cc | 31 +++++++++++++++++++++++++++++++ 3 files changed, 36 insertions(+), 4 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index b66fa34a9..114e7f858 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -17,7 +17,6 @@ package kernel import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/bpf" - "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -166,7 +165,7 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { // privileges over the remaining namespaces created by the call." - // user_namespaces(7) creds := t.Credentials() - var userns *auth.UserNamespace + userns := creds.UserNamespace if opts.NewUserNamespace { var err error // "EPERM (since Linux 3.9): CLONE_NEWUSER was specified in flags and @@ -182,7 +181,7 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { return 0, nil, err } } - if (opts.NewPIDNamespace || opts.NewNetworkNamespace || opts.NewUTSNamespace) && !creds.HasCapability(linux.CAP_SYS_ADMIN) { + if (opts.NewPIDNamespace || opts.NewNetworkNamespace || opts.NewUTSNamespace) && !creds.HasCapabilityIn(linux.CAP_SYS_ADMIN, userns) { return 0, nil, syserror.EPERM } @@ -287,7 +286,7 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { nt.SetSignalStack(t.SignalStack()) } - if userns != nil { + if userns != creds.UserNamespace { if err := nt.SetUserNamespace(userns); err != nil { // This shouldn't be possible: userns was created from nt.creds, so // nt should have CAP_SYS_ADMIN in userns. diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index 3c61c48ef..e7f5ea998 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -732,7 +732,9 @@ cc_binary( srcs = ["fork.cc"], linkstatic = 1, deps = [ + "//test/util:capability_util", "//test/util:logging", + "//test/util:memory_util", "//test/util:test_main", "//test/util:test_util", "//test/util:thread_util", diff --git a/test/syscalls/linux/fork.cc b/test/syscalls/linux/fork.cc index 1bff5e50f..73ac885b5 100644 --- a/test/syscalls/linux/fork.cc +++ b/test/syscalls/linux/fork.cc @@ -21,11 +21,14 @@ #include #include #include +#include #include "gtest/gtest.h" #include "absl/time/clock.h" #include "absl/time/time.h" +#include "test/util/capability_util.h" #include "test/util/logging.h" +#include "test/util/memory_util.h" #include "test/util/test_util.h" #include "test/util/thread_util.h" @@ -393,6 +396,34 @@ TEST_F(ForkTest, Affinity) { EXPECT_THAT(Wait(child), SyscallSucceedsWithValue(0)); } +TEST(CloneTest, NewUserNamespacePermitsAllOtherNamespaces) { + // "If CLONE_NEWUSER is specified along with other CLONE_NEW* flags in a + // single clone(2) or unshare(2) call, the user namespace is guaranteed to be + // created first, giving the child (clone(2)) or caller (unshare(2)) + // privileges over the remaining namespaces created by the call. Thus, it is + // possible for an unprivileged caller to specify this combination of flags." + // - user_namespaces(7) + SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(CanCreateUserNamespace())); + Mapping child_stack = ASSERT_NO_ERRNO_AND_VALUE( + MmapAnon(kPageSize, PROT_READ | PROT_WRITE, MAP_PRIVATE)); + int child_pid; + // We only test with CLONE_NEWIPC, CLONE_NEWNET, and CLONE_NEWUTS since these + // namespaces were implemented in Linux before user namespaces. + ASSERT_THAT( + child_pid = clone( + +[](void*) { return 0; }, + reinterpret_cast(child_stack.addr() + kPageSize), + CLONE_NEWUSER | CLONE_NEWIPC | CLONE_NEWNET | CLONE_NEWUTS | SIGCHLD, + /* arg = */ nullptr), + SyscallSucceeds()); + + int status; + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFEXITED(status) && WEXITSTATUS(status) == 0) + << "status = " << status; +} + #ifdef __x86_64__ // Clone with CLONE_SETTLS and a non-canonical TLS address is rejected. TEST(CloneTest, NonCanonicalTLS) { -- cgit v1.2.3 From bed6f8534b1bedaad031682fe052b5a46d9cb3ee Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Tue, 19 Feb 2019 15:48:39 -0800 Subject: Set rax to syscall number on SECCOMP_RET_TRAP. PiperOrigin-RevId: 234690475 Change-Id: I1cbfb5aecd4697a4a26ec8524354aa8656cc3ba1 --- pkg/sentry/kernel/seccomp.go | 3 +++ test/syscalls/linux/seccomp.cc | 53 ++++++++++++++++++++++++++---------------- 2 files changed, 36 insertions(+), 20 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/seccomp.go b/pkg/sentry/kernel/seccomp.go index cec179246..4bed4d373 100644 --- a/pkg/sentry/kernel/seccomp.go +++ b/pkg/sentry/kernel/seccomp.go @@ -75,6 +75,9 @@ func (t *Task) checkSeccompSyscall(sysno int32, args arch.SyscallArguments, ip u // portion of the return value will be passed as si_errno." - // Documentation/prctl/seccomp_filter.txt t.SendSignal(seccompSiginfo(t, int32(result.Data()), sysno, ip)) + // "The return value register will contain an arch-dependent value." In + // practice, it's ~always the syscall number. + t.Arch().SetReturn(uintptr(sysno)) case linux.SECCOMP_RET_ERRNO: // "Results in the lower 16-bits of the return value being passed to diff --git a/test/syscalls/linux/seccomp.cc b/test/syscalls/linux/seccomp.cc index ac416b75f..27740d7ef 100644 --- a/test/syscalls/linux/seccomp.cc +++ b/test/syscalls/linux/seccomp.cc @@ -23,6 +23,7 @@ #include #include #include +#include #include #include @@ -161,16 +162,21 @@ TEST(SeccompTest, RetTrapCausesSIGSYS) { pid_t const pid = fork(); if (pid == 0) { constexpr uint16_t kTrapValue = 0xdead; - RegisterSignalHandler(SIGSYS, +[](int signo, siginfo_t* info, void*) { - // This is a signal handler, so we must stay async-signal-safe. - TEST_CHECK(info->si_signo == SIGSYS); - TEST_CHECK(info->si_code == SYS_SECCOMP); - TEST_CHECK(info->si_errno == kTrapValue); - TEST_CHECK(info->si_call_addr != nullptr); - TEST_CHECK(info->si_syscall == kFilteredSyscall); - TEST_CHECK(info->si_arch == AUDIT_ARCH_X86_64); - _exit(0); - }); + RegisterSignalHandler( + SIGSYS, +[](int signo, siginfo_t* info, void* ucv) { + ucontext_t* uc = static_cast(ucv); + // This is a signal handler, so we must stay async-signal-safe. + TEST_CHECK(info->si_signo == SIGSYS); + TEST_CHECK(info->si_code == SYS_SECCOMP); + TEST_CHECK(info->si_errno == kTrapValue); + TEST_CHECK(info->si_call_addr != nullptr); + TEST_CHECK(info->si_syscall == kFilteredSyscall); +#ifdef __x86_64__ + TEST_CHECK(info->si_arch == AUDIT_ARCH_X86_64); + TEST_CHECK(uc->uc_mcontext.gregs[REG_RAX] == kFilteredSyscall); +#endif // defined(__x86_64__) + _exit(0); + }); ApplySeccompFilter(kFilteredSyscall, SECCOMP_RET_TRAP | kTrapValue); syscall(kFilteredSyscall); TEST_CHECK_MSG(false, "Survived invocation of test syscall"); @@ -182,6 +188,8 @@ TEST(SeccompTest, RetTrapCausesSIGSYS) { << "status " << status; } +#ifdef __x86_64__ + constexpr uint64_t kVsyscallTimeEntry = 0xffffffffff600400; time_t vsyscall_time(time_t* t) { @@ -194,16 +202,19 @@ TEST(SeccompTest, SeccompAppliesToVsyscall) { pid_t const pid = fork(); if (pid == 0) { constexpr uint16_t kTrapValue = 0xdead; - RegisterSignalHandler(SIGSYS, +[](int signo, siginfo_t* info, void*) { - // This is a signal handler, so we must stay async-signal-safe. - TEST_CHECK(info->si_signo == SIGSYS); - TEST_CHECK(info->si_code == SYS_SECCOMP); - TEST_CHECK(info->si_errno == kTrapValue); - TEST_CHECK(info->si_call_addr != nullptr); - TEST_CHECK(info->si_syscall == SYS_time); - TEST_CHECK(info->si_arch == AUDIT_ARCH_X86_64); - _exit(0); - }); + RegisterSignalHandler( + SIGSYS, +[](int signo, siginfo_t* info, void* ucv) { + ucontext_t* uc = static_cast(ucv); + // This is a signal handler, so we must stay async-signal-safe. + TEST_CHECK(info->si_signo == SIGSYS); + TEST_CHECK(info->si_code == SYS_SECCOMP); + TEST_CHECK(info->si_errno == kTrapValue); + TEST_CHECK(info->si_call_addr != nullptr); + TEST_CHECK(info->si_syscall == SYS_time); + TEST_CHECK(info->si_arch == AUDIT_ARCH_X86_64); + TEST_CHECK(uc->uc_mcontext.gregs[REG_RAX] == SYS_time); + _exit(0); + }); ApplySeccompFilter(SYS_time, SECCOMP_RET_TRAP | kTrapValue); vsyscall_time(nullptr); // Should result in death. TEST_CHECK_MSG(false, "Survived invocation of test syscall"); @@ -234,6 +245,8 @@ TEST(SeccompTest, RetKillVsyscallCausesDeathBySIGSYS) { << "status " << status; } +#endif // defined(__x86_64__) + TEST(SeccompTest, RetTraceWithoutPtracerReturnsENOSYS) { pid_t const pid = fork(); if (pid == 0) { -- cgit v1.2.3 From 15d3189884c2e8050992381ff2a1f0521eae0ba2 Mon Sep 17 00:00:00 2001 From: Haibo Xu Date: Wed, 20 Feb 2019 15:09:50 -0800 Subject: Make some ptrace commands x86-only Signed-off-by: Haibo Xu Change-Id: I9751f859332d433ca772d6b9733f5a5a64398ec7 PiperOrigin-RevId: 234877624 --- pkg/sentry/kernel/BUILD | 2 + pkg/sentry/kernel/ptrace.go | 63 ++------------------------- pkg/sentry/kernel/ptrace_amd64.go | 89 +++++++++++++++++++++++++++++++++++++++ pkg/sentry/kernel/ptrace_arm64.go | 28 ++++++++++++ 4 files changed, 122 insertions(+), 60 deletions(-) create mode 100644 pkg/sentry/kernel/ptrace_amd64.go create mode 100644 pkg/sentry/kernel/ptrace_arm64.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 773cb8c91..d9bbfb556 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -95,6 +95,8 @@ go_library( "posixtimer.go", "process_group_list.go", "ptrace.go", + "ptrace_amd64.go", + "ptrace_arm64.go", "rseq.go", "seccomp.go", "seqatomic_taskgoroutineschedinfo.go", diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index 20bac2b70..fa7a0d141 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -863,42 +863,6 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { }) return err - case linux.PTRACE_PEEKUSR: // aka PTRACE_PEEKUSER - n, err := target.Arch().PtracePeekUser(uintptr(addr)) - if err != nil { - return err - } - _, err = t.CopyOut(data, n) - return err - - case linux.PTRACE_POKEUSR: // aka PTRACE_POKEUSER - return target.Arch().PtracePokeUser(uintptr(addr), uintptr(data)) - - case linux.PTRACE_GETREGS: - // "Copy the tracee's general-purpose ... registers ... to the address - // data in the tracer. ... (addr is ignored.) Note that SPARC systems - // have the meaning of data and addr reversed ..." - _, err := target.Arch().PtraceGetRegs(&usermem.IOReadWriter{ - Ctx: t, - IO: t.MemoryManager(), - Addr: data, - Opts: usermem.IOOpts{ - AddressSpaceActive: true, - }, - }) - return err - - case linux.PTRACE_GETFPREGS: - _, err := target.Arch().PtraceGetFPRegs(&usermem.IOReadWriter{ - Ctx: t, - IO: t.MemoryManager(), - Addr: data, - Opts: usermem.IOOpts{ - AddressSpaceActive: true, - }, - }) - return err - case linux.PTRACE_GETREGSET: // "Read the tracee's registers. addr specifies, in an // architecture-dependent way, the type of registers to be read. ... @@ -930,28 +894,6 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { ar.End = end return t.CopyOutIovecs(data, usermem.AddrRangeSeqOf(ar)) - case linux.PTRACE_SETREGS: - _, err := target.Arch().PtraceSetRegs(&usermem.IOReadWriter{ - Ctx: t, - IO: t.MemoryManager(), - Addr: data, - Opts: usermem.IOOpts{ - AddressSpaceActive: true, - }, - }) - return err - - case linux.PTRACE_SETFPREGS: - _, err := target.Arch().PtraceSetFPRegs(&usermem.IOReadWriter{ - Ctx: t, - IO: t.MemoryManager(), - Addr: data, - Opts: usermem.IOOpts{ - AddressSpaceActive: true, - }, - }) - return err - case linux.PTRACE_SETREGSET: ars, err := t.CopyInIovecs(data, 1) if err != nil { @@ -1047,8 +989,9 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { _, err := t.CopyOut(usermem.Addr(data), target.ptraceEventMsg) return err + // PEEKSIGINFO is unimplemented but seems to have no users anywhere. + default: - // PEEKSIGINFO is unimplemented but seems to have no users anywhere. - return syserror.EIO + return t.ptraceArch(target, req, addr, data) } } diff --git a/pkg/sentry/kernel/ptrace_amd64.go b/pkg/sentry/kernel/ptrace_amd64.go new file mode 100644 index 000000000..1f88efca3 --- /dev/null +++ b/pkg/sentry/kernel/ptrace_amd64.go @@ -0,0 +1,89 @@ +// Copyright 2019 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// +build amd64 + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// ptraceArch implements arch-specific ptrace commands. +func (t *Task) ptraceArch(target *Task, req int64, addr, data usermem.Addr) error { + switch req { + case linux.PTRACE_PEEKUSR: // aka PTRACE_PEEKUSER + n, err := target.Arch().PtracePeekUser(uintptr(addr)) + if err != nil { + return err + } + _, err = t.CopyOut(data, n) + return err + + case linux.PTRACE_POKEUSR: // aka PTRACE_POKEUSER + return target.Arch().PtracePokeUser(uintptr(addr), uintptr(data)) + + case linux.PTRACE_GETREGS: + // "Copy the tracee's general-purpose ... registers ... to the address + // data in the tracer. ... (addr is ignored.) Note that SPARC systems + // have the meaning of data and addr reversed ..." + _, err := target.Arch().PtraceGetRegs(&usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: data, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }) + return err + + case linux.PTRACE_GETFPREGS: + _, err := target.Arch().PtraceGetFPRegs(&usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: data, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }) + return err + + case linux.PTRACE_SETREGS: + _, err := target.Arch().PtraceSetRegs(&usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: data, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }) + return err + + case linux.PTRACE_SETFPREGS: + _, err := target.Arch().PtraceSetFPRegs(&usermem.IOReadWriter{ + Ctx: t, + IO: t.MemoryManager(), + Addr: data, + Opts: usermem.IOOpts{ + AddressSpaceActive: true, + }, + }) + return err + + default: + return syserror.EIO + } +} diff --git a/pkg/sentry/kernel/ptrace_arm64.go b/pkg/sentry/kernel/ptrace_arm64.go new file mode 100644 index 000000000..4636405e6 --- /dev/null +++ b/pkg/sentry/kernel/ptrace_arm64.go @@ -0,0 +1,28 @@ +// Copyright 2019 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// +build arm64 + +package kernel + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// ptraceArch implements arch-specific ptrace commands. +func (t *Task) ptraceArch(target *Task, req int64, addr, data usermem.Addr) error { + return syserror.EIO +} -- cgit v1.2.3 From 3dbd4a16f8ae4da967f69fd93870462d1b3554f5 Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Fri, 1 Mar 2019 10:55:22 -0800 Subject: Add semctl(GETPID) syscall Also added unimplemented notification for semctl(2) commands. PiperOrigin-RevId: 236340672 Change-Id: I0795e3bd2e6d41d7936fabb731884df426a42478 --- pkg/abi/linux/sem.go | 5 ++-- pkg/sentry/kernel/semaphore/semaphore.go | 34 +++++++++++++++++---- pkg/sentry/kernel/semaphore/semaphore_test.go | 6 ++-- pkg/sentry/syscalls/linux/sys_sem.go | 43 +++++++++++++++++++++++++-- runsc/boot/compat.go | 4 +++ test/syscalls/linux/semaphore.cc | 29 ++++++++++++++++++ 6 files changed, 107 insertions(+), 14 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/sem.go b/pkg/abi/linux/sem.go index d1a0bdb32..b80c93daf 100644 --- a/pkg/abi/linux/sem.go +++ b/pkg/abi/linux/sem.go @@ -27,8 +27,9 @@ const ( // ipcs ctl cmds. Source: include/uapi/linux/sem.h const ( - SEM_STAT = 18 - SEM_INFO = 19 + SEM_STAT = 18 + SEM_INFO = 19 + SEM_STAT_ANY = 20 ) const SEM_UNDO = 0x1000 diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index c134931cd..29a2eb804 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -92,6 +92,7 @@ type Set struct { type sem struct { value int16 waiters waiterList `state:"zerovalue"` + pid int32 } // waiter represents a caller that is waiting for the semaphore value to @@ -283,7 +284,7 @@ func (s *Set) Change(ctx context.Context, creds *auth.Credentials, owner fs.File } // SetVal overrides a semaphore value, waking up waiters as needed. -func (s *Set) SetVal(ctx context.Context, num int32, val int16, creds *auth.Credentials) error { +func (s *Set) SetVal(ctx context.Context, num int32, val int16, creds *auth.Credentials, pid int32) error { if val < 0 || val > valueMax { return syserror.ERANGE } @@ -303,15 +304,17 @@ func (s *Set) SetVal(ctx context.Context, num int32, val int16, creds *auth.Cred // TODO: Clear undo entries in all processes sem.value = val + sem.pid = pid s.changeTime = ktime.NowFromContext(ctx) sem.wakeWaiters() return nil } -// SetValAll overrides all semaphores values, waking up waiters as needed. +// SetValAll overrides all semaphores values, waking up waiters as needed. It also +// sets semaphore's PID which was fixed in Linux 4.6. // // 'len(vals)' must be equal to 's.Size()'. -func (s *Set) SetValAll(ctx context.Context, vals []uint16, creds *auth.Credentials) error { +func (s *Set) SetValAll(ctx context.Context, vals []uint16, creds *auth.Credentials, pid int32) error { if len(vals) != s.Size() { panic(fmt.Sprintf("vals length (%d) different that Set.Size() (%d)", len(vals), s.Size())) } @@ -335,6 +338,7 @@ func (s *Set) SetValAll(ctx context.Context, vals []uint16, creds *auth.Credenti // TODO: Clear undo entries in all processes sem.value = int16(val) + sem.pid = pid sem.wakeWaiters() } s.changeTime = ktime.NowFromContext(ctx) @@ -375,12 +379,29 @@ func (s *Set) GetValAll(creds *auth.Credentials) ([]uint16, error) { return vals, nil } +// GetPID returns the PID set when performing operations in the semaphore. +func (s *Set) GetPID(num int32, creds *auth.Credentials) (int32, error) { + s.mu.Lock() + defer s.mu.Unlock() + + // "The calling process must have read permission on the semaphore set." + if !s.checkPerms(creds, fs.PermMask{Read: true}) { + return 0, syserror.EACCES + } + + sem := s.findSem(num) + if sem == nil { + return 0, syserror.ERANGE + } + return sem.pid, nil +} + // ExecuteOps attempts to execute a list of operations to the set. It only // succeeds when all operations can be applied. No changes are made if it fails. // // On failure, it may return an error (retries are hopeless) or it may return // a channel that can be waited on before attempting again. -func (s *Set) ExecuteOps(ctx context.Context, ops []linux.Sembuf, creds *auth.Credentials) (chan struct{}, int32, error) { +func (s *Set) ExecuteOps(ctx context.Context, ops []linux.Sembuf, creds *auth.Credentials, pid int32) (chan struct{}, int32, error) { s.mu.Lock() defer s.mu.Unlock() @@ -404,14 +425,14 @@ func (s *Set) ExecuteOps(ctx context.Context, ops []linux.Sembuf, creds *auth.Cr return nil, 0, syserror.EACCES } - ch, num, err := s.executeOps(ctx, ops) + ch, num, err := s.executeOps(ctx, ops, pid) if err != nil { return nil, 0, err } return ch, num, nil } -func (s *Set) executeOps(ctx context.Context, ops []linux.Sembuf) (chan struct{}, int32, error) { +func (s *Set) executeOps(ctx context.Context, ops []linux.Sembuf, pid int32) (chan struct{}, int32, error) { // Changes to semaphores go to this slice temporarily until they all succeed. tmpVals := make([]int16, len(s.sems)) for i := range s.sems { @@ -464,6 +485,7 @@ func (s *Set) executeOps(ctx context.Context, ops []linux.Sembuf) (chan struct{} for i, v := range tmpVals { s.sems[i].value = v s.sems[i].wakeWaiters() + s.sems[i].pid = pid } s.opTime = ktime.NowFromContext(ctx) return nil, 0, nil diff --git a/pkg/sentry/kernel/semaphore/semaphore_test.go b/pkg/sentry/kernel/semaphore/semaphore_test.go index 5f886bf31..2e51e6ee5 100644 --- a/pkg/sentry/kernel/semaphore/semaphore_test.go +++ b/pkg/sentry/kernel/semaphore/semaphore_test.go @@ -25,7 +25,7 @@ import ( ) func executeOps(ctx context.Context, t *testing.T, set *Set, ops []linux.Sembuf, block bool) chan struct{} { - ch, _, err := set.executeOps(ctx, ops) + ch, _, err := set.executeOps(ctx, ops, 123) if err != nil { t.Fatalf("ExecuteOps(ops) failed, err: %v, ops: %+v", err, ops) } @@ -123,13 +123,13 @@ func TestNoWait(t *testing.T) { ops[0].SemOp = -2 ops[0].SemFlg = linux.IPC_NOWAIT - if _, _, err := set.executeOps(ctx, ops); err != syserror.ErrWouldBlock { + if _, _, err := set.executeOps(ctx, ops, 123); err != syserror.ErrWouldBlock { t.Fatalf("ExecuteOps(ops) wrong result, got: %v, expected: %v", err, syserror.ErrWouldBlock) } ops[0].SemOp = 0 ops[0].SemFlg = linux.IPC_NOWAIT - if _, _, err := set.executeOps(ctx, ops); err != syserror.ErrWouldBlock { + if _, _, err := set.executeOps(ctx, ops, 123); err != syserror.ErrWouldBlock { t.Fatalf("ExecuteOps(ops) wrong result, got: %v, expected: %v", err, syserror.ErrWouldBlock) } } diff --git a/pkg/sentry/syscalls/linux/sys_sem.go b/pkg/sentry/syscalls/linux/sys_sem.go index 6775725ca..86f850ef1 100644 --- a/pkg/sentry/syscalls/linux/sys_sem.go +++ b/pkg/sentry/syscalls/linux/sys_sem.go @@ -71,8 +71,9 @@ func Semop(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall } creds := auth.CredentialsFromContext(t) + pid := t.Kernel().GlobalInit().PIDNamespace().IDOfThreadGroup(t.ThreadGroup()) for { - ch, num, err := set.ExecuteOps(t, ops, creds) + ch, num, err := set.ExecuteOps(t, ops, creds, int32(pid)) if ch == nil || err != nil { // We're done (either on success or a failure). return 0, nil, err @@ -123,6 +124,21 @@ func Semctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal perms := fs.FilePermsFromMode(linux.FileMode(s.SemPerm.Mode & 0777)) return 0, nil, ipcSet(t, id, auth.UID(s.SemPerm.UID), auth.GID(s.SemPerm.GID), perms) + case linux.GETPID: + v, err := getPID(t, id, num) + return uintptr(v), nil, err + + case linux.IPC_INFO, + linux.SEM_INFO, + linux.IPC_STAT, + linux.SEM_STAT, + linux.SEM_STAT_ANY, + linux.GETNCNT, + linux.GETZCNT: + + t.Kernel().EmitUnimplementedEvent(t) + fallthrough + default: return 0, nil, syserror.EINVAL } @@ -161,7 +177,8 @@ func setVal(t *kernel.Task, id int32, num int32, val int16) error { return syserror.EINVAL } creds := auth.CredentialsFromContext(t) - return set.SetVal(t, num, val, creds) + pid := t.Kernel().GlobalInit().PIDNamespace().IDOfThreadGroup(t.ThreadGroup()) + return set.SetVal(t, num, val, creds, int32(pid)) } func setValAll(t *kernel.Task, id int32, array usermem.Addr) error { @@ -175,7 +192,8 @@ func setValAll(t *kernel.Task, id int32, array usermem.Addr) error { return err } creds := auth.CredentialsFromContext(t) - return set.SetValAll(t, vals, creds) + pid := t.Kernel().GlobalInit().PIDNamespace().IDOfThreadGroup(t.ThreadGroup()) + return set.SetValAll(t, vals, creds, int32(pid)) } func getVal(t *kernel.Task, id int32, num int32) (int16, error) { @@ -202,3 +220,22 @@ func getValAll(t *kernel.Task, id int32, array usermem.Addr) error { _, err = t.CopyOut(array, vals) return err } + +func getPID(t *kernel.Task, id int32, num int32) (int32, error) { + r := t.IPCNamespace().SemaphoreRegistry() + set := r.FindByID(id) + if set == nil { + return 0, syserror.EINVAL + } + creds := auth.CredentialsFromContext(t) + gpid, err := set.GetPID(num, creds) + if err != nil { + return 0, err + } + // Convert pid from init namespace to the caller's namespace. + tg := t.PIDNamespace().ThreadGroupWithID(kernel.ThreadID(gpid)) + if tg == nil { + return 0, nil + } + return int32(tg.ID()), nil +} diff --git a/runsc/boot/compat.go b/runsc/boot/compat.go index c2a77ebf5..572b5b472 100644 --- a/runsc/boot/compat.go +++ b/runsc/boot/compat.go @@ -100,6 +100,10 @@ func (c *compatEmitter) Emit(msg proto.Message) (hangup bool, err error) { // args: fd, level, name, ... tr = newArgsTracker(1, 2) + case syscall.SYS_SEMCTL: + // args: semid, semnum, cmd, ... + tr = newArgsTracker(2) + default: tr = &onceTracker{} } diff --git a/test/syscalls/linux/semaphore.cc b/test/syscalls/linux/semaphore.cc index da3d2c6fe..1c47b6851 100644 --- a/test/syscalls/linux/semaphore.cc +++ b/test/syscalls/linux/semaphore.cc @@ -431,6 +431,35 @@ TEST(SemaphoreTest, SemCtlValAll) { SyscallFailsWithErrno(EFAULT)); } +TEST(SemaphoreTest, SemCtlGetPid) { + AutoSem sem(semget(IPC_PRIVATE, 1, 0600 | IPC_CREAT)); + ASSERT_THAT(sem.get(), SyscallSucceeds()); + + ASSERT_THAT(semctl(sem.get(), 0, SETVAL, 1), SyscallSucceeds()); + EXPECT_THAT(semctl(sem.get(), 0, GETPID), SyscallSucceedsWithValue(getpid())); +} + +TEST(SemaphoreTest, SemCtlGetPidFork) { + AutoSem sem(semget(IPC_PRIVATE, 1, 0600 | IPC_CREAT)); + ASSERT_THAT(sem.get(), SyscallSucceeds()); + + const pid_t child_pid = fork(); + if (child_pid == 0) { + ASSERT_THAT(semctl(sem.get(), 0, SETVAL, 1), SyscallSucceeds()); + ASSERT_THAT(semctl(sem.get(), 0, GETPID), + SyscallSucceedsWithValue(getpid())); + + _exit(0); + } + ASSERT_THAT(child_pid, SyscallSucceeds()); + + int status; + ASSERT_THAT(RetryEINTR(waitpid)(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFEXITED(status) && WEXITSTATUS(status) == 0) + << " status " << status; +} + TEST(SemaphoreTest, SemIpcSet) { // Drop CAP_IPC_OWNER which allows us to bypass semaphore permissions. ASSERT_NO_ERRNO(SetCapability(CAP_IPC_OWNER, false)); -- cgit v1.2.3 From 0b76887147820a809beaa497ede8dc4f7b7b120a Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Tue, 5 Mar 2019 23:39:14 -0800 Subject: Priority-inheritance futex implementation It is Implemented without the priority inheritance part given that gVisor defers scheduling decisions to Go runtime and doesn't have control over it. PiperOrigin-RevId: 236989545 Change-Id: I714c8ca0798743ecf3167b14ffeb5cd834302560 --- pkg/abi/linux/futex.go | 6 + pkg/sentry/kernel/futex/BUILD | 2 + pkg/sentry/kernel/futex/futex.go | 215 +++++++++++++++++++++--- pkg/sentry/kernel/futex/futex_test.go | 4 + pkg/sentry/kernel/task_futex.go | 7 + pkg/sentry/mm/io.go | 45 +++++ pkg/sentry/platform/platform.go | 10 ++ pkg/sentry/platform/safecopy/BUILD | 4 +- pkg/sentry/platform/safecopy/atomic_amd64.s | 28 +++ pkg/sentry/platform/safecopy/atomic_arm64.s | 28 +++ pkg/sentry/platform/safecopy/safecopy.go | 4 + pkg/sentry/platform/safecopy/safecopy_unsafe.go | 20 +++ pkg/sentry/platform/safecopy/sighandler_amd64.s | 9 + pkg/sentry/platform/safecopy/sighandler_arm64.s | 11 ++ pkg/sentry/safemem/block_unsafe.go | 10 ++ pkg/sentry/syscalls/linux/sys_futex.go | 68 +++++++- pkg/sentry/usermem/bytes_io_unsafe.go | 8 + pkg/sentry/usermem/usermem.go | 7 + pkg/syserror/syserror.go | 1 + runsc/boot/compat.go | 4 +- test/syscalls/linux/BUILD | 1 + test/syscalls/linux/futex.cc | 115 ++++++++++++- 22 files changed, 578 insertions(+), 29 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/futex.go b/pkg/abi/linux/futex.go index 5dff01fba..afdf4123b 100644 --- a/pkg/abi/linux/futex.go +++ b/pkg/abi/linux/futex.go @@ -54,3 +54,9 @@ const ( // FUTEX_TID_MASK is the TID portion of a PI futex word. const FUTEX_TID_MASK = 0x3fffffff + +// Constants used for priority-inheritance futexes. +const ( + FUTEX_WAITERS = 0x80000000 + FUTEX_OWNER_DIED = 0x40000000 +) diff --git a/pkg/sentry/kernel/futex/BUILD b/pkg/sentry/kernel/futex/BUILD index 91feeb5ed..b6af5b20b 100644 --- a/pkg/sentry/kernel/futex/BUILD +++ b/pkg/sentry/kernel/futex/BUILD @@ -37,6 +37,8 @@ go_library( visibility = ["//pkg/sentry:internal"], deps = [ "//pkg/abi/linux", + "//pkg/log", + "//pkg/sentry/context", "//pkg/sentry/memmap", "//pkg/sentry/usermem", "//pkg/syserror", diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index b3e628fd4..cd7d51621 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -95,12 +95,15 @@ func (k *Key) matches(k2 *Key) bool { // Target abstracts memory accesses and keys. type Target interface { - // SwapUint32 gives access to usermem.SwapUint32. + // SwapUint32 gives access to usermem.IO.SwapUint32. SwapUint32(addr usermem.Addr, new uint32) (uint32, error) - // CompareAndSwap gives access to usermem.CompareAndSwapUint32. + // CompareAndSwap gives access to usermem.IO.CompareAndSwapUint32. CompareAndSwapUint32(addr usermem.Addr, old, new uint32) (uint32, error) + // LoadUint32 gives access to usermem.IO.LoadUint32. + LoadUint32(addr usermem.Addr) (uint32, error) + // GetSharedKey returns a Key with kind KindSharedPrivate or // KindSharedMappable corresponding to the memory mapped at address addr. // @@ -112,11 +115,11 @@ type Target interface { // check performs a basic equality check on the given address. func check(t Target, addr usermem.Addr, val uint32) error { - prev, err := t.CompareAndSwapUint32(addr, val, val) + cur, err := t.LoadUint32(addr) if err != nil { return err } - if prev != val { + if cur != val { return syserror.EAGAIN } return nil @@ -140,11 +143,14 @@ func atomicOp(t Target, addr usermem.Addr, opIn uint32) (bool, error) { ) if opType == linux.FUTEX_OP_SET { oldVal, err = t.SwapUint32(addr, opArg) + if err != nil { + return false, err + } } else { for { - oldVal, err = t.CompareAndSwapUint32(addr, 0, 0) + oldVal, err = t.LoadUint32(addr) if err != nil { - break + return false, err } var newVal uint32 switch opType { @@ -161,7 +167,7 @@ func atomicOp(t Target, addr usermem.Addr, opIn uint32) (bool, error) { } prev, err := t.CompareAndSwapUint32(addr, oldVal, newVal) if err != nil { - break + return false, err } if prev == oldVal { break // Success. @@ -222,6 +228,9 @@ type Waiter struct { // The bitmask we're waiting on. // This is used the case of a FUTEX_WAKE_BITSET. bitmask uint32 + + // tid is the thread ID for the waiter in case this is a PI mutex. + tid uint32 } // NewWaiter returns a new unqueued Waiter. @@ -262,23 +271,28 @@ func (b *bucket) wakeLocked(key *Key, bitmask uint32, n int) int { // Remove from the bucket and wake the waiter. woke := w w = w.Next() // Next iteration. - b.waiters.Remove(woke) - woke.C <- struct{}{} - - // NOTE: The above channel write establishes a write barrier according - // to the memory model, so nothing may be ordered around it. Since - // we've dequeued woke and will never touch it again, we can safely - // store nil to woke.bucket here and allow the WaitComplete() to - // short-circuit grabbing the bucket lock. If they somehow miss the - // store, we are still holding the lock, so we can know that they won't - // dequeue woke, assume it's free and have the below operation - // afterwards. - woke.bucket.Store(nil) + b.wakeWaiterLocked(woke) done++ } return done } +func (b *bucket) wakeWaiterLocked(w *Waiter) { + // Remove from the bucket and wake the waiter. + b.waiters.Remove(w) + w.C <- struct{}{} + + // NOTE: The above channel write establishes a write barrier according + // to the memory model, so nothing may be ordered around it. Since + // we've dequeued w and will never touch it again, we can safely + // store nil to w.bucket here and allow the WaitComplete() to + // short-circuit grabbing the bucket lock. If they somehow miss the + // store, we are still holding the lock, so we can know that they won't + // dequeue w, assume it's free and have the below operation + // afterwards. + w.bucket.Store(nil) +} + // requeueLocked takes n waiters from the bucket and moves them to naddr on the // bucket "to". // @@ -596,7 +610,7 @@ func (m *Manager) WaitComplete(w *Waiter) { continue } - // Remove w from b. + // Remove waiter from bucket. b.waiters.Remove(w) w.bucket.Store(nil) b.mu.Unlock() @@ -606,3 +620,164 @@ func (m *Manager) WaitComplete(w *Waiter) { // Release references held by the waiter. w.key.release() } + +// LockPI attempts to lock the futex following the Priority-inheritance futex +// rules. The lock is acquired only when 'addr' points to 0. The TID of the +// calling task is set to 'addr' to indicate the futex is owned. It returns true +// if the futex was successfully acquired. +// +// FUTEX_OWNER_DIED is only set by the Linux when robust lists are in use (see +// exit_robust_list()). Given we don't support robust lists, although handled +// below, it's never set. +func (m *Manager) LockPI(w *Waiter, t Target, addr usermem.Addr, tid uint32, private, try bool) (bool, error) { + k, err := getKey(t, addr, private) + if err != nil { + return false, err + } + // Ownership of k is transferred to w below. + + // Prepare the Waiter before taking the bucket lock. + select { + case <-w.C: + default: + } + w.key = k + w.tid = tid + + b := m.lockBucket(&k) + // Hot function: avoid defers. + + success, err := m.lockPILocked(w, t, addr, tid, b, try) + if err != nil { + w.key.release() + b.mu.Unlock() + return false, err + } + if success || try { + // Release waiter if it's not going to be a wait. + w.key.release() + } + b.mu.Unlock() + return success, nil +} + +func (m *Manager) lockPILocked(w *Waiter, t Target, addr usermem.Addr, tid uint32, b *bucket, try bool) (bool, error) { + for { + cur, err := t.LoadUint32(addr) + if err != nil { + return false, err + } + if (cur & linux.FUTEX_TID_MASK) == tid { + return false, syserror.EDEADLK + } + + if (cur & linux.FUTEX_TID_MASK) == 0 { + // No owner and no waiters, try to acquire the futex. + + // Set TID and preserve owner died status. + val := tid + val |= cur & linux.FUTEX_OWNER_DIED + prev, err := t.CompareAndSwapUint32(addr, cur, val) + if err != nil { + return false, err + } + if prev != cur { + // CAS failed, retry... + // Linux reacquires the bucket lock on retries, which will re-lookup the + // mapping at the futex address. However, retrying while holding the + // lock is more efficient and reduces the chance of another conflict. + continue + } + // Futex acquired. + return true, nil + } + + // Futex is already owned, prepare to wait. + + if try { + // Caller doesn't want to wait. + return false, nil + } + + // Set waiters bit if not set yet. + if cur&linux.FUTEX_WAITERS == 0 { + prev, err := t.CompareAndSwapUint32(addr, cur, cur|linux.FUTEX_WAITERS) + if err != nil { + return false, err + } + if prev != cur { + // CAS failed, retry... + continue + } + } + + // Add the waiter to the bucket. + b.waiters.PushBack(w) + w.bucket.Store(b) + return false, nil + } +} + +// UnlockPI unlock the futex following the Priority-inheritance futex +// rules. The address provided must contain the caller's TID. If there are +// waiters, TID of the next waiter (FIFO) is set to the given address, and the +// waiter woken up. If there are no waiters, 0 is set to the address. +func (m *Manager) UnlockPI(t Target, addr usermem.Addr, tid uint32, private bool) error { + k, err := getKey(t, addr, private) + if err != nil { + return err + } + b := m.lockBucket(&k) + + err = m.unlockPILocked(t, addr, tid, b) + + k.release() + b.mu.Unlock() + return err +} + +func (m *Manager) unlockPILocked(t Target, addr usermem.Addr, tid uint32, b *bucket) error { + cur, err := t.LoadUint32(addr) + if err != nil { + return err + } + + if (cur & linux.FUTEX_TID_MASK) != tid { + return syserror.EPERM + } + + if b.waiters.Empty() { + // It's safe to set 0 because there are no waiters, no new owner, and the + // executing task is the current owner (no owner died bit). + prev, err := t.CompareAndSwapUint32(addr, cur, 0) + if err != nil { + return err + } + if prev != cur { + // Let user mode handle CAS races. This is different than lock, which + // retries when CAS fails. + return syserror.EAGAIN + } + return nil + } + + next := b.waiters.Front() + + // Set next owner's TID, waiters if there are any. Resets owner died bit, if + // set, because the executing task takes over as the owner. + val := next.tid + if next.Next() != nil { + val |= linux.FUTEX_WAITERS + } + + prev, err := t.CompareAndSwapUint32(addr, cur, val) + if err != nil { + return err + } + if prev != cur { + return syserror.EINVAL + } + + b.wakeWaiterLocked(next) + return nil +} diff --git a/pkg/sentry/kernel/futex/futex_test.go b/pkg/sentry/kernel/futex/futex_test.go index a7ab9f229..9d44ee8e5 100644 --- a/pkg/sentry/kernel/futex/futex_test.go +++ b/pkg/sentry/kernel/futex/futex_test.go @@ -49,6 +49,10 @@ func (t testData) CompareAndSwapUint32(addr usermem.Addr, old, new uint32) (uint return atomic.LoadUint32((*uint32)(unsafe.Pointer(&t[addr]))), nil } +func (t testData) LoadUint32(addr usermem.Addr) (uint32, error) { + return atomic.LoadUint32((*uint32)(unsafe.Pointer(&t[addr]))), nil +} + func (t testData) GetSharedKey(addr usermem.Addr) (Key, error) { return Key{ Kind: KindSharedMappable, diff --git a/pkg/sentry/kernel/task_futex.go b/pkg/sentry/kernel/task_futex.go index 921f7bdbc..351cf47d7 100644 --- a/pkg/sentry/kernel/task_futex.go +++ b/pkg/sentry/kernel/task_futex.go @@ -41,6 +41,13 @@ func (t *Task) CompareAndSwapUint32(addr usermem.Addr, old, new uint32) (uint32, }) } +// LoadUint32 implemets futex.Target.LoadUint32. +func (t *Task) LoadUint32(addr usermem.Addr) (uint32, error) { + return t.MemoryManager().LoadUint32(t, addr, usermem.IOOpts{ + AddressSpaceActive: true, + }) +} + // GetSharedKey implements futex.Target.GetSharedKey. func (t *Task) GetSharedKey(addr usermem.Addr) (futex.Key, error) { return t.MemoryManager().GetSharedFutexKey(t, addr) diff --git a/pkg/sentry/mm/io.go b/pkg/sentry/mm/io.go index 6600ddd78..e0cebef84 100644 --- a/pkg/sentry/mm/io.go +++ b/pkg/sentry/mm/io.go @@ -346,6 +346,7 @@ func (mm *MemoryManager) SwapUint32(ctx context.Context, addr usermem.Addr, new if err != nil { return 0, translateIOError(ctx, err) } + // Return the number of bytes read. return 4, nil }) return old, err @@ -388,11 +389,55 @@ func (mm *MemoryManager) CompareAndSwapUint32(ctx context.Context, addr usermem. if err != nil { return 0, translateIOError(ctx, err) } + // Return the number of bytes read. return 4, nil }) return prev, err } +// LoadUint32 implements usermem.IO.LoadUint32. +func (mm *MemoryManager) LoadUint32(ctx context.Context, addr usermem.Addr, opts usermem.IOOpts) (uint32, error) { + ar, ok := mm.CheckIORange(addr, 4) + if !ok { + return 0, syserror.EFAULT + } + + // Do AddressSpace IO if applicable. + if mm.haveASIO && opts.AddressSpaceActive && !opts.IgnorePermissions { + for { + val, err := mm.as.LoadUint32(addr) + if err == nil { + return val, nil + } + if f, ok := err.(platform.SegmentationFault); ok { + if err := mm.handleASIOFault(ctx, f.Addr, ar, usermem.Read); err != nil { + return 0, err + } + continue + } + return 0, translateIOError(ctx, err) + } + } + + // Go through internal mappings. + var val uint32 + _, err := mm.withInternalMappings(ctx, ar, usermem.Read, opts.IgnorePermissions, func(ims safemem.BlockSeq) (uint64, error) { + if ims.NumBlocks() != 1 || ims.NumBytes() != 4 { + // Atomicity is unachievable across mappings. + return 0, syserror.EFAULT + } + im := ims.Head() + var err error + val, err = safemem.LoadUint32(im) + if err != nil { + return 0, translateIOError(ctx, err) + } + // Return the number of bytes read. + return 4, nil + }) + return val, err +} + // handleASIOFault handles a page fault at address addr for an AddressSpaceIO // operation spanning ioar. // diff --git a/pkg/sentry/platform/platform.go b/pkg/sentry/platform/platform.go index f16588e6e..a9e76bd45 100644 --- a/pkg/sentry/platform/platform.go +++ b/pkg/sentry/platform/platform.go @@ -254,6 +254,11 @@ type AddressSpaceIO interface { // // Preconditions: addr must be aligned to a 4-byte boundary. CompareAndSwapUint32(addr usermem.Addr, old, new uint32) (uint32, error) + + // LoadUint32 atomically loads the uint32 value at addr and returns it. + // + // Preconditions: addr must be aligned to a 4-byte boundary. + LoadUint32(addr usermem.Addr) (uint32, error) } // NoAddressSpaceIO implements AddressSpaceIO methods by panicing. @@ -284,6 +289,11 @@ func (NoAddressSpaceIO) CompareAndSwapUint32(addr usermem.Addr, old, new uint32) panic("This platform does not support AddressSpaceIO") } +// LoadUint32 implements AddressSpaceIO.LoadUint32. +func (NoAddressSpaceIO) LoadUint32(addr usermem.Addr) (uint32, error) { + panic("This platform does not support AddressSpaceIO") +} + // SegmentationFault is an error returned by AddressSpaceIO methods when IO // fails due to access of an unmapped page, or a mapped page with insufficient // permissions. diff --git a/pkg/sentry/platform/safecopy/BUILD b/pkg/sentry/platform/safecopy/BUILD index 05a6a61ae..d97a40297 100644 --- a/pkg/sentry/platform/safecopy/BUILD +++ b/pkg/sentry/platform/safecopy/BUILD @@ -18,9 +18,7 @@ go_library( ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/platform/safecopy", visibility = ["//pkg/sentry:internal"], - deps = [ - "//pkg/syserror", - ], + deps = ["//pkg/syserror"], ) go_test( diff --git a/pkg/sentry/platform/safecopy/atomic_amd64.s b/pkg/sentry/platform/safecopy/atomic_amd64.s index 873ffa046..f90b4bfd1 100644 --- a/pkg/sentry/platform/safecopy/atomic_amd64.s +++ b/pkg/sentry/platform/safecopy/atomic_amd64.s @@ -106,3 +106,31 @@ TEXT ·compareAndSwapUint32(SB), NOSPLIT, $0-24 CMPXCHGL DX, 0(DI) MOVL AX, prev+16(FP) RET + +// handleLoadUint32Fault returns the value stored in DI. Control is transferred +// to it when LoadUint32 below receives SIGSEGV or SIGBUS, with the signal +// number stored in DI. +// +// It must have the same frame configuration as loadUint32 so that it can undo +// any potential call frame set up by the assembler. +TEXT handleLoadUint32Fault(SB), NOSPLIT, $0-16 + MOVL DI, sig+12(FP) + RET + +// loadUint32 atomically loads *addr and returns it. If a SIGSEGV or SIGBUS +// signal is received, the value returned is unspecified, and sig is the number +// of the signal that was received. +// +// Preconditions: addr must be aligned to a 4-byte boundary. +// +//func loadUint32(ptr unsafe.Pointer) (val uint32, sig int32) +TEXT ·loadUint32(SB), NOSPLIT, $0-16 + // Store 0 as the returned signal number. If we run to completion, + // this is the value the caller will see; if a signal is received, + // handleLoadUint32Fault will store a different value in this address. + MOVL $0, sig+12(FP) + + MOVQ addr+0(FP), AX + MOVL (AX), BX + MOVL BX, val+8(FP) + RET diff --git a/pkg/sentry/platform/safecopy/atomic_arm64.s b/pkg/sentry/platform/safecopy/atomic_arm64.s index 554a5c1e1..d58ed71f7 100644 --- a/pkg/sentry/platform/safecopy/atomic_arm64.s +++ b/pkg/sentry/platform/safecopy/atomic_arm64.s @@ -96,3 +96,31 @@ again: done: MOVW R3, prev+16(FP) RET + +// handleLoadUint32Fault returns the value stored in DI. Control is transferred +// to it when LoadUint32 below receives SIGSEGV or SIGBUS, with the signal +// number stored in DI. +// +// It must have the same frame configuration as loadUint32 so that it can undo +// any potential call frame set up by the assembler. +TEXT handleLoadUint32Fault(SB), NOSPLIT, $0-16 + MOVW R1, sig+12(FP) + RET + +// loadUint32 atomically loads *addr and returns it. If a SIGSEGV or SIGBUS +// signal is received, the value returned is unspecified, and sig is the number +// of the signal that was received. +// +// Preconditions: addr must be aligned to a 4-byte boundary. +// +//func loadUint32(ptr unsafe.Pointer) (val uint32, sig int32) +TEXT ·loadUint32(SB), NOSPLIT, $0-16 + // Store 0 as the returned signal number. If we run to completion, + // this is the value the caller will see; if a signal is received, + // handleLoadUint32Fault will store a different value in this address. + MOVW $0, sig+12(FP) + + MOVD addr+0(FP), R0 + LDARW (R0), R1 + MOVW R1, val+8(FP) + RET diff --git a/pkg/sentry/platform/safecopy/safecopy.go b/pkg/sentry/platform/safecopy/safecopy.go index c60f73103..69c66a3b7 100644 --- a/pkg/sentry/platform/safecopy/safecopy.go +++ b/pkg/sentry/platform/safecopy/safecopy.go @@ -75,6 +75,8 @@ var ( swapUint64End uintptr compareAndSwapUint32Begin uintptr compareAndSwapUint32End uintptr + loadUint32Begin uintptr + loadUint32End uintptr // savedSigSegVHandler is a pointer to the SIGSEGV handler that was // configured before we replaced it with our own. We still call into it @@ -119,6 +121,8 @@ func initializeAddresses() { swapUint64End = FindEndAddress(swapUint64Begin) compareAndSwapUint32Begin = reflect.ValueOf(compareAndSwapUint32).Pointer() compareAndSwapUint32End = FindEndAddress(compareAndSwapUint32Begin) + loadUint32Begin = reflect.ValueOf(loadUint32).Pointer() + loadUint32End = FindEndAddress(loadUint32Begin) } func init() { diff --git a/pkg/sentry/platform/safecopy/safecopy_unsafe.go b/pkg/sentry/platform/safecopy/safecopy_unsafe.go index e78a6714e..f84527484 100644 --- a/pkg/sentry/platform/safecopy/safecopy_unsafe.go +++ b/pkg/sentry/platform/safecopy/safecopy_unsafe.go @@ -79,6 +79,14 @@ func swapUint64(ptr unsafe.Pointer, new uint64) (old uint64, sig int32) //go:noescape func compareAndSwapUint32(ptr unsafe.Pointer, old, new uint32) (prev uint32, sig int32) +// LoadUint32 is like sync/atomic.LoadUint32, but operates with user memory. It +// may fail with SIGSEGV or SIGBUS if it is received while reading from ptr. +// +// Preconditions: ptr must be aligned to a 4-byte boundary. +// +//go:noescape +func loadUint32(ptr unsafe.Pointer) (val uint32, sig int32) + // CopyIn copies len(dst) bytes from src to dst. It returns the number of bytes // copied and an error if SIGSEGV or SIGBUS is received while reading from src. func CopyIn(dst []byte, src unsafe.Pointer) (int, error) { @@ -260,6 +268,18 @@ func CompareAndSwapUint32(ptr unsafe.Pointer, old, new uint32) (uint32, error) { return prev, errorFromFaultSignal(ptr, sig) } +// LoadUint32 is like sync/atomic.LoadUint32, but operates with user memory. It +// may fail with SIGSEGV or SIGBUS if it is received while reading from ptr. +// +// Preconditions: ptr must be aligned to a 4-byte boundary. +func LoadUint32(ptr unsafe.Pointer) (uint32, error) { + if addr := uintptr(ptr); addr&3 != 0 { + return 0, AlignmentError{addr, 4} + } + val, sig := loadUint32(ptr) + return val, errorFromFaultSignal(ptr, sig) +} + func errorFromFaultSignal(addr unsafe.Pointer, sig int32) error { switch sig { case 0: diff --git a/pkg/sentry/platform/safecopy/sighandler_amd64.s b/pkg/sentry/platform/safecopy/sighandler_amd64.s index 06614f1b4..db7701a29 100644 --- a/pkg/sentry/platform/safecopy/sighandler_amd64.s +++ b/pkg/sentry/platform/safecopy/sighandler_amd64.s @@ -101,6 +101,15 @@ not_swapuint64: JMP handle_fault not_casuint32: + CMPQ CX, ·loadUint32Begin(SB) + JB not_loaduint32 + CMPQ CX, ·loadUint32End(SB) + JAE not_loaduint32 + + LEAQ handleLoadUint32Fault(SB), CX + JMP handle_fault + +not_loaduint32: original_handler: // Jump to the previous signal handler, which is likely the golang one. XORQ CX, CX diff --git a/pkg/sentry/platform/safecopy/sighandler_arm64.s b/pkg/sentry/platform/safecopy/sighandler_arm64.s index 5e8e193e7..cdfca8207 100644 --- a/pkg/sentry/platform/safecopy/sighandler_arm64.s +++ b/pkg/sentry/platform/safecopy/sighandler_arm64.s @@ -110,6 +110,17 @@ not_swapuint64: B handle_fault not_casuint32: + MOVD ·loadUint32Begin(SB), R8 + CMP R8, R7 + BLO not_loaduint32 + MOVD ·loadUint32End(SB), R8 + CMP R8, R7 + BHS not_loaduint32 + + MOVD $handleLoadUint32Fault(SB), R7 + B handle_fault + +not_loaduint32: original_handler: // Jump to the previous signal handler, which is likely the golang one. MOVD ·savedSigBusHandler(SB), R7 diff --git a/pkg/sentry/safemem/block_unsafe.go b/pkg/sentry/safemem/block_unsafe.go index e91ff66ae..c3a9780d2 100644 --- a/pkg/sentry/safemem/block_unsafe.go +++ b/pkg/sentry/safemem/block_unsafe.go @@ -267,3 +267,13 @@ func CompareAndSwapUint32(b Block, old, new uint32) (uint32, error) { } return safecopy.CompareAndSwapUint32(b.start, old, new) } + +// LoadUint32 invokes safecopy.LoadUint32 on the first 4 bytes of b. +// +// Preconditions: b.Len() >= 4. +func LoadUint32(b Block) (uint32, error) { + if b.length < 4 { + panic(fmt.Sprintf("insufficient length: %d", b.length)) + } + return safecopy.LoadUint32(b.start) +} diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index 7a1d396ec..f0c89cba4 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -124,6 +124,46 @@ func futexWaitDuration(t *kernel.Task, duration time.Duration, forever bool, add return 0, kernel.ERESTART_RESTARTBLOCK } +func futexLockPI(t *kernel.Task, ts linux.Timespec, forever bool, addr usermem.Addr, private bool) error { + w := t.FutexWaiter() + locked, err := t.Futex().LockPI(w, t, addr, uint32(t.ThreadID()), private, false) + if err != nil { + return err + } + if locked { + // Futex acquired, we're done! + return nil + } + + if forever { + err = t.Block(w.C) + } else { + notifier, tchan := ktime.NewChannelNotifier() + timer := ktime.NewTimer(t.Kernel().RealtimeClock(), notifier) + timer.Swap(ktime.Setting{ + Enabled: true, + Next: ktime.FromTimespec(ts), + }) + err = t.BlockWithTimer(w.C, tchan) + timer.Destroy() + } + + t.Futex().WaitComplete(w) + return syserror.ConvertIntr(err, kernel.ERESTARTSYS) +} + +func tryLockPI(t *kernel.Task, addr usermem.Addr, private bool) error { + w := t.FutexWaiter() + locked, err := t.Futex().LockPI(w, t, addr, uint32(t.ThreadID()), private, true) + if err != nil { + return err + } + if !locked { + return syserror.EWOULDBLOCK + } + return nil +} + // Futex implements linux syscall futex(2). // It provides a method for a program to wait for a value at a given address to // change, and a method to wake up anyone waiting on a particular address. @@ -144,7 +184,7 @@ func Futex(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall switch cmd { case linux.FUTEX_WAIT, linux.FUTEX_WAIT_BITSET: // WAIT{_BITSET} wait forever if the timeout isn't passed. - forever := timeout == 0 + forever := (timeout == 0) var timespec linux.Timespec if !forever { @@ -205,8 +245,30 @@ func Futex(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall n, err := t.Futex().WakeOp(t, addr, naddr, private, val, nreq, op) return uintptr(n), nil, err - case linux.FUTEX_LOCK_PI, linux.FUTEX_UNLOCK_PI, linux.FUTEX_TRYLOCK_PI, linux.FUTEX_WAIT_REQUEUE_PI, linux.FUTEX_CMP_REQUEUE_PI: - // We don't support any priority inversion futexes. + case linux.FUTEX_LOCK_PI: + forever := (timeout == 0) + + var timespec linux.Timespec + if !forever { + var err error + timespec, err = copyTimespecIn(t, timeout) + if err != nil { + return 0, nil, err + } + } + err := futexLockPI(t, timespec, forever, addr, private) + return 0, nil, err + + case linux.FUTEX_TRYLOCK_PI: + err := tryLockPI(t, addr, private) + return 0, nil, err + + case linux.FUTEX_UNLOCK_PI: + err := t.Futex().UnlockPI(t, addr, uint32(t.ThreadID()), private) + return 0, nil, err + + case linux.FUTEX_WAIT_REQUEUE_PI, linux.FUTEX_CMP_REQUEUE_PI: + t.Kernel().EmitUnimplementedEvent(t) return 0, nil, syserror.ENOSYS default: diff --git a/pkg/sentry/usermem/bytes_io_unsafe.go b/pkg/sentry/usermem/bytes_io_unsafe.go index 8bdf3a508..7add8bc82 100644 --- a/pkg/sentry/usermem/bytes_io_unsafe.go +++ b/pkg/sentry/usermem/bytes_io_unsafe.go @@ -37,3 +37,11 @@ func (b *BytesIO) CompareAndSwapUint32(ctx context.Context, addr Addr, old, new } return atomicbitops.CompareAndSwapUint32((*uint32)(unsafe.Pointer(&b.Bytes[int(addr)])), old, new), nil } + +// LoadUint32 implements IO.LoadUint32. +func (b *BytesIO) LoadUint32(ctx context.Context, addr Addr, opts IOOpts) (uint32, error) { + if _, err := b.rangeCheck(addr, 4); err != nil { + return 0, err + } + return atomic.LoadUint32((*uint32)(unsafe.Pointer(&b.Bytes[int(addr)]))), nil +} diff --git a/pkg/sentry/usermem/usermem.go b/pkg/sentry/usermem/usermem.go index 75ac4d22d..c3c9c153b 100644 --- a/pkg/sentry/usermem/usermem.go +++ b/pkg/sentry/usermem/usermem.go @@ -103,6 +103,13 @@ type IO interface { // any following locks in the lock order. addr must be aligned to a 4-byte // boundary. CompareAndSwapUint32(ctx context.Context, addr Addr, old, new uint32, opts IOOpts) (uint32, error) + + // LoadUint32 atomically loads the uint32 value at addr and returns it. + // + // Preconditions: The caller must not hold mm.MemoryManager.mappingMu or + // any following locks in the lock order. addr must be aligned to a 4-byte + // boundary. + LoadUint32(ctx context.Context, addr Addr, opts IOOpts) (uint32, error) } // IOOpts contains options applicable to all IO methods. diff --git a/pkg/syserror/syserror.go b/pkg/syserror/syserror.go index 4228707f4..5558cccff 100644 --- a/pkg/syserror/syserror.go +++ b/pkg/syserror/syserror.go @@ -33,6 +33,7 @@ var ( ECHILD = error(syscall.ECHILD) ECONNREFUSED = error(syscall.ECONNREFUSED) ECONNRESET = error(syscall.ECONNRESET) + EDEADLK = error(syscall.EDEADLK) EEXIST = error(syscall.EEXIST) EFAULT = error(syscall.EFAULT) EFBIG = error(syscall.EFBIG) diff --git a/runsc/boot/compat.go b/runsc/boot/compat.go index 37d0c31fd..b3499bcde 100644 --- a/runsc/boot/compat.go +++ b/runsc/boot/compat.go @@ -99,8 +99,8 @@ func (c *compatEmitter) emitUnimplementedSyscall(us *spb.UnimplementedSyscall) { // args: cmd, ... tr = newArgsTracker(0) - case syscall.SYS_IOCTL, syscall.SYS_EPOLL_CTL, syscall.SYS_SHMCTL: - // args: fd, cmd, ... + case syscall.SYS_IOCTL, syscall.SYS_EPOLL_CTL, syscall.SYS_SHMCTL, syscall.SYS_FUTEX: + // args: fd/addr, cmd, ... tr = newArgsTracker(1) case syscall.SYS_GETSOCKOPT, syscall.SYS_SETSOCKOPT: diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index 4c818238b..2c214925e 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -808,6 +808,7 @@ cc_binary( "//test/util:cleanup", "//test/util:file_descriptor", "//test/util:memory_util", + "//test/util:save_util", "//test/util:temp_path", "//test/util:test_main", "//test/util:test_util", diff --git a/test/syscalls/linux/futex.cc b/test/syscalls/linux/futex.cc index 6fa284013..35933b660 100644 --- a/test/syscalls/linux/futex.cc +++ b/test/syscalls/linux/futex.cc @@ -32,6 +32,7 @@ #include "test/util/cleanup.h" #include "test/util/file_descriptor.h" #include "test/util/memory_util.h" +#include "test/util/save_util.h" #include "test/util/temp_path.h" #include "test/util/test_util.h" #include "test/util/thread_util.h" @@ -118,6 +119,30 @@ int futex_wake_op(bool priv, std::atomic* uaddr1, std::atomic* uaddr2, return syscall(SYS_futex, uaddr1, op, nwake1, nwake2, uaddr2, sub_op); } +int futex_lock_pi(bool priv, std::atomic* uaddr) { + int op = FUTEX_LOCK_PI; + if (priv) { + op |= FUTEX_PRIVATE_FLAG; + } + return RetryEINTR(syscall)(SYS_futex, uaddr, op, nullptr, nullptr); +} + +int futex_trylock_pi(bool priv, std::atomic* uaddr) { + int op = FUTEX_TRYLOCK_PI; + if (priv) { + op |= FUTEX_PRIVATE_FLAG; + } + return RetryEINTR(syscall)(SYS_futex, uaddr, op, nullptr, nullptr); +} + +int futex_unlock_pi(bool priv, std::atomic* uaddr) { + int op = FUTEX_UNLOCK_PI; + if (priv) { + op |= FUTEX_PRIVATE_FLAG; + } + return RetryEINTR(syscall)(SYS_futex, uaddr, op, nullptr, nullptr); +} + // Fixture for futex tests parameterized by whether to use private or shared // futexes. class PrivateAndSharedFutexTest : public ::testing::TestWithParam { @@ -589,7 +614,95 @@ TEST(SharedFutexTest, WakeInterprocessFile_NoRandomSave) { << " status " << status; } -} // namespace +TEST_P(PrivateAndSharedFutexTest, PIBasic) { + std::atomic a = ATOMIC_VAR_INIT(0); + + ASSERT_THAT(futex_lock_pi(IsPrivate(), &a), SyscallSucceeds()); + EXPECT_EQ(a.load(), gettid()); + EXPECT_THAT(futex_lock_pi(IsPrivate(), &a), SyscallFailsWithErrno(EDEADLK)); + ASSERT_THAT(futex_unlock_pi(IsPrivate(), &a), SyscallSucceeds()); + EXPECT_EQ(a.load(), 0); + EXPECT_THAT(futex_unlock_pi(IsPrivate(), &a), SyscallFailsWithErrno(EPERM)); +} + +TEST_P(PrivateAndSharedFutexTest, PIConcurrency_NoRandomSave) { + DisableSave ds; // Too many syscalls. + + std::atomic a = ATOMIC_VAR_INIT(0); + const bool is_priv = IsPrivate(); + + std::unique_ptr threads[100]; + for (size_t i = 0; i < ABSL_ARRAYSIZE(threads); ++i) { + threads[i] = absl::make_unique([is_priv, &a] { + for (size_t j = 0; j < 10; ++j) { + ASSERT_THAT(futex_lock_pi(is_priv, &a), SyscallSucceeds()); + EXPECT_EQ(a.load() & FUTEX_TID_MASK, gettid()); + SleepSafe(absl::Milliseconds(5)); + ASSERT_THAT(futex_unlock_pi(is_priv, &a), SyscallSucceeds()); + } + }); + } +} + +TEST_P(PrivateAndSharedFutexTest, PIWaiters) { + std::atomic a = ATOMIC_VAR_INIT(0); + const bool is_priv = IsPrivate(); + + ASSERT_THAT(futex_lock_pi(is_priv, &a), SyscallSucceeds()); + EXPECT_EQ(a.load(), gettid()); + + ScopedThread th([is_priv, &a] { + ASSERT_THAT(futex_lock_pi(is_priv, &a), SyscallSucceeds()); + ASSERT_THAT(futex_unlock_pi(is_priv, &a), SyscallSucceeds()); + }); + + // Wait until the thread blocks on the futex, setting the waiters bit. + auto start = absl::Now(); + while (a.load() != (FUTEX_WAITERS | gettid())) { + ASSERT_LT(absl::Now() - start, absl::Seconds(5)); + absl::SleepFor(absl::Milliseconds(100)); + } + ASSERT_THAT(futex_unlock_pi(is_priv, &a), SyscallSucceeds()); +} + +TEST_P(PrivateAndSharedFutexTest, PITryLock) { + std::atomic a = ATOMIC_VAR_INIT(0); + const bool is_priv = IsPrivate(); + + ASSERT_THAT(futex_trylock_pi(IsPrivate(), &a), SyscallSucceeds()); + EXPECT_EQ(a.load(), gettid()); + + EXPECT_THAT(futex_trylock_pi(is_priv, &a), SyscallFailsWithErrno(EDEADLK)); + ScopedThread th([is_priv, &a] { + EXPECT_THAT(futex_trylock_pi(is_priv, &a), SyscallFailsWithErrno(EAGAIN)); + }); + th.Join(); + + ASSERT_THAT(futex_unlock_pi(IsPrivate(), &a), SyscallSucceeds()); +} + +TEST_P(PrivateAndSharedFutexTest, PITryLockConcurrency_NoRandomSave) { + DisableSave ds; // Too many syscalls. + + std::atomic a = ATOMIC_VAR_INIT(0); + const bool is_priv = IsPrivate(); + + std::unique_ptr threads[100]; + for (size_t i = 0; i < ABSL_ARRAYSIZE(threads); ++i) { + threads[i] = absl::make_unique([is_priv, &a] { + for (size_t j = 0; j < 10;) { + if (futex_trylock_pi(is_priv, &a) >= 0) { + ++j; + EXPECT_EQ(a.load() & FUTEX_TID_MASK, gettid()); + SleepSafe(absl::Milliseconds(5)); + ASSERT_THAT(futex_unlock_pi(is_priv, &a), SyscallSucceeds()); + } + } + }); + } +} + +} // namespace } // namespace testing } // namespace gvisor -- cgit v1.2.3 From 8f4634997bd97810a85a70b71f000378d9db2e55 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Thu, 14 Mar 2019 08:11:36 -0700 Subject: Decouple filemem from platform and move it to pgalloc.MemoryFile. This is in preparation for improved page cache reclaim, which requires greater integration between the page cache and page allocator. PiperOrigin-RevId: 238444706 Change-Id: Id24141b3678d96c7d7dc24baddd9be555bffafe4 --- pkg/sentry/context/contexttest/BUILD | 2 + pkg/sentry/context/contexttest/contexttest.go | 25 + pkg/sentry/fs/ashmem/BUILD | 1 - pkg/sentry/fs/binder/BUILD | 1 + pkg/sentry/fs/binder/binder.go | 21 +- pkg/sentry/fs/dev/BUILD | 2 +- pkg/sentry/fs/dev/null.go | 4 +- pkg/sentry/fs/fsutil/BUILD | 1 + pkg/sentry/fs/fsutil/README.md | 11 +- pkg/sentry/fs/fsutil/file_range_set.go | 23 +- pkg/sentry/fs/fsutil/inode_cached.go | 42 +- pkg/sentry/fs/proc/meminfo.go | 6 +- pkg/sentry/fs/tmpfs/inode_file.go | 24 +- pkg/sentry/fs/tmpfs/tmpfs.go | 2 +- pkg/sentry/kernel/BUILD | 3 +- pkg/sentry/kernel/contexttest/BUILD | 1 + pkg/sentry/kernel/contexttest/contexttest.go | 2 + pkg/sentry/kernel/kernel.go | 57 +- pkg/sentry/kernel/memevent/memory_events.go | 2 +- pkg/sentry/kernel/shm/BUILD | 1 + pkg/sentry/kernel/shm/shm.go | 19 +- pkg/sentry/kernel/task.go | 5 + pkg/sentry/kernel/task_context.go | 2 +- pkg/sentry/kernel/timekeeper.go | 5 +- pkg/sentry/kernel/timekeeper_test.go | 8 +- pkg/sentry/kernel/vdso.go | 17 +- pkg/sentry/loader/BUILD | 2 +- pkg/sentry/loader/vdso.go | 21 +- pkg/sentry/memutil/memutil_unsafe.go | 14 +- pkg/sentry/mm/BUILD | 2 + pkg/sentry/mm/README.md | 4 +- pkg/sentry/mm/aio_context.go | 17 +- pkg/sentry/mm/lifecycle.go | 5 +- pkg/sentry/mm/mm.go | 20 +- pkg/sentry/mm/mm_test.go | 4 +- pkg/sentry/mm/pma.go | 20 +- pkg/sentry/mm/save_restore.go | 10 +- pkg/sentry/mm/special_mappable.go | 36 +- pkg/sentry/mm/syscalls.go | 8 +- pkg/sentry/pgalloc/BUILD | 57 ++ pkg/sentry/pgalloc/context.go | 48 ++ pkg/sentry/pgalloc/pgalloc.go | 922 ++++++++++++++++++++++++++ pkg/sentry/pgalloc/pgalloc_test.go | 168 +++++ pkg/sentry/pgalloc/pgalloc_unsafe.go | 40 ++ pkg/sentry/pgalloc/save_restore.go | 205 ++++++ pkg/sentry/platform/filemem/BUILD | 56 -- pkg/sentry/platform/filemem/filemem.go | 879 ------------------------ pkg/sentry/platform/filemem/filemem_state.go | 194 ------ pkg/sentry/platform/filemem/filemem_test.go | 168 ----- pkg/sentry/platform/filemem/filemem_unsafe.go | 40 -- pkg/sentry/platform/kvm/BUILD | 1 - pkg/sentry/platform/kvm/address_space.go | 4 - pkg/sentry/platform/kvm/kvm.go | 17 - pkg/sentry/platform/kvm/kvm_test.go | 1 - pkg/sentry/platform/platform.go | 90 +-- pkg/sentry/platform/ptrace/BUILD | 1 - pkg/sentry/platform/ptrace/ptrace.go | 14 +- pkg/sentry/state/BUILD | 1 - pkg/sentry/state/state.go | 5 +- pkg/sentry/syscalls/linux/sys_sysinfo.go | 6 +- pkg/sentry/usage/memory.go | 3 - runsc/boot/BUILD | 2 + runsc/boot/controller.go | 9 +- runsc/boot/events.go | 2 +- runsc/boot/loader.go | 26 +- 65 files changed, 1743 insertions(+), 1666 deletions(-) create mode 100644 pkg/sentry/pgalloc/BUILD create mode 100644 pkg/sentry/pgalloc/context.go create mode 100644 pkg/sentry/pgalloc/pgalloc.go create mode 100644 pkg/sentry/pgalloc/pgalloc_test.go create mode 100644 pkg/sentry/pgalloc/pgalloc_unsafe.go create mode 100644 pkg/sentry/pgalloc/save_restore.go delete mode 100644 pkg/sentry/platform/filemem/BUILD delete mode 100644 pkg/sentry/platform/filemem/filemem.go delete mode 100644 pkg/sentry/platform/filemem/filemem_state.go delete mode 100644 pkg/sentry/platform/filemem/filemem_test.go delete mode 100644 pkg/sentry/platform/filemem/filemem_unsafe.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/context/contexttest/BUILD b/pkg/sentry/context/contexttest/BUILD index bed156b70..ce4f1e42c 100644 --- a/pkg/sentry/context/contexttest/BUILD +++ b/pkg/sentry/context/contexttest/BUILD @@ -13,6 +13,8 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/limits", + "//pkg/sentry/memutil", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", "//pkg/sentry/platform/ptrace", "//pkg/sentry/uniqueid", diff --git a/pkg/sentry/context/contexttest/contexttest.go b/pkg/sentry/context/contexttest/contexttest.go index d5fd9f165..a29087775 100644 --- a/pkg/sentry/context/contexttest/contexttest.go +++ b/pkg/sentry/context/contexttest/contexttest.go @@ -16,6 +16,7 @@ package contexttest import ( + "os" "sync/atomic" "testing" "time" @@ -24,6 +25,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" + "gvisor.googlesource.com/gvisor/pkg/sentry/memutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/platform/ptrace" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" @@ -35,6 +38,17 @@ import ( // Note that some filesystems may require a minimal kernel for testing, which // this test context does not provide. For such tests, see kernel/contexttest. func Context(tb testing.TB) context.Context { + const memfileName = "contexttest-memory" + memfd, err := memutil.CreateMemFD(memfileName, 0) + if err != nil { + tb.Fatalf("error creating application memory file: %v", err) + } + memfile := os.NewFile(uintptr(memfd), memfileName) + mf, err := pgalloc.NewMemoryFile(memfile) + if err != nil { + memfile.Close() + tb.Fatalf("error creating pgalloc.MemoryFile: %v", err) + } p, err := ptrace.New() if err != nil { tb.Fatal(err) @@ -43,6 +57,7 @@ func Context(tb testing.TB) context.Context { return &TestContext{ Context: context.Background(), l: limits.NewLimitSet(), + mf: mf, platform: p, otherValues: make(map[interface{}]interface{}), } @@ -53,6 +68,7 @@ func Context(tb testing.TB) context.Context { type TestContext struct { context.Context l *limits.LimitSet + mf *pgalloc.MemoryFile platform platform.Platform otherValues map[interface{}]interface{} } @@ -94,6 +110,10 @@ func (t *TestContext) Value(key interface{}) interface{} { switch key { case limits.CtxLimits: return t.l + case pgalloc.CtxMemoryFile: + return t.mf + case pgalloc.CtxMemoryFileProvider: + return t case platform.CtxPlatform: return t.platform case uniqueid.CtxGlobalUniqueID: @@ -112,6 +132,11 @@ func (t *TestContext) Value(key interface{}) interface{} { } } +// MemoryFile implements pgalloc.MemoryFileProvider.MemoryFile. +func (t *TestContext) MemoryFile() *pgalloc.MemoryFile { + return t.mf +} + // RootContext returns a Context that may be used in tests that need root // credentials. Uses ptrace as the platform.Platform. func RootContext(tb testing.TB) context.Context { diff --git a/pkg/sentry/fs/ashmem/BUILD b/pkg/sentry/fs/ashmem/BUILD index dcf620dca..ef1c31a3e 100644 --- a/pkg/sentry/fs/ashmem/BUILD +++ b/pkg/sentry/fs/ashmem/BUILD @@ -23,7 +23,6 @@ go_library( "//pkg/sentry/fs/tmpfs", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", - "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", "//pkg/syserror", diff --git a/pkg/sentry/fs/binder/BUILD b/pkg/sentry/fs/binder/BUILD index 8a448175f..3710664d3 100644 --- a/pkg/sentry/fs/binder/BUILD +++ b/pkg/sentry/fs/binder/BUILD @@ -17,6 +17,7 @@ go_library( "//pkg/sentry/fs/fsutil", "//pkg/sentry/kernel", "//pkg/sentry/memmap", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index 19cd55e65..16fb4806f 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -25,6 +25,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -74,9 +75,9 @@ func NewDevice(ctx context.Context, owner fs.FileOwner, fp fs.FilePermissions) * // ioctl. func (bd *Device) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) (*fs.File, error) { return fs.NewFile(ctx, d, flags, &Proc{ - bd: bd, - task: kernel.TaskFromContext(ctx), - platform: platform.FromContext(ctx), + bd: bd, + task: kernel.TaskFromContext(ctx), + mfp: pgalloc.MemoryFileProviderFromContext(ctx), }), nil } @@ -88,14 +89,14 @@ type Proc struct { fsutil.FileNoFsync `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` - bd *Device - task *kernel.Task - platform platform.Platform + bd *Device + task *kernel.Task + mfp pgalloc.MemoryFileProvider // mu protects fr. mu sync.Mutex `state:"nosave"` - // mapped is memory allocated from platform.Memory() by AddMapping. + // mapped is memory allocated from mfp.MemoryFile() by AddMapping. mapped platform.FileRange } @@ -104,7 +105,7 @@ func (bp *Proc) Release() { bp.mu.Lock() defer bp.mu.Unlock() if bp.mapped.Length() != 0 { - bp.platform.Memory().DecRef(bp.mapped) + bp.mfp.MemoryFile().DecRef(bp.mapped) } } @@ -204,7 +205,7 @@ func (bp *Proc) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar userm } // Binder only allocates and maps a single page up-front // (drivers/android/binder.c:binder_mmap() => binder_update_page_range()). - fr, err := bp.platform.Memory().Allocate(usermem.PageSize, usage.Anonymous) + fr, err := bp.mfp.MemoryFile().Allocate(usermem.PageSize, usage.Anonymous) if err != nil { return err } @@ -241,7 +242,7 @@ func (bp *Proc) Translate(ctx context.Context, required, optional memmap.Mappabl return []memmap.Translation{ { Source: memmap.MappableRange{0, usermem.PageSize}, - File: bp.platform.Memory(), + File: bp.mfp.MemoryFile(), Offset: bp.mapped.Start, }, }, err diff --git a/pkg/sentry/fs/dev/BUILD b/pkg/sentry/fs/dev/BUILD index e5b962c8c..6c4fdaba9 100644 --- a/pkg/sentry/fs/dev/BUILD +++ b/pkg/sentry/fs/dev/BUILD @@ -27,7 +27,7 @@ go_library( "//pkg/sentry/fs/tmpfs", "//pkg/sentry/memmap", "//pkg/sentry/mm", - "//pkg/sentry/platform", + "//pkg/sentry/pgalloc", "//pkg/sentry/safemem", "//pkg/sentry/usermem", "//pkg/syserror", diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 73fd09058..83f43c203 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -21,7 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/mm" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -115,7 +115,7 @@ var _ fs.FileOperations = (*zeroFileOperations)(nil) // ConfigureMMap implements fs.FileOperations.ConfigureMMap. func (*zeroFileOperations) ConfigureMMap(ctx context.Context, file *fs.File, opts *memmap.MMapOpts) error { - m, err := mm.NewSharedAnonMappable(opts.Length, platform.FromContext(ctx)) + m, err := mm.NewSharedAnonMappable(opts.Length, pgalloc.MemoryFileProviderFromContext(ctx)) if err != nil { return err } diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index d41fc17cc..01098675d 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -85,6 +85,7 @@ go_library( "//pkg/sentry/fs", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", "//pkg/sentry/safemem", "//pkg/sentry/socket/unix/transport", diff --git a/pkg/sentry/fs/fsutil/README.md b/pkg/sentry/fs/fsutil/README.md index 6e677890c..8be367334 100644 --- a/pkg/sentry/fs/fsutil/README.md +++ b/pkg/sentry/fs/fsutil/README.md @@ -112,11 +112,12 @@ finds the file that was mapped and its `CachingInodeOperations`. It then calls It may choose to allocate more memory (i.e. do "readahead") to minimize subsequent faults. -Memory that is allocated comes from a host tmpfs file (see `filemem.FileMem`). -The host tmpfs file memory is brought up to date with the contents of the mapped -file on its filesystem. The region of the host tmpfs file that reflects the -mapped file is then mapped into the host address space of the application so -that subsequent memory accesses do not repeatedly generate a `SIGSEGV`. +Memory that is allocated comes from a host tmpfs file (see +`pgalloc.MemoryFile`). The host tmpfs file memory is brought up to date with the +contents of the mapped file on its filesystem. The region of the host tmpfs file +that reflects the mapped file is then mapped into the host address space of the +application so that subsequent memory accesses do not repeatedly generate a +`SIGSEGV`. The range that was allocated, including any extra memory allocation to minimize faults, is marked dirty due to the write fault. This overcounts dirty memory if diff --git a/pkg/sentry/fs/fsutil/file_range_set.go b/pkg/sentry/fs/fsutil/file_range_set.go index dd7ab4b4a..32ebf64ff 100644 --- a/pkg/sentry/fs/fsutil/file_range_set.go +++ b/pkg/sentry/fs/fsutil/file_range_set.go @@ -21,6 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" @@ -77,7 +78,7 @@ func (seg FileRangeIterator) FileRangeOf(mr memmap.MappableRange) platform.FileR } // Fill attempts to ensure that all memmap.Mappable offsets in required are -// mapped to a platform.File offset, by allocating from mem with the given +// mapped to a platform.File offset, by allocating from mf with the given // memory usage kind and invoking readAt to store data into memory. (If readAt // returns a successful partial read, Fill will call it repeatedly until all // bytes have been read.) EOF is handled consistently with the requirements of @@ -90,7 +91,7 @@ func (seg FileRangeIterator) FileRangeOf(mr memmap.MappableRange) platform.FileR // // Preconditions: required.Length() > 0. optional.IsSupersetOf(required). // required and optional must be page-aligned. -func (frs *FileRangeSet) Fill(ctx context.Context, required, optional memmap.MappableRange, mem platform.Memory, kind usage.MemoryKind, readAt func(ctx context.Context, dsts safemem.BlockSeq, offset uint64) (uint64, error)) error { +func (frs *FileRangeSet) Fill(ctx context.Context, required, optional memmap.MappableRange, mf *pgalloc.MemoryFile, kind usage.MemoryKind, readAt func(ctx context.Context, dsts safemem.BlockSeq, offset uint64) (uint64, error)) error { gap := frs.LowerBoundGap(required.Start) for gap.Ok() && gap.Start() < required.End { if gap.Range().Length() == 0 { @@ -100,7 +101,7 @@ func (frs *FileRangeSet) Fill(ctx context.Context, required, optional memmap.Map gr := gap.Range().Intersect(optional) // Read data into the gap. - fr, err := platform.AllocateAndFill(mem, gr.Length(), kind, safemem.ReaderFunc(func(dsts safemem.BlockSeq) (uint64, error) { + fr, err := mf.AllocateAndFill(gr.Length(), kind, safemem.ReaderFunc(func(dsts safemem.BlockSeq) (uint64, error) { var done uint64 for !dsts.IsEmpty() { n, err := readAt(ctx, dsts, gr.Start+done) @@ -108,7 +109,7 @@ func (frs *FileRangeSet) Fill(ctx context.Context, required, optional memmap.Map dsts = dsts.DropFirst64(n) if err != nil { if err == io.EOF { - // platform.AllocateAndFill truncates down to a page + // MemoryFile.AllocateAndFill truncates down to a page // boundary, but FileRangeSet.Fill is supposed to // zero-fill to the end of the page in this case. donepgaddr, ok := usermem.Addr(done).RoundUp() @@ -143,20 +144,20 @@ func (frs *FileRangeSet) Fill(ctx context.Context, required, optional memmap.Map // corresponding platform.FileRanges. // // Preconditions: mr must be page-aligned. -func (frs *FileRangeSet) Drop(mr memmap.MappableRange, mem platform.Memory) { +func (frs *FileRangeSet) Drop(mr memmap.MappableRange, mf *pgalloc.MemoryFile) { seg := frs.LowerBoundSegment(mr.Start) for seg.Ok() && seg.Start() < mr.End { seg = frs.Isolate(seg, mr) - mem.DecRef(seg.FileRange()) + mf.DecRef(seg.FileRange()) seg = frs.Remove(seg).NextSegment() } } // DropAll removes all segments in mr, freeing the corresponding // platform.FileRanges. -func (frs *FileRangeSet) DropAll(mem platform.Memory) { +func (frs *FileRangeSet) DropAll(mf *pgalloc.MemoryFile) { for seg := frs.FirstSegment(); seg.Ok(); seg = seg.NextSegment() { - mem.DecRef(seg.FileRange()) + mf.DecRef(seg.FileRange()) } frs.RemoveAll() } @@ -164,7 +165,7 @@ func (frs *FileRangeSet) DropAll(mem platform.Memory) { // Truncate updates frs to reflect Mappable truncation to the given length: // bytes after the new EOF on the same page are zeroed, and pages after the new // EOF are freed. -func (frs *FileRangeSet) Truncate(end uint64, mem platform.Memory) { +func (frs *FileRangeSet) Truncate(end uint64, mf *pgalloc.MemoryFile) { pgendaddr, ok := usermem.Addr(end).RoundUp() if ok { pgend := uint64(pgendaddr) @@ -173,7 +174,7 @@ func (frs *FileRangeSet) Truncate(end uint64, mem platform.Memory) { frs.SplitAt(pgend) seg := frs.LowerBoundSegment(pgend) for seg.Ok() { - mem.DecRef(seg.FileRange()) + mf.DecRef(seg.FileRange()) seg = frs.Remove(seg).NextSegment() } @@ -189,7 +190,7 @@ func (frs *FileRangeSet) Truncate(end uint64, mem platform.Memory) { if seg.Ok() { fr := seg.FileRange() fr.Start += end - seg.Start() - ims, err := mem.MapInternal(fr, usermem.Write) + ims, err := mf.MapInternal(fr, usermem.Write) if err != nil { // There's no good recourse from here. This means // that we can't keep cached memory consistent with diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index ef11676b8..9bd923678 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -25,6 +25,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" @@ -62,8 +63,8 @@ type CachingInodeOperations struct { // backingFile is a handle to a cached file object. backingFile CachedFileObject - // platform is used to allocate memory that caches backingFile's contents. - platform platform.Platform + // mfp is used to allocate memory that caches backingFile's contents. + mfp pgalloc.MemoryFileProvider // forcePageCache indicates the sentry page cache should be used regardless // of whether the platform supports host mapped I/O or not. This must not be @@ -96,7 +97,7 @@ type CachingInodeOperations struct { dataMu sync.RWMutex `state:"nosave"` // cache maps offsets into the cached file to offsets into - // platform.Memory() that store the file's data. + // mfp.MemoryFile() that store the file's data. // // cache is protected by dataMu. cache FileRangeSet @@ -148,13 +149,13 @@ type CachedFileObject interface { // NewCachingInodeOperations returns a new CachingInodeOperations backed by // a CachedFileObject and its initial unstable attributes. func NewCachingInodeOperations(ctx context.Context, backingFile CachedFileObject, uattr fs.UnstableAttr, forcePageCache bool) *CachingInodeOperations { - p := platform.FromContext(ctx) - if p == nil { - panic(fmt.Sprintf("context.Context %T lacks non-nil value for key %T", ctx, platform.CtxPlatform)) + mfp := pgalloc.MemoryFileProviderFromContext(ctx) + if mfp == nil { + panic(fmt.Sprintf("context.Context %T lacks non-nil value for key %T", ctx, pgalloc.CtxMemoryFileProvider)) } return &CachingInodeOperations{ backingFile: backingFile, - platform: p, + mfp: mfp, forcePageCache: forcePageCache, attr: uattr, hostFileMapper: NewHostFileMapper(), @@ -311,7 +312,7 @@ func (c *CachingInodeOperations) Truncate(ctx context.Context, inode *fs.Inode, // written back. c.dataMu.Lock() defer c.dataMu.Unlock() - c.cache.Truncate(uint64(size), c.platform.Memory()) + c.cache.Truncate(uint64(size), c.mfp.MemoryFile()) c.dirty.KeepClean(memmap.MappableRange{uint64(size), oldpgend}) return nil @@ -323,7 +324,7 @@ func (c *CachingInodeOperations) WriteOut(ctx context.Context, inode *fs.Inode) // Write dirty pages back. c.dataMu.Lock() - err := SyncDirtyAll(ctx, &c.cache, &c.dirty, uint64(c.attr.Size), c.platform.Memory(), c.backingFile.WriteFromBlocksAt) + err := SyncDirtyAll(ctx, &c.cache, &c.dirty, uint64(c.attr.Size), c.mfp.MemoryFile(), c.backingFile.WriteFromBlocksAt) c.dataMu.Unlock() if err != nil { c.attrMu.Unlock() @@ -527,7 +528,7 @@ func (rw *inodeReadWriter) ReadToBlocks(dsts safemem.BlockSeq) (uint64, error) { return 0, nil } - mem := rw.c.platform.Memory() + mem := rw.c.mfp.MemoryFile() var done uint64 seg, gap := rw.c.cache.Find(uint64(rw.offset)) for rw.offset < end { @@ -613,7 +614,7 @@ func (rw *inodeReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error return 0, nil } - mem := rw.c.platform.Memory() + mf := rw.c.mfp.MemoryFile() var done uint64 seg, gap := rw.c.cache.Find(uint64(rw.offset)) for rw.offset < end { @@ -622,7 +623,7 @@ func (rw *inodeReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error case seg.Ok() && seg.Start() < mr.End: // Get internal mappings from the cache. segMR := seg.Range().Intersect(mr) - ims, err := mem.MapInternal(seg.FileRangeOf(segMR), usermem.Write) + ims, err := mf.MapInternal(seg.FileRangeOf(segMR), usermem.Write) if err != nil { rw.maybeGrowFile() rw.c.dataMu.Unlock() @@ -711,13 +712,13 @@ func (c *CachingInodeOperations) RemoveMapping(ctx context.Context, ms memmap.Ma // Writeback dirty mapped memory now that there are no longer any // mappings that reference it. This is our naive memory eviction // strategy. - mem := c.platform.Memory() + mf := c.mfp.MemoryFile() c.dataMu.Lock() for _, r := range unmapped { - if err := SyncDirty(ctx, r, &c.cache, &c.dirty, uint64(c.attr.Size), c.platform.Memory(), c.backingFile.WriteFromBlocksAt); err != nil { + if err := SyncDirty(ctx, r, &c.cache, &c.dirty, uint64(c.attr.Size), mf, c.backingFile.WriteFromBlocksAt); err != nil { log.Warningf("Failed to writeback cached data %v: %v", r, err) } - c.cache.Drop(r, mem) + c.cache.Drop(r, mf) c.dirty.KeepClean(r) } c.dataMu.Unlock() @@ -760,8 +761,8 @@ func (c *CachingInodeOperations) Translate(ctx context.Context, required, option optional.End = pgend } - mem := c.platform.Memory() - cerr := c.cache.Fill(ctx, required, maxFillRange(required, optional), mem, usage.PageCache, c.backingFile.ReadToBlocksAt) + mf := c.mfp.MemoryFile() + cerr := c.cache.Fill(ctx, required, maxFillRange(required, optional), mf, usage.PageCache, c.backingFile.ReadToBlocksAt) var ts []memmap.Translation var translatedEnd uint64 @@ -769,7 +770,7 @@ func (c *CachingInodeOperations) Translate(ctx context.Context, required, option segMR := seg.Range().Intersect(optional) ts = append(ts, memmap.Translation{ Source: segMR, - File: mem, + File: mf, Offset: seg.FileRangeOf(segMR).Start, }) if at.Write { @@ -820,16 +821,17 @@ func (c *CachingInodeOperations) InvalidateUnsavable(ctx context.Context) error // Sync the cache's contents so that if we have a host fd after restore, // the remote file's contents are coherent. + mf := c.mfp.MemoryFile() c.dataMu.Lock() defer c.dataMu.Unlock() - if err := SyncDirtyAll(ctx, &c.cache, &c.dirty, uint64(c.attr.Size), c.platform.Memory(), c.backingFile.WriteFromBlocksAt); err != nil { + if err := SyncDirtyAll(ctx, &c.cache, &c.dirty, uint64(c.attr.Size), mf, c.backingFile.WriteFromBlocksAt); err != nil { return err } // Discard the cache so that it's not stored in saved state. This is safe // because per InvalidateUnsavable invariants, no new translations can have // been returned after we invalidated all existing translations above. - c.cache.DropAll(c.platform.Memory()) + c.cache.DropAll(mf) c.dirty.RemoveAll() return nil diff --git a/pkg/sentry/fs/proc/meminfo.go b/pkg/sentry/fs/proc/meminfo.go index b31258eed..620e93ce3 100644 --- a/pkg/sentry/fs/proc/meminfo.go +++ b/pkg/sentry/fs/proc/meminfo.go @@ -44,10 +44,10 @@ func (d *meminfoData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) return nil, 0 } - mem := d.k.Platform.Memory() - mem.UpdateUsage() + mf := d.k.MemoryFile() + mf.UpdateUsage() snapshot, totalUsage := usage.MemoryAccounting.Copy() - totalSize := usage.TotalMemory(mem.TotalSize(), totalUsage) + totalSize := usage.TotalMemory(mf.TotalSize(), totalUsage) anon := snapshot.Anonymous + snapshot.Tmpfs file := snapshot.PageCache + snapshot.Mapped // We don't actually have active/inactive LRUs, so just make up numbers. diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 13d06684d..a98fbf0f1 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -52,7 +52,7 @@ type fileInodeOperations struct { fsutil.InodeSimpleExtendedAttributes - // kernel is used to allocate platform memory that stores the file's contents. + // kernel is used to allocate memory that stores the file's contents. kernel *kernel.Kernel // memUsage is the default memory usage that will be reported by this file. @@ -85,7 +85,7 @@ type fileInodeOperations struct { var _ fs.InodeOperations = (*fileInodeOperations)(nil) -// NewInMemoryFile returns a new file backed by p.Memory(). +// NewInMemoryFile returns a new file backed by Kernel.MemoryFile(). func NewInMemoryFile(ctx context.Context, usage usage.MemoryKind, uattr fs.UnstableAttr) fs.InodeOperations { return &fileInodeOperations{ attr: uattr, @@ -98,7 +98,7 @@ func NewInMemoryFile(ctx context.Context, usage usage.MemoryKind, uattr fs.Unsta func (f *fileInodeOperations) Release(context.Context) { f.dataMu.Lock() defer f.dataMu.Unlock() - f.data.DropAll(f.kernel.Platform.Memory()) + f.data.DropAll(f.kernel.MemoryFile()) } // Mappable implements fs.InodeOperations.Mappable. @@ -202,7 +202,7 @@ func (f *fileInodeOperations) Truncate(ctx context.Context, _ *fs.Inode, size in // and can remove them. f.dataMu.Lock() defer f.dataMu.Unlock() - f.data.Truncate(uint64(size), f.kernel.Platform.Memory()) + f.data.Truncate(uint64(size), f.kernel.MemoryFile()) return nil } @@ -312,7 +312,7 @@ func (rw *fileReadWriter) ReadToBlocks(dsts safemem.BlockSeq) (uint64, error) { return 0, nil } - mem := rw.f.kernel.Platform.Memory() + mf := rw.f.kernel.MemoryFile() var done uint64 seg, gap := rw.f.data.Find(uint64(rw.offset)) for rw.offset < end { @@ -320,7 +320,7 @@ func (rw *fileReadWriter) ReadToBlocks(dsts safemem.BlockSeq) (uint64, error) { switch { case seg.Ok(): // Get internal mappings. - ims, err := mem.MapInternal(seg.FileRangeOf(seg.Range().Intersect(mr)), usermem.Read) + ims, err := mf.MapInternal(seg.FileRangeOf(seg.Range().Intersect(mr)), usermem.Read) if err != nil { return done, err } @@ -378,7 +378,7 @@ func (rw *fileReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error) } }() - mem := rw.f.kernel.Platform.Memory() + mf := rw.f.kernel.MemoryFile() // Page-aligned mr for when we need to allocate memory. RoundUp can't // overflow since end is an int64. pgstartaddr := usermem.Addr(rw.offset).RoundDown() @@ -392,7 +392,7 @@ func (rw *fileReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error) switch { case seg.Ok(): // Get internal mappings. - ims, err := mem.MapInternal(seg.FileRangeOf(seg.Range().Intersect(mr)), usermem.Write) + ims, err := mf.MapInternal(seg.FileRangeOf(seg.Range().Intersect(mr)), usermem.Write) if err != nil { return done, err } @@ -412,7 +412,7 @@ func (rw *fileReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error) case gap.Ok(): // Allocate memory for the write. gapMR := gap.Range().Intersect(pgMR) - fr, err := mem.Allocate(gapMR.Length(), rw.f.memUsage) + fr, err := mf.Allocate(gapMR.Length(), rw.f.memUsage) if err != nil { return done, err } @@ -467,8 +467,8 @@ func (f *fileInodeOperations) Translate(ctx context.Context, required, optional optional.End = pgend } - mem := f.kernel.Platform.Memory() - cerr := f.data.Fill(ctx, required, optional, mem, f.memUsage, func(_ context.Context, dsts safemem.BlockSeq, _ uint64) (uint64, error) { + mf := f.kernel.MemoryFile() + cerr := f.data.Fill(ctx, required, optional, mf, f.memUsage, func(_ context.Context, dsts safemem.BlockSeq, _ uint64) (uint64, error) { // Newly-allocated pages are zeroed, so we don't need to do anything. return dsts.NumBytes(), nil }) @@ -479,7 +479,7 @@ func (f *fileInodeOperations) Translate(ctx context.Context, required, optional segMR := seg.Range().Intersect(optional) ts = append(ts, memmap.Translation{ Source: segMR, - File: mem, + File: mf, Offset: seg.FileRangeOf(segMR).Start, }) translatedEnd = segMR.End diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 4b1762ce4..1a9d12c0b 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -74,7 +74,7 @@ type Dir struct { // InodeOperation methods to it. ramfsDir *ramfs.Dir - // kernel is used to allocate platform memory as storage for tmpfs Files. + // kernel is used to allocate memory as storage for tmpfs Files. kernel *kernel.Kernel } diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index d9bbfb556..4d34bc733 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -173,6 +173,7 @@ go_library( "//pkg/sentry/loader", "//pkg/sentry/memmap", "//pkg/sentry/mm", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", "//pkg/sentry/safemem", "//pkg/sentry/socket/netlink/port", @@ -212,7 +213,7 @@ go_test( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/kernel/sched", "//pkg/sentry/limits", - "//pkg/sentry/platform", + "//pkg/sentry/pgalloc", "//pkg/sentry/time", "//pkg/sentry/usage", "//pkg/sentry/usermem", diff --git a/pkg/sentry/kernel/contexttest/BUILD b/pkg/sentry/kernel/contexttest/BUILD index 5769a3b28..bfb2a0b73 100644 --- a/pkg/sentry/kernel/contexttest/BUILD +++ b/pkg/sentry/kernel/contexttest/BUILD @@ -12,6 +12,7 @@ go_library( "//pkg/sentry/context", "//pkg/sentry/context/contexttest", "//pkg/sentry/kernel", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", ], ) diff --git a/pkg/sentry/kernel/contexttest/contexttest.go b/pkg/sentry/kernel/contexttest/contexttest.go index 9eb18e7e8..eb56a6a07 100644 --- a/pkg/sentry/kernel/contexttest/contexttest.go +++ b/pkg/sentry/kernel/contexttest/contexttest.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" ) @@ -33,6 +34,7 @@ func Context(tb testing.TB) context.Context { k := &kernel.Kernel{ Platform: platform.FromContext(ctx), } + k.SetMemoryFile(pgalloc.MemoryFileFromContext(ctx)) ctx.(*contexttest.TestContext).RegisterValue(kernel.CtxKernel, k) return ctx } diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index c6afae2e6..3533fd8f7 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -58,6 +58,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/loader" "gvisor.googlesource.com/gvisor/pkg/sentry/mm" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/socket/netlink/port" sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" @@ -89,12 +90,14 @@ type Kernel struct { // All of the following fields are immutable unless otherwise specified. - // Platform is the platform that is used to execute tasks in the - // created Kernel. It is embedded so that Kernel can directly serve as - // Platform in mm logic and also serve as platform.MemoryProvider in - // filemem S/R logic. + // Platform is the platform that is used to execute tasks in the created + // Kernel. See comment on pgalloc.MemoryFileProvider for why Platform is + // embedded anonymously (the same issue applies). platform.Platform `state:"nosave"` + // mf provides application memory. + mf *pgalloc.MemoryFile `state:"nosave"` + // See InitKernelArgs for the meaning of these fields. featureSet *cpuid.FeatureSet timekeeper *Timekeeper @@ -229,7 +232,8 @@ type InitKernelArgs struct { // Init initialize the Kernel with no tasks. // -// Callers must manually set Kernel.Platform before caling Init. +// Callers must manually set Kernel.Platform and call Kernel.SetMemoryFile +// before calling Init. func (k *Kernel) Init(args InitKernelArgs) error { if args.FeatureSet == nil { return fmt.Errorf("FeatureSet is nil") @@ -332,15 +336,9 @@ func (k *Kernel) SaveTo(w io.Writer) error { log.Infof("Kernel save stats: %s", &stats) log.Infof("Kernel save took [%s].", time.Since(kernelStart)) - // Save the memory state. - // - // FIXME: In the future, this should not be dispatched via - // an abstract memory type. This should be dispatched to a single - // memory implementation that belongs to the kernel. (There is - // currently a single implementation anyways, it just needs to be - // "unabstracted" and reparented appropriately.) + // Save the memory file's state. memoryStart := time.Now() - if err := k.Platform.Memory().SaveTo(w); err != nil { + if err := k.mf.SaveTo(w); err != nil { return err } log.Infof("Memory save took [%s].", time.Since(memoryStart)) @@ -418,13 +416,9 @@ func (ts *TaskSet) unregisterEpollWaiters() { } // LoadFrom returns a new Kernel loaded from args. -func (k *Kernel) LoadFrom(r io.Reader, p platform.Platform, net inet.Stack) error { +func (k *Kernel) LoadFrom(r io.Reader, net inet.Stack) error { loadStart := time.Now() - if p == nil { - return fmt.Errorf("Platform is nil") - } - k.Platform = p k.networkStack = net initAppCores := k.applicationCores @@ -438,11 +432,9 @@ func (k *Kernel) LoadFrom(r io.Reader, p platform.Platform, net inet.Stack) erro log.Infof("Kernel load stats: %s", &stats) log.Infof("Kernel load took [%s].", time.Since(kernelStart)) - // Load the memory state. - // - // See the note in SaveTo. + // Load the memory file's state. memoryStart := time.Now() - if err := k.Platform.Memory().LoadFrom(r); err != nil { + if err := k.mf.LoadFrom(r); err != nil { return err } log.Infof("Memory load took [%s].", time.Since(memoryStart)) @@ -597,6 +589,10 @@ func (ctx *createProcessContext) Value(key interface{}) interface{} { return ctx.k.RealtimeClock() case limits.CtxLimits: return ctx.args.Limits + case pgalloc.CtxMemoryFile: + return ctx.k.mf + case pgalloc.CtxMemoryFileProvider: + return ctx.k case platform.CtxPlatform: return ctx.k case uniqueid.CtxGlobalUniqueID: @@ -1018,6 +1014,17 @@ func (k *Kernel) NowMonotonic() int64 { return now } +// SetMemoryFile sets Kernel.mf. SetMemoryFile must be called before Init or +// LoadFrom. +func (k *Kernel) SetMemoryFile(mf *pgalloc.MemoryFile) { + k.mf = mf +} + +// MemoryFile implements pgalloc.MemoryFileProvider.MemoryFile. +func (k *Kernel) MemoryFile() *pgalloc.MemoryFile { + return k.mf +} + // SupervisorContext returns a Context with maximum privileges in k. It should // only be used by goroutines outside the control of the emulated kernel // defined by e. @@ -1083,7 +1090,7 @@ func (k *Kernel) ListSockets(family int) []*refs.WeakRef { socks := []*refs.WeakRef{} if table, ok := k.socketTable[family]; ok { socks = make([]*refs.WeakRef, 0, len(table)) - for s, _ := range table { + for s := range table { socks = append(socks, s) } } @@ -1123,6 +1130,10 @@ func (ctx supervisorContext) Value(key interface{}) interface{} { case limits.CtxLimits: // No limits apply. return limits.NewLimitSet() + case pgalloc.CtxMemoryFile: + return ctx.k.mf + case pgalloc.CtxMemoryFileProvider: + return ctx.k case platform.CtxPlatform: return ctx.k case uniqueid.CtxGlobalUniqueID: diff --git a/pkg/sentry/kernel/memevent/memory_events.go b/pkg/sentry/kernel/memevent/memory_events.go index b6283c5d1..d09d6debf 100644 --- a/pkg/sentry/kernel/memevent/memory_events.go +++ b/pkg/sentry/kernel/memevent/memory_events.go @@ -95,7 +95,7 @@ func (m *MemoryEvents) run() { } func (m *MemoryEvents) emit() { - totalPlatform, err := m.k.Platform.Memory().TotalUsage() + totalPlatform, err := m.k.MemoryFile().TotalUsage() if err != nil { log.Warningf("Failed to fetch memory usage for memory events: %v", err) return diff --git a/pkg/sentry/kernel/shm/BUILD b/pkg/sentry/kernel/shm/BUILD index f45770eef..bc2089872 100644 --- a/pkg/sentry/kernel/shm/BUILD +++ b/pkg/sentry/kernel/shm/BUILD @@ -20,6 +20,7 @@ go_library( "//pkg/sentry/kernel/auth", "//pkg/sentry/kernel/time", "//pkg/sentry/memmap", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", "//pkg/sentry/usage", "//pkg/sentry/usermem", diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 96414d060..4525aabf4 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -45,6 +45,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -199,19 +200,19 @@ func (r *Registry) FindOrCreate(ctx context.Context, pid int32, key Key, size ui // // Precondition: Caller must hold r.mu. func (r *Registry) newShm(ctx context.Context, pid int32, key Key, creator fs.FileOwner, perms fs.FilePermissions, size uint64) (*Shm, error) { - p := platform.FromContext(ctx) - if p == nil { - panic(fmt.Sprintf("context.Context %T lacks non-nil value for key %T", ctx, platform.CtxPlatform)) + mfp := pgalloc.MemoryFileProviderFromContext(ctx) + if mfp == nil { + panic(fmt.Sprintf("context.Context %T lacks non-nil value for key %T", ctx, pgalloc.CtxMemoryFileProvider)) } effectiveSize := uint64(usermem.Addr(size).MustRoundUp()) - fr, err := p.Memory().Allocate(effectiveSize, usage.Anonymous) + fr, err := mfp.MemoryFile().Allocate(effectiveSize, usage.Anonymous) if err != nil { return nil, err } shm := &Shm{ - p: p, + mfp: mfp, registry: r, creator: creator, size: size, @@ -312,7 +313,7 @@ type Shm struct { // destruction. refs.AtomicRefCount - p platform.Platform + mfp pgalloc.MemoryFileProvider // registry points to the shm registry containing this segment. Immutable. registry *Registry @@ -333,7 +334,7 @@ type Shm struct { // Invariant: effectiveSize must be a multiple of usermem.PageSize. effectiveSize uint64 - // fr is the offset into platform.Memory() that backs this contents of this + // fr is the offset into mfp.MemoryFile() that backs this contents of this // segment. Immutable. fr platform.FileRange @@ -452,7 +453,7 @@ func (s *Shm) Translate(ctx context.Context, required, optional memmap.MappableR return []memmap.Translation{ { Source: source, - File: s.p.Memory(), + File: s.mfp.MemoryFile(), Offset: s.fr.Start + source.Start, }, }, err @@ -599,7 +600,7 @@ func (s *Shm) Set(ctx context.Context, ds *linux.ShmidDS) error { } func (s *Shm) destroy() { - s.p.Memory().DecRef(s.fr) + s.mfp.MemoryFile().DecRef(s.fr) s.registry.remove(s) } diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 702e40cce..e9f133c0b 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -29,6 +29,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/sched" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/unimpl" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" @@ -587,6 +588,10 @@ func (t *Task) Value(key interface{}) interface{} { return t.k.RealtimeClock() case limits.CtxLimits: return t.tg.limits + case pgalloc.CtxMemoryFile: + return t.k.mf + case pgalloc.CtxMemoryFileProvider: + return t.k case platform.CtxPlatform: return t.k case uniqueid.CtxGlobalUniqueID: diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index ee3e49d17..d1c82f2aa 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -144,7 +144,7 @@ func (t *Task) Stack() *arch.Stack { // * fs: Binary FeatureSet func (k *Kernel) LoadTaskImage(ctx context.Context, mounts *fs.MountNamespace, root, wd *fs.Dirent, maxTraversals *uint, filename string, argv, envv []string, fs *cpuid.FeatureSet) (*TaskContext, *syserr.Error) { // Prepare a new user address space to load into. - m := mm.NewMemoryManager(k) + m := mm.NewMemoryManager(k, k) defer m.DecUsers(ctx) os, ac, name, err := loader.Load(ctx, m, mounts, root, wd, maxTraversals, fs, filename, argv, envv, k.extraAuxv, k.vdso) diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go index 6bff80f13..d7bd85e78 100644 --- a/pkg/sentry/kernel/timekeeper.go +++ b/pkg/sentry/kernel/timekeeper.go @@ -21,6 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/log" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" ) @@ -85,9 +86,9 @@ type Timekeeper struct { // NewTimekeeper does not take ownership of paramPage. // // SetClocks must be called on the returned Timekeeper before it is usable. -func NewTimekeeper(platform platform.Platform, paramPage platform.FileRange) (*Timekeeper, error) { +func NewTimekeeper(mfp pgalloc.MemoryFileProvider, paramPage platform.FileRange) (*Timekeeper, error) { return &Timekeeper{ - params: NewVDSOParamPage(platform, paramPage), + params: NewVDSOParamPage(mfp, paramPage), }, nil } diff --git a/pkg/sentry/kernel/timekeeper_test.go b/pkg/sentry/kernel/timekeeper_test.go index 71674c21c..6084bcb18 100644 --- a/pkg/sentry/kernel/timekeeper_test.go +++ b/pkg/sentry/kernel/timekeeper_test.go @@ -18,7 +18,7 @@ import ( "testing" "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" sentrytime "gvisor.googlesource.com/gvisor/pkg/sentry/time" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -53,13 +53,13 @@ func (c *mockClocks) GetTime(id sentrytime.ClockID) (int64, error) { // SetClocks called. func stateTestClocklessTimekeeper(tb testing.TB) *Timekeeper { ctx := contexttest.Context(tb) - p := platform.FromContext(ctx) - fr, err := p.Memory().Allocate(usermem.PageSize, usage.Anonymous) + mfp := pgalloc.MemoryFileProviderFromContext(ctx) + fr, err := mfp.MemoryFile().Allocate(usermem.PageSize, usage.Anonymous) if err != nil { tb.Fatalf("failed to allocate memory: %v", err) } return &Timekeeper{ - params: NewVDSOParamPage(p, fr), + params: NewVDSOParamPage(mfp, fr), } } diff --git a/pkg/sentry/kernel/vdso.go b/pkg/sentry/kernel/vdso.go index 0ec858a4a..3a35f1d00 100644 --- a/pkg/sentry/kernel/vdso.go +++ b/pkg/sentry/kernel/vdso.go @@ -18,6 +18,7 @@ import ( "fmt" "gvisor.googlesource.com/gvisor/pkg/binary" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -55,9 +56,9 @@ type vdsoParams struct { // // +stateify savable type VDSOParamPage struct { - // The parameter page is fr, allocated from platform.Memory(). - platform platform.Platform - fr platform.FileRange + // The parameter page is fr, allocated from mfp.MemoryFile(). + mfp pgalloc.MemoryFileProvider + fr platform.FileRange // seq is the current sequence count written to the page. // @@ -73,20 +74,20 @@ type VDSOParamPage struct { // // Preconditions: // -// * fr is a single page allocated from platform.Memory(). VDSOParamPage does +// * fr is a single page allocated from mfp.MemoryFile(). VDSOParamPage does // not take ownership of fr; it must remain allocated for the lifetime of the // VDSOParamPage. // // * VDSOParamPage must be the only writer to fr. // -// * platform.Memory().MapInternal(fr) must return a single safemem.Block. -func NewVDSOParamPage(platform platform.Platform, fr platform.FileRange) *VDSOParamPage { - return &VDSOParamPage{platform: platform, fr: fr} +// * mfp.MemoryFile().MapInternal(fr) must return a single safemem.Block. +func NewVDSOParamPage(mfp pgalloc.MemoryFileProvider, fr platform.FileRange) *VDSOParamPage { + return &VDSOParamPage{mfp: mfp, fr: fr} } // access returns a mapping of the param page. func (v *VDSOParamPage) access() (safemem.Block, error) { - bs, err := v.platform.Memory().MapInternal(v.fr, usermem.ReadWrite) + bs, err := v.mfp.MemoryFile().MapInternal(v.fr, usermem.ReadWrite) if err != nil { return safemem.Block{}, err } diff --git a/pkg/sentry/loader/BUILD b/pkg/sentry/loader/BUILD index 1ea260a4e..66300f25a 100644 --- a/pkg/sentry/loader/BUILD +++ b/pkg/sentry/loader/BUILD @@ -39,7 +39,7 @@ go_library( "//pkg/sentry/limits", "//pkg/sentry/memmap", "//pkg/sentry/mm", - "//pkg/sentry/platform", + "//pkg/sentry/pgalloc", "//pkg/sentry/safemem", "//pkg/sentry/uniqueid", "//pkg/sentry/usage", diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index c070c7316..273f6b5b9 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -28,7 +28,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/fs/fsutil" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/mm" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/uniqueid" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" @@ -217,7 +217,7 @@ type VDSO struct { // PrepareVDSO validates the system VDSO and returns a VDSO, containing the // param page for updating by the kernel. -func PrepareVDSO(p platform.Platform) (*VDSO, error) { +func PrepareVDSO(mfp pgalloc.MemoryFileProvider) (*VDSO, error) { vdsoFile := newByteReaderFile(vdsoBin) // First make sure the VDSO is valid. vdsoFile does not use ctx, so a @@ -234,35 +234,36 @@ func PrepareVDSO(p platform.Platform) (*VDSO, error) { return nil, fmt.Errorf("VDSO size overflows? %#x", len(vdsoBin)) } - vdso, err := p.Memory().Allocate(uint64(size), usage.System) + mf := mfp.MemoryFile() + vdso, err := mf.Allocate(uint64(size), usage.System) if err != nil { return nil, fmt.Errorf("unable to allocate VDSO memory: %v", err) } - ims, err := p.Memory().MapInternal(vdso, usermem.ReadWrite) + ims, err := mf.MapInternal(vdso, usermem.ReadWrite) if err != nil { - p.Memory().DecRef(vdso) + mf.DecRef(vdso) return nil, fmt.Errorf("unable to map VDSO memory: %v", err) } _, err = safemem.CopySeq(ims, safemem.BlockSeqOf(safemem.BlockFromSafeSlice(vdsoBin))) if err != nil { - p.Memory().DecRef(vdso) + mf.DecRef(vdso) return nil, fmt.Errorf("unable to copy VDSO into memory: %v", err) } // Finally, allocate a param page for this VDSO. - paramPage, err := p.Memory().Allocate(usermem.PageSize, usage.System) + paramPage, err := mf.Allocate(usermem.PageSize, usage.System) if err != nil { - p.Memory().DecRef(vdso) + mf.DecRef(vdso) return nil, fmt.Errorf("unable to allocate VDSO param page: %v", err) } return &VDSO{ - ParamPage: mm.NewSpecialMappable("[vvar]", p, paramPage), + ParamPage: mm.NewSpecialMappable("[vvar]", mfp, paramPage), // TODO: Don't advertise the VDSO, as some applications may // not be able to handle multiple [vdso] hints. - vdso: mm.NewSpecialMappable("", p, vdso), + vdso: mm.NewSpecialMappable("", mfp, vdso), phdrs: info.phdrs, }, nil } diff --git a/pkg/sentry/memutil/memutil_unsafe.go b/pkg/sentry/memutil/memutil_unsafe.go index 8d9fc64fb..bc2c72f55 100644 --- a/pkg/sentry/memutil/memutil_unsafe.go +++ b/pkg/sentry/memutil/memutil_unsafe.go @@ -15,6 +15,7 @@ package memutil import ( + "fmt" "syscall" "unsafe" @@ -22,14 +23,17 @@ import ( ) // CreateMemFD creates a memfd file and returns the fd. -func CreateMemFD(name string, flags int) (fd int, err error) { +func CreateMemFD(name string, flags int) (int, error) { p, err := syscall.BytePtrFromString(name) if err != nil { return -1, err } - r0, _, e0 := syscall.Syscall(unix.SYS_MEMFD_CREATE, uintptr(unsafe.Pointer(p)), uintptr(flags), 0) - if e0 != 0 { - return -1, e0 + fd, _, e := syscall.Syscall(unix.SYS_MEMFD_CREATE, uintptr(unsafe.Pointer(p)), uintptr(flags), 0) + if e != 0 { + if e == syscall.ENOSYS { + return -1, fmt.Errorf("memfd_create(2) is not implemented. Check that you have Linux 3.17 or higher") + } + return -1, e } - return int(r0), nil + return int(fd), nil } diff --git a/pkg/sentry/mm/BUILD b/pkg/sentry/mm/BUILD index a85ffdef8..c78cb4280 100644 --- a/pkg/sentry/mm/BUILD +++ b/pkg/sentry/mm/BUILD @@ -111,6 +111,7 @@ go_library( "//pkg/sentry/kernel/shm", "//pkg/sentry/limits", "//pkg/sentry/memmap", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", "//pkg/sentry/platform/safecopy", "//pkg/sentry/safemem", @@ -133,6 +134,7 @@ go_test( "//pkg/sentry/context/contexttest", "//pkg/sentry/limits", "//pkg/sentry/memmap", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", "//pkg/sentry/usermem", "//pkg/syserror", diff --git a/pkg/sentry/mm/README.md b/pkg/sentry/mm/README.md index e485a5ca5..e6efbf565 100644 --- a/pkg/sentry/mm/README.md +++ b/pkg/sentry/mm/README.md @@ -153,7 +153,7 @@ manner, and the sentry handles the fault: represented by a host file descriptor and offset, since (as noted in "Background") this is the memory mapping primitive provided by the host kernel. In general, memory is allocated from a temporary host file using the - `filemem` package. Supposing that the sentry allocates offset 0x3000 from + `pgalloc` package. Supposing that the sentry allocates offset 0x3000 from host file "memory-file", the resulting state is: Sentry VMA: VA:0x400000 -> /tmp/foo:0x0 @@ -274,7 +274,7 @@ In the sentry: methods [`platform.AddressSpace.MapFile` and `platform.AddressSpace.Unmap`][platform]. -[filemem]: https://gvisor.googlesource.com/gvisor/+/master/pkg/sentry/platform/filemem/filemem.go [memmap]: https://gvisor.googlesource.com/gvisor/+/master/pkg/sentry/memmap/memmap.go [mm]: https://gvisor.googlesource.com/gvisor/+/master/pkg/sentry/mm/mm.go +[pgalloc]: https://gvisor.googlesource.com/gvisor/+/master/pkg/sentry/pgalloc/pgalloc.go [platform]: https://gvisor.googlesource.com/gvisor/+/master/pkg/sentry/platform/platform.go diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index 5e86d3b49..6cec6387a 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -21,6 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -201,24 +202,24 @@ func (ctx *AIOContext) WaitChannel() (chan struct{}, bool) { type aioMappable struct { refs.AtomicRefCount - p platform.Platform - fr platform.FileRange + mfp pgalloc.MemoryFileProvider + fr platform.FileRange } var aioRingBufferSize = uint64(usermem.Addr(linux.AIORingSize).MustRoundUp()) -func newAIOMappable(p platform.Platform) (*aioMappable, error) { - fr, err := p.Memory().Allocate(aioRingBufferSize, usage.Anonymous) +func newAIOMappable(mfp pgalloc.MemoryFileProvider) (*aioMappable, error) { + fr, err := mfp.MemoryFile().Allocate(aioRingBufferSize, usage.Anonymous) if err != nil { return nil, err } - return &aioMappable{p: p, fr: fr}, nil + return &aioMappable{mfp: mfp, fr: fr}, nil } // DecRef implements refs.RefCounter.DecRef. func (m *aioMappable) DecRef() { m.AtomicRefCount.DecRefWithDestructor(func() { - m.p.Memory().DecRef(m.fr) + m.mfp.MemoryFile().DecRef(m.fr) }) } @@ -299,7 +300,7 @@ func (m *aioMappable) Translate(ctx context.Context, required, optional memmap.M return []memmap.Translation{ { Source: source, - File: m.p.Memory(), + File: m.mfp.MemoryFile(), Offset: m.fr.Start + source.Start, }, }, err @@ -320,7 +321,7 @@ func (mm *MemoryManager) NewAIOContext(ctx context.Context, events uint32) (uint // libaio peeks inside looking for a magic number. This function allocates // a page per context and keeps it set to zeroes to ensure it will not // match AIO_RING_MAGIC and make libaio happy. - m, err := newAIOMappable(mm.p) + m, err := newAIOMappable(mm.mfp) if err != nil { return 0, err } diff --git a/pkg/sentry/mm/lifecycle.go b/pkg/sentry/mm/lifecycle.go index 1ee8ae74e..a71286f14 100644 --- a/pkg/sentry/mm/lifecycle.go +++ b/pkg/sentry/mm/lifecycle.go @@ -23,14 +23,16 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) // NewMemoryManager returns a new MemoryManager with no mappings and 1 user. -func NewMemoryManager(p platform.Platform) *MemoryManager { +func NewMemoryManager(p platform.Platform, mfp pgalloc.MemoryFileProvider) *MemoryManager { return &MemoryManager{ p: p, + mfp: mfp, haveASIO: p.SupportsAddressSpaceIO(), privateRefs: &privateRefs{}, users: 1, @@ -60,6 +62,7 @@ func (mm *MemoryManager) Fork(ctx context.Context) (*MemoryManager, error) { defer mm.mappingMu.RUnlock() mm2 := &MemoryManager{ p: mm.p, + mfp: mm.mfp, haveASIO: mm.haveASIO, layout: mm.layout, privateRefs: mm.privateRefs, diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index e2c636f38..6ed838d64 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -40,6 +40,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -50,10 +51,9 @@ import ( // // +stateify savable type MemoryManager struct { - // p is the platform. - // - // p is immutable. - p platform.Platform + // p and mfp are immutable. + p platform.Platform + mfp pgalloc.MemoryFileProvider // haveASIO is the cached result of p.SupportsAddressSpaceIO(). Aside from // eliminating an indirect call in the hot I/O path, this makes @@ -369,8 +369,8 @@ func (v *vma) loadRealPerms(b int) { // +stateify savable type pma struct { // file is the file mapped by this pma. Only pmas for which file == - // platform.Platform.Memory() may be saved. pmas hold a reference to the - // corresponding file range while they exist. + // MemoryManager.mfp.MemoryFile() may be saved. pmas hold a reference to + // the corresponding file range while they exist. file platform.File `state:"nosave"` // off is the offset into file at which this pma begins. @@ -387,7 +387,7 @@ type pma struct { // private is true if this pma represents private memory. // - // If private is true, file must be platform.Platform.Memory(), the pma + // If private is true, file must be MemoryManager.mfp.MemoryFile(), the pma // holds a reference on the mapped memory that is tracked in privateRefs, // and calls to Invalidate for which // memmap.InvalidateOpts.InvalidatePrivate is false should ignore the pma. @@ -405,9 +405,9 @@ type pma struct { type privateRefs struct { mu sync.Mutex `state:"nosave"` - // refs maps offsets into Platform.Memory() to the number of pmas (or, - // equivalently, MemoryManagers) that share ownership of the memory at that - // offset. + // refs maps offsets into MemoryManager.mfp.MemoryFile() to the number of + // pmas (or, equivalently, MemoryManagers) that share ownership of the + // memory at that offset. refs fileRefcountSet } diff --git a/pkg/sentry/mm/mm_test.go b/pkg/sentry/mm/mm_test.go index f2db43196..e12cb3bd1 100644 --- a/pkg/sentry/mm/mm_test.go +++ b/pkg/sentry/mm/mm_test.go @@ -22,6 +22,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/context/contexttest" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" @@ -29,7 +30,8 @@ import ( func testMemoryManager(ctx context.Context) *MemoryManager { p := platform.FromContext(ctx) - mm := NewMemoryManager(p) + mfp := pgalloc.MemoryFileProviderFromContext(ctx) + mm := NewMemoryManager(p, mfp) mm.layout = arch.MmapLayout{ MinAddr: p.MinUserAddress(), MaxAddr: p.MaxUserAddress(), diff --git a/pkg/sentry/mm/pma.go b/pkg/sentry/mm/pma.go index d102035d8..bb779a45b 100644 --- a/pkg/sentry/mm/pma.go +++ b/pkg/sentry/mm/pma.go @@ -328,8 +328,8 @@ func (mm *MemoryManager) insertPMAsLocked(ctx context.Context, vseg vmaIterator, // Limit the range we allocate to ar, aligned to privateAllocUnit. maskAR := privateAligned(ar) allocAR := optAR.Intersect(maskAR) - mem := mm.p.Memory() - fr, err := mem.Allocate(uint64(allocAR.Length()), usage.Anonymous) + mf := mm.mfp.MemoryFile() + fr, err := mf.Allocate(uint64(allocAR.Length()), usage.Anonymous) if err != nil { return pgap, err } @@ -342,10 +342,10 @@ func (mm *MemoryManager) insertPMAsLocked(ctx context.Context, vseg vmaIterator, } mm.addRSSLocked(allocAR) - mem.IncRef(fr) + mf.IncRef(fr) return mm.pmas.Insert(pgap, allocAR, pma{ - file: mem, + file: mf, off: fr.Start, vmaEffectivePerms: vma.effectivePerms, vmaMaxPerms: vma.maxPerms, @@ -426,7 +426,7 @@ func (mm *MemoryManager) breakCopyOnWriteLocked(pseg pmaIterator, ar usermem.Add // Limit the range we copy to ar, aligned to privateAllocUnit. maskAR := privateAligned(ar) var invalidatedIterators, didUnmapAS bool - mem := mm.p.Memory() + mf := mm.mfp.MemoryFile() for { if mm.isPMACopyOnWriteLocked(pseg) { // Determine the range to copy. @@ -438,7 +438,7 @@ func (mm *MemoryManager) breakCopyOnWriteLocked(pseg pmaIterator, ar usermem.Add } // Copy contents. - fr, err := platform.AllocateAndFill(mem, uint64(copyAR.Length()), usage.Anonymous, &safemem.BlockSeqReader{mm.internalMappingsLocked(pseg, copyAR)}) + fr, err := mf.AllocateAndFill(uint64(copyAR.Length()), usage.Anonymous, &safemem.BlockSeqReader{mm.internalMappingsLocked(pseg, copyAR)}) if _, ok := err.(safecopy.BusError); ok { // If we got SIGBUS during the copy, deliver SIGBUS to // userspace (instead of SIGSEGV) if we're breaking @@ -449,7 +449,7 @@ func (mm *MemoryManager) breakCopyOnWriteLocked(pseg pmaIterator, ar usermem.Add return pseg.PrevGap(), invalidatedIterators, err } mm.incPrivateRef(fr) - mem.IncRef(fr) + mf.IncRef(fr) // Unmap all of maskAR, not just copyAR, to minimize host syscalls. // AddressSpace mappings must be removed before mm.decPrivateRef(). @@ -471,7 +471,7 @@ func (mm *MemoryManager) breakCopyOnWriteLocked(pseg pmaIterator, ar usermem.Add } pma.file.DecRef(pseg.fileRange()) - pma.file = mem + pma.file = mf pma.off = fr.Start pma.private = true pma.needCOW = false @@ -881,9 +881,9 @@ func (mm *MemoryManager) decPrivateRef(fr platform.FileRange) { refSet.MergeAdjacent(fr) mm.privateRefs.mu.Unlock() - mem := mm.p.Memory() + mf := mm.mfp.MemoryFile() for _, fr := range freed { - mem.DecRef(fr) + mf.DecRef(fr) } } diff --git a/pkg/sentry/mm/save_restore.go b/pkg/sentry/mm/save_restore.go index 6e7080a84..46e0e0754 100644 --- a/pkg/sentry/mm/save_restore.go +++ b/pkg/sentry/mm/save_restore.go @@ -37,12 +37,12 @@ func (mm *MemoryManager) InvalidateUnsavable(ctx context.Context) error { // beforeSave is invoked by stateify. func (mm *MemoryManager) beforeSave() { - mem := mm.p.Memory() + mf := mm.mfp.MemoryFile() for pseg := mm.pmas.FirstSegment(); pseg.Ok(); pseg = pseg.NextSegment() { - if pma := pseg.ValuePtr(); pma.file != mem { + if pma := pseg.ValuePtr(); pma.file != mf { // InvalidateUnsavable should have caused all such pmas to be // invalidated. - panic(fmt.Sprintf("Can't save pma %#v with non-Memory file of type %T:\n%s", pseg.Range(), pma.file, mm)) + panic(fmt.Sprintf("Can't save pma %#v with non-MemoryFile of type %T:\n%s", pseg.Range(), pma.file, mm)) } } } @@ -50,8 +50,8 @@ func (mm *MemoryManager) beforeSave() { // afterLoad is invoked by stateify. func (mm *MemoryManager) afterLoad() { mm.haveASIO = mm.p.SupportsAddressSpaceIO() - mem := mm.p.Memory() + mf := mm.mfp.MemoryFile() for pseg := mm.pmas.FirstSegment(); pseg.Ok(); pseg = pseg.NextSegment() { - pseg.ValuePtr().file = mem + pseg.ValuePtr().file = mf } } diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index 64d0dd3f6..aa94d7d6a 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -18,6 +18,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -33,24 +34,24 @@ import ( type SpecialMappable struct { refs.AtomicRefCount - p platform.Platform + mfp pgalloc.MemoryFileProvider fr platform.FileRange name string } // NewSpecialMappable returns a SpecialMappable that owns fr, which represents -// offsets in p.Memory() that contain the SpecialMappable's data. The +// offsets in mfp.MemoryFile() that contain the SpecialMappable's data. The // SpecialMappable will use the given name in /proc/[pid]/maps. // // Preconditions: fr.Length() != 0. -func NewSpecialMappable(name string, p platform.Platform, fr platform.FileRange) *SpecialMappable { - return &SpecialMappable{p: p, fr: fr, name: name} +func NewSpecialMappable(name string, mfp pgalloc.MemoryFileProvider, fr platform.FileRange) *SpecialMappable { + return &SpecialMappable{mfp: mfp, fr: fr, name: name} } // DecRef implements refs.RefCounter.DecRef. func (m *SpecialMappable) DecRef() { m.AtomicRefCount.DecRefWithDestructor(func() { - m.p.Memory().DecRef(m.fr) + m.mfp.MemoryFile().DecRef(m.fr) }) } @@ -99,7 +100,7 @@ func (m *SpecialMappable) Translate(ctx context.Context, required, optional memm return []memmap.Translation{ { Source: source, - File: m.p.Memory(), + File: m.mfp.MemoryFile(), Offset: m.fr.Start + source.Start, }, }, err @@ -109,19 +110,19 @@ func (m *SpecialMappable) Translate(ctx context.Context, required, optional memm // InvalidateUnsavable implements memmap.Mappable.InvalidateUnsavable. func (m *SpecialMappable) InvalidateUnsavable(ctx context.Context) error { - // Since data is stored in platform.Platform.Memory(), the contents of - // which are preserved across save/restore, we don't need to do anything. + // Since data is stored in pgalloc.MemoryFile, the contents of which are + // preserved across save/restore, we don't need to do anything. return nil } -// Platform returns the Platform whose Memory stores the SpecialMappable's -// contents. -func (m *SpecialMappable) Platform() platform.Platform { - return m.p +// MemoryFileProvider returns the MemoryFileProvider whose MemoryFile stores +// the SpecialMappable's contents. +func (m *SpecialMappable) MemoryFileProvider() pgalloc.MemoryFileProvider { + return m.mfp } -// FileRange returns the offsets into Platform().Memory() that store the -// SpecialMappable's contents. +// FileRange returns the offsets into MemoryFileProvider().MemoryFile() that +// store the SpecialMappable's contents. func (m *SpecialMappable) FileRange() platform.FileRange { return m.fr } @@ -137,7 +138,7 @@ func (m *SpecialMappable) Length() uint64 { // TODO: The use of SpecialMappable is a lazy code reuse hack. Linux // uses an ephemeral file created by mm/shmem.c:shmem_zero_setup(); we should // do the same to get non-zero device and inode IDs. -func NewSharedAnonMappable(length uint64, p platform.Platform) (*SpecialMappable, error) { +func NewSharedAnonMappable(length uint64, mfp pgalloc.MemoryFileProvider) (*SpecialMappable, error) { if length == 0 { return nil, syserror.EINVAL } @@ -145,10 +146,9 @@ func NewSharedAnonMappable(length uint64, p platform.Platform) (*SpecialMappable if !ok { return nil, syserror.EINVAL } - - fr, err := p.Memory().Allocate(uint64(alignedLen), usage.Anonymous) + fr, err := mfp.MemoryFile().Allocate(uint64(alignedLen), usage.Anonymous) if err != nil { return nil, err } - return NewSpecialMappable("/dev/zero (deleted)", p, fr), nil + return NewSpecialMappable("/dev/zero (deleted)", mfp, fr), nil } diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index fd6929e08..b56e0d3b9 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -24,7 +24,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/futex" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" "gvisor.googlesource.com/gvisor/pkg/syserror" ) @@ -99,7 +99,7 @@ func (mm *MemoryManager) MMap(ctx context.Context, opts memmap.MMapOpts) (userme if opts.MappingIdentity != nil { return 0, syserror.EINVAL } - m, err := NewSharedAnonMappable(opts.Length, platform.FromContext(ctx)) + m, err := NewSharedAnonMappable(opts.Length, pgalloc.MemoryFileProviderFromContext(ctx)) if err != nil { return 0, err } @@ -965,7 +965,7 @@ func (mm *MemoryManager) Decommit(addr usermem.Addr, length uint64) error { // ensures that Decommit immediately reduces host memory usage. var didUnmapAS bool pseg := mm.pmas.LowerBoundSegment(ar.Start) - mem := mm.p.Memory() + mf := mm.mfp.MemoryFile() for vseg := mm.vmas.LowerBoundSegment(ar.Start); vseg.Ok() && vseg.Start() < ar.End; vseg = vseg.NextSegment() { vma := vseg.ValuePtr() if vma.mlockMode != memmap.MLockNone { @@ -984,7 +984,7 @@ func (mm *MemoryManager) Decommit(addr usermem.Addr, length uint64) error { if pma.private && !mm.isPMACopyOnWriteLocked(pseg) { psegAR := pseg.Range().Intersect(ar) if vsegAR.IsSupersetOf(psegAR) && vma.mappable == nil { - if err := mem.Decommit(pseg.fileRangeOf(psegAR)); err == nil { + if err := mf.Decommit(pseg.fileRangeOf(psegAR)); err == nil { pseg = pseg.NextSegment() continue } diff --git a/pkg/sentry/pgalloc/BUILD b/pkg/sentry/pgalloc/BUILD new file mode 100644 index 000000000..7efa55c20 --- /dev/null +++ b/pkg/sentry/pgalloc/BUILD @@ -0,0 +1,57 @@ +package(licenses = ["notice"]) + +load("//tools/go_generics:defs.bzl", "go_template_instance") +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") + +go_template_instance( + name = "usage_set", + out = "usage_set.go", + consts = { + "minDegree": "10", + }, + imports = { + "platform": "gvisor.googlesource.com/gvisor/pkg/sentry/platform", + }, + package = "pgalloc", + prefix = "usage", + template = "//pkg/segment:generic_set", + types = { + "Key": "uint64", + "Range": "platform.FileRange", + "Value": "usageInfo", + "Functions": "usageSetFunctions", + }, +) + +go_library( + name = "pgalloc", + srcs = [ + "context.go", + "pgalloc.go", + "pgalloc_unsafe.go", + "save_restore.go", + "usage_set.go", + ], + importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc", + visibility = ["//pkg/sentry:internal"], + deps = [ + "//pkg/log", + "//pkg/sentry/arch", + "//pkg/sentry/context", + "//pkg/sentry/memutil", + "//pkg/sentry/platform", + "//pkg/sentry/safemem", + "//pkg/sentry/usage", + "//pkg/sentry/usermem", + "//pkg/state", + "//pkg/syserror", + ], +) + +go_test( + name = "pgalloc_test", + size = "small", + srcs = ["pgalloc_test.go"], + embed = [":pgalloc"], + deps = ["//pkg/sentry/usermem"], +) diff --git a/pkg/sentry/pgalloc/context.go b/pkg/sentry/pgalloc/context.go new file mode 100644 index 000000000..adc97e78f --- /dev/null +++ b/pkg/sentry/pgalloc/context.go @@ -0,0 +1,48 @@ +// Copyright 2019 Google Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pgalloc + +import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/context" +) + +// contextID is this package's type for context.Context.Value keys. +type contextID int + +const ( + // CtxMemoryFile is a Context.Value key for a MemoryFile. + CtxMemoryFile contextID = iota + + // CtxMemoryFileProvider is a Context.Value key for a MemoryFileProvider. + CtxMemoryFileProvider +) + +// MemoryFileFromContext returns the MemoryFile used by ctx, or nil if no such +// MemoryFile exists. +func MemoryFileFromContext(ctx context.Context) *MemoryFile { + if v := ctx.Value(CtxMemoryFile); v != nil { + return v.(*MemoryFile) + } + return nil +} + +// MemoryFileProviderFromContext returns the MemoryFileProvider used by ctx, or nil if no such +// MemoryFileProvider exists. +func MemoryFileProviderFromContext(ctx context.Context) MemoryFileProvider { + if v := ctx.Value(CtxMemoryFileProvider); v != nil { + return v.(MemoryFileProvider) + } + return nil +} diff --git a/pkg/sentry/pgalloc/pgalloc.go b/pkg/sentry/pgalloc/pgalloc.go new file mode 100644 index 000000000..0754e608f --- /dev/null +++ b/pkg/sentry/pgalloc/pgalloc.go @@ -0,0 +1,922 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// Package pgalloc contains the page allocator subsystem, which manages memory +// that may be mapped into application address spaces. +// +// Lock order: +// +// pgalloc.MemoryFile.mu +// pgalloc.MemoryFile.mappingsMu +package pgalloc + +import ( + "fmt" + "math" + "os" + "sync" + "sync/atomic" + "syscall" + "time" + + "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/platform" + "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" + "gvisor.googlesource.com/gvisor/pkg/sentry/usage" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// MemoryFile is a platform.File whose pages may be allocated to arbitrary +// users. +type MemoryFile struct { + // MemoryFile owns a single backing file, which is modeled as follows: + // + // Each page in the file can be committed or uncommitted. A page is + // committed if the host kernel is spending resources to store its contents + // and uncommitted otherwise. This definition includes pages that the host + // kernel has swapped; this is intentional, to ensure that accounting does + // not change even if host kernel swapping behavior changes, and that + // memory used by pseudo-swap mechanisms like zswap is still accounted. + // + // The initial contents of uncommitted pages are implicitly zero bytes. A + // read or write to the contents of an uncommitted page causes it to be + // committed. This is the only event that can cause a uncommitted page to + // be committed. + // + // fallocate(FALLOC_FL_PUNCH_HOLE) (MemoryFile.Decommit) causes committed + // pages to be uncommitted. This is the only event that can cause a + // committed page to be uncommitted. + // + // Memory accounting is based on identifying the set of committed pages. + // Since we do not have direct access to the MMU, tracking reads and writes + // to uncommitted pages to detect commitment would introduce additional + // page faults, which would be prohibitively expensive. Instead, we query + // the host kernel to determine which pages are committed. + + // file is the backing file. The file pointer is immutable. + file *os.File + + mu sync.Mutex + + // usage maps each page in the file to metadata for that page. Pages for + // which no segment exists in usage are both unallocated (not in use) and + // uncommitted. + // + // Since usage stores usageInfo objects by value, clients should usually + // use usageIterator.ValuePtr() instead of usageIterator.Value() to get a + // pointer to the usageInfo rather than a copy. + // + // usage must be kept maximally merged (that is, there should never be two + // adjacent segments with the same values). At least markReclaimed depends + // on this property. + // + // usage is protected by mu. + usage usageSet + + // The UpdateUsage function scans all segments with knownCommitted set + // to false, sees which pages are committed and creates corresponding + // segments with knownCommitted set to true. + // + // In order to avoid unnecessary scans, usageExpected tracks the total + // file blocks expected. This is used to elide the scan when this + // matches the underlying file blocks. + // + // To track swapped pages, usageSwapped tracks the discrepency between + // what is observed in core and what is reported by the file. When + // usageSwapped is non-zero, a sweep will be performed at least every + // second. The start of the last sweep is recorded in usageLast. + // + // All usage attributes are all protected by mu. + usageExpected uint64 + usageSwapped uint64 + usageLast time.Time + + // minUnallocatedPage is the minimum page that may be unallocated. + // i.e., there are no unallocated pages below minUnallocatedPage. + // + // minUnallocatedPage is protected by mu. + minUnallocatedPage uint64 + + // fileSize is the size of the backing memory file in bytes. fileSize is + // always a power-of-two multiple of chunkSize. + // + // fileSize is protected by mu. + fileSize int64 + + // destroyed is set by Destroy to instruct the reclaimer goroutine to + // release resources and exit. destroyed is protected by mu. + destroyed bool + + // reclaimable is true if usage may contain reclaimable pages. reclaimable + // is protected by mu. + reclaimable bool + + // minReclaimablePage is the minimum page that may be reclaimable. + // i.e., all reclaimable pages are >= minReclaimablePage. + // + // minReclaimablePage is protected by mu. + minReclaimablePage uint64 + + // reclaimCond is signaled (with mu locked) when reclaimable or destroyed + // transitions from false to true. + reclaimCond sync.Cond + + // Pages from the backing file are mapped into the local address space on + // the granularity of large pieces called chunks. mappings is a []uintptr + // that stores, for each chunk, the start address of a mapping of that + // chunk in the current process' address space, or 0 if no such mapping + // exists. Once a chunk is mapped, it is never remapped or unmapped until + // the MemoryFile is destroyed. + // + // Mutating the mappings slice or its contents requires both holding + // mappingsMu and using atomic memory operations. (The slice is mutated + // whenever the file is expanded. Per the above, the only permitted + // mutation of the slice's contents is the assignment of a mapping to a + // chunk that was previously unmapped.) Reading the slice or its contents + // only requires *either* holding mappingsMu or using atomic memory + // operations. This allows MemoryFile.MapInternal to avoid locking in the + // common case where chunk mappings already exist. + mappingsMu sync.Mutex + mappings atomic.Value +} + +// usage tracks usage information. +// +// +stateify savable +type usageInfo struct { + // kind is the usage kind. + kind usage.MemoryKind + + // knownCommitted is true if the tracked region is definitely committed. + // (If it is false, the tracked region may or may not be committed.) + knownCommitted bool + + refs uint64 +} + +const ( + chunkShift = 24 + chunkSize = 1 << chunkShift // 16 MB + chunkMask = chunkSize - 1 + + initialSize = chunkSize + + // maxPage is the highest 64-bit page. + maxPage = math.MaxUint64 &^ (usermem.PageSize - 1) +) + +// NewMemoryFile creates a MemoryFile backed by the given file. If +// NewMemoryFile succeeds, ownership of file is transferred to the returned +// MemoryFile. +func NewMemoryFile(file *os.File) (*MemoryFile, error) { + // Truncate the file to 0 bytes first to ensure that it's empty. + if err := file.Truncate(0); err != nil { + return nil, err + } + if err := file.Truncate(initialSize); err != nil { + return nil, err + } + f := &MemoryFile{ + fileSize: initialSize, + file: file, + // No pages are reclaimable. DecRef will always be able to + // decrease minReclaimablePage from this point. + minReclaimablePage: maxPage, + } + f.reclaimCond.L = &f.mu + f.mappings.Store(make([]uintptr, initialSize/chunkSize)) + go f.runReclaim() // S/R-SAFE: f.mu + + // The Linux kernel contains an optional feature called "Integrity + // Measurement Architecture" (IMA). If IMA is enabled, it will checksum + // binaries the first time they are mapped PROT_EXEC. This is bad news for + // executable pages mapped from our backing file, which can grow to + // terabytes in (sparse) size. If IMA attempts to checksum a file that + // large, it will allocate all of the sparse pages and quickly exhaust all + // memory. + // + // Work around IMA by immediately creating a temporary PROT_EXEC mapping, + // while the backing file is still small. IMA will ignore any future + // mappings. + m, _, errno := syscall.Syscall6( + syscall.SYS_MMAP, + 0, + usermem.PageSize, + syscall.PROT_EXEC, + syscall.MAP_SHARED, + file.Fd(), + 0) + if errno != 0 { + // This isn't fatal (IMA may not even be in use). Log the error, but + // don't return it. + log.Warningf("Failed to pre-map MemoryFile PROT_EXEC: %v", errno) + } else { + if _, _, errno := syscall.Syscall( + syscall.SYS_MUNMAP, + m, + usermem.PageSize, + 0); errno != 0 { + panic(fmt.Sprintf("failed to unmap PROT_EXEC MemoryFile mapping: %v", errno)) + } + } + + return f, nil +} + +// Destroy releases all resources used by f. +// +// Preconditions: All pages allocated by f have been freed. +// +// Postconditions: None of f's methods may be called after Destroy. +func (f *MemoryFile) Destroy() { + f.mu.Lock() + defer f.mu.Unlock() + f.destroyed = true + f.reclaimCond.Signal() +} + +// Allocate returns a range of initially-zeroed pages of the given length with +// the given accounting kind and a single reference held by the caller. When +// the last reference on an allocated page is released, ownership of the page +// is returned to the MemoryFile, allowing it to be returned by a future call +// to Allocate. +// +// Preconditions: length must be page-aligned and non-zero. +func (f *MemoryFile) Allocate(length uint64, kind usage.MemoryKind) (platform.FileRange, error) { + if length == 0 || length%usermem.PageSize != 0 { + panic(fmt.Sprintf("invalid allocation length: %#x", length)) + } + + f.mu.Lock() + defer f.mu.Unlock() + + // Align hugepage-and-larger allocations on hugepage boundaries to try + // to take advantage of hugetmpfs. + alignment := uint64(usermem.PageSize) + if length >= usermem.HugePageSize { + alignment = usermem.HugePageSize + } + + start, minUnallocatedPage := findUnallocatedRange(&f.usage, f.minUnallocatedPage, length, alignment) + end := start + length + // File offsets are int64s. Since length must be strictly positive, end + // cannot legitimately be 0. + if end < start || int64(end) <= 0 { + return platform.FileRange{}, syserror.ENOMEM + } + + // Expand the file if needed. Double the file size on each expansion; + // uncommitted pages have effectively no cost. + fileSize := f.fileSize + for int64(end) > fileSize { + if fileSize >= 2*fileSize { + // fileSize overflow. + return platform.FileRange{}, syserror.ENOMEM + } + fileSize *= 2 + } + if fileSize > f.fileSize { + if err := f.file.Truncate(fileSize); err != nil { + return platform.FileRange{}, err + } + f.fileSize = fileSize + f.mappingsMu.Lock() + oldMappings := f.mappings.Load().([]uintptr) + newMappings := make([]uintptr, fileSize>>chunkShift) + copy(newMappings, oldMappings) + f.mappings.Store(newMappings) + f.mappingsMu.Unlock() + } + + // Mark selected pages as in use. + fr := platform.FileRange{start, end} + if !f.usage.Add(fr, usageInfo{ + kind: kind, + refs: 1, + }) { + panic(fmt.Sprintf("allocating %v: failed to insert into usage set:\n%v", fr, &f.usage)) + } + + if minUnallocatedPage < start { + f.minUnallocatedPage = minUnallocatedPage + } else { + // start was the first unallocated page. The next must be + // somewhere beyond end. + f.minUnallocatedPage = end + } + + return fr, nil +} + +// findUnallocatedRange returns the first unallocated page in usage of the +// specified length and alignment beginning at page start and the first single +// unallocated page. +func findUnallocatedRange(usage *usageSet, start, length, alignment uint64) (uint64, uint64) { + // Only searched until the first page is found. + firstPage := start + foundFirstPage := false + alignMask := alignment - 1 + for seg := usage.LowerBoundSegment(start); seg.Ok(); seg = seg.NextSegment() { + r := seg.Range() + + if !foundFirstPage && r.Start > firstPage { + foundFirstPage = true + } + + if start >= r.End { + // start was rounded up to an alignment boundary from the end + // of a previous segment and is now beyond r.End. + continue + } + // This segment represents allocated or reclaimable pages; only the + // range from start to the segment's beginning is allocatable, and the + // next allocatable range begins after the segment. + if r.Start > start && r.Start-start >= length { + break + } + start = (r.End + alignMask) &^ alignMask + if !foundFirstPage { + firstPage = r.End + } + } + return start, firstPage +} + +// AllocateAndFill allocates memory of the given kind and fills it by calling +// r.ReadToBlocks() repeatedly until either length bytes are read or a non-nil +// error is returned. It returns the memory filled by r, truncated down to the +// nearest page. If this is shorter than length bytes due to an error returned +// by r.ReadToBlocks(), it returns that error. +// +// Preconditions: length > 0. length must be page-aligned. +func (f *MemoryFile) AllocateAndFill(length uint64, kind usage.MemoryKind, r safemem.Reader) (platform.FileRange, error) { + fr, err := f.Allocate(length, kind) + if err != nil { + return platform.FileRange{}, err + } + dsts, err := f.MapInternal(fr, usermem.Write) + if err != nil { + f.DecRef(fr) + return platform.FileRange{}, err + } + n, err := safemem.ReadFullToBlocks(r, dsts) + un := uint64(usermem.Addr(n).RoundDown()) + if un < length { + // Free unused memory and update fr to contain only the memory that is + // still allocated. + f.DecRef(platform.FileRange{fr.Start + un, fr.End}) + fr.End = fr.Start + un + } + return fr, err +} + +// fallocate(2) modes, defined in Linux's include/uapi/linux/falloc.h. +const ( + _FALLOC_FL_KEEP_SIZE = 1 + _FALLOC_FL_PUNCH_HOLE = 2 +) + +// Decommit releases resources associated with maintaining the contents of the +// given pages. If Decommit succeeds, future accesses of the decommitted pages +// will read zeroes. +// +// Preconditions: fr.Length() > 0. +func (f *MemoryFile) Decommit(fr platform.FileRange) error { + if !fr.WellFormed() || fr.Length() == 0 || fr.Start%usermem.PageSize != 0 || fr.End%usermem.PageSize != 0 { + panic(fmt.Sprintf("invalid range: %v", fr)) + } + + // "After a successful call, subsequent reads from this range will + // return zeroes. The FALLOC_FL_PUNCH_HOLE flag must be ORed with + // FALLOC_FL_KEEP_SIZE in mode ..." - fallocate(2) + err := syscall.Fallocate( + int(f.file.Fd()), + _FALLOC_FL_PUNCH_HOLE|_FALLOC_FL_KEEP_SIZE, + int64(fr.Start), + int64(fr.Length())) + if err != nil { + return err + } + f.markDecommitted(fr) + return nil +} + +func (f *MemoryFile) markDecommitted(fr platform.FileRange) { + f.mu.Lock() + defer f.mu.Unlock() + // Since we're changing the knownCommitted attribute, we need to merge + // across the entire range to ensure that the usage tree is minimal. + gap := f.usage.ApplyContiguous(fr, func(seg usageIterator) { + val := seg.ValuePtr() + if val.knownCommitted { + // Drop the usageExpected appropriately. + amount := seg.Range().Length() + usage.MemoryAccounting.Dec(amount, val.kind) + f.usageExpected -= amount + val.knownCommitted = false + } + }) + if gap.Ok() { + panic(fmt.Sprintf("Decommit(%v): attempted to decommit unallocated pages %v:\n%v", fr, gap.Range(), &f.usage)) + } + f.usage.MergeRange(fr) +} + +// runReclaim implements the reclaimer goroutine, which continuously decommits +// reclaimable pages in order to reduce memory usage and make them available +// for allocation. +func (f *MemoryFile) runReclaim() { + for { + fr, ok := f.findReclaimable() + if !ok { + break + } + + if err := f.Decommit(fr); err != nil { + log.Warningf("Reclaim failed to decommit %v: %v", fr, err) + // Zero the pages manually. This won't reduce memory usage, but at + // least ensures that the pages will be zero when reallocated. + f.forEachMappingSlice(fr, func(bs []byte) { + for i := range bs { + bs[i] = 0 + } + }) + // Pretend the pages were decommitted even though they weren't, + // since the memory accounting implementation has no idea how to + // deal with this. + f.markDecommitted(fr) + } + f.markReclaimed(fr) + } + // We only get here if findReclaimable finds f.destroyed set and returns + // false. + f.mu.Lock() + defer f.mu.Unlock() + if !f.destroyed { + panic("findReclaimable broke out of reclaim loop, but destroyed is no longer set") + } + f.file.Close() + // Ensure that any attempts to use f.file.Fd() fail instead of getting a fd + // that has possibly been reassigned. + f.file = nil + mappings := f.mappings.Load().([]uintptr) + for i, m := range mappings { + if m != 0 { + _, _, errno := syscall.Syscall(syscall.SYS_MUNMAP, m, chunkSize, 0) + if errno != 0 { + log.Warningf("Failed to unmap mapping %#x for MemoryFile chunk %d: %v", m, i, errno) + } + } + } + // Similarly, invalidate f.mappings. (atomic.Value.Store(nil) panics.) + f.mappings.Store([]uintptr{}) +} + +func (f *MemoryFile) findReclaimable() (platform.FileRange, bool) { + f.mu.Lock() + defer f.mu.Unlock() + for { + for { + if f.destroyed { + return platform.FileRange{}, false + } + if f.reclaimable { + break + } + f.reclaimCond.Wait() + } + // Allocate returns the first usable range in offset order and is + // currently a linear scan, so reclaiming from the beginning of the + // file minimizes the expected latency of Allocate. + for seg := f.usage.LowerBoundSegment(f.minReclaimablePage); seg.Ok(); seg = seg.NextSegment() { + if seg.ValuePtr().refs == 0 { + f.minReclaimablePage = seg.End() + return seg.Range(), true + } + } + // No pages are reclaimable. + f.reclaimable = false + f.minReclaimablePage = maxPage + } +} + +func (f *MemoryFile) markReclaimed(fr platform.FileRange) { + f.mu.Lock() + defer f.mu.Unlock() + seg := f.usage.FindSegment(fr.Start) + // All of fr should be mapped to a single uncommitted reclaimable segment + // accounted to System. + if !seg.Ok() { + panic(fmt.Sprintf("reclaimed pages %v include unreferenced pages:\n%v", fr, &f.usage)) + } + if !seg.Range().IsSupersetOf(fr) { + panic(fmt.Sprintf("reclaimed pages %v are not entirely contained in segment %v with state %v:\n%v", fr, seg.Range(), seg.Value(), &f.usage)) + } + if got, want := seg.Value(), (usageInfo{ + kind: usage.System, + knownCommitted: false, + refs: 0, + }); got != want { + panic(fmt.Sprintf("reclaimed pages %v in segment %v has incorrect state %v, wanted %v:\n%v", fr, seg.Range(), got, want, &f.usage)) + } + // Deallocate reclaimed pages. Even though all of seg is reclaimable, the + // caller of markReclaimed may not have decommitted it, so we can only mark + // fr as reclaimed. + f.usage.Remove(f.usage.Isolate(seg, fr)) + if fr.Start < f.minUnallocatedPage { + // We've deallocated at least one lower page. + f.minUnallocatedPage = fr.Start + } +} + +// IncRef implements platform.File.IncRef. +func (f *MemoryFile) IncRef(fr platform.FileRange) { + if !fr.WellFormed() || fr.Length() == 0 || fr.Start%usermem.PageSize != 0 || fr.End%usermem.PageSize != 0 { + panic(fmt.Sprintf("invalid range: %v", fr)) + } + + f.mu.Lock() + defer f.mu.Unlock() + + gap := f.usage.ApplyContiguous(fr, func(seg usageIterator) { + seg.ValuePtr().refs++ + }) + if gap.Ok() { + panic(fmt.Sprintf("IncRef(%v): attempted to IncRef on unallocated pages %v:\n%v", fr, gap.Range(), &f.usage)) + } + + f.usage.MergeAdjacent(fr) +} + +// DecRef implements platform.File.DecRef. +func (f *MemoryFile) DecRef(fr platform.FileRange) { + if !fr.WellFormed() || fr.Length() == 0 || fr.Start%usermem.PageSize != 0 || fr.End%usermem.PageSize != 0 { + panic(fmt.Sprintf("invalid range: %v", fr)) + } + + var freed bool + + f.mu.Lock() + defer f.mu.Unlock() + + for seg := f.usage.FindSegment(fr.Start); seg.Ok() && seg.Start() < fr.End; seg = seg.NextSegment() { + seg = f.usage.Isolate(seg, fr) + val := seg.ValuePtr() + if val.refs == 0 { + panic(fmt.Sprintf("DecRef(%v): 0 existing references on %v:\n%v", fr, seg.Range(), &f.usage)) + } + val.refs-- + if val.refs == 0 { + freed = true + // Reclassify memory as System, until it's freed by the reclaim + // goroutine. + if val.knownCommitted { + usage.MemoryAccounting.Move(seg.Range().Length(), usage.System, val.kind) + } + val.kind = usage.System + } + } + f.usage.MergeAdjacent(fr) + + if freed { + if fr.Start < f.minReclaimablePage { + // We've freed at least one lower page. + f.minReclaimablePage = fr.Start + } + f.reclaimable = true + f.reclaimCond.Signal() + } +} + +// MapInternal implements platform.File.MapInternal. +func (f *MemoryFile) MapInternal(fr platform.FileRange, at usermem.AccessType) (safemem.BlockSeq, error) { + if !fr.WellFormed() || fr.Length() == 0 { + panic(fmt.Sprintf("invalid range: %v", fr)) + } + if at.Execute { + return safemem.BlockSeq{}, syserror.EACCES + } + + chunks := ((fr.End + chunkMask) >> chunkShift) - (fr.Start >> chunkShift) + if chunks == 1 { + // Avoid an unnecessary slice allocation. + var seq safemem.BlockSeq + err := f.forEachMappingSlice(fr, func(bs []byte) { + seq = safemem.BlockSeqOf(safemem.BlockFromSafeSlice(bs)) + }) + return seq, err + } + blocks := make([]safemem.Block, 0, chunks) + err := f.forEachMappingSlice(fr, func(bs []byte) { + blocks = append(blocks, safemem.BlockFromSafeSlice(bs)) + }) + return safemem.BlockSeqFromSlice(blocks), err +} + +// forEachMappingSlice invokes fn on a sequence of byte slices that +// collectively map all bytes in fr. +func (f *MemoryFile) forEachMappingSlice(fr platform.FileRange, fn func([]byte)) error { + mappings := f.mappings.Load().([]uintptr) + for chunkStart := fr.Start &^ chunkMask; chunkStart < fr.End; chunkStart += chunkSize { + chunk := int(chunkStart >> chunkShift) + m := atomic.LoadUintptr(&mappings[chunk]) + if m == 0 { + var err error + mappings, m, err = f.getChunkMapping(chunk) + if err != nil { + return err + } + } + startOff := uint64(0) + if chunkStart < fr.Start { + startOff = fr.Start - chunkStart + } + endOff := uint64(chunkSize) + if chunkStart+chunkSize > fr.End { + endOff = fr.End - chunkStart + } + fn(unsafeSlice(m, chunkSize)[startOff:endOff]) + } + return nil +} + +func (f *MemoryFile) getChunkMapping(chunk int) ([]uintptr, uintptr, error) { + f.mappingsMu.Lock() + defer f.mappingsMu.Unlock() + // Another thread may have replaced f.mappings altogether due to file + // expansion. + mappings := f.mappings.Load().([]uintptr) + // Another thread may have already mapped the chunk. + if m := mappings[chunk]; m != 0 { + return mappings, m, nil + } + m, _, errno := syscall.Syscall6( + syscall.SYS_MMAP, + 0, + chunkSize, + syscall.PROT_READ|syscall.PROT_WRITE, + syscall.MAP_SHARED, + f.file.Fd(), + uintptr(chunk<>chunkShift) + f.mappings.Store(newMappings) + if err := state.Load(r, &f.usage, nil); err != nil { + return err + } + + // Try to map committed chunks concurrently: For any given chunk, either + // this loop or the following one will mmap the chunk first and cache it in + // f.mappings for the other, but this loop is likely to run ahead of the + // other since it doesn't do any work between mmaps. The rest of this + // function doesn't mutate f.usage, so it's safe to iterate concurrently. + mapperDone := make(chan struct{}) + mapperCanceled := int32(0) + go func() { // S/R-SAFE: see comment + defer func() { close(mapperDone) }() + for seg := f.usage.FirstSegment(); seg.Ok(); seg = seg.NextSegment() { + if atomic.LoadInt32(&mapperCanceled) != 0 { + return + } + if seg.Value().knownCommitted { + f.forEachMappingSlice(seg.Range(), func(s []byte) {}) + } + } + }() + defer func() { + atomic.StoreInt32(&mapperCanceled, 1) + <-mapperDone + }() + + // Load committed pages. + for seg := f.usage.FirstSegment(); seg.Ok(); seg = seg.NextSegment() { + if !seg.Value().knownCommitted { + continue + } + // Verify header. + length, object, err := state.ReadHeader(r) + if err != nil { + return err + } + if object { + // Not expected. + return fmt.Errorf("unexpected object") + } + if expected := uint64(seg.Range().Length()); length != expected { + // Size mismatch. + return fmt.Errorf("mismatched segment: expected %d, got %d", expected, length) + } + // Read data. + var ioErr error + err = f.forEachMappingSlice(seg.Range(), func(s []byte) { + if ioErr != nil { + return + } + _, ioErr = io.ReadFull(r, s) + }) + if ioErr != nil { + return ioErr + } + if err != nil { + return err + } + + // Update accounting for restored pages. We need to do this here since + // these segments are marked as "known committed", and will be skipped + // over on accounting scans. + usage.MemoryAccounting.Inc(seg.End()-seg.Start(), seg.Value().kind) + } + + return nil +} + +// MemoryFileProvider provides the MemoryFile method. +// +// This type exists to work around a save/restore defect. The only object in a +// saved object graph that S/R allows to be replaced at time of restore is the +// starting point of the restore, kernel.Kernel. However, the MemoryFile +// changes between save and restore as well, so objects that need persistent +// access to the MemoryFile must instead store a pointer to the Kernel and call +// Kernel.MemoryFile() as required. In most cases, depending on the kernel +// package directly would create a package dependency loop, so the stored +// pointer must instead be a MemoryProvider interface object. Correspondingly, +// kernel.Kernel is the only implementation of this interface. +type MemoryFileProvider interface { + // MemoryFile returns the Kernel MemoryFile. + MemoryFile() *MemoryFile +} diff --git a/pkg/sentry/platform/filemem/BUILD b/pkg/sentry/platform/filemem/BUILD deleted file mode 100644 index 1a61cfaa5..000000000 --- a/pkg/sentry/platform/filemem/BUILD +++ /dev/null @@ -1,56 +0,0 @@ -package(licenses = ["notice"]) - -load("//tools/go_generics:defs.bzl", "go_template_instance") -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") - -go_template_instance( - name = "usage_set", - out = "usage_set.go", - consts = { - "minDegree": "10", - }, - imports = { - "platform": "gvisor.googlesource.com/gvisor/pkg/sentry/platform", - }, - package = "filemem", - prefix = "usage", - template = "//pkg/segment:generic_set", - types = { - "Key": "uint64", - "Range": "platform.FileRange", - "Value": "usageInfo", - "Functions": "usageSetFunctions", - }, -) - -go_library( - name = "filemem", - srcs = [ - "filemem.go", - "filemem_state.go", - "filemem_unsafe.go", - "usage_set.go", - ], - importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/platform/filemem", - visibility = ["//pkg/sentry:internal"], - deps = [ - "//pkg/log", - "//pkg/sentry/arch", - "//pkg/sentry/context", - "//pkg/sentry/memutil", - "//pkg/sentry/platform", - "//pkg/sentry/safemem", - "//pkg/sentry/usage", - "//pkg/sentry/usermem", - "//pkg/state", - "//pkg/syserror", - ], -) - -go_test( - name = "filemem_test", - size = "small", - srcs = ["filemem_test.go"], - embed = [":filemem"], - deps = ["//pkg/sentry/usermem"], -) diff --git a/pkg/sentry/platform/filemem/filemem.go b/pkg/sentry/platform/filemem/filemem.go deleted file mode 100644 index f41c70ba5..000000000 --- a/pkg/sentry/platform/filemem/filemem.go +++ /dev/null @@ -1,879 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// Package filemem provides a reusable implementation of platform.Memory. -// -// It enables memory to be sourced from a memfd file. -// -// Lock order: -// -// filemem.FileMem.mu -// filemem.FileMem.mappingsMu -package filemem - -import ( - "fmt" - "math" - "os" - "sync" - "sync/atomic" - "syscall" - "time" - - "gvisor.googlesource.com/gvisor/pkg/log" - "gvisor.googlesource.com/gvisor/pkg/sentry/memutil" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" - "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" - "gvisor.googlesource.com/gvisor/pkg/sentry/usage" - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" -) - -// FileMem is a platform.Memory that allocates from a host file that it owns. -type FileMem struct { - // Filemem models the backing file as follows: - // - // Each page in the file can be committed or uncommitted. A page is - // committed if the host kernel is spending resources to store its contents - // and uncommitted otherwise. This definition includes pages that the host - // kernel has swapped; this is intentional, to ensure that accounting does - // not change even if host kernel swapping behavior changes, and that - // memory used by pseudo-swap mechanisms like zswap is still accounted. - // - // The initial contents of uncommitted pages are implicitly zero bytes. A - // read or write to the contents of an uncommitted page causes it to be - // committed. This is the only event that can cause a uncommitted page to - // be committed. - // - // fallocate(FALLOC_FL_PUNCH_HOLE) (FileMem.Decommit) causes committed - // pages to be uncommitted. This is the only event that can cause a - // committed page to be uncommitted. - // - // Filemem's accounting is based on identifying the set of committed pages. - // Since filemem does not have direct access to the MMU, tracking reads and - // writes to uncommitted pages to detect commitment would introduce - // additional page faults, which would be prohibitively expensive. Instead, - // filemem queries the host kernel to determine which pages are committed. - - // file is the backing memory file. The file pointer is immutable. - file *os.File - - mu sync.Mutex - - // usage maps each page in the file to metadata for that page. Pages for - // which no segment exists in usage are both unallocated (not in use) and - // uncommitted. - // - // Since usage stores usageInfo objects by value, clients should usually - // use usageIterator.ValuePtr() instead of usageIterator.Value() to get a - // pointer to the usageInfo rather than a copy. - // - // usage must be kept maximally merged (that is, there should never be two - // adjacent segments with the same values). At least markReclaimed depends - // on this property. - // - // usage is protected by mu. - usage usageSet - - // The UpdateUsage function scans all segments with knownCommitted set - // to false, sees which pages are committed and creates corresponding - // segments with knownCommitted set to true. - // - // In order to avoid unnecessary scans, usageExpected tracks the total - // file blocks expected. This is used to elide the scan when this - // matches the underlying file blocks. - // - // To track swapped pages, usageSwapped tracks the discrepency between - // what is observed in core and what is reported by the file. When - // usageSwapped is non-zero, a sweep will be performed at least every - // second. The start of the last sweep is recorded in usageLast. - // - // All usage attributes are all protected by mu. - usageExpected uint64 - usageSwapped uint64 - usageLast time.Time - - // minUnallocatedPage is the minimum page that may be unallocated. - // i.e., there are no unallocated pages below minUnallocatedPage. - // - // minUnallocatedPage is protected by mu. - minUnallocatedPage uint64 - - // fileSize is the size of the backing memory file in bytes. fileSize is - // always a power-of-two multiple of chunkSize. - // - // fileSize is protected by mu. - fileSize int64 - - // destroyed is set by Destroy to instruct the reclaimer goroutine to - // release resources and exit. destroyed is protected by mu. - destroyed bool - - // reclaimable is true if usage may contain reclaimable pages. reclaimable - // is protected by mu. - reclaimable bool - - // minReclaimablePage is the minimum page that may be reclaimable. - // i.e., all reclaimable pages are >= minReclaimablePage. - // - // minReclaimablePage is protected by mu. - minReclaimablePage uint64 - - // reclaimCond is signaled (with mu locked) when reclaimable or destroyed - // transitions from false to true. - reclaimCond sync.Cond - - // Filemem pages are mapped into the local address space on the granularity - // of large pieces called chunks. mappings is a []uintptr that stores, for - // each chunk, the start address of a mapping of that chunk in the current - // process' address space, or 0 if no such mapping exists. Once a chunk is - // mapped, it is never remapped or unmapped until the filemem is destroyed. - // - // Mutating the mappings slice or its contents requires both holding - // mappingsMu and using atomic memory operations. (The slice is mutated - // whenever the file is expanded. Per the above, the only permitted - // mutation of the slice's contents is the assignment of a mapping to a - // chunk that was previously unmapped.) Reading the slice or its contents - // only requires *either* holding mappingsMu or using atomic memory - // operations. This allows FileMem.AccessPhysical to avoid locking in the - // common case where chunk mappings already exist. - - mappingsMu sync.Mutex - mappings atomic.Value -} - -// usage tracks usage information. -// -// +stateify savable -type usageInfo struct { - // kind is the usage kind. - kind usage.MemoryKind - - // knownCommitted indicates whether this region is known to be - // committed. If this is false, then the region may or may not have - // been touched. If it is true however, then mincore (below) has - // indicated that the page is present at least once. - knownCommitted bool - - refs uint64 -} - -const ( - chunkShift = 24 - chunkSize = 1 << chunkShift // 16 MB - chunkMask = chunkSize - 1 - - initialSize = chunkSize - - // maxPage is the highest 64-bit page. - maxPage = math.MaxUint64 &^ (usermem.PageSize - 1) -) - -// newFromFile creates a FileMem backed by the given file. -func newFromFile(file *os.File) (*FileMem, error) { - if err := file.Truncate(initialSize); err != nil { - return nil, err - } - f := &FileMem{ - fileSize: initialSize, - file: file, - // No pages are reclaimable. DecRef will always be able to - // decrease minReclaimablePage from this point. - minReclaimablePage: maxPage, - } - f.reclaimCond.L = &f.mu - f.mappings.Store(make([]uintptr, initialSize/chunkSize)) - go f.runReclaim() // S/R-SAFE: f.mu - - // The Linux kernel contains an optional feature called "Integrity - // Measurement Architecture" (IMA). If IMA is enabled, it will checksum - // binaries the first time they are mapped PROT_EXEC. This is bad news for - // executable pages mapped from FileMem, which can grow to terabytes in - // (sparse) size. If IMA attempts to checksum a file that large, it will - // allocate all of the sparse pages and quickly exhaust all memory. - // - // Work around IMA by immediately creating a temporary PROT_EXEC mapping, - // while FileMem is still small. IMA will ignore any future mappings. - m, _, errno := syscall.Syscall6( - syscall.SYS_MMAP, - 0, - usermem.PageSize, - syscall.PROT_EXEC, - syscall.MAP_SHARED, - f.file.Fd(), - 0) - if errno != 0 { - // This isn't fatal to filemem (IMA may not even be in use). Log the - // error, but don't return it. - log.Warningf("Failed to pre-map FileMem PROT_EXEC: %v", errno) - } else { - syscall.Syscall( - syscall.SYS_MUNMAP, - m, - usermem.PageSize, - 0) - } - - return f, nil -} - -// New creates a FileMem backed by a memfd file. -func New(name string) (*FileMem, error) { - fd, err := memutil.CreateMemFD(name, 0) - if err != nil { - if e, ok := err.(syscall.Errno); ok && e == syscall.ENOSYS { - return nil, fmt.Errorf("memfd_create(2) is not implemented. Check that you have Linux 3.17 or higher") - } - return nil, err - } - return newFromFile(os.NewFile(uintptr(fd), name)) -} - -// Destroy implements platform.Memory.Destroy. -func (f *FileMem) Destroy() { - f.mu.Lock() - defer f.mu.Unlock() - f.destroyed = true - f.reclaimCond.Signal() -} - -// Allocate implements platform.Memory.Allocate. -func (f *FileMem) Allocate(length uint64, kind usage.MemoryKind) (platform.FileRange, error) { - if length == 0 || length%usermem.PageSize != 0 { - panic(fmt.Sprintf("invalid allocation length: %#x", length)) - } - - f.mu.Lock() - defer f.mu.Unlock() - - // Align hugepage-and-larger allocations on hugepage boundaries to try - // to take advantage of hugetmpfs. - alignment := uint64(usermem.PageSize) - if length >= usermem.HugePageSize { - alignment = usermem.HugePageSize - } - - start, minUnallocatedPage := findUnallocatedRange(&f.usage, f.minUnallocatedPage, length, alignment) - end := start + length - // File offsets are int64s. Since length must be strictly positive, end - // cannot legitimately be 0. - if end < start || int64(end) <= 0 { - return platform.FileRange{}, syserror.ENOMEM - } - - // Expand the file if needed. Double the file size on each expansion; - // uncommitted pages have effectively no cost. - fileSize := f.fileSize - for int64(end) > fileSize { - if fileSize >= 2*fileSize { - // fileSize overflow. - return platform.FileRange{}, syserror.ENOMEM - } - fileSize *= 2 - } - if fileSize > f.fileSize { - if err := f.file.Truncate(fileSize); err != nil { - return platform.FileRange{}, err - } - f.fileSize = fileSize - f.mappingsMu.Lock() - oldMappings := f.mappings.Load().([]uintptr) - newMappings := make([]uintptr, fileSize>>chunkShift) - copy(newMappings, oldMappings) - f.mappings.Store(newMappings) - f.mappingsMu.Unlock() - } - - // Mark selected pages as in use. - fr := platform.FileRange{start, end} - if !f.usage.Add(fr, usageInfo{ - kind: kind, - refs: 1, - }) { - panic(fmt.Sprintf("allocating %v: failed to insert into f.usage:\n%v", fr, &f.usage)) - } - - if minUnallocatedPage < start { - f.minUnallocatedPage = minUnallocatedPage - } else { - // start was the first unallocated page. The next must be - // somewhere beyond end. - f.minUnallocatedPage = end - } - - return fr, nil -} - -// findUnallocatedRange returns the first unallocated page in usage of the -// specified length and alignment beginning at page start and the first single -// unallocated page. -func findUnallocatedRange(usage *usageSet, start, length, alignment uint64) (uint64, uint64) { - // Only searched until the first page is found. - firstPage := start - foundFirstPage := false - alignMask := alignment - 1 - for seg := usage.LowerBoundSegment(start); seg.Ok(); seg = seg.NextSegment() { - r := seg.Range() - - if !foundFirstPage && r.Start > firstPage { - foundFirstPage = true - } - - if start >= r.End { - // start was rounded up to an alignment boundary from the end - // of a previous segment and is now beyond r.End. - continue - } - // This segment represents allocated or reclaimable pages; only the - // range from start to the segment's beginning is allocatable, and the - // next allocatable range begins after the segment. - if r.Start > start && r.Start-start >= length { - break - } - start = (r.End + alignMask) &^ alignMask - if !foundFirstPage { - firstPage = r.End - } - } - return start, firstPage -} - -// fallocate(2) modes, defined in Linux's include/uapi/linux/falloc.h. -const ( - _FALLOC_FL_KEEP_SIZE = 1 - _FALLOC_FL_PUNCH_HOLE = 2 -) - -// Decommit implements platform.Memory.Decommit. -func (f *FileMem) Decommit(fr platform.FileRange) error { - if !fr.WellFormed() || fr.Length() == 0 || fr.Start%usermem.PageSize != 0 || fr.End%usermem.PageSize != 0 { - panic(fmt.Sprintf("invalid range: %v", fr)) - } - - // "After a successful call, subsequent reads from this range will - // return zeroes. The FALLOC_FL_PUNCH_HOLE flag must be ORed with - // FALLOC_FL_KEEP_SIZE in mode ..." - fallocate(2) - err := syscall.Fallocate( - int(f.file.Fd()), - _FALLOC_FL_PUNCH_HOLE|_FALLOC_FL_KEEP_SIZE, - int64(fr.Start), - int64(fr.Length())) - if err != nil { - return err - } - f.markDecommitted(fr) - return nil -} - -func (f *FileMem) markDecommitted(fr platform.FileRange) { - f.mu.Lock() - defer f.mu.Unlock() - // Since we're changing the knownCommitted attribute, we need to merge - // across the entire range to ensure that the usage tree is minimal. - gap := f.usage.ApplyContiguous(fr, func(seg usageIterator) { - val := seg.ValuePtr() - if val.knownCommitted { - // Drop the usageExpected appropriately. - amount := seg.Range().Length() - usage.MemoryAccounting.Dec(amount, val.kind) - f.usageExpected -= amount - val.knownCommitted = false - } - }) - if gap.Ok() { - panic(fmt.Sprintf("Decommit(%v): attempted to decommit unallocated pages %v:\n%v", fr, gap.Range(), &f.usage)) - } - f.usage.MergeRange(fr) -} - -// runReclaim implements the reclaimer goroutine, which continuously decommits -// reclaimable frames in order to reduce memory usage. -func (f *FileMem) runReclaim() { - for { - fr, ok := f.findReclaimable() - if !ok { - break - } - - if err := f.Decommit(fr); err != nil { - log.Warningf("Reclaim failed to decommit %v: %v", fr, err) - // Zero the frames manually. This won't reduce memory usage, but at - // least ensures that the frames will be zero when reallocated. - f.forEachMappingSlice(fr, func(bs []byte) { - for i := range bs { - bs[i] = 0 - } - }) - // Pretend the frames were decommitted even though they weren't, - // since the memory accounting implementation has no idea how to - // deal with this. - f.markDecommitted(fr) - } - f.markReclaimed(fr) - } - // We only get here if findReclaimable finds f.destroyed set and returns - // false. - f.mu.Lock() - defer f.mu.Unlock() - if !f.destroyed { - panic("findReclaimable broke out of reclaim loop, but f.destroyed is no longer set") - } - f.file.Close() - // Ensure that any attempts to use f.file.Fd() fail instead of getting a fd - // that has possibly been reassigned. - f.file = nil - mappings := f.mappings.Load().([]uintptr) - for i, m := range mappings { - if m != 0 { - _, _, errno := syscall.Syscall(syscall.SYS_MUNMAP, m, chunkSize, 0) - if errno != 0 { - log.Warningf("Failed to unmap mapping %#x for filemem chunk %d: %v", m, i, errno) - } - } - } - // Similarly, invalidate f.mappings. (atomic.Value.Store(nil) panics.) - f.mappings.Store([]uintptr{}) -} - -func (f *FileMem) findReclaimable() (platform.FileRange, bool) { - f.mu.Lock() - defer f.mu.Unlock() - for { - for { - if f.destroyed { - return platform.FileRange{}, false - } - if f.reclaimable { - break - } - f.reclaimCond.Wait() - } - // Allocate returns the first usable range in offset order and is - // currently a linear scan, so reclaiming from the beginning of the - // file minimizes the expected latency of Allocate. - for seg := f.usage.LowerBoundSegment(f.minReclaimablePage); seg.Ok(); seg = seg.NextSegment() { - if seg.ValuePtr().refs == 0 { - f.minReclaimablePage = seg.End() - return seg.Range(), true - } - } - f.reclaimable = false - // No pages are reclaimable. - f.minReclaimablePage = maxPage - } -} - -func (f *FileMem) markReclaimed(fr platform.FileRange) { - f.mu.Lock() - defer f.mu.Unlock() - seg := f.usage.FindSegment(fr.Start) - // All of fr should be mapped to a single uncommitted reclaimable segment - // accounted to System. - if !seg.Ok() { - panic(fmt.Sprintf("Reclaimed pages %v include unreferenced pages:\n%v", fr, &f.usage)) - } - if !seg.Range().IsSupersetOf(fr) { - panic(fmt.Sprintf("Reclaimed pages %v are not entirely contained in segment %v with state %v:\n%v", fr, seg.Range(), seg.Value(), &f.usage)) - } - if got, want := seg.Value(), (usageInfo{ - kind: usage.System, - knownCommitted: false, - refs: 0, - }); got != want { - panic(fmt.Sprintf("Reclaimed pages %v in segment %v has incorrect state %v, wanted %v:\n%v", fr, seg.Range(), got, want, &f.usage)) - } - // Deallocate reclaimed pages. Even though all of seg is reclaimable, the - // caller of markReclaimed may not have decommitted it, so we can only mark - // fr as reclaimed. - f.usage.Remove(f.usage.Isolate(seg, fr)) - if fr.Start < f.minUnallocatedPage { - // We've deallocated at least one lower page. - f.minUnallocatedPage = fr.Start - } -} - -// IncRef implements platform.File.IncRef. -func (f *FileMem) IncRef(fr platform.FileRange) { - if !fr.WellFormed() || fr.Length() == 0 || fr.Start%usermem.PageSize != 0 || fr.End%usermem.PageSize != 0 { - panic(fmt.Sprintf("invalid range: %v", fr)) - } - - f.mu.Lock() - defer f.mu.Unlock() - - gap := f.usage.ApplyContiguous(fr, func(seg usageIterator) { - seg.ValuePtr().refs++ - }) - if gap.Ok() { - panic(fmt.Sprintf("IncRef(%v): attempted to IncRef on unallocated pages %v:\n%v", fr, gap.Range(), &f.usage)) - } - - f.usage.MergeAdjacent(fr) -} - -// DecRef implements platform.File.DecRef. -func (f *FileMem) DecRef(fr platform.FileRange) { - if !fr.WellFormed() || fr.Length() == 0 || fr.Start%usermem.PageSize != 0 || fr.End%usermem.PageSize != 0 { - panic(fmt.Sprintf("invalid range: %v", fr)) - } - - var freed bool - - f.mu.Lock() - defer f.mu.Unlock() - - for seg := f.usage.FindSegment(fr.Start); seg.Ok() && seg.Start() < fr.End; seg = seg.NextSegment() { - seg = f.usage.Isolate(seg, fr) - val := seg.ValuePtr() - if val.refs == 0 { - panic(fmt.Sprintf("DecRef(%v): 0 existing references on %v:\n%v", fr, seg.Range(), &f.usage)) - } - val.refs-- - if val.refs == 0 { - freed = true - // Reclassify memory as System, until it's freed by the reclaim - // goroutine. - if val.knownCommitted { - usage.MemoryAccounting.Move(seg.Range().Length(), usage.System, val.kind) - } - val.kind = usage.System - } - } - f.usage.MergeAdjacent(fr) - - if freed { - if fr.Start < f.minReclaimablePage { - // We've freed at least one lower page. - f.minReclaimablePage = fr.Start - } - f.reclaimable = true - f.reclaimCond.Signal() - } -} - -// MapInternal implements platform.File.MapInternal. -func (f *FileMem) MapInternal(fr platform.FileRange, at usermem.AccessType) (safemem.BlockSeq, error) { - if !fr.WellFormed() || fr.Length() == 0 { - panic(fmt.Sprintf("invalid range: %v", fr)) - } - if at.Execute { - return safemem.BlockSeq{}, syserror.EACCES - } - - chunks := ((fr.End + chunkMask) >> chunkShift) - (fr.Start >> chunkShift) - if chunks == 1 { - // Avoid an unnecessary slice allocation. - var seq safemem.BlockSeq - err := f.forEachMappingSlice(fr, func(bs []byte) { - seq = safemem.BlockSeqOf(safemem.BlockFromSafeSlice(bs)) - }) - return seq, err - } - blocks := make([]safemem.Block, 0, chunks) - err := f.forEachMappingSlice(fr, func(bs []byte) { - blocks = append(blocks, safemem.BlockFromSafeSlice(bs)) - }) - return safemem.BlockSeqFromSlice(blocks), err -} - -// forEachMappingSlice invokes fn on a sequence of byte slices that -// collectively map all bytes in fr. -func (f *FileMem) forEachMappingSlice(fr platform.FileRange, fn func([]byte)) error { - mappings := f.mappings.Load().([]uintptr) - for chunkStart := fr.Start &^ chunkMask; chunkStart < fr.End; chunkStart += chunkSize { - chunk := int(chunkStart >> chunkShift) - m := atomic.LoadUintptr(&mappings[chunk]) - if m == 0 { - var err error - mappings, m, err = f.getChunkMapping(chunk) - if err != nil { - return err - } - } - startOff := uint64(0) - if chunkStart < fr.Start { - startOff = fr.Start - chunkStart - } - endOff := uint64(chunkSize) - if chunkStart+chunkSize > fr.End { - endOff = fr.End - chunkStart - } - fn(unsafeSlice(m, chunkSize)[startOff:endOff]) - } - return nil -} - -func (f *FileMem) getChunkMapping(chunk int) ([]uintptr, uintptr, error) { - f.mappingsMu.Lock() - defer f.mappingsMu.Unlock() - // Another thread may have replaced f.mappings altogether due to file - // expansion. - mappings := f.mappings.Load().([]uintptr) - // Another thread may have already mapped the chunk. - if m := mappings[chunk]; m != 0 { - return mappings, m, nil - } - m, _, errno := syscall.Syscall6( - syscall.SYS_MMAP, - 0, - chunkSize, - syscall.PROT_READ|syscall.PROT_WRITE, - syscall.MAP_SHARED, - f.file.Fd(), - uintptr(chunk<>chunkShift) - f.mappings.Store(newMappings) - if err := state.Load(r, &f.usage, nil); err != nil { - return err - } - - // Try to map committed chunks concurrently: For any given chunk, either - // this loop or the following one will mmap the chunk first and cache it in - // f.mappings for the other, but this loop is likely to run ahead of the - // other since it doesn't do any work between mmaps. The rest of this - // function doesn't mutate f.usage, so it's safe to iterate concurrently. - mapperDone := make(chan struct{}) - mapperCanceled := int32(0) - go func() { // S/R-SAFE: see comment - defer func() { close(mapperDone) }() - for seg := f.usage.FirstSegment(); seg.Ok(); seg = seg.NextSegment() { - if atomic.LoadInt32(&mapperCanceled) != 0 { - return - } - if seg.Value().knownCommitted { - f.forEachMappingSlice(seg.Range(), func(s []byte) {}) - } - } - }() - defer func() { - atomic.StoreInt32(&mapperCanceled, 1) - <-mapperDone - }() - - // Load committed pages. - for seg := f.usage.FirstSegment(); seg.Ok(); seg = seg.NextSegment() { - if !seg.Value().knownCommitted { - continue - } - // Verify header. - length, object, err := state.ReadHeader(r) - if err != nil { - return err - } - if object { - // Not expected. - return fmt.Errorf("unexpected object") - } - if expected := uint64(seg.Range().Length()); length != expected { - // Size mismatch. - return fmt.Errorf("mismatched segment: expected %d, got %d", expected, length) - } - // Read data. - var ioErr error - err = f.forEachMappingSlice(seg.Range(), func(s []byte) { - if ioErr != nil { - return - } - _, ioErr = io.ReadFull(r, s) - }) - if ioErr != nil { - return ioErr - } - if err != nil { - return err - } - - // Update accounting for restored pages. We need to do this here since - // these segments are marked as "known committed", and will be skipped - // over on accounting scans. - usage.MemoryAccounting.Inc(seg.End()-seg.Start(), seg.Value().kind) - } - - return nil -} diff --git a/pkg/sentry/platform/filemem/filemem_test.go b/pkg/sentry/platform/filemem/filemem_test.go deleted file mode 100644 index 9becec25f..000000000 --- a/pkg/sentry/platform/filemem/filemem_test.go +++ /dev/null @@ -1,168 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package filemem - -import ( - "testing" - - "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" -) - -const ( - page = usermem.PageSize - hugepage = usermem.HugePageSize -) - -func TestFindUnallocatedRange(t *testing.T) { - for _, test := range []struct { - desc string - usage *usageSegmentDataSlices - start uint64 - length uint64 - alignment uint64 - unallocated uint64 - minUnallocated uint64 - }{ - { - desc: "Initial allocation succeeds", - usage: &usageSegmentDataSlices{}, - start: 0, - length: page, - alignment: page, - unallocated: 0, - minUnallocated: 0, - }, - { - desc: "Allocation begins at start of file", - usage: &usageSegmentDataSlices{ - Start: []uint64{page}, - End: []uint64{2 * page}, - Values: []usageInfo{{refs: 1}}, - }, - start: 0, - length: page, - alignment: page, - unallocated: 0, - minUnallocated: 0, - }, - { - desc: "In-use frames are not allocatable", - usage: &usageSegmentDataSlices{ - Start: []uint64{0, page}, - End: []uint64{page, 2 * page}, - Values: []usageInfo{{refs: 1}, {refs: 2}}, - }, - start: 0, - length: page, - alignment: page, - unallocated: 2 * page, - minUnallocated: 2 * page, - }, - { - desc: "Reclaimable frames are not allocatable", - usage: &usageSegmentDataSlices{ - Start: []uint64{0, page, 2 * page}, - End: []uint64{page, 2 * page, 3 * page}, - Values: []usageInfo{{refs: 1}, {refs: 0}, {refs: 1}}, - }, - start: 0, - length: page, - alignment: page, - unallocated: 3 * page, - minUnallocated: 3 * page, - }, - { - desc: "Gaps between in-use frames are allocatable", - usage: &usageSegmentDataSlices{ - Start: []uint64{0, 2 * page}, - End: []uint64{page, 3 * page}, - Values: []usageInfo{{refs: 1}, {refs: 1}}, - }, - start: 0, - length: page, - alignment: page, - unallocated: page, - minUnallocated: page, - }, - { - desc: "Inadequately-sized gaps are rejected", - usage: &usageSegmentDataSlices{ - Start: []uint64{0, 2 * page}, - End: []uint64{page, 3 * page}, - Values: []usageInfo{{refs: 1}, {refs: 1}}, - }, - start: 0, - length: 2 * page, - alignment: page, - unallocated: 3 * page, - minUnallocated: page, - }, - { - desc: "Hugepage alignment is honored", - usage: &usageSegmentDataSlices{ - Start: []uint64{0, hugepage + page}, - // Hugepage-sized gap here that shouldn't be allocated from - // since it's incorrectly aligned. - End: []uint64{page, hugepage + 2*page}, - Values: []usageInfo{{refs: 1}, {refs: 1}}, - }, - start: 0, - length: hugepage, - alignment: hugepage, - unallocated: 2 * hugepage, - minUnallocated: page, - }, - { - desc: "Pages before start ignored", - usage: &usageSegmentDataSlices{ - Start: []uint64{page, 3 * page}, - End: []uint64{2 * page, 4 * page}, - Values: []usageInfo{{refs: 1}, {refs: 2}}, - }, - start: page, - length: page, - alignment: page, - unallocated: 2 * page, - minUnallocated: 2 * page, - }, - { - desc: "start may be in the middle of segment", - usage: &usageSegmentDataSlices{ - Start: []uint64{0, 3 * page}, - End: []uint64{2 * page, 4 * page}, - Values: []usageInfo{{refs: 1}, {refs: 2}}, - }, - start: page, - length: page, - alignment: page, - unallocated: 2 * page, - minUnallocated: 2 * page, - }, - } { - t.Run(test.desc, func(t *testing.T) { - var usage usageSet - if err := usage.ImportSortedSlices(test.usage); err != nil { - t.Fatalf("Failed to initialize usage from %v: %v", test.usage, err) - } - unallocated, minUnallocated := findUnallocatedRange(&usage, test.start, test.length, test.alignment) - if unallocated != test.unallocated { - t.Errorf("findUnallocatedRange(%v, %x, %x, %x): got unallocated %x, wanted %x", test.usage, test.start, test.length, test.alignment, unallocated, test.unallocated) - } - if minUnallocated != test.minUnallocated { - t.Errorf("findUnallocatedRange(%v, %x, %x, %x): got minUnallocated %x, wanted %x", test.usage, test.start, test.length, test.alignment, minUnallocated, test.minUnallocated) - } - }) - } -} diff --git a/pkg/sentry/platform/filemem/filemem_unsafe.go b/pkg/sentry/platform/filemem/filemem_unsafe.go deleted file mode 100644 index 776aed74d..000000000 --- a/pkg/sentry/platform/filemem/filemem_unsafe.go +++ /dev/null @@ -1,40 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package filemem - -import ( - "reflect" - "syscall" - "unsafe" -) - -func unsafeSlice(addr uintptr, length int) (slice []byte) { - sh := (*reflect.SliceHeader)(unsafe.Pointer(&slice)) - sh.Data = addr - sh.Len = length - sh.Cap = length - return -} - -func mincore(s []byte, buf []byte) error { - if _, _, errno := syscall.RawSyscall( - syscall.SYS_MINCORE, - uintptr(unsafe.Pointer(&s[0])), - uintptr(len(s)), - uintptr(unsafe.Pointer(&buf[0]))); errno != 0 { - return errno - } - return nil -} diff --git a/pkg/sentry/platform/kvm/BUILD b/pkg/sentry/platform/kvm/BUILD index b7bf88249..9999e58f4 100644 --- a/pkg/sentry/platform/kvm/BUILD +++ b/pkg/sentry/platform/kvm/BUILD @@ -34,7 +34,6 @@ go_library( "//pkg/log", "//pkg/sentry/arch", "//pkg/sentry/platform", - "//pkg/sentry/platform/filemem", "//pkg/sentry/platform/interrupt", "//pkg/sentry/platform/procid", "//pkg/sentry/platform/ring0", diff --git a/pkg/sentry/platform/kvm/address_space.go b/pkg/sentry/platform/kvm/address_space.go index 6d8d8e65b..f2f7ab1e8 100644 --- a/pkg/sentry/platform/kvm/address_space.go +++ b/pkg/sentry/platform/kvm/address_space.go @@ -20,7 +20,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/atomicbitops" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform/filemem" "gvisor.googlesource.com/gvisor/pkg/sentry/platform/ring0/pagetables" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) @@ -76,9 +75,6 @@ type addressSpace struct { // Note that the page tables themselves are not locked. mu sync.Mutex - // filemem is the memory instance. - filemem *filemem.FileMem - // machine is the underlying machine. machine *machine diff --git a/pkg/sentry/platform/kvm/kvm.go b/pkg/sentry/platform/kvm/kvm.go index d4f50024d..c5a4435b1 100644 --- a/pkg/sentry/platform/kvm/kvm.go +++ b/pkg/sentry/platform/kvm/kvm.go @@ -23,7 +23,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/cpuid" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform/filemem" "gvisor.googlesource.com/gvisor/pkg/sentry/platform/ring0" "gvisor.googlesource.com/gvisor/pkg/sentry/platform/ring0/pagetables" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -33,9 +32,6 @@ import ( type KVM struct { platform.NoCPUPreemptionDetection - // filemem is our memory source. - *filemem.FileMem - // machine is the backing VM. machine *machine } @@ -56,12 +52,6 @@ func OpenDevice() (*os.File, error) { // New returns a new KVM-based implementation of the platform interface. func New(deviceFile *os.File) (*KVM, error) { - // Allocate physical memory for the vCPUs. - fm, err := filemem.New("kvm-memory") - if err != nil { - return nil, err - } - fd := deviceFile.Fd() // Ensure global initialization is done. @@ -90,7 +80,6 @@ func New(deviceFile *os.File) (*KVM, error) { // All set. return &KVM{ - FileMem: fm, machine: machine, }, nil } @@ -140,7 +129,6 @@ func (k *KVM) NewAddressSpace(_ interface{}) (platform.AddressSpace, <-chan stru // Return the new address space. return &addressSpace{ - filemem: k.FileMem, machine: k.machine, pageTables: pageTables, dirtySet: k.machine.newDirtySet(), @@ -153,8 +141,3 @@ func (k *KVM) NewContext() platform.Context { machine: k.machine, } } - -// Memory returns the platform memory used to do allocations. -func (k *KVM) Memory() platform.Memory { - return k.FileMem -} diff --git a/pkg/sentry/platform/kvm/kvm_test.go b/pkg/sentry/platform/kvm/kvm_test.go index fff463a6e..361200622 100644 --- a/pkg/sentry/platform/kvm/kvm_test.go +++ b/pkg/sentry/platform/kvm/kvm_test.go @@ -48,7 +48,6 @@ func kvmTest(t testHarness, setup func(*KVM), fn func(*vCPU) bool) { t.Fatalf("error creating KVM instance: %v", err) } defer k.machine.Destroy() - defer k.FileMem.Destroy() // Call additional setup. if setup != nil { diff --git a/pkg/sentry/platform/platform.go b/pkg/sentry/platform/platform.go index b2ce851da..d1c9458ea 100644 --- a/pkg/sentry/platform/platform.go +++ b/pkg/sentry/platform/platform.go @@ -19,17 +19,15 @@ package platform import ( "fmt" - "io" "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" - "gvisor.googlesource.com/gvisor/pkg/sentry/usage" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// Platform provides abstractions for execution contexts (Context) and memory -// management (Memory, AddressSpace). +// Platform provides abstractions for execution contexts (Context, +// AddressSpace). type Platform interface { // SupportsAddressSpaceIO returns true if AddressSpaces returned by this // Platform support AddressSpaceIO methods. @@ -87,9 +85,6 @@ type Platform interface { // NewContext returns a new execution context. NewContext() Context - // Memory returns memory for allocations. - Memory() Memory - // PreemptAllCPUs causes all concurrent calls to Context.Switch(), as well // as the first following call to Context.Switch() for each Context, to // return ErrContextCPUPreempted. @@ -352,84 +347,3 @@ type File interface { func (fr FileRange) String() string { return fmt.Sprintf("[%#x, %#x)", fr.Start, fr.End) } - -// Memory represents an allocatable File that may be mapped into any -// AddressSpace associated with the same Platform. -type Memory interface { - File - - // Allocate returns a range of initially-zeroed pages of the given length - // with the given accounting kind and a single reference held by the - // caller. When the last reference on an allocated page is released, - // ownership of the page is returned to the Memory, allowing it to be - // returned by a future call to Allocate. - // - // Preconditions: length must be page-aligned and non-zero. - Allocate(length uint64, kind usage.MemoryKind) (FileRange, error) - - // Decommit releases resources associated with maintaining the contents of - // the given frames. If Decommit succeeds, future accesses of the - // decommitted frames will read zeroes. - // - // Preconditions: fr.Length() > 0. - Decommit(fr FileRange) error - - // UpdateUsage updates the memory usage statistics. This must be called - // before the relevant memory statistics in usage.MemoryAccounting can - // be considered accurate. - UpdateUsage() error - - // TotalUsage returns an aggregate usage for all memory statistics - // except Mapped (which is external to the Memory implementation). This - // is generally much cheaper than UpdateUsage, but will not provide a - // fine-grained breakdown. - TotalUsage() (uint64, error) - - // TotalSize returns the current maximum size of the Memory in bytes. The - // value returned by TotalSize is permitted to change. - TotalSize() uint64 - - // Destroy releases all resources associated with the Memory. - // - // Preconditions: There are no remaining uses of any of the freed memory's - // frames. - // - // Postconditions: None of the Memory's methods may be called after Destroy. - Destroy() - - // SaveTo saves the memory state to the given stream, which will - // generally be a statefile. - SaveTo(w io.Writer) error - - // LoadFrom loads the memory state from the given stream, which will - // generally be a statefile. - LoadFrom(r io.Reader) error -} - -// AllocateAndFill allocates memory of the given kind from mem and fills it by -// calling r.ReadToBlocks() repeatedly until either length bytes are read or a -// non-nil error is returned. It returns the memory filled by r, truncated down -// to the nearest page. If this is shorter than length bytes due to an error -// returned by r.ReadToBlocks(), it returns that error. -// -// Preconditions: length > 0. length must be page-aligned. -func AllocateAndFill(mem Memory, length uint64, kind usage.MemoryKind, r safemem.Reader) (FileRange, error) { - fr, err := mem.Allocate(length, kind) - if err != nil { - return FileRange{}, err - } - dsts, err := mem.MapInternal(fr, usermem.Write) - if err != nil { - mem.DecRef(fr) - return FileRange{}, err - } - n, err := safemem.ReadFullToBlocks(r, dsts) - un := uint64(usermem.Addr(n).RoundDown()) - if un < length { - // Free unused memory and update fr to contain only the memory that is - // still allocated. - mem.DecRef(FileRange{fr.Start + un, fr.End}) - fr.End = fr.Start + un - } - return fr, err -} diff --git a/pkg/sentry/platform/ptrace/BUILD b/pkg/sentry/platform/ptrace/BUILD index f86790942..e9e4a0d16 100644 --- a/pkg/sentry/platform/ptrace/BUILD +++ b/pkg/sentry/platform/ptrace/BUILD @@ -23,7 +23,6 @@ go_library( "//pkg/seccomp", "//pkg/sentry/arch", "//pkg/sentry/platform", - "//pkg/sentry/platform/filemem", "//pkg/sentry/platform/interrupt", "//pkg/sentry/platform/procid", "//pkg/sentry/platform/safecopy", diff --git a/pkg/sentry/platform/ptrace/ptrace.go b/pkg/sentry/platform/ptrace/ptrace.go index 8d3f6ac9a..3c0713e95 100644 --- a/pkg/sentry/platform/ptrace/ptrace.go +++ b/pkg/sentry/platform/ptrace/ptrace.go @@ -50,7 +50,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform/filemem" "gvisor.googlesource.com/gvisor/pkg/sentry/platform/interrupt" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) @@ -181,7 +180,6 @@ func (c *context) Interrupt() { type PTrace struct { platform.MMapMinAddr platform.NoCPUPreemptionDetection - *filemem.FileMem } // New returns a new ptrace-based implementation of the platform interface. @@ -202,12 +200,7 @@ func New() (*PTrace, error) { globalPool.master = master }) - fm, err := filemem.New("ptrace-memory") - if err != nil { - return nil, err - } - - return &PTrace{FileMem: fm}, nil + return &PTrace{}, nil } // SupportsAddressSpaceIO implements platform.Platform.SupportsAddressSpaceIO. @@ -243,8 +236,3 @@ func (p *PTrace) NewAddressSpace(_ interface{}) (platform.AddressSpace, <-chan s func (*PTrace) NewContext() platform.Context { return &context{} } - -// Memory returns the platform memory used to do allocations. -func (p *PTrace) Memory() platform.Memory { - return p.FileMem -} diff --git a/pkg/sentry/state/BUILD b/pkg/sentry/state/BUILD index 42c459acc..69385e23c 100644 --- a/pkg/sentry/state/BUILD +++ b/pkg/sentry/state/BUILD @@ -16,7 +16,6 @@ go_library( "//pkg/log", "//pkg/sentry/inet", "//pkg/sentry/kernel", - "//pkg/sentry/platform", "//pkg/sentry/watchdog", "//pkg/state/statefile", ], diff --git a/pkg/sentry/state/state.go b/pkg/sentry/state/state.go index 70b33f190..67db78a56 100644 --- a/pkg/sentry/state/state.go +++ b/pkg/sentry/state/state.go @@ -22,7 +22,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/inet" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" - "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/watchdog" "gvisor.googlesource.com/gvisor/pkg/state/statefile" ) @@ -95,7 +94,7 @@ type LoadOpts struct { } // Load loads the given kernel, setting the provided platform and stack. -func (opts LoadOpts) Load(k *kernel.Kernel, p platform.Platform, n inet.Stack) error { +func (opts LoadOpts) Load(k *kernel.Kernel, n inet.Stack) error { // Open the file. r, m, err := statefile.NewReader(opts.Source, opts.Key) if err != nil { @@ -105,5 +104,5 @@ func (opts LoadOpts) Load(k *kernel.Kernel, p platform.Platform, n inet.Stack) e previousMetadata = m // Restore the Kernel object graph. - return k.LoadFrom(r, p, n) + return k.LoadFrom(r, n) } diff --git a/pkg/sentry/syscalls/linux/sys_sysinfo.go b/pkg/sentry/syscalls/linux/sys_sysinfo.go index 5eeb3ba58..6f7acf98f 100644 --- a/pkg/sentry/syscalls/linux/sys_sysinfo.go +++ b/pkg/sentry/syscalls/linux/sys_sysinfo.go @@ -25,10 +25,10 @@ import ( func Sysinfo(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { addr := args[0].Pointer() - mem := t.Kernel().Platform.Memory() - mem.UpdateUsage() + mf := t.Kernel().MemoryFile() + mf.UpdateUsage() _, totalUsage := usage.MemoryAccounting.Copy() - totalSize := usage.TotalMemory(mem.TotalSize(), totalUsage) + totalSize := usage.TotalMemory(mf.TotalSize(), totalUsage) // Only a subset of the fields in sysinfo_t make sense to return. si := linux.Sysinfo{ diff --git a/pkg/sentry/usage/memory.go b/pkg/sentry/usage/memory.go index 7e065cb76..5be9ed9c6 100644 --- a/pkg/sentry/usage/memory.go +++ b/pkg/sentry/usage/memory.go @@ -122,9 +122,6 @@ func Init() error { const name = "memory-usage" fd, err := memutil.CreateMemFD(name, 0) if err != nil { - if e, ok := err.(syscall.Errno); ok && e == syscall.ENOSYS { - return fmt.Errorf("memfd_create(2) is not implemented. Check that you have Linux 3.17 or higher") - } return fmt.Errorf("error creating usage file: %v", err) } file := os.NewFile(uintptr(fd), name) diff --git a/runsc/boot/BUILD b/runsc/boot/BUILD index daa197437..df9907e52 100644 --- a/runsc/boot/BUILD +++ b/runsc/boot/BUILD @@ -51,6 +51,8 @@ go_library( "//pkg/sentry/kernel/kdefs", "//pkg/sentry/limits", "//pkg/sentry/loader", + "//pkg/sentry/memutil", + "//pkg/sentry/pgalloc", "//pkg/sentry/platform", "//pkg/sentry/platform/kvm", "//pkg/sentry/platform/ptrace", diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index a864be720..14e1eba5b 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -332,6 +332,11 @@ func (cm *containerManager) Restore(o *RestoreOpts, _ *struct{}) error { k := &kernel.Kernel{ Platform: p, } + mf, err := createMemoryFile() + if err != nil { + return fmt.Errorf("creating memory file: %v", err) + } + k.SetMemoryFile(mf) cm.l.k = k // Set up the restore environment. @@ -362,7 +367,7 @@ func (cm *containerManager) Restore(o *RestoreOpts, _ *struct{}) error { loadOpts := state.LoadOpts{ Source: o.FilePayload.Files[0], } - if err := loadOpts.Load(k, p, networkStack); err != nil { + if err := loadOpts.Load(k, networkStack); err != nil { return err } @@ -384,7 +389,7 @@ func (cm *containerManager) Restore(o *RestoreOpts, _ *struct{}) error { cm.l.mu.Lock() eid := execID{cid: o.SandboxID} cm.l.processes = map[execID]*execProcess{ - eid: &execProcess{ + eid: { tg: cm.l.k.GlobalInit(), }, } diff --git a/runsc/boot/events.go b/runsc/boot/events.go index f954b8c0b..717adfedd 100644 --- a/runsc/boot/events.go +++ b/runsc/boot/events.go @@ -68,7 +68,7 @@ func (cm *containerManager) Event(_ *struct{}, out *Event) error { } func (s *Stats) populateMemory(k *kernel.Kernel) { - mem := k.Platform.Memory() + mem := k.MemoryFile() mem.UpdateUsage() _, totalUsage := usage.MemoryAccounting.Copy() s.Memory.Usage = MemoryEntry{ diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 9ebe64dce..56cb137f0 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -37,6 +37,8 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/loader" + "gvisor.googlesource.com/gvisor/pkg/sentry/memutil" + "gvisor.googlesource.com/gvisor/pkg/sentry/pgalloc" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/platform/kvm" "gvisor.googlesource.com/gvisor/pkg/sentry/platform/ptrace" @@ -189,6 +191,13 @@ func New(args Args) (*Loader, error) { Platform: p, } + // Create memory file. + mf, err := createMemoryFile() + if err != nil { + return nil, fmt.Errorf("creating memory file: %v", err) + } + k.SetMemoryFile(mf) + // Create VDSO. // // Pass k as the platform since it is savable, unlike the actual platform. @@ -297,7 +306,7 @@ func New(args Args) (*Loader, error) { stdioFDs: args.StdioFDs, rootProcArgs: procArgs, sandboxID: args.ID, - processes: map[execID]*execProcess{eid: &execProcess{}}, + processes: map[execID]*execProcess{eid: {}}, } // We don't care about child signals; some platforms can generate a @@ -404,6 +413,21 @@ func createPlatform(conf *Config, deviceFD int) (platform.Platform, error) { } } +func createMemoryFile() (*pgalloc.MemoryFile, error) { + const memfileName = "runsc-memory" + memfd, err := memutil.CreateMemFD(memfileName, 0) + if err != nil { + return nil, fmt.Errorf("error creating memfd: %v", err) + } + memfile := os.NewFile(uintptr(memfd), memfileName) + mf, err := pgalloc.NewMemoryFile(memfile) + if err != nil { + memfile.Close() + return nil, fmt.Errorf("error creating pgalloc.MemoryFile: %v", err) + } + return mf, nil +} + // Run runs the root container.. func (l *Loader) Run() error { err := l.run() -- cgit v1.2.3 From cea1dd7d21b976ad5cb145b94be7b1bf879235be Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Mon, 18 Mar 2019 10:47:59 -0700 Subject: Remove racy access to shm fields. PiperOrigin-RevId: 239016776 Change-Id: Ia7af4258e7c69b16a4630a6f3278aa8e6b627746 --- pkg/sentry/kernel/shm/shm.go | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 4525aabf4..a7f0758ec 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -286,7 +286,7 @@ func (r *Registry) remove(s *Shm) { defer s.mu.Unlock() if s.key != linux.IPC_PRIVATE { - panic(fmt.Sprintf("Attempted to remove shm segment %+v from the registry whose key is still associated", s)) + panic(fmt.Sprintf("Attempted to remove shm segment %d (key=%d) from the registry whose key is still associated", s.ID, s.key)) } delete(r.shms, s.ID) -- cgit v1.2.3 From 064fda1a759fa3e73d25da3fd535d256ac8ccfb0 Mon Sep 17 00:00:00 2001 From: Andrei Vagin Date: Wed, 20 Mar 2019 18:39:57 -0700 Subject: gvisor: don't allocate a new credential object on fork A credential object is immutable, so we don't need to copy it for a new task. PiperOrigin-RevId: 239519266 Change-Id: I0632f641fdea9554779ac25d84bee4231d0d18f2 --- pkg/sentry/kernel/task_clone.go | 2 +- pkg/sentry/kernel/task_identity.go | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index 114e7f858..daf974920 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -252,7 +252,7 @@ func (t *Task) Clone(opts *CloneOptions) (ThreadID, *SyscallControl, error) { TaskContext: tc, FSContext: fsc, FDMap: fds, - Credentials: creds.Fork(), + Credentials: creds, Niceness: t.Niceness(), NetworkNamespaced: t.netns, AllowedCPUMask: t.CPUMask(), diff --git a/pkg/sentry/kernel/task_identity.go b/pkg/sentry/kernel/task_identity.go index 8f90ed786..e105eba13 100644 --- a/pkg/sentry/kernel/task_identity.go +++ b/pkg/sentry/kernel/task_identity.go @@ -372,6 +372,7 @@ func (t *Task) DropBoundingCapability(cp linux.Capability) error { if !t.creds.HasCapability(linux.CAP_SETPCAP) { return syserror.EPERM } + t.creds = t.creds.Fork() // See doc for creds. t.creds.BoundingCaps &^= auth.CapabilitySetOf(cp) return nil } -- cgit v1.2.3 From 3d0b960112c94379e4974fd9b60d4632548a4389 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Fri, 22 Mar 2019 08:54:30 -0700 Subject: Implement PTRACE_SEIZE, PTRACE_INTERRUPT, and PTRACE_LISTEN. PiperOrigin-RevId: 239803092 Change-Id: I42d612ed6a889e011e8474538958c6de90c6fcab --- pkg/sentry/kernel/ptrace.go | 242 +++++++++++++++++++++++++++----------- pkg/sentry/kernel/sessions.go | 2 +- pkg/sentry/kernel/task.go | 42 +++++-- pkg/sentry/kernel/task_exit.go | 26 +--- pkg/sentry/kernel/task_signals.go | 220 ++++++++++++++++++++-------------- pkg/sentry/kernel/thread_group.go | 47 ++++---- test/syscalls/linux/ptrace.cc | 135 +++++++++++++++++++-- 7 files changed, 496 insertions(+), 218 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index fa7a0d141..e8043bf8a 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -193,6 +193,10 @@ type ptraceStop struct { // If frozen is true, the stopped task's tracer is currently operating on // it, so Task.Kill should not remove the stop. frozen bool + + // If listen is true, the stopped task's tracer invoked PTRACE_LISTEN, so + // ptraceFreeze should fail. + listen bool } // Killable implements TaskStop.Killable. @@ -216,11 +220,11 @@ func (t *Task) beginPtraceStopLocked() bool { // is what prevents tasks from entering ptrace-stops after being killed. // Note that if t was SIGKILLed and beingPtraceStopLocked is being called // for PTRACE_EVENT_EXIT, the task will have dequeued the signal before - // entering the exit path, so t.killable() will no longer return true. This - // is consistent with Linux: "Bugs: ... A SIGKILL signal may still cause a - // PTRACE_EVENT_EXIT stop before actual signal death. This may be changed - // in the future; SIGKILL is meant to always immediately kill tasks even - // under ptrace. Last confirmed on Linux 3.13." - ptrace(2) + // entering the exit path, so t.killedLocked() will no longer return true. + // This is consistent with Linux: "Bugs: ... A SIGKILL signal may still + // cause a PTRACE_EVENT_EXIT stop before actual signal death. This may be + // changed in the future; SIGKILL is meant to always immediately kill tasks + // even under ptrace. Last confirmed on Linux 3.13." - ptrace(2) if t.killedLocked() { return false } @@ -230,6 +234,10 @@ func (t *Task) beginPtraceStopLocked() bool { // Preconditions: The TaskSet mutex must be locked. func (t *Task) ptraceTrapLocked(code int32) { + // This is unconditional in ptrace_stop(). + t.tg.signalHandlers.mu.Lock() + t.trapStopPending = false + t.tg.signalHandlers.mu.Unlock() t.ptraceCode = code t.ptraceSiginfo = &arch.SignalInfo{ Signo: int32(linux.SIGTRAP), @@ -260,6 +268,9 @@ func (t *Task) ptraceFreeze() bool { if !ok { return false } + if s.listen { + return false + } s.frozen = true return true } @@ -273,6 +284,12 @@ func (t *Task) ptraceUnfreeze() { // preventing its thread group from completing execve. t.tg.signalHandlers.mu.Lock() defer t.tg.signalHandlers.mu.Unlock() + t.ptraceUnfreezeLocked() +} + +// Preconditions: t must be in a frozen ptraceStop. t's signal mutex must be +// locked. +func (t *Task) ptraceUnfreezeLocked() { // Do this even if the task has been killed to ensure a panic if t.stop is // nil or not a ptraceStop. t.stop.(*ptraceStop).frozen = false @@ -336,8 +353,9 @@ func (t *Task) ptraceTraceme() error { return nil } -// ptraceAttach implements ptrace(PTRACE_ATTACH, target). t is the caller. -func (t *Task) ptraceAttach(target *Task) error { +// ptraceAttach implements ptrace(PTRACE_ATTACH, target) if seize is false, and +// ptrace(PTRACE_SEIZE, target, 0, opts) if seize is true. t is the caller. +func (t *Task) ptraceAttach(target *Task, seize bool, opts uintptr) error { if t.tg == target.tg { return syserror.EPERM } @@ -355,19 +373,31 @@ func (t *Task) ptraceAttach(target *Task) error { if target.exitState >= TaskExitZombie { return syserror.EPERM } + if seize { + if err := t.ptraceSetOptionsLocked(opts); err != nil { + return syserror.EIO + } + } target.ptraceTracer.Store(t) t.ptraceTracees[target] = struct{}{} + target.ptraceSeized = seize target.tg.signalHandlers.mu.Lock() - target.sendSignalLocked(&arch.SignalInfo{ - Signo: int32(linux.SIGSTOP), - Code: arch.SignalInfoUser, - }, false /* group */) + // "Unlike PTRACE_ATTACH, PTRACE_SEIZE does not stop the process." - + // ptrace(2) + if !seize { + target.sendSignalLocked(&arch.SignalInfo{ + Signo: int32(linux.SIGSTOP), + Code: arch.SignalInfoUser, + }, false /* group */) + } // Undocumented Linux feature: If the tracee is already group-stopped (and // consequently will not report the SIGSTOP just sent), force it to leave // and re-enter the stop so that it will switch to a ptrace-stop. if target.stop == (*groupStop)(nil) { - target.groupStopRequired = true + target.trapStopPending = true target.endInternalStopLocked() + // TODO: Linux blocks ptrace_attach() until the task has + // entered the ptrace-stop (or exited) via JOBCTL_TRAPPING. } target.tg.signalHandlers.mu.Unlock() return nil @@ -418,6 +448,7 @@ func (t *Task) exitPtrace() { // // Preconditions: The TaskSet mutex must be locked for writing. func (t *Task) forgetTracerLocked() { + t.ptraceSeized = false t.ptraceOpts = ptraceOptions{} t.ptraceSyscallMode = ptraceSyscallNone t.ptraceSinglestep = false @@ -426,21 +457,25 @@ func (t *Task) forgetTracerLocked() { t.exitTracerAcked = true t.exitNotifyLocked(true) } - // If t is ptrace-stopped, but its thread group is in a group stop and t is - // eligible to participate, make it do so. This is essentially the reverse - // of the special case in ptraceAttach, which converts a group stop to a - // ptrace stop. ("Handling of restart from group-stop is currently buggy, - // but the "as planned" behavior is to leave tracee stopped and waiting for - // SIGCONT." - ptrace(2)) t.tg.signalHandlers.mu.Lock() defer t.tg.signalHandlers.mu.Unlock() - if t.stop == nil { - return + // Unset t.trapStopPending, which might have been set by PTRACE_INTERRUPT. If + // it wasn't, it will be reset via t.groupStopPending after the following. + t.trapStopPending = false + // If t's thread group is in a group stop and t is eligible to participate, + // make it do so. This is essentially the reverse of the special case in + // ptraceAttach, which converts a group stop to a ptrace stop. ("Handling + // of restart from group-stop is currently buggy, but the "as planned" + // behavior is to leave tracee stopped and waiting for SIGCONT." - + // ptrace(2)) + if (t.tg.groupStopComplete || t.tg.groupStopPendingCount != 0) && !t.groupStopPending && t.exitState < TaskExitInitiated { + t.groupStopPending = true + // t already participated in the group stop when it unset + // groupStopPending. + t.groupStopAcknowledged = true + t.interrupt() } if _, ok := t.stop.(*ptraceStop); ok { - if t.exitState < TaskExitInitiated && t.tg.groupStopPhase >= groupStopInitiated { - t.groupStopRequired = true - } t.endInternalStopLocked() } } @@ -460,9 +495,9 @@ func (t *Task) ptraceSignalLocked(info *arch.SignalInfo) bool { // The tracer might change this signal into a stop signal, in which case // any SIGCONT received after the signal was originally dequeued should // cancel it. This is consistent with Linux. - if t.tg.groupStopPhase == groupStopNone { - t.tg.groupStopPhase = groupStopDequeued - } + t.tg.groupStopDequeued = true + // This is unconditional in ptrace_stop(). + t.trapStopPending = false // Can't lock the TaskSet mutex while holding a signal mutex. t.tg.signalHandlers.mu.Unlock() defer t.tg.signalHandlers.mu.Lock() @@ -612,22 +647,27 @@ func (t *Task) ptraceClone(kind ptraceCloneKind, child *Task, opts *CloneOptions if tracer != nil { child.ptraceTracer.Store(tracer) tracer.ptraceTracees[child] = struct{}{} + // "The "seized" behavior ... is inherited by children that are + // automatically attached using PTRACE_O_TRACEFORK, + // PTRACE_O_TRACEVFORK, and PTRACE_O_TRACECLONE." - ptrace(2) + child.ptraceSeized = t.ptraceSeized // "Flags are inherited by new tracees created and "auto-attached" // via active PTRACE_O_TRACEFORK, PTRACE_O_TRACEVFORK, or - // PTRACE_O_TRACECLONE options." + // PTRACE_O_TRACECLONE options." - ptrace(2) child.ptraceOpts = t.ptraceOpts child.tg.signalHandlers.mu.Lock() - // If the child is PT_SEIZED (currently not possible in the sentry - // because PTRACE_SEIZE is unimplemented, but for future - // reference), Linux just sets JOBCTL_TRAP_STOP instead, so the - // child skips signal-delivery-stop and goes directly to - // group-stop. - // - // The child will self-t.interrupt() when its task goroutine starts + // "PTRACE_SEIZE: ... Automatically attached children stop with + // PTRACE_EVENT_STOP and WSTOPSIG(status) returns SIGTRAP instead + // of having SIGSTOP signal delivered to them." - ptrace(2) + if child.ptraceSeized { + child.trapStopPending = true + } else { + child.pendingSignals.enqueue(&arch.SignalInfo{ + Signo: int32(linux.SIGSTOP), + }, nil) + } + // The child will self-interrupt() when its task goroutine starts // running, so we don't have to. - child.pendingSignals.enqueue(&arch.SignalInfo{ - Signo: int32(linux.SIGSTOP), - }, nil) child.tg.signalHandlers.mu.Unlock() } } @@ -681,6 +721,9 @@ func (t *Task) ptraceExec(oldTID ThreadID) { // Employing PTRACE_GETSIGINFO for this signal returns si_code set to 0 // (SI_USER). This signal may be blocked by signal mask, and thus may be // delivered (much) later." - ptrace(2) + if t.ptraceSeized { + return + } t.tg.signalHandlers.mu.Lock() defer t.tg.signalHandlers.mu.Unlock() t.sendSignalLocked(&arch.SignalInfo{ @@ -749,6 +792,57 @@ func (t *Task) ptraceKill(target *Task) error { return nil } +func (t *Task) ptraceInterrupt(target *Task) error { + t.tg.pidns.owner.mu.Lock() + defer t.tg.pidns.owner.mu.Unlock() + if target.Tracer() != t { + return syserror.ESRCH + } + if !target.ptraceSeized { + return syserror.EIO + } + target.tg.signalHandlers.mu.Lock() + defer target.tg.signalHandlers.mu.Unlock() + if target.killedLocked() || target.exitState >= TaskExitInitiated { + return nil + } + target.trapStopPending = true + if s, ok := target.stop.(*ptraceStop); ok && s.listen { + target.endInternalStopLocked() + } + target.interrupt() + return nil +} + +// Preconditions: The TaskSet mutex must be locked for writing. t must have a +// tracer. +func (t *Task) ptraceSetOptionsLocked(opts uintptr) error { + const valid = uintptr(linux.PTRACE_O_EXITKILL | + linux.PTRACE_O_TRACESYSGOOD | + linux.PTRACE_O_TRACECLONE | + linux.PTRACE_O_TRACEEXEC | + linux.PTRACE_O_TRACEEXIT | + linux.PTRACE_O_TRACEFORK | + linux.PTRACE_O_TRACESECCOMP | + linux.PTRACE_O_TRACEVFORK | + linux.PTRACE_O_TRACEVFORKDONE) + if opts&^valid != 0 { + return syserror.EINVAL + } + t.ptraceOpts = ptraceOptions{ + ExitKill: opts&linux.PTRACE_O_EXITKILL != 0, + SysGood: opts&linux.PTRACE_O_TRACESYSGOOD != 0, + TraceClone: opts&linux.PTRACE_O_TRACECLONE != 0, + TraceExec: opts&linux.PTRACE_O_TRACEEXEC != 0, + TraceExit: opts&linux.PTRACE_O_TRACEEXIT != 0, + TraceFork: opts&linux.PTRACE_O_TRACEFORK != 0, + TraceSeccomp: opts&linux.PTRACE_O_TRACESECCOMP != 0, + TraceVfork: opts&linux.PTRACE_O_TRACEVFORK != 0, + TraceVforkDone: opts&linux.PTRACE_O_TRACEVFORKDONE != 0, + } + return nil +} + // Ptrace implements the ptrace system call. func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { // PTRACE_TRACEME ignores all other arguments. @@ -762,16 +856,23 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { return syserror.ESRCH } - // PTRACE_ATTACH (and PTRACE_SEIZE, which is unimplemented) do not require - // that target is not already a tracee. - if req == linux.PTRACE_ATTACH { - return t.ptraceAttach(target) + // PTRACE_ATTACH and PTRACE_SEIZE do not require that target is not already + // a tracee. + if req == linux.PTRACE_ATTACH || req == linux.PTRACE_SEIZE { + seize := req == linux.PTRACE_SEIZE + if seize && addr != 0 { + return syserror.EIO + } + return t.ptraceAttach(target, seize, uintptr(data)) } - // PTRACE_KILL (and PTRACE_INTERRUPT, which is unimplemented) require that - // the target is a tracee, but does not require that it is ptrace-stopped. + // PTRACE_KILL and PTRACE_INTERRUPT require that the target is a tracee, + // but does not require that it is ptrace-stopped. if req == linux.PTRACE_KILL { return t.ptraceKill(target) } + if req == linux.PTRACE_INTERRUPT { + return t.ptraceInterrupt(target) + } // All other ptrace requests require that the target is a ptrace-stopped // tracee, and freeze the ptrace-stop so the tracee can be operated on. t.tg.pidns.owner.mu.RLock() @@ -801,6 +902,8 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { t.UninterruptibleSleepFinish(true) // Resuming commands end the ptrace stop, but only if successful. + // PTRACE_LISTEN ends the ptrace stop if trapNotifyPending is already set on the + // target. switch req { case linux.PTRACE_DETACH: if err := t.ptraceDetach(target, linux.Signal(data)); err != nil { @@ -808,37 +911,65 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { return err } return nil + case linux.PTRACE_CONT: if err := target.ptraceUnstop(ptraceSyscallNone, false, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil + case linux.PTRACE_SYSCALL: if err := target.ptraceUnstop(ptraceSyscallIntercept, false, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil + case linux.PTRACE_SINGLESTEP: if err := target.ptraceUnstop(ptraceSyscallNone, true, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil + case linux.PTRACE_SYSEMU: if err := target.ptraceUnstop(ptraceSyscallEmu, false, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil + case linux.PTRACE_SYSEMU_SINGLESTEP: if err := target.ptraceUnstop(ptraceSyscallEmu, true, linux.Signal(data)); err != nil { target.ptraceUnfreeze() return err } return nil + + case linux.PTRACE_LISTEN: + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() + if !target.ptraceSeized { + return syserror.EIO + } + if target.ptraceSiginfo == nil { + return syserror.EIO + } + if target.ptraceSiginfo.Code>>8 != linux.PTRACE_EVENT_STOP { + return syserror.EIO + } + target.tg.signalHandlers.mu.Lock() + defer target.tg.signalHandlers.mu.Unlock() + if target.trapNotifyPending { + target.endInternalStopLocked() + } else { + target.stop.(*ptraceStop).listen = true + target.ptraceUnfreezeLocked() + } + return nil } + // All other ptrace requests expect us to unfreeze the stop. defer target.ptraceUnfreeze() @@ -958,30 +1089,7 @@ func (t *Task) Ptrace(req int64, pid ThreadID, addr, data usermem.Addr) error { case linux.PTRACE_SETOPTIONS: t.tg.pidns.owner.mu.Lock() defer t.tg.pidns.owner.mu.Unlock() - validOpts := uintptr(linux.PTRACE_O_EXITKILL | - linux.PTRACE_O_TRACESYSGOOD | - linux.PTRACE_O_TRACECLONE | - linux.PTRACE_O_TRACEEXEC | - linux.PTRACE_O_TRACEEXIT | - linux.PTRACE_O_TRACEFORK | - linux.PTRACE_O_TRACESECCOMP | - linux.PTRACE_O_TRACEVFORK | - linux.PTRACE_O_TRACEVFORKDONE) - if uintptr(data)&^validOpts != 0 { - return syserror.EINVAL - } - target.ptraceOpts = ptraceOptions{ - ExitKill: data&linux.PTRACE_O_EXITKILL != 0, - SysGood: data&linux.PTRACE_O_TRACESYSGOOD != 0, - TraceClone: data&linux.PTRACE_O_TRACECLONE != 0, - TraceExec: data&linux.PTRACE_O_TRACEEXEC != 0, - TraceExit: data&linux.PTRACE_O_TRACEEXIT != 0, - TraceFork: data&linux.PTRACE_O_TRACEFORK != 0, - TraceSeccomp: data&linux.PTRACE_O_TRACESECCOMP != 0, - TraceVfork: data&linux.PTRACE_O_TRACEVFORK != 0, - TraceVforkDone: data&linux.PTRACE_O_TRACEVFORKDONE != 0, - } - return nil + return target.ptraceSetOptionsLocked(uintptr(data)) case linux.PTRACE_GETEVENTMSG: t.tg.pidns.owner.mu.RLock() diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index 6fd65f2b0..ae6daac60 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -204,7 +204,7 @@ func (pg *ProcessGroup) handleOrphan() { return } tg.signalHandlers.mu.Lock() - if tg.groupStopPhase == groupStopComplete { + if tg.groupStopComplete { hasStopped = true } tg.signalHandlers.mu.Unlock() diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index e9f133c0b..f958aba26 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -133,28 +133,42 @@ type Task struct { // signalStack is exclusive to the task goroutine. signalStack arch.SignalStack - // If groupStopRequired is true, the task should enter a group stop in the - // interrupt path. groupStopRequired is not redundant with - // tg.groupStopPhase != groupStopNone, because ptrace allows tracers to - // resume individual tasks from a group stop without ending the group stop - // as a whole. + // If groupStopPending is true, the task should participate in a group + // stop in the interrupt path. // - // groupStopRequired is analogous to JOBCTL_TRAP_STOP in Linux, except that - // Linux only uses that flag for ptraced tasks. + // groupStopPending is analogous to JOBCTL_STOP_PENDING in Linux. // - // groupStopRequired is protected by the signal mutex. - groupStopRequired bool + // groupStopPending is protected by the signal mutex. + groupStopPending bool // If groupStopAcknowledged is true, the task has already acknowledged that // it is entering the most recent group stop that has been initiated on its - // thread group. groupStopAcknowledged is only meaningful if - // tg.groupStopPhase == groupStopInitiated. + // thread group. // // groupStopAcknowledged is analogous to !JOBCTL_STOP_CONSUME in Linux. // // groupStopAcknowledged is protected by the signal mutex. groupStopAcknowledged bool + // If trapStopPending is true, the task goroutine should enter a + // PTRACE_INTERRUPT-induced stop from the interrupt path. + // + // trapStopPending is analogous to JOBCTL_TRAP_STOP in Linux, except that + // Linux also sets JOBCTL_TRAP_STOP when a ptraced task detects + // JOBCTL_STOP_PENDING. + // + // trapStopPending is protected by the signal mutex. + trapStopPending bool + + // If trapNotifyPending is true, this task is PTRACE_SEIZEd, and a group + // stop has begun or ended since the last time the task entered a + // ptrace-stop from the group-stop path. + // + // trapNotifyPending is analogous to JOBCTL_TRAP_NOTIFY in Linux. + // + // trapNotifyPending is protected by the signal mutex. + trapNotifyPending bool + // If stop is not nil, it is the internally-initiated condition that // currently prevents the task goroutine from running. // @@ -296,6 +310,12 @@ type Task struct { // ptraceTracees is protected by the TaskSet mutex. ptraceTracees map[*Task]struct{} + // ptraceSeized is true if ptraceTracer attached to this task with + // PTRACE_SEIZE. + // + // ptraceSeized is protected by the TaskSet mutex. + ptraceSeized bool + // ptraceOpts contains ptrace options explicitly set by the tracer. If // ptraceTracer is nil, ptraceOpts is expected to be the zero value. // diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index 791cc9831..b9c558ccb 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -122,7 +122,6 @@ func (t *Task) killLocked() { if t.stop != nil && t.stop.Killable() { t.endInternalStopLocked() } - t.groupStopRequired = false t.pendingSignals.enqueue(&arch.SignalInfo{ Signo: int32(linux.SIGKILL), // Linux just sets SIGKILL in the pending signal bitmask without @@ -304,33 +303,16 @@ func (t *Task) exitThreadGroup() bool { t.setSignalMaskLocked(^linux.SignalSet(0)) // Check if this task's exit interacts with an initiated group stop. - if t.tg.groupStopPhase != groupStopInitiated { + if !t.groupStopPending { t.tg.signalHandlers.mu.Unlock() return last } - if t.groupStopAcknowledged { - // Un-acknowledge the group stop. - t.tg.groupStopCount-- - t.groupStopAcknowledged = false - // If the group stop wasn't complete before, then there is still at - // least one other task that hasn't acknowledged the group stop, so - // it is still not complete now. - t.tg.signalHandlers.mu.Unlock() - return last - } - if t.tg.groupStopCount != t.tg.activeTasks { - t.tg.signalHandlers.mu.Unlock() - return last - } - t.Debugf("Completing group stop") - t.tg.groupStopPhase = groupStopComplete - t.tg.groupStopWaitable = true + t.groupStopPending = false sig := t.tg.groupStopSignal - t.tg.groupContNotify = false - t.tg.groupContWaitable = false + notifyParent := t.participateGroupStopLocked() // signalStop must be called with t's signal mutex unlocked. t.tg.signalHandlers.mu.Unlock() - if t.tg.leader.parent != nil { + if notifyParent && t.tg.leader.parent != nil { t.tg.leader.parent.signalStop(t, arch.CLD_STOPPED, int32(sig)) t.tg.leader.parent.tg.eventQueue.Notify(EventChildGroupStop) } diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 583acddb1..6a204aa59 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -748,48 +748,21 @@ type groupStop struct{} // Killable implements TaskStop.Killable. func (*groupStop) Killable() bool { return true } -type groupStopPhase int - -const ( - // groupStopNone indicates that a thread group is not in, or attempting to - // enter or leave, a group stop. - groupStopNone groupStopPhase = iota - - // groupStopDequeued indicates that at least one task in a thread group has - // dequeued a stop signal (or dequeued any signal and entered a - // signal-delivery-stop as a result, which allows ptrace to change the - // signal into a stop signal), but temporarily dropped the signal mutex - // without initiating the group stop. - // - // groupStopDequeued is analogous to JOBCTL_STOP_DEQUEUED in Linux. - groupStopDequeued - - // groupStopInitiated indicates that a task in a thread group has initiated - // a group stop, but not all tasks in the thread group have acknowledged - // entering the group stop. - // - // groupStopInitiated is represented by JOBCTL_STOP_PENDING && - // !SIGNAL_STOP_STOPPED in Linux. - groupStopInitiated - - // groupStopComplete indicates that all tasks in a thread group have - // acknowledged entering the group stop, and the last one to do so has - // notified the thread group's parent. - // - // groupStopComplete is represented by JOBCTL_STOP_PENDING && - // SIGNAL_STOP_STOPPED in Linux. - groupStopComplete -) - // initiateGroupStop attempts to initiate a group stop based on a // previously-dequeued stop signal. // // Preconditions: The caller must be running on the task goroutine. func (t *Task) initiateGroupStop(info *arch.SignalInfo) { + t.tg.pidns.owner.mu.RLock() + defer t.tg.pidns.owner.mu.RUnlock() t.tg.signalHandlers.mu.Lock() defer t.tg.signalHandlers.mu.Unlock() - if t.tg.groupStopPhase != groupStopDequeued { - t.Debugf("Signal %d: not stopping thread group: lost to racing signal", info.Signo) + if t.groupStopPending { + t.Debugf("Signal %d: not stopping thread group: lost to racing stop signal", info.Signo) + return + } + if !t.tg.groupStopDequeued { + t.Debugf("Signal %d: not stopping thread group: lost to racing SIGCONT", info.Signo) return } if t.tg.exiting { @@ -800,15 +773,27 @@ func (t *Task) initiateGroupStop(info *arch.SignalInfo) { t.Debugf("Signal %d: not stopping thread group: lost to racing execve", info.Signo) return } - t.Debugf("Signal %d: stopping thread group", info.Signo) - t.tg.groupStopPhase = groupStopInitiated - t.tg.groupStopSignal = linux.Signal(info.Signo) - t.tg.groupStopCount = 0 + if !t.tg.groupStopComplete { + t.tg.groupStopSignal = linux.Signal(info.Signo) + } + t.tg.groupStopPendingCount = 0 for t2 := t.tg.tasks.Front(); t2 != nil; t2 = t2.Next() { - t2.groupStopRequired = true + if t2.killedLocked() || t2.exitState >= TaskExitInitiated { + t2.groupStopPending = false + continue + } + t2.groupStopPending = true t2.groupStopAcknowledged = false + if t2.ptraceSeized { + t2.trapNotifyPending = true + if s, ok := t2.stop.(*ptraceStop); ok && s.listen { + t2.endInternalStopLocked() + } + } t2.interrupt() + t.tg.groupStopPendingCount++ } + t.Debugf("Signal %d: stopping %d threads in thread group", info.Signo, t.tg.groupStopPendingCount) } // endGroupStopLocked ensures that all prior stop signals received by tg are @@ -820,37 +805,77 @@ func (tg *ThreadGroup) endGroupStopLocked(broadcast bool) { // Discard all previously-queued stop signals. linux.ForEachSignal(StopSignals, tg.discardSpecificLocked) - if tg.groupStopPhase != groupStopNone { - tg.leader.Debugf("Ending group stop currently in phase %d", tg.groupStopPhase) - if tg.groupStopPhase == groupStopInitiated || tg.groupStopPhase == groupStopComplete { - tg.groupStopSignal = 0 - for t := tg.tasks.Front(); t != nil; t = t.Next() { - if _, ok := t.stop.(*groupStop); ok { - t.endInternalStopLocked() - } + if tg.groupStopPendingCount == 0 && !tg.groupStopComplete { + return + } + + completeStr := "incomplete" + if tg.groupStopComplete { + completeStr = "complete" + } + tg.leader.Debugf("Ending %s group stop with %d threads pending", completeStr, tg.groupStopPendingCount) + for t := tg.tasks.Front(); t != nil; t = t.Next() { + t.groupStopPending = false + if t.ptraceSeized { + t.trapNotifyPending = true + if s, ok := t.stop.(*ptraceStop); ok && s.listen { + t.endInternalStopLocked() } - if broadcast { - // Instead of notifying the parent here, set groupContNotify so - // that one of the continuing tasks does so. (Linux does - // something similar.) The reason we do this is to keep locking - // sane. In order to send a signal to the parent, we need to - // lock its signal mutex, but we're already holding tg's signal - // mutex, and the TaskSet mutex must be locked for writing for - // us to hold two signal mutexes. Since we don't want to - // require this for endGroupStopLocked (which is called from - // signal-sending paths), nor do we want to lose atomicity by - // releasing the mutexes we're already holding, just let the - // continuing thread group deal with it. - tg.groupContNotify = true - tg.groupContInterrupted = tg.groupStopPhase == groupStopInitiated - tg.groupContWaitable = true + } else { + if _, ok := t.stop.(*groupStop); ok { + t.endInternalStopLocked() } } - // If groupStopPhase was groupStopDequeued, setting it to groupStopNone - // will cause following calls to initiateGroupStop to recognize that - // the group stop has been cancelled. - tg.groupStopPhase = groupStopNone } + if broadcast { + // Instead of notifying the parent here, set groupContNotify so that + // one of the continuing tasks does so. (Linux does something similar.) + // The reason we do this is to keep locking sane. In order to send a + // signal to the parent, we need to lock its signal mutex, but we're + // already holding tg's signal mutex, and the TaskSet mutex must be + // locked for writing for us to hold two signal mutexes. Since we don't + // want to require this for endGroupStopLocked (which is called from + // signal-sending paths), nor do we want to lose atomicity by releasing + // the mutexes we're already holding, just let the continuing thread + // group deal with it. + tg.groupContNotify = true + tg.groupContInterrupted = !tg.groupStopComplete + tg.groupContWaitable = true + } + // Unsetting groupStopDequeued will cause racing calls to initiateGroupStop + // to recognize that the group stop has been cancelled. + tg.groupStopDequeued = false + tg.groupStopSignal = 0 + tg.groupStopPendingCount = 0 + tg.groupStopComplete = false + tg.groupStopWaitable = false +} + +// participateGroupStopLocked is called to handle thread group side effects +// after t unsets t.groupStopPending. The caller must handle task side effects +// (e.g. placing the task goroutine into the group stop). It returns true if +// the caller must notify t.tg.leader's parent of a completed group stop (which +// participateGroupStopLocked cannot do due to holding the wrong locks). +// +// Preconditions: The signal mutex must be locked. +func (t *Task) participateGroupStopLocked() bool { + if t.groupStopAcknowledged { + return false + } + t.groupStopAcknowledged = true + t.tg.groupStopPendingCount-- + if t.tg.groupStopPendingCount != 0 { + return false + } + if t.tg.groupStopComplete { + return false + } + t.Debugf("Completing group stop") + t.tg.groupStopComplete = true + t.tg.groupStopWaitable = true + t.tg.groupContNotify = false + t.tg.groupContWaitable = false + return true } // signalStop sends a signal to t's thread group of a new group stop, group @@ -899,7 +924,7 @@ func (*runInterrupt) execute(t *Task) taskRunState { // leader's) tracer are in the same thread group, deduplicate // notifications. notifyParent := t.tg.leader.parent != nil - if tracer := t.tg.leader.ptraceTracer.Load().(*Task); tracer != nil { + if tracer := t.tg.leader.Tracer(); tracer != nil { if notifyParent && tracer.tg == t.tg.leader.parent.tg { notifyParent = false } @@ -938,23 +963,21 @@ func (*runInterrupt) execute(t *Task) taskRunState { return (*runInterrupt)(nil) } - // Do we need to enter a group stop? - if t.groupStopRequired { - t.groupStopRequired = false + // Do we need to enter a group stop or related ptrace stop? This path is + // analogous to Linux's kernel/signal.c:get_signal() => do_signal_stop() + // (with ptrace enabled) and do_jobctl_trap(). + if t.groupStopPending || t.trapStopPending || t.trapNotifyPending { sig := t.tg.groupStopSignal notifyParent := false - if !t.groupStopAcknowledged { - t.groupStopAcknowledged = true - t.tg.groupStopCount++ - if t.tg.groupStopCount == t.tg.activeTasks { - t.Debugf("Completing group stop") - notifyParent = true - t.tg.groupStopPhase = groupStopComplete - t.tg.groupStopWaitable = true - t.tg.groupContNotify = false - t.tg.groupContWaitable = false - } + if t.groupStopPending { + t.groupStopPending = false + // We care about t.tg.groupStopSignal (for tracer notification) + // even if this doesn't complete a group stop, so keep the + // value of sig we've already read. + notifyParent = t.participateGroupStopLocked() } + t.trapStopPending = false + t.trapNotifyPending = false // Drop the signal mutex so we can take the TaskSet mutex. t.tg.signalHandlers.mu.Unlock() @@ -963,8 +986,26 @@ func (*runInterrupt) execute(t *Task) taskRunState { notifyParent = false } if tracer := t.Tracer(); tracer != nil { - t.ptraceCode = int32(sig) - t.ptraceSiginfo = nil + if t.ptraceSeized { + if sig == 0 { + sig = linux.SIGTRAP + } + // "If tracee was attached using PTRACE_SEIZE, group-stop is + // indicated by PTRACE_EVENT_STOP: status>>16 == + // PTRACE_EVENT_STOP. This allows detection of group-stops + // without requiring an extra PTRACE_GETSIGINFO call." - + // "Group-stop", ptrace(2) + t.ptraceCode = int32(sig) | linux.PTRACE_EVENT_STOP<<8 + t.ptraceSiginfo = &arch.SignalInfo{ + Signo: int32(sig), + Code: t.ptraceCode, + } + t.ptraceSiginfo.SetPid(int32(t.tg.pidns.tids[t])) + t.ptraceSiginfo.SetUid(int32(t.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow())) + } else { + t.ptraceCode = int32(sig) + t.ptraceSiginfo = nil + } if t.beginPtraceStopLocked() { tracer.signalStop(t, arch.CLD_STOPPED, int32(sig)) // For consistency with Linux, if the parent and tracer are in the @@ -994,12 +1035,11 @@ func (*runInterrupt) execute(t *Task) taskRunState { // Are there signals pending? if info := t.dequeueSignalLocked(t.signalMask); info != nil { - if linux.SignalSetOf(linux.Signal(info.Signo))&StopSignals != 0 && t.tg.groupStopPhase == groupStopNone { - // Indicate that we've dequeued a stop signal before - // unlocking the signal mutex; initiateGroupStop will check - // that the phase hasn't changed (or is at least another - // "stop signal dequeued" phase) after relocking it. - t.tg.groupStopPhase = groupStopDequeued + if linux.SignalSetOf(linux.Signal(info.Signo))&StopSignals != 0 { + // Indicate that we've dequeued a stop signal before unlocking the + // signal mutex; initiateGroupStop will check for races with + // endGroupStopLocked after relocking it. + t.tg.groupStopDequeued = true } if t.ptraceSignalLocked(info) { // Dequeueing the signal action must wait until after the diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index d7652f57c..1b7b74319 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -60,25 +60,35 @@ type ThreadGroup struct { // pendingSignals is protected by the signal mutex. pendingSignals pendingSignals - // groupStopPhase indicates the state of a group stop in progress on the - // thread group, if any. + // If groupStopDequeued is true, a task in the thread group has dequeued a + // stop signal, but has not yet initiated the group stop. // - // groupStopPhase is protected by the signal mutex. - groupStopPhase groupStopPhase + // groupStopDequeued is analogous to Linux's JOBCTL_STOP_DEQUEUED. + // + // groupStopDequeued is protected by the signal mutex. + groupStopDequeued bool // groupStopSignal is the signal that caused a group stop to be initiated. - // groupStopSignal is only meaningful if groupStopPhase is - // groupStopInitiated or groupStopComplete. // // groupStopSignal is protected by the signal mutex. groupStopSignal linux.Signal - // groupStopCount is the number of non-exited tasks in the thread group - // that have acknowledged an initiated group stop. groupStopCount is only - // meaningful if groupStopPhase is groupStopInitiated. + // groupStopPendingCount is the number of active tasks in the thread group + // for which Task.groupStopPending is set. + // + // groupStopPendingCount is analogous to Linux's + // signal_struct::group_stop_count. // - // groupStopCount is protected by the signal mutex. - groupStopCount int + // groupStopPendingCount is protected by the signal mutex. + groupStopPendingCount int + + // If groupStopComplete is true, groupStopPendingCount transitioned from + // non-zero to zero without an intervening SIGCONT. + // + // groupStopComplete is analogous to Linux's SIGNAL_STOP_STOPPED. + // + // groupStopComplete is protected by the signal mutex. + groupStopComplete bool // If groupStopWaitable is true, the thread group is indicating a waitable // group stop event (as defined by EventChildGroupStop). @@ -91,14 +101,9 @@ type ThreadGroup struct { // If groupContNotify is true, then a SIGCONT has recently ended a group // stop on this thread group, and the first task to observe it should - // notify its parent. - // - // groupContNotify is protected by the signal mutex. - groupContNotify bool - - // If groupContNotify is true, groupContInterrupted is true iff SIGCONT - // ended a group stop in phase groupStopInitiated. If groupContNotify is - // false, groupContInterrupted is meaningless. + // notify its parent. groupContInterrupted is true iff SIGCONT ended an + // incomplete group stop. If groupContNotify is false, groupContInterrupted is + // meaningless. // // Analogues in Linux: // @@ -110,7 +115,9 @@ type ThreadGroup struct { // // - !groupContNotify is represented by neither flag being set. // - // groupContInterrupted is protected by the signal mutex. + // groupContNotify and groupContInterrupted are protected by the signal + // mutex. + groupContNotify bool groupContInterrupted bool // If groupContWaitable is true, the thread group is indicating a waitable diff --git a/test/syscalls/linux/ptrace.cc b/test/syscalls/linux/ptrace.cc index 6f1701aef..6d5c425d8 100644 --- a/test/syscalls/linux/ptrace.cc +++ b/test/syscalls/linux/ptrace.cc @@ -44,6 +44,20 @@ namespace testing { namespace { +// PTRACE_GETSIGMASK and PTRACE_SETSIGMASK are not defined until glibc 2.23 +// (fb53a27c5741 "Add new header definitions from Linux 4.4 (plus older ptrace +// definitions)"). +constexpr auto kPtraceGetSigMask = static_cast<__ptrace_request>(0x420a); +constexpr auto kPtraceSetSigMask = static_cast<__ptrace_request>(0x420b); + +// PTRACE_SYSEMU is not defined until glibc 2.27 (c48831d0eebf "linux/x86: sync +// sys/ptrace.h with Linux 4.14 [BZ #22433]"). +constexpr auto kPtraceSysemu = static_cast<__ptrace_request>(31); + +// PTRACE_EVENT_STOP is not defined until glibc 2.26 (3f67d1a7021e "Add Linux +// PTRACE_EVENT_STOP"). +constexpr int kPtraceEventStop = 128; + // Sends sig to the current process with tgkill(2). // // glibc's raise(2) may change the signal mask before sending the signal. These @@ -146,10 +160,6 @@ TEST(PtraceTest, AttachParent_PeekData_PokeData_SignalSuppression) { } TEST(PtraceTest, GetSigMask) { - // doesn't define these until Linux 4.4, even though the features - // were added in 3.11. - constexpr auto kPtraceGetSigMask = static_cast(0x420a); - constexpr auto kPtraceSetSigMask = static_cast(0x420b); // glibc and the Linux kernel define a sigset_t with different sizes. To avoid // creating a kernel_sigset_t and recreating all the modification functions // (sigemptyset, etc), we just hardcode the kernel sigset size. @@ -878,9 +888,7 @@ TEST(PtraceTest, Sysemu_PokeUser) { << " status " << status; // Suppress the SIGSTOP and wait for the child to enter syscall-enter-stop - // for its first exit_group syscall. glibc doesn't necessarily define - // PTRACE_SYSEMU. - constexpr auto kPtraceSysemu = static_cast<__ptrace_request>(31); + // for its first exit_group syscall. ASSERT_THAT(ptrace(kPtraceSysemu, child_pid, 0, 0), SyscallSucceeds()); ASSERT_THAT(waitpid(child_pid, &status, 0), SyscallSucceedsWithValue(child_pid)); @@ -999,6 +1007,119 @@ TEST(PtraceTest, ERESTART_NoRandomSave) { } #endif // defined(__x86_64__) +TEST(PtraceTest, Seize_Interrupt_Listen) { + volatile long child_should_spin = 1; + pid_t const child_pid = fork(); + if (child_pid == 0) { + // In child process. + while (child_should_spin) { + SleepSafe(absl::Seconds(1)); + } + _exit(1); + } + + // In parent process. + ASSERT_THAT(child_pid, SyscallSucceeds()); + + // Attach to the child with PTRACE_SEIZE; doing so should not stop the child. + ASSERT_THAT(ptrace(PTRACE_SEIZE, child_pid, 0, 0), SyscallSucceeds()); + int status; + EXPECT_THAT(waitpid(child_pid, &status, WNOHANG), + SyscallSucceedsWithValue(0)); + + // Stop the child with PTRACE_INTERRUPT. + ASSERT_THAT(ptrace(PTRACE_INTERRUPT, child_pid, 0, 0), SyscallSucceeds()); + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_EQ(SIGTRAP | (kPtraceEventStop << 8), status >> 8); + + // Unset child_should_spin to verify that the child never leaves the spin + // loop. + ASSERT_THAT(ptrace(PTRACE_POKEDATA, child_pid, &child_should_spin, 0), + SyscallSucceeds()); + + // Send SIGSTOP to the child, then resume it, allowing it to proceed to + // signal-delivery-stop. + ASSERT_THAT(kill(child_pid, SIGSTOP), SyscallSucceeds()); + ASSERT_THAT(ptrace(PTRACE_CONT, child_pid, 0, 0), SyscallSucceeds()); + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFSTOPPED(status) && WSTOPSIG(status) == SIGSTOP) + << " status " << status; + + // Release the child from signal-delivery-stop without suppressing the + // SIGSTOP, causing it to enter group-stop. + ASSERT_THAT(ptrace(PTRACE_CONT, child_pid, 0, SIGSTOP), SyscallSucceeds()); + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_EQ(SIGSTOP | (kPtraceEventStop << 8), status >> 8); + + // "The state of the tracee after PTRACE_LISTEN is somewhat of a gray area: it + // is not in any ptrace-stop (ptrace commands won't work on it, and it will + // deliver waitpid(2) notifications), but it also may be considered 'stopped' + // because it is not executing instructions (is not scheduled), and if it was + // in group-stop before PTRACE_LISTEN, it will not respond to signals until + // SIGCONT is received." - ptrace(2). + ASSERT_THAT(ptrace(PTRACE_LISTEN, child_pid, 0, 0), SyscallSucceeds()); + EXPECT_THAT(ptrace(PTRACE_CONT, child_pid, 0, 0), + SyscallFailsWithErrno(ESRCH)); + EXPECT_THAT(waitpid(child_pid, &status, WNOHANG), + SyscallSucceedsWithValue(0)); + EXPECT_THAT(kill(child_pid, SIGTERM), SyscallSucceeds()); + absl::SleepFor(absl::Seconds(1)); + EXPECT_THAT(waitpid(child_pid, &status, WNOHANG), + SyscallSucceedsWithValue(0)); + + // Send SIGCONT to the child, causing it to leave group-stop and re-trap due + // to PTRACE_LISTEN. + EXPECT_THAT(kill(child_pid, SIGCONT), SyscallSucceeds()); + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_EQ(SIGTRAP | (kPtraceEventStop << 8), status >> 8); + + // Detach the child and expect it to exit due to the SIGTERM we sent while + // it was stopped by PTRACE_LISTEN. + ASSERT_THAT(ptrace(PTRACE_DETACH, child_pid, 0, 0), SyscallSucceeds()); + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFSIGNALED(status) && WTERMSIG(status) == SIGTERM) + << " status " << status; +} + +TEST(PtraceTest, Interrupt_Listen_RequireSeize) { + pid_t const child_pid = fork(); + if (child_pid == 0) { + // In child process. + TEST_PCHECK(ptrace(PTRACE_TRACEME, 0, 0, 0) == 0); + MaybeSave(); + raise(SIGSTOP); + _exit(0); + } + // In parent process. + ASSERT_THAT(child_pid, SyscallSucceeds()); + + // Wait for the child to send itself SIGSTOP and enter signal-delivery-stop. + int status; + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFSTOPPED(status) && WSTOPSIG(status) == SIGSTOP) + << " status " << status; + + // PTRACE_INTERRUPT and PTRACE_LISTEN should fail since the child wasn't + // attached with PTRACE_SEIZE, leaving the child in signal-delivery-stop. + EXPECT_THAT(ptrace(PTRACE_INTERRUPT, child_pid, 0, 0), + SyscallFailsWithErrno(EIO)); + EXPECT_THAT(ptrace(PTRACE_LISTEN, child_pid, 0, 0), + SyscallFailsWithErrno(EIO)); + + // Suppress SIGSTOP and detach from the child, expecting it to exit normally. + ASSERT_THAT(ptrace(PTRACE_DETACH, child_pid, 0, 0), SyscallSucceeds()); + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFEXITED(status) && WEXITSTATUS(status) == 0) + << " status " << status; +} + } // namespace } // namespace testing -- cgit v1.2.3 From ddc05e3053e387be9c81aa98c621b6fc92b01000 Mon Sep 17 00:00:00 2001 From: Andrei Vagin Date: Mon, 25 Mar 2019 11:40:49 -0700 Subject: epoll: use ilist:generic_list instead of ilist:ilist ilist:generic_list works faster than ilist:ilist. Here is a beanchmark test to measure performance of epoll_wait, when readyList isn't empty. It shows about 30% better performance with these changes. Benchmark Time(ns) CPU(ns) Iterations Before: BM_EpollAllEvents 46725 46899 14286 After: BM_EpollAllEvents 33167 33300 18919 PiperOrigin-RevId: 240185278 Change-Id: I3e33f9b214db13ab840b91613400525de5b58d18 --- pkg/sentry/kernel/epoll/BUILD | 15 ++++++++++++++- pkg/sentry/kernel/epoll/epoll.go | 17 ++++++++--------- pkg/sentry/kernel/epoll/epoll_state.go | 14 ++++++-------- 3 files changed, 28 insertions(+), 18 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/epoll/BUILD b/pkg/sentry/kernel/epoll/BUILD index 1567d5050..3ac59e13e 100644 --- a/pkg/sentry/kernel/epoll/BUILD +++ b/pkg/sentry/kernel/epoll/BUILD @@ -1,17 +1,30 @@ package(licenses = ["notice"]) +load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +go_template_instance( + name = "epoll_list", + out = "epoll_list.go", + package = "epoll", + prefix = "pollEntry", + template = "//pkg/ilist:generic_list", + types = { + "Element": "*pollEntry", + "Linker": "*pollEntry", + }, +) + go_library( name = "epoll", srcs = [ "epoll.go", + "epoll_list.go", "epoll_state.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/epoll", visibility = ["//pkg/sentry:internal"], deps = [ - "//pkg/ilist", "//pkg/refs", "//pkg/sentry/context", "//pkg/sentry/fs", diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index 502395f18..61c0fb7c5 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -21,7 +21,6 @@ import ( "sync" "syscall" - "gvisor.googlesource.com/gvisor/pkg/ilist" "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" @@ -70,7 +69,7 @@ type FileIdentifier struct { // // +stateify savable type pollEntry struct { - ilist.Entry + pollEntryEntry file *refs.WeakRef `state:"manual"` id FileIdentifier `state:"wait"` userData [2]int32 @@ -84,7 +83,7 @@ type pollEntry struct { // struct, while state framework currently does not support such // in-struct pointers. Instead, EventPoll will properly set this field // in its loading logic. - curList *ilist.List `state:"nosave"` + curList *pollEntryList `state:"nosave"` } // WeakRefGone implements refs.WeakRefUser.WeakRefGone. @@ -133,9 +132,9 @@ type EventPoll struct { // disabledList -- when the entry is disabled. This happens when // a one-shot entry gets delivered via readEvents(). listsMu sync.Mutex `state:"nosave"` - readyList ilist.List - waitingList ilist.List - disabledList ilist.List + readyList pollEntryList + waitingList pollEntryList + disabledList pollEntryList } // cycleMu is used to serialize all the cycle checks. This is only used when @@ -189,7 +188,7 @@ func (e *EventPoll) eventsAvailable() bool { e.listsMu.Lock() for it := e.readyList.Front(); it != nil; { - entry := it.(*pollEntry) + entry := it it = it.Next() // If the entry is ready, we know 'e' has at least one entry @@ -225,14 +224,14 @@ func (e *EventPoll) Readiness(mask waiter.EventMask) waiter.EventMask { // ReadEvents returns up to max available events. func (e *EventPoll) ReadEvents(max int) []Event { - var local ilist.List + var local pollEntryList var ret []Event e.listsMu.Lock() // Go through all entries we believe may be ready. for it := e.readyList.Front(); it != nil && len(ret) < max; { - entry := it.(*pollEntry) + entry := it it = it.Next() // Check the entry's readiness. It it's not really ready, we diff --git a/pkg/sentry/kernel/epoll/epoll_state.go b/pkg/sentry/kernel/epoll/epoll_state.go index 7f3e2004a..f6e3e4825 100644 --- a/pkg/sentry/kernel/epoll/epoll_state.go +++ b/pkg/sentry/kernel/epoll/epoll_state.go @@ -15,7 +15,6 @@ package epoll import ( - "gvisor.googlesource.com/gvisor/pkg/ilist" "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/waiter" ) @@ -33,18 +32,17 @@ func (e *EventPoll) afterLoad() { e.listsMu.Lock() defer e.listsMu.Unlock() - for _, ls := range []*ilist.List{&e.waitingList, &e.readyList, &e.disabledList} { + for _, ls := range []*pollEntryList{&e.waitingList, &e.readyList, &e.disabledList} { for it := ls.Front(); it != nil; it = it.Next() { - it.(*pollEntry).curList = ls + it.curList = ls } } for it := e.waitingList.Front(); it != nil; it = it.Next() { - p := it.(*pollEntry) - if p.id.File.Readiness(p.mask) != 0 { - e.waitingList.Remove(p) - e.readyList.PushBack(p) - p.curList = &e.readyList + if it.id.File.Readiness(it.mask) != 0 { + e.waitingList.Remove(it) + e.readyList.PushBack(it) + it.curList = &e.readyList e.Notify(waiter.EventIn) } } -- cgit v1.2.3 From f3723f805989d0c5956fbb6aa88b1cd9ac20753c Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Mon, 25 Mar 2019 12:41:36 -0700 Subject: Call memmap.Mappable.Translate with more conservative usermem.AccessType. MM.insertPMAsLocked() passes vma.maxPerms to memmap.Mappable.Translate (although it unsets AccessType.Write if the vma is private). This somewhat simplifies handling of pmas, since it means only COW-break needs to replace existing pmas. However, it also means that a MAP_SHARED mapping of a file opened O_RDWR dirties the file, regardless of the mapping's permissions and whether or not the mapping is ever actually written to with I/O that ignores permissions (e.g. ptrace(PTRACE_POKEDATA)). To fix this: - Change the pma-getting path to request only the permissions that are required for the calling access. - Change memmap.Mappable.Translate to take requested permissions, and return allowed permissions. This preserves the existing behavior in the common cases where the memmap.Mappable isn't fsutil.CachingInodeOperations and doesn't care if the translated platform.File pages are written to. - Change the MM.getPMAsLocked path to support permission upgrading of pmas outside of copy-on-write. PiperOrigin-RevId: 240196979 Change-Id: Ie0147c62c1fbc409467a6fa16269a413f3d7d571 --- pkg/sentry/fs/binder/binder.go | 1 + pkg/sentry/fs/fsutil/host_mappable.go | 1 + pkg/sentry/fs/fsutil/inode_cached.go | 19 +- pkg/sentry/fs/tmpfs/inode_file.go | 1 + pkg/sentry/kernel/shm/shm.go | 1 + pkg/sentry/memmap/memmap.go | 17 +- pkg/sentry/mm/address_space.go | 2 +- pkg/sentry/mm/aio_context.go | 1 + pkg/sentry/mm/debug.go | 6 +- pkg/sentry/mm/io.go | 20 +- pkg/sentry/mm/lifecycle.go | 4 +- pkg/sentry/mm/mm.go | 27 +- pkg/sentry/mm/pma.go | 664 +++++++++++++++++----------------- pkg/sentry/mm/special_mappable.go | 1 + pkg/sentry/mm/syscalls.go | 22 +- pkg/sentry/usermem/access_type.go | 9 + 16 files changed, 416 insertions(+), 380 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index 16fb4806f..188353961 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -244,6 +244,7 @@ func (bp *Proc) Translate(ctx context.Context, required, optional memmap.Mappabl Source: memmap.MappableRange{0, usermem.PageSize}, File: bp.mfp.MemoryFile(), Offset: bp.mapped.Start, + Perms: usermem.AnyAccess, }, }, err } diff --git a/pkg/sentry/fs/fsutil/host_mappable.go b/pkg/sentry/fs/fsutil/host_mappable.go index 1bb5c6b6e..4a182baa1 100644 --- a/pkg/sentry/fs/fsutil/host_mappable.go +++ b/pkg/sentry/fs/fsutil/host_mappable.go @@ -94,6 +94,7 @@ func (h *HostMappable) Translate(ctx context.Context, required, optional memmap. Source: optional, File: h, Offset: optional.Start, + Perms: usermem.AnyAccess, }, }, nil } diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index 9bd923678..6ca51ab0d 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -739,6 +739,7 @@ func (c *CachingInodeOperations) Translate(ctx context.Context, required, option Source: optional, File: c, Offset: optional.Start, + Perms: usermem.AnyAccess, }, }, nil } @@ -768,16 +769,24 @@ func (c *CachingInodeOperations) Translate(ctx context.Context, required, option var translatedEnd uint64 for seg := c.cache.FindSegment(required.Start); seg.Ok() && seg.Start() < required.End; seg, _ = seg.NextNonEmpty() { segMR := seg.Range().Intersect(optional) - ts = append(ts, memmap.Translation{ - Source: segMR, - File: mf, - Offset: seg.FileRangeOf(segMR).Start, - }) + // TODO: Make Translations writable even if writability is + // not required if already kept-dirty by another writable translation. + perms := usermem.AccessType{ + Read: true, + Execute: true, + } if at.Write { // From this point forward, this memory can be dirtied through the // mapping at any time. c.dirty.KeepDirty(segMR) + perms.Write = true } + ts = append(ts, memmap.Translation{ + Source: segMR, + File: mf, + Offset: seg.FileRangeOf(segMR).Start, + Perms: perms, + }) translatedEnd = segMR.End } diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 3c84b2977..25bf2b9dd 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -481,6 +481,7 @@ func (f *fileInodeOperations) Translate(ctx context.Context, required, optional Source: segMR, File: mf, Offset: seg.FileRangeOf(segMR).Start, + Perms: usermem.AnyAccess, }) translatedEnd = segMR.End } diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index a7f0758ec..2b291e4f8 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -455,6 +455,7 @@ func (s *Shm) Translate(ctx context.Context, required, optional memmap.MappableR Source: source, File: s.mfp.MemoryFile(), Offset: s.fr.Start + source.Start, + Perms: usermem.AnyAccess, }, }, err } diff --git a/pkg/sentry/memmap/memmap.go b/pkg/sentry/memmap/memmap.go index 70cdf428b..1ef1f0dd8 100644 --- a/pkg/sentry/memmap/memmap.go +++ b/pkg/sentry/memmap/memmap.go @@ -105,13 +105,22 @@ type Translation struct { // Offset is the offset into File at which this Translation begins. Offset uint64 + + // Perms is the set of permissions for which platform.AddressSpace.MapFile + // and platform.AddressSpace.MapInternal on this Translation is permitted. + Perms usermem.AccessType +} + +// FileRange returns the platform.FileRange represented by t. +func (t Translation) FileRange() platform.FileRange { + return platform.FileRange{t.Offset, t.Offset + t.Source.Length()} } // CheckTranslateResult returns an error if (ts, terr) does not satisfy all -// postconditions for Mappable.Translate(required, optional). +// postconditions for Mappable.Translate(required, optional, at). // // Preconditions: As for Mappable.Translate. -func CheckTranslateResult(required, optional MappableRange, ts []Translation, terr error) error { +func CheckTranslateResult(required, optional MappableRange, at usermem.AccessType, ts []Translation, terr error) error { // Verify that the inputs to Mappable.Translate were valid. if !required.WellFormed() || required.Length() <= 0 { panic(fmt.Sprintf("invalid required range: %v", required)) @@ -156,6 +165,10 @@ func CheckTranslateResult(required, optional MappableRange, ts []Translation, te if !optional.IsSupersetOf(t.Source) { return fmt.Errorf("Translation %+v lies outside optional range %v", t, optional) } + // Each Translation must permit a superset of requested accesses. + if !t.Perms.SupersetOf(at) { + return fmt.Errorf("Translation %+v does not permit all requested access types %v", t, at) + } } // If the set of Translations does not cover the entire required range, // Translate must return a non-nil error explaining why. diff --git a/pkg/sentry/mm/address_space.go b/pkg/sentry/mm/address_space.go index 90cfef746..4dddcf7b5 100644 --- a/pkg/sentry/mm/address_space.go +++ b/pkg/sentry/mm/address_space.go @@ -179,7 +179,7 @@ func (mm *MemoryManager) mapASLocked(pseg pmaIterator, ar usermem.AddrRange, pre pma := pseg.ValuePtr() pmaAR := pseg.Range() pmaMapAR := pmaAR.Intersect(mapAR) - perms := pma.vmaEffectivePerms + perms := pma.effectivePerms if pma.needCOW { perms.Write = false } diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index 6cec6387a..f7ff06de0 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -302,6 +302,7 @@ func (m *aioMappable) Translate(ctx context.Context, required, optional memmap.M Source: source, File: m.mfp.MemoryFile(), Offset: m.fr.Start + source.Start, + Perms: usermem.AnyAccess, }, }, err } diff --git a/pkg/sentry/mm/debug.go b/pkg/sentry/mm/debug.go index d341b9c07..d075ee1ca 100644 --- a/pkg/sentry/mm/debug.go +++ b/pkg/sentry/mm/debug.go @@ -68,12 +68,12 @@ func (pseg pmaIterator) debugStringEntryLocked() []byte { fmt.Fprintf(&b, "%08x-%08x ", pseg.Start(), pseg.End()) pma := pseg.ValuePtr() - if pma.vmaEffectivePerms.Read { + if pma.effectivePerms.Read { b.WriteByte('r') } else { b.WriteByte('-') } - if pma.vmaEffectivePerms.Write { + if pma.effectivePerms.Write { if pma.needCOW { b.WriteByte('c') } else { @@ -82,7 +82,7 @@ func (pseg pmaIterator) debugStringEntryLocked() []byte { } else { b.WriteByte('-') } - if pma.vmaEffectivePerms.Execute { + if pma.effectivePerms.Execute { b.WriteByte('x') } else { b.WriteByte('-') diff --git a/pkg/sentry/mm/io.go b/pkg/sentry/mm/io.go index e0cebef84..81787a6fd 100644 --- a/pkg/sentry/mm/io.go +++ b/pkg/sentry/mm/io.go @@ -466,9 +466,7 @@ func (mm *MemoryManager) handleASIOFault(ctx context.Context, addr usermem.Addr, // Ensure that we have usable pmas. mm.activeMu.Lock() - pseg, pend, err := mm.getPMAsLocked(ctx, vseg, ar, pmaOpts{ - breakCOW: at.Write, - }) + pseg, pend, err := mm.getPMAsLocked(ctx, vseg, ar, at) mm.mappingMu.RUnlock() if pendaddr := pend.Start(); pendaddr < ar.End { if pendaddr <= ar.Start { @@ -498,14 +496,10 @@ func (mm *MemoryManager) handleASIOFault(ctx context.Context, addr usermem.Addr, // // Preconditions: 0 < ar.Length() <= math.MaxInt64. func (mm *MemoryManager) withInternalMappings(ctx context.Context, ar usermem.AddrRange, at usermem.AccessType, ignorePermissions bool, f func(safemem.BlockSeq) (uint64, error)) (int64, error) { - po := pmaOpts{ - breakCOW: at.Write, - } - // If pmas are already available, we can do IO without touching mm.vmas or // mm.mappingMu. mm.activeMu.RLock() - if pseg := mm.existingPMAsLocked(ar, at, ignorePermissions, po, true /* needInternalMappings */); pseg.Ok() { + if pseg := mm.existingPMAsLocked(ar, at, ignorePermissions, true /* needInternalMappings */); pseg.Ok() { n, err := f(mm.internalMappingsLocked(pseg, ar)) mm.activeMu.RUnlock() // Do not convert errors returned by f to EFAULT. @@ -526,7 +520,7 @@ func (mm *MemoryManager) withInternalMappings(ctx context.Context, ar usermem.Ad // Ensure that we have usable pmas. mm.activeMu.Lock() - pseg, pend, perr := mm.getPMAsLocked(ctx, vseg, ar, po) + pseg, pend, perr := mm.getPMAsLocked(ctx, vseg, ar, at) mm.mappingMu.RUnlock() if pendaddr := pend.Start(); pendaddr < ar.End { if pendaddr <= ar.Start { @@ -578,14 +572,10 @@ func (mm *MemoryManager) withVecInternalMappings(ctx context.Context, ars userme return mm.withInternalMappings(ctx, ars.Head(), at, ignorePermissions, f) } - po := pmaOpts{ - breakCOW: at.Write, - } - // If pmas are already available, we can do IO without touching mm.vmas or // mm.mappingMu. mm.activeMu.RLock() - if mm.existingVecPMAsLocked(ars, at, ignorePermissions, po, true /* needInternalMappings */) { + if mm.existingVecPMAsLocked(ars, at, ignorePermissions, true /* needInternalMappings */) { n, err := f(mm.vecInternalMappingsLocked(ars)) mm.activeMu.RUnlock() // Do not convert errors returned by f to EFAULT. @@ -603,7 +593,7 @@ func (mm *MemoryManager) withVecInternalMappings(ctx context.Context, ars userme // Ensure that we have usable pmas. mm.activeMu.Lock() - pars, perr := mm.getVecPMAsLocked(ctx, vars, po) + pars, perr := mm.getVecPMAsLocked(ctx, vars, at) mm.mappingMu.RUnlock() if pars.NumBytes() == 0 { mm.activeMu.Unlock() diff --git a/pkg/sentry/mm/lifecycle.go b/pkg/sentry/mm/lifecycle.go index a71286f14..2fe03172c 100644 --- a/pkg/sentry/mm/lifecycle.go +++ b/pkg/sentry/mm/lifecycle.go @@ -124,7 +124,7 @@ func (mm *MemoryManager) Fork(ctx context.Context) (*MemoryManager, error) { } if !pma.needCOW { pma.needCOW = true - if pma.vmaEffectivePerms.Write { + if pma.effectivePerms.Write { // We don't want to unmap the whole address space, even though // doing so would reduce calls to unmapASLocked(), because mm // will most likely continue to be used after the fork, so @@ -139,7 +139,9 @@ func (mm *MemoryManager) Fork(ctx context.Context) (*MemoryManager, error) { } unmapAR = srcpseg.Range() } + pma.effectivePerms.Write = false } + pma.maxPerms.Write = false } fr := srcpseg.fileRange() mm2.incPrivateRef(fr) diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index 6ed838d64..a3417a46e 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -71,9 +71,6 @@ type MemoryManager struct { // ownership is shared by one or more pmas instead of being owned by a // memmap.Mappable). // - // NOTE: This should be replaced using refcounts on - // platform.File. - // // privateRefs is immutable. privateRefs *privateRefs @@ -374,13 +371,27 @@ type pma struct { file platform.File `state:"nosave"` // off is the offset into file at which this pma begins. + // + // Note that pmas do *not* hold references on offsets in file! If private + // is true, MemoryManager.privateRefs holds the reference instead. If + // private is false, the corresponding memmap.Mappable holds the reference + // instead (per memmap.Mappable.Translate requirement). off uint64 - // vmaEffectivePerms and vmaMaxPerms are duplicated from the - // corresponding vma so that the IO implementation can avoid iterating - // mm.vmas when pmas already exist. - vmaEffectivePerms usermem.AccessType - vmaMaxPerms usermem.AccessType + // translatePerms is the permissions returned by memmap.Mappable.Translate. + // If private is true, translatePerms is usermem.AnyAccess. + translatePerms usermem.AccessType + + // effectivePerms is the permissions allowed for non-ignorePermissions + // accesses. maxPerms is the permissions allowed for ignorePermissions + // accesses. These are vma.effectivePerms and vma.maxPerms respectively, + // masked by pma.translatePerms and with Write disallowed if pma.needCOW is + // true. + // + // These are stored in the pma so that the IO implementation can avoid + // iterating mm.vmas when pmas already exist. + effectivePerms usermem.AccessType + maxPerms usermem.AccessType // needCOW is true if writes to the mapping must be propagated to a copy. needCOW bool diff --git a/pkg/sentry/mm/pma.go b/pkg/sentry/mm/pma.go index bb779a45b..e090537cc 100644 --- a/pkg/sentry/mm/pma.go +++ b/pkg/sentry/mm/pma.go @@ -27,18 +27,13 @@ import ( "gvisor.googlesource.com/gvisor/pkg/syserror" ) -type pmaOpts struct { - // If breakCOW is true, pmas must not be copy-on-write. - breakCOW bool -} - // existingPMAsLocked checks that pmas exist for all addresses in ar, and // support access of type (at, ignorePermissions). If so, it returns an // iterator to the pma containing ar.Start. Otherwise it returns a terminal // iterator. // // Preconditions: mm.activeMu must be locked. ar.Length() != 0. -func (mm *MemoryManager) existingPMAsLocked(ar usermem.AddrRange, at usermem.AccessType, ignorePermissions bool, opts pmaOpts, needInternalMappings bool) pmaIterator { +func (mm *MemoryManager) existingPMAsLocked(ar usermem.AddrRange, at usermem.AccessType, ignorePermissions bool, needInternalMappings bool) pmaIterator { if checkInvariants { if !ar.WellFormed() || ar.Length() <= 0 { panic(fmt.Sprintf("invalid ar: %v", ar)) @@ -49,16 +44,11 @@ func (mm *MemoryManager) existingPMAsLocked(ar usermem.AddrRange, at usermem.Acc pseg := first for pseg.Ok() { pma := pseg.ValuePtr() - perms := pma.vmaEffectivePerms + perms := pma.effectivePerms if ignorePermissions { - perms = pma.vmaMaxPerms + perms = pma.maxPerms } if !perms.SupersetOf(at) { - // These are the vma's permissions, so the caller will get an error - // when they try to get new pmas. - return pmaIterator{} - } - if opts.breakCOW && pma.needCOW { return pmaIterator{} } if needInternalMappings && pma.internalMappings.IsEmpty() { @@ -79,17 +69,17 @@ func (mm *MemoryManager) existingPMAsLocked(ar usermem.AddrRange, at usermem.Acc // and support access of type (at, ignorePermissions). // // Preconditions: mm.activeMu must be locked. -func (mm *MemoryManager) existingVecPMAsLocked(ars usermem.AddrRangeSeq, at usermem.AccessType, ignorePermissions bool, opts pmaOpts, needInternalMappings bool) bool { +func (mm *MemoryManager) existingVecPMAsLocked(ars usermem.AddrRangeSeq, at usermem.AccessType, ignorePermissions bool, needInternalMappings bool) bool { for ; !ars.IsEmpty(); ars = ars.Tail() { - if ar := ars.Head(); ar.Length() != 0 && !mm.existingPMAsLocked(ar, at, ignorePermissions, opts, needInternalMappings).Ok() { + if ar := ars.Head(); ar.Length() != 0 && !mm.existingPMAsLocked(ar, at, ignorePermissions, needInternalMappings).Ok() { return false } } return true } -// getPMAsLocked ensures that pmas exist for all addresses in ar, subject to -// opts. It returns: +// getPMAsLocked ensures that pmas exist for all addresses in ar, and support +// access of type at. It returns: // // - An iterator to the pma containing ar.Start. If no pma contains ar.Start, // the iterator is unspecified. @@ -102,8 +92,9 @@ func (mm *MemoryManager) existingVecPMAsLocked(ars usermem.AddrRangeSeq, at user // // Preconditions: mm.mappingMu must be locked. mm.activeMu must be locked for // writing. ar.Length() != 0. vseg.Range().Contains(ar.Start). vmas must exist -// for all addresses in ar. -func (mm *MemoryManager) getPMAsLocked(ctx context.Context, vseg vmaIterator, ar usermem.AddrRange, opts pmaOpts) (pmaIterator, pmaGapIterator, error) { +// for all addresses in ar, and support accesses of type at (i.e. permission +// checks must have been performed against vmas). +func (mm *MemoryManager) getPMAsLocked(ctx context.Context, vseg vmaIterator, ar usermem.AddrRange, at usermem.AccessType) (pmaIterator, pmaGapIterator, error) { if checkInvariants { if !ar.WellFormed() || ar.Length() <= 0 { panic(fmt.Sprintf("invalid ar: %v", ar)) @@ -125,54 +116,40 @@ func (mm *MemoryManager) getPMAsLocked(ctx context.Context, vseg vmaIterator, ar } ar = usermem.AddrRange{ar.Start.RoundDown(), end} - pstart, pend, perr := mm.ensurePMAsLocked(ctx, vseg, ar) + pstart, pend, perr := mm.getPMAsInternalLocked(ctx, vseg, ar, at) if pend.Start() <= ar.Start { return pmaIterator{}, pend, perr } - // ensurePMAsLocked may not have pstart due to iterator invalidation. We - // need it, either to return it immediately or to pass to - // breakCopyOnWriteLocked. + // getPMAsInternalLocked may not have returned pstart due to iterator + // invalidation. if !pstart.Ok() { pstart = mm.findOrSeekPrevUpperBoundPMA(ar.Start, pend) } - - var cowerr error - if opts.breakCOW { - if pend.Start() < ar.End { - // Adjust ar to reflect missing pmas. - ar.End = pend.Start() - } - var invalidated bool - pend, invalidated, cowerr = mm.breakCopyOnWriteLocked(pstart, ar) - if pend.Start() <= ar.Start { - return pmaIterator{}, pend, cowerr - } - if invalidated { - pstart = mm.findOrSeekPrevUpperBoundPMA(ar.Start, pend) - } - } - - if cowerr != nil { - return pstart, pend, cowerr - } if perr != nil { return pstart, pend, perr } return pstart, pend, alignerr } -// getVecPMAsLocked ensures that pmas exist for all addresses in ars. It -// returns the subset of ars for which pmas exist. If this is not equal to ars, -// it returns a non-nil error explaining why. +// getVecPMAsLocked ensures that pmas exist for all addresses in ars, and +// support access of type at. It returns the subset of ars for which pmas +// exist. If this is not equal to ars, it returns a non-nil error explaining +// why. // // Preconditions: mm.mappingMu must be locked. mm.activeMu must be locked for -// writing. vmas must exist for all addresses in ars. -func (mm *MemoryManager) getVecPMAsLocked(ctx context.Context, ars usermem.AddrRangeSeq, opts pmaOpts) (usermem.AddrRangeSeq, error) { +// writing. vmas must exist for all addresses in ars, and support accesses of +// type at (i.e. permission checks must have been performed against vmas). +func (mm *MemoryManager) getVecPMAsLocked(ctx context.Context, ars usermem.AddrRangeSeq, at usermem.AccessType) (usermem.AddrRangeSeq, error) { for arsit := ars; !arsit.IsEmpty(); arsit = arsit.Tail() { ar := arsit.Head() if ar.Length() == 0 { continue } + if checkInvariants { + if !ar.WellFormed() { + panic(fmt.Sprintf("invalid ar: %v", ar)) + } + } // Page-align ar so that all AddrRanges are aligned. end, ok := ar.End.RoundUp() @@ -183,26 +160,7 @@ func (mm *MemoryManager) getVecPMAsLocked(ctx context.Context, ars usermem.AddrR } ar = usermem.AddrRange{ar.Start.RoundDown(), end} - pstart, pend, perr := mm.ensurePMAsLocked(ctx, mm.vmas.FindSegment(ar.Start), ar) - if pend.Start() <= ar.Start { - return truncatedAddrRangeSeq(ars, arsit, pend.Start()), perr - } - - var cowerr error - if opts.breakCOW { - if !pstart.Ok() { - pstart = mm.findOrSeekPrevUpperBoundPMA(ar.Start, pend) - } - if pend.Start() < ar.End { - // Adjust ar to reflect missing pmas. - ar.End = pend.Start() - } - pend, _, cowerr = mm.breakCopyOnWriteLocked(pstart, ar) - } - - if cowerr != nil { - return truncatedAddrRangeSeq(ars, arsit, pend.Start()), cowerr - } + _, pend, perr := mm.getPMAsInternalLocked(ctx, mm.vmas.FindSegment(ar.Start), ar, at) if perr != nil { return truncatedAddrRangeSeq(ars, arsit, pend.Start()), perr } @@ -214,64 +172,303 @@ func (mm *MemoryManager) getVecPMAsLocked(ctx context.Context, ars usermem.AddrR return ars, nil } -// ensurePMAsLocked ensures that pmas exist for all addresses in ar. It returns: +// getPMAsInternalLocked is equivalent to getPMAsLocked, with the following +// exceptions: // -// - An iterator to the pma containing ar.Start, on a best-effort basis (that -// is, the returned iterator may be terminal, even if such a pma exists). -// Returning this iterator on a best-effort basis allows callers that require -// it to use it when it's cheaply available, while also avoiding the overhead -// of retrieving it when it's not. -// -// - An iterator to the gap after the last pma containing an address in ar. If -// pmas exist for no addresses in ar, the iterator is to a gap that begins -// before ar.Start. +// - getPMAsInternalLocked returns a pmaIterator on a best-effort basis (that +// is, the returned iterator may be terminal, even if a pma that contains +// ar.Start exists). Returning this iterator on a best-effort basis allows +// callers that require it to use it when it's cheaply available, while also +// avoiding the overhead of retrieving it when it's not. // -// - An error that is non-nil if pmas exist for only a subset of ar. +// - getPMAsInternalLocked additionally requires that ar is page-aligned. // -// Preconditions: mm.mappingMu must be locked. mm.activeMu must be locked for -// writing. ar.Length() != 0. ar must be page-aligned. -// vseg.Range().Contains(ar.Start). vmas must exist for all addresses in ar. -func (mm *MemoryManager) ensurePMAsLocked(ctx context.Context, vseg vmaIterator, ar usermem.AddrRange) (pmaIterator, pmaGapIterator, error) { +// getPMAsInternalLocked is an implementation helper for getPMAsLocked and +// getVecPMAsLocked; other clients should call one of those instead. +func (mm *MemoryManager) getPMAsInternalLocked(ctx context.Context, vseg vmaIterator, ar usermem.AddrRange, at usermem.AccessType) (pmaIterator, pmaGapIterator, error) { if checkInvariants { if !ar.WellFormed() || ar.Length() <= 0 || !ar.IsPageAligned() { panic(fmt.Sprintf("invalid ar: %v", ar)) } + if !vseg.Ok() { + panic("terminal vma iterator") + } if !vseg.Range().Contains(ar.Start) { panic(fmt.Sprintf("initial vma %v does not cover start of ar %v", vseg.Range(), ar)) } } - pstart, pgap := mm.pmas.Find(ar.Start) - if pstart.Ok() { - pgap = pstart.NextGap() - } - for pgap.Start() < ar.End { - if pgap.Range().Length() == 0 { - pgap = pgap.NextGap() - continue - } - // A single pgap might be spanned by multiple vmas. Insert pmas to - // cover the first (vma, pgap) pair. - pgapAR := pgap.Range().Intersect(ar) - vseg = vseg.seekNextLowerBound(pgapAR.Start) - if checkInvariants { - if !vseg.Ok() { - panic(fmt.Sprintf("no vma after %#x", pgapAR.Start)) - } - if pgapAR.Start < vseg.Start() { - panic(fmt.Sprintf("no vma in [%#x, %#x)", pgapAR.Start, vseg.Start())) + mf := mm.mfp.MemoryFile() + // Limit the range we allocate to ar, aligned to privateAllocUnit. + maskAR := privateAligned(ar) + didUnmapAS := false + // The range in which we iterate vmas and pmas is still limited to ar, to + // ensure that we don't allocate or COW-break a pma we don't need. + pseg, pgap := mm.pmas.Find(ar.Start) + pstart := pseg + for { + // Get pmas for this vma. + vsegAR := vseg.Range().Intersect(ar) + vma := vseg.ValuePtr() + pmaLoop: + for { + switch { + case pgap.Ok() && pgap.Start() < vsegAR.End: + // Need a pma here. + optAR := vseg.Range().Intersect(pgap.Range()) + if checkInvariants { + if optAR.Length() <= 0 { + panic(fmt.Sprintf("vseg %v and pgap %v do not overlap", vseg, pgap)) + } + } + if vma.mappable == nil { + // Private anonymous mappings get pmas by allocating. + allocAR := optAR.Intersect(maskAR) + fr, err := mf.Allocate(uint64(allocAR.Length()), usage.Anonymous) + if err != nil { + return pstart, pgap, err + } + if checkInvariants { + if !fr.WellFormed() || fr.Length() != uint64(allocAR.Length()) { + panic(fmt.Sprintf("Allocate(%v) returned invalid FileRange %v", allocAR.Length(), fr)) + } + } + mm.addRSSLocked(allocAR) + mm.incPrivateRef(fr) + mf.IncRef(fr) + pseg, pgap = mm.pmas.Insert(pgap, allocAR, pma{ + file: mf, + off: fr.Start, + translatePerms: usermem.AnyAccess, + effectivePerms: vma.effectivePerms, + maxPerms: vma.maxPerms, + // Since we just allocated this memory and have the + // only reference, the new pma does not need + // copy-on-write. + private: true, + }).NextNonEmpty() + pstart = pmaIterator{} // iterators invalidated + } else { + // Other mappings get pmas by translating. + optMR := vseg.mappableRangeOf(optAR) + reqAR := optAR.Intersect(ar) + reqMR := vseg.mappableRangeOf(reqAR) + perms := at + if vma.private { + // This pma will be copy-on-write; don't require write + // permission, but do require read permission to + // facilitate the copy. + // + // If at.Write is true, we will need to break + // copy-on-write immediately, which occurs after + // translation below. + perms.Read = true + perms.Write = false + } + ts, err := vma.mappable.Translate(ctx, reqMR, optMR, perms) + if checkInvariants { + if err := memmap.CheckTranslateResult(reqMR, optMR, perms, ts, err); err != nil { + panic(fmt.Sprintf("Mappable(%T).Translate(%v, %v, %v): %v", vma.mappable, reqMR, optMR, perms, err)) + } + } + // Install a pma for each translation. + if len(ts) == 0 { + return pstart, pgap, err + } + pstart = pmaIterator{} // iterators invalidated + for _, t := range ts { + newpmaAR := vseg.addrRangeOf(t.Source) + newpma := pma{ + file: t.File, + off: t.Offset, + translatePerms: t.Perms, + effectivePerms: vma.effectivePerms.Intersect(t.Perms), + maxPerms: vma.maxPerms.Intersect(t.Perms), + } + if vma.private { + newpma.effectivePerms.Write = false + newpma.maxPerms.Write = false + newpma.needCOW = true + } + mm.addRSSLocked(newpmaAR) + t.File.IncRef(t.FileRange()) + // This is valid because memmap.Mappable.Translate is + // required to return Translations in increasing + // Translation.Source order. + pseg = mm.pmas.Insert(pgap, newpmaAR, newpma) + pgap = pseg.NextGap() + } + // The error returned by Translate is only significant if + // it occurred before ar.End. + if err != nil && vseg.addrRangeOf(ts[len(ts)-1].Source).End < ar.End { + return pstart, pgap, err + } + // Rewind pseg to the first pma inserted and continue the + // loop to check if we need to break copy-on-write. + pseg, pgap = mm.findOrSeekPrevUpperBoundPMA(vseg.addrRangeOf(ts[0].Source).Start, pgap), pmaGapIterator{} + continue + } + + case pseg.Ok() && pseg.Start() < vsegAR.End: + oldpma := pseg.ValuePtr() + if at.Write && mm.isPMACopyOnWriteLocked(vseg, pseg) { + // Break copy-on-write by copying. + if checkInvariants { + if !oldpma.maxPerms.Read { + panic(fmt.Sprintf("pma %v needs to be copied for writing, but is not readable: %v", pseg.Range(), oldpma)) + } + } + copyAR := pseg.Range().Intersect(maskAR) + // Get internal mappings from the pma to copy from. + if err := pseg.getInternalMappingsLocked(); err != nil { + return pstart, pseg.PrevGap(), err + } + // Copy contents. + fr, err := mf.AllocateAndFill(uint64(copyAR.Length()), usage.Anonymous, &safemem.BlockSeqReader{mm.internalMappingsLocked(pseg, copyAR)}) + if _, ok := err.(safecopy.BusError); ok { + // If we got SIGBUS during the copy, deliver SIGBUS to + // userspace (instead of SIGSEGV) if we're breaking + // copy-on-write due to application page fault. + err = &memmap.BusError{err} + } + if fr.Length() == 0 { + return pstart, pseg.PrevGap(), err + } + // Unmap all of maskAR, not just copyAR, to minimize host + // syscalls. AddressSpace mappings must be removed before + // mm.decPrivateRef(). + if !didUnmapAS { + mm.unmapASLocked(maskAR) + didUnmapAS = true + } + // Replace the pma with a copy in the part of the address + // range where copying was successful. This doesn't change + // RSS. + copyAR.End = copyAR.Start + usermem.Addr(fr.Length()) + if copyAR != pseg.Range() { + pseg = mm.pmas.Isolate(pseg, copyAR) + pstart = pmaIterator{} // iterators invalidated + } + oldpma = pseg.ValuePtr() + if oldpma.private { + mm.decPrivateRef(pseg.fileRange()) + } + oldpma.file.DecRef(pseg.fileRange()) + mm.incPrivateRef(fr) + mf.IncRef(fr) + oldpma.file = mf + oldpma.off = fr.Start + oldpma.translatePerms = usermem.AnyAccess + oldpma.effectivePerms = vma.effectivePerms + oldpma.maxPerms = vma.maxPerms + oldpma.needCOW = false + oldpma.private = true + oldpma.internalMappings = safemem.BlockSeq{} + // Try to merge the pma with its neighbors. + if prev := pseg.PrevSegment(); prev.Ok() { + if merged := mm.pmas.Merge(prev, pseg); merged.Ok() { + pseg = merged + pstart = pmaIterator{} // iterators invalidated + } + } + if next := pseg.NextSegment(); next.Ok() { + if merged := mm.pmas.Merge(pseg, next); merged.Ok() { + pseg = merged + pstart = pmaIterator{} // iterators invalidated + } + } + // The error returned by AllocateAndFill is only + // significant if it occurred before ar.End. + if err != nil && pseg.End() < ar.End { + return pstart, pseg.NextGap(), err + } + // Ensure pseg and pgap are correct for the next iteration + // of the loop. + pseg, pgap = pseg.NextNonEmpty() + } else if !oldpma.translatePerms.SupersetOf(at) { + // Get new pmas (with sufficient permissions) by calling + // memmap.Mappable.Translate again. + if checkInvariants { + if oldpma.private { + panic(fmt.Sprintf("private pma %v has non-maximal pma.translatePerms: %v", pseg.Range(), oldpma)) + } + } + // Allow the entire pma to be replaced. + optAR := pseg.Range() + optMR := vseg.mappableRangeOf(optAR) + reqAR := optAR.Intersect(ar) + reqMR := vseg.mappableRangeOf(reqAR) + perms := oldpma.translatePerms.Union(at) + ts, err := vma.mappable.Translate(ctx, reqMR, optMR, perms) + if checkInvariants { + if err := memmap.CheckTranslateResult(reqMR, optMR, perms, ts, err); err != nil { + panic(fmt.Sprintf("Mappable(%T).Translate(%v, %v, %v): %v", vma.mappable, reqMR, optMR, perms, err)) + } + } + // Remove the part of the existing pma covered by new + // Translations, then insert new pmas. This doesn't change + // RSS. Note that we don't need to call unmapASLocked: any + // existing AddressSpace mappings are still valid (though + // less permissive than the new pmas indicate) until + // Invalidate is called, and will be replaced by future + // calls to mapASLocked. + if len(ts) == 0 { + return pstart, pseg.PrevGap(), err + } + transMR := memmap.MappableRange{ts[0].Source.Start, ts[len(ts)-1].Source.End} + transAR := vseg.addrRangeOf(transMR) + pseg = mm.pmas.Isolate(pseg, transAR) + pseg.ValuePtr().file.DecRef(pseg.fileRange()) + pgap = mm.pmas.Remove(pseg) + pstart = pmaIterator{} // iterators invalidated + for _, t := range ts { + newpmaAR := vseg.addrRangeOf(t.Source) + newpma := pma{ + file: t.File, + off: t.Offset, + translatePerms: t.Perms, + effectivePerms: vma.effectivePerms.Intersect(t.Perms), + maxPerms: vma.maxPerms.Intersect(t.Perms), + } + if vma.private { + newpma.effectivePerms.Write = false + newpma.maxPerms.Write = false + newpma.needCOW = true + } + t.File.IncRef(t.FileRange()) + pseg = mm.pmas.Insert(pgap, newpmaAR, newpma) + pgap = pseg.NextGap() + } + // The error returned by Translate is only significant if + // it occurred before ar.End. + if err != nil && pseg.End() < ar.End { + return pstart, pgap, err + } + // Ensure pseg and pgap are correct for the next iteration + // of the loop. + if pgap.Range().Length() == 0 { + pseg, pgap = pgap.NextSegment(), pmaGapIterator{} + } else { + pseg = pmaIterator{} + } + } else { + // We have a usable pma; continue. + pseg, pgap = pseg.NextNonEmpty() + } + + default: + break pmaLoop } } - var err error - pgap, err = mm.insertPMAsLocked(ctx, vseg, pgap, ar) - // insertPMAsLocked most likely invalidated iterators, so pstart is now - // unknown. - pstart = pmaIterator{} - if err != nil { - return pstart, pgap, err + // Go to the next vma. + if ar.End <= vseg.End() { + if pgap.Ok() { + return pstart, pgap, nil + } + return pstart, pseg.PrevGap(), nil } + vseg = vseg.NextSegment() } - return pstart, pgap, nil } const ( @@ -299,215 +496,16 @@ func privateAligned(ar usermem.AddrRange) usermem.AddrRange { return aligned } -// insertPMAsLocked inserts pmas into pgap corresponding to the vma iterated by -// vseg, spanning at least ar. It returns: -// -// - An iterator to the gap after the last pma containing an address in ar. If -// pmas exist for no addresses in ar, the iterator is to a gap that begins -// before ar.Start. -// -// - An error that is non-nil if pmas exist for only a subset of ar. -// -// Preconditions: mm.mappingMu must be locked. mm.activeMu must be locked for -// writing. vseg.Range().Intersect(pgap.Range()).Intersect(ar).Length() != 0. -// ar must be page-aligned. -func (mm *MemoryManager) insertPMAsLocked(ctx context.Context, vseg vmaIterator, pgap pmaGapIterator, ar usermem.AddrRange) (pmaGapIterator, error) { - optAR := vseg.Range().Intersect(pgap.Range()) - if checkInvariants { - if optAR.Length() <= 0 { - panic(fmt.Sprintf("vseg %v and pgap %v do not overlap", vseg, pgap)) - } - if !ar.WellFormed() || ar.Length() <= 0 || !ar.IsPageAligned() { - panic(fmt.Sprintf("invalid ar %v", ar)) - } - } - vma := vseg.ValuePtr() - - // Private anonymous mappings get pmas by allocating. - if vma.mappable == nil { - // Limit the range we allocate to ar, aligned to privateAllocUnit. - maskAR := privateAligned(ar) - allocAR := optAR.Intersect(maskAR) - mf := mm.mfp.MemoryFile() - fr, err := mf.Allocate(uint64(allocAR.Length()), usage.Anonymous) - if err != nil { - return pgap, err - } - mm.incPrivateRef(fr) - - if checkInvariants { - if !fr.WellFormed() || fr.Length() != uint64(allocAR.Length()) { - panic(fmt.Sprintf("Allocate(%v) returned invalid FileRange %v", allocAR.Length(), fr)) - } - } - - mm.addRSSLocked(allocAR) - mf.IncRef(fr) - - return mm.pmas.Insert(pgap, allocAR, pma{ - file: mf, - off: fr.Start, - vmaEffectivePerms: vma.effectivePerms, - vmaMaxPerms: vma.maxPerms, - private: true, - // Since we just allocated this memory and have the only reference, - // the new pma does not need copy-on-write. - }).NextGap(), nil - } - - // Other mappings get pmas by translating. Limit the required range - // to ar. - optMR := vseg.mappableRangeOf(optAR) - reqAR := optAR.Intersect(ar) - reqMR := vseg.mappableRangeOf(reqAR) - perms := vma.maxPerms - if vma.private { - perms.Write = false - } - ts, err := vma.mappable.Translate(ctx, reqMR, optMR, perms) - if checkInvariants { - if err := memmap.CheckTranslateResult(reqMR, optMR, ts, err); err != nil { - panic(fmt.Sprintf("Mappable(%T).Translate(%v, %v): %v", vma.mappable, reqMR, optMR, err)) - } - } - - // Install a pma for each Translation. - for _, t := range ts { - // This is valid because memmap.Mappable.Translate is required to - // return Translations in increasing Translation.Source order. - addrRange := vseg.addrRangeOf(t.Source) - mm.addRSSLocked(addrRange) - pseg := mm.pmas.Insert(pgap, addrRange, pma{ - file: t.File, - off: t.Offset, - vmaEffectivePerms: vma.effectivePerms, - vmaMaxPerms: vma.maxPerms, - needCOW: vma.private, - }) - // The new pseg may have been merged with existing segments, only take a - // ref on the inserted range. - t.File.IncRef(pseg.fileRangeOf(addrRange)) - pgap = pseg.NextGap() - } - - // Even if Translate returned an error, if we got to ar.End, - // insertPMAsLocked succeeded. - if ar.End <= pgap.Start() { - return pgap, nil - } - return pgap, err -} - -// breakCopyOnWriteLocked ensures that pmas in ar are not copy-on-write. It -// returns: +// isPMACopyOnWriteLocked returns true if the contents of the pma represented +// by pseg must be copied to a new private pma to be written to. // -// - An iterator to the gap after the last non-COW pma containing an address in -// ar. If non-COW pmas exist for no addresses in ar, the iterator is to a gap -// that begins before ar.Start. +// If the pma is a copy-on-write private pma, and holds the only reference on +// the memory it maps, isPMACopyOnWriteLocked will take ownership of the memory +// and update the pma to indicate that it does not require copy-on-write. // -// - A boolean that is true if iterators into mm.pmas may have been -// invalidated. -// -// - An error that is non-nil if non-COW pmas exist for only a subset of ar. -// -// Preconditions: mm.activeMu must be locked for writing. ar.Length() != 0. ar -// must be page-aligned. pseg.Range().Contains(ar.Start). pmas must exist for -// all addresses in ar. -func (mm *MemoryManager) breakCopyOnWriteLocked(pseg pmaIterator, ar usermem.AddrRange) (pmaGapIterator, bool, error) { - if checkInvariants { - if !ar.WellFormed() || ar.Length() <= 0 || !ar.IsPageAligned() { - panic(fmt.Sprintf("invalid ar: %v", ar)) - } - if !pseg.Range().Contains(ar.Start) { - panic(fmt.Sprintf("initial pma %v does not cover start of ar %v", pseg.Range(), ar)) - } - } - - // Limit the range we copy to ar, aligned to privateAllocUnit. - maskAR := privateAligned(ar) - var invalidatedIterators, didUnmapAS bool - mf := mm.mfp.MemoryFile() - for { - if mm.isPMACopyOnWriteLocked(pseg) { - // Determine the range to copy. - copyAR := pseg.Range().Intersect(maskAR) - - // Get internal mappings from the pma to copy from. - if err := pseg.getInternalMappingsLocked(); err != nil { - return pseg.PrevGap(), invalidatedIterators, err - } - - // Copy contents. - fr, err := mf.AllocateAndFill(uint64(copyAR.Length()), usage.Anonymous, &safemem.BlockSeqReader{mm.internalMappingsLocked(pseg, copyAR)}) - if _, ok := err.(safecopy.BusError); ok { - // If we got SIGBUS during the copy, deliver SIGBUS to - // userspace (instead of SIGSEGV) if we're breaking - // copy-on-write due to application page fault. - err = &memmap.BusError{err} - } - if fr.Length() == 0 { - return pseg.PrevGap(), invalidatedIterators, err - } - mm.incPrivateRef(fr) - mf.IncRef(fr) - - // Unmap all of maskAR, not just copyAR, to minimize host syscalls. - // AddressSpace mappings must be removed before mm.decPrivateRef(). - if !didUnmapAS { - mm.unmapASLocked(maskAR) - didUnmapAS = true - } - - // Replace the pma with a copy in the part of the address range - // where copying was successful. - copyAR.End = copyAR.Start + usermem.Addr(fr.Length()) - if copyAR != pseg.Range() { - pseg = mm.pmas.Isolate(pseg, copyAR) - invalidatedIterators = true - } - pma := pseg.ValuePtr() - if pma.private { - mm.decPrivateRef(pseg.fileRange()) - } - pma.file.DecRef(pseg.fileRange()) - - pma.file = mf - pma.off = fr.Start - pma.private = true - pma.needCOW = false - pma.internalMappings = safemem.BlockSeq{} - - // Try to merge pma with its neighbors. - if prev := pseg.PrevSegment(); prev.Ok() { - if merged := mm.pmas.Merge(prev, pseg); merged.Ok() { - pseg = merged - invalidatedIterators = true - } - } - if next := pseg.NextSegment(); next.Ok() { - if merged := mm.pmas.Merge(pseg, next); merged.Ok() { - pseg = merged - invalidatedIterators = true - } - } - - // If an error occurred after ar.End, breakCopyOnWriteLocked still - // did its job, so discard the error. - if err != nil && pseg.End() < ar.End { - return pseg.NextGap(), invalidatedIterators, err - } - } - // This checks against ar.End, not maskAR.End, so we will never break - // COW on a pma that does not intersect ar. - if ar.End <= pseg.End() { - return pseg.NextGap(), invalidatedIterators, nil - } - pseg = pseg.NextSegment() - } -} - -// Preconditions: mm.activeMu must be locked for writing. -func (mm *MemoryManager) isPMACopyOnWriteLocked(pseg pmaIterator) bool { +// Preconditions: vseg.Range().IsSupersetOf(pseg.Range()). mm.mappingMu must be +// locked. mm.activeMu must be locked for writing. +func (mm *MemoryManager) isPMACopyOnWriteLocked(vseg vmaIterator, pseg pmaIterator) bool { pma := pseg.ValuePtr() if !pma.needCOW { return false @@ -526,6 +524,10 @@ func (mm *MemoryManager) isPMACopyOnWriteLocked(pseg pmaIterator) bool { rseg := mm.privateRefs.refs.FindSegment(fr.Start) if rseg.Ok() && rseg.Value() == 1 && fr.End <= rseg.End() { pma.needCOW = false + // pma.private => pma.translatePerms == usermem.AnyAccess + vma := vseg.ValuePtr() + pma.effectivePerms = vma.effectivePerms + pma.maxPerms = vma.maxPerms return false } return true @@ -617,9 +619,7 @@ func (mm *MemoryManager) Pin(ctx context.Context, ar usermem.AddrRange, at userm // Ensure that we have usable pmas. mm.activeMu.Lock() - pseg, pend, perr := mm.getPMAsLocked(ctx, vseg, ar, pmaOpts{ - breakCOW: at.Write, - }) + pseg, pend, perr := mm.getPMAsLocked(ctx, vseg, ar, at) mm.mappingMu.RUnlock() if pendaddr := pend.Start(); pendaddr < ar.End { if pendaddr <= ar.Start { @@ -925,8 +925,9 @@ func (pmaSetFunctions) ClearValue(pma *pma) { func (pmaSetFunctions) Merge(ar1 usermem.AddrRange, pma1 pma, ar2 usermem.AddrRange, pma2 pma) (pma, bool) { if pma1.file != pma2.file || pma1.off+uint64(ar1.Length()) != pma2.off || - pma1.vmaEffectivePerms != pma2.vmaEffectivePerms || - pma1.vmaMaxPerms != pma2.vmaMaxPerms || + pma1.translatePerms != pma2.translatePerms || + pma1.effectivePerms != pma2.effectivePerms || + pma1.maxPerms != pma2.maxPerms || pma1.needCOW != pma2.needCOW || pma1.private != pma2.private { return pma{}, false @@ -979,20 +980,13 @@ func (mm *MemoryManager) findOrSeekPrevUpperBoundPMA(addr usermem.Addr, pgap pma func (pseg pmaIterator) getInternalMappingsLocked() error { pma := pseg.ValuePtr() if pma.internalMappings.IsEmpty() { - // Internal mappings are used for ignorePermissions accesses, - // so we need to use vma.maxPerms instead of - // vma.effectivePerms. However, we will never execute - // application code through an internal mapping, and we don't - // actually need a writable mapping if copy-on-write is in - // effect. (But get a writable mapping anyway if the pma is - // private, so that if breakCopyOnWriteLocked => - // isPMACopyOnWriteLocked takes ownership of the pma instead of - // copying, it doesn't need to get a new mapping.) - perms := pma.vmaMaxPerms + // This must use maxPerms (instead of perms) because some permission + // constraints are only visible to vmas; for example, mappings of + // read-only files have vma.maxPerms.Write unset, but this may not be + // visible to the memmap.Mappable. + perms := pma.maxPerms + // We will never execute application code through an internal mapping. perms.Execute = false - if pma.needCOW && !pma.private { - perms.Write = false - } ims, err := pma.file.MapInternal(pseg.fileRange(), perms) if err != nil { return err diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index aa94d7d6a..cfbf7a104 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -102,6 +102,7 @@ func (m *SpecialMappable) Translate(ctx context.Context, required, optional memm Source: source, File: m.mfp.MemoryFile(), Offset: m.fr.Start + source.Start, + Perms: usermem.AnyAccess, }, }, err } diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index b56e0d3b9..3725c98aa 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -54,9 +54,7 @@ func (mm *MemoryManager) HandleUserFault(ctx context.Context, addr usermem.Addr, // Ensure that we have a usable pma. mm.activeMu.Lock() - pseg, _, err := mm.getPMAsLocked(ctx, vseg, ar, pmaOpts{ - breakCOW: at.Write, - }) + pseg, _, err := mm.getPMAsLocked(ctx, vseg, ar, at) mm.mappingMu.RUnlock() if err != nil { mm.activeMu.Unlock() @@ -186,7 +184,7 @@ func (mm *MemoryManager) populateVMA(ctx context.Context, vseg vmaIterator, ar u } // Ensure that we have usable pmas. - pseg, _, err := mm.getPMAsLocked(ctx, vseg, ar, pmaOpts{}) + pseg, _, err := mm.getPMAsLocked(ctx, vseg, ar, usermem.NoAccess) if err != nil { // mm/util.c:vm_mmap_pgoff() ignores the error, if any, from // mm/gup.c:mm_populate(). If it matters, we'll get it again when @@ -231,7 +229,7 @@ func (mm *MemoryManager) populateVMAAndUnlock(ctx context.Context, vseg vmaItera // mm.mappingMu doesn't need to be write-locked for getPMAsLocked, and it // isn't needed at all for mapASLocked. mm.mappingMu.DowngradeLock() - pseg, _, err := mm.getPMAsLocked(ctx, vseg, ar, pmaOpts{}) + pseg, _, err := mm.getPMAsLocked(ctx, vseg, ar, usermem.NoAccess) mm.mappingMu.RUnlock() if err != nil { mm.activeMu.Unlock() @@ -651,13 +649,17 @@ func (mm *MemoryManager) MProtect(addr usermem.Addr, length uint64, realPerms us for pseg.Ok() && pseg.Start() < vseg.End() { if pseg.Range().Overlaps(vseg.Range()) { pseg = mm.pmas.Isolate(pseg, vseg.Range()) - if !effectivePerms.SupersetOf(pseg.ValuePtr().vmaEffectivePerms) && !didUnmapAS { + pma := pseg.ValuePtr() + if !effectivePerms.SupersetOf(pma.effectivePerms) && !didUnmapAS { // Unmap all of ar, not just vseg.Range(), to minimize host // syscalls. mm.unmapASLocked(ar) didUnmapAS = true } - pseg.ValuePtr().vmaEffectivePerms = effectivePerms + pma.effectivePerms = effectivePerms.Intersect(pma.translatePerms) + if pma.needCOW { + pma.effectivePerms.Write = false + } } pseg = pseg.NextSegment() } @@ -828,7 +830,7 @@ func (mm *MemoryManager) MLock(ctx context.Context, addr usermem.Addr, length ui mm.mappingMu.RUnlock() return syserror.ENOMEM } - _, _, err := mm.getPMAsLocked(ctx, vseg, vseg.Range().Intersect(ar), pmaOpts{}) + _, _, err := mm.getPMAsLocked(ctx, vseg, vseg.Range().Intersect(ar), usermem.NoAccess) if err != nil { mm.activeMu.Unlock() mm.mappingMu.RUnlock() @@ -923,7 +925,7 @@ func (mm *MemoryManager) MLockAll(ctx context.Context, opts MLockAllOpts) error mm.mappingMu.DowngradeLock() for vseg := mm.vmas.FirstSegment(); vseg.Ok(); vseg = vseg.NextSegment() { if vseg.ValuePtr().effectivePerms.Any() { - mm.getPMAsLocked(ctx, vseg, vseg.Range(), pmaOpts{}) + mm.getPMAsLocked(ctx, vseg, vseg.Range(), usermem.NoAccess) } } @@ -981,7 +983,7 @@ func (mm *MemoryManager) Decommit(addr usermem.Addr, length uint64) error { } for pseg.Ok() && pseg.Start() < vsegAR.End { pma := pseg.ValuePtr() - if pma.private && !mm.isPMACopyOnWriteLocked(pseg) { + if pma.private && !mm.isPMACopyOnWriteLocked(vseg, pseg) { psegAR := pseg.Range().Intersect(ar) if vsegAR.IsSupersetOf(psegAR) && vma.mappable == nil { if err := mf.Decommit(pseg.fileRangeOf(psegAR)); err == nil { diff --git a/pkg/sentry/usermem/access_type.go b/pkg/sentry/usermem/access_type.go index c71d05afe..9e6a27bcf 100644 --- a/pkg/sentry/usermem/access_type.go +++ b/pkg/sentry/usermem/access_type.go @@ -93,6 +93,15 @@ func (a AccessType) Intersect(other AccessType) AccessType { } } +// Union returns the access types set in either a or other. +func (a AccessType) Union(other AccessType) AccessType { + return AccessType{ + Read: a.Read || other.Read, + Write: a.Write || other.Write, + Execute: a.Execute || other.Execute, + } +} + // Effective returns the set of effective access types allowed by a, even if // some types are not explicitly allowed. func (a AccessType) Effective() AccessType { -- cgit v1.2.3 From 31c2236e97db6a57ca9b6ab3771876cd2231fd85 Mon Sep 17 00:00:00 2001 From: "chris.zn" Date: Thu, 28 Mar 2019 18:04:32 -0700 Subject: set task's name when fork When fork a child process, the name filed of TaskContext is not set. It results in that when we cat /proc/{pid}/status, the name filed is null. Like this: Name: State: S (sleeping) Tgid: 28 Pid: 28 PPid: 26 TracerPid: 0 FDSize: 8 VmSize: 89712 kB VmRSS: 6648 kB Threads: 1 CapInh: 00000000a93d35fb CapPrm: 0000000000000000 CapEff: 0000000000000000 CapBnd: 00000000a93d35fb Seccomp: 0 Change-Id: I5d469098c37cedd19da16b7ffab2e546a28a321e PiperOrigin-RevId: 240893304 --- pkg/sentry/kernel/task_context.go | 1 + test/syscalls/linux/prctl.cc | 24 ++++++++++++++++++++++++ 2 files changed, 25 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index d1c82f2aa..1b4d4cf2f 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -74,6 +74,7 @@ func (tc *TaskContext) release() { // of the original's. func (tc *TaskContext) Fork(ctx context.Context, k *Kernel, shareAddressSpace bool) (*TaskContext, error) { newTC := &TaskContext{ + Name: tc.Name, Arch: tc.Arch.Fork(), st: tc.st, } diff --git a/test/syscalls/linux/prctl.cc b/test/syscalls/linux/prctl.cc index 44f3df6a3..854dec714 100644 --- a/test/syscalls/linux/prctl.cc +++ b/test/syscalls/linux/prctl.cc @@ -52,6 +52,30 @@ TEST(PrctlTest, SetNameLongName) { ASSERT_EQ(long_name.substr(0, truncated_length), std::string(truncated_name)); } +TEST(PrctlTest, ChildProcessName) { + constexpr size_t kMaxNameLength = 15; + + char parent_name[kMaxNameLength + 1] = {}; + memset(parent_name, 'a', kMaxNameLength); + + ASSERT_THAT(prctl(PR_SET_NAME, parent_name), SyscallSucceeds()); + + pid_t child_pid = fork(); + TEST_PCHECK(child_pid >= 0); + if (child_pid == 0) { + char child_name[kMaxNameLength + 1] = {}; + TEST_PCHECK(prctl(PR_GET_NAME, child_name) >= 0); + TEST_CHECK(memcmp(parent_name, child_name, sizeof(parent_name)) == 0); + _exit(0); + } + + int status; + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFEXITED(status) && WEXITSTATUS(status) == 0) + << "status =" << status; +} + // Offset added to exit code from test child to distinguish from other abnormal // exits. constexpr int kPrctlNoNewPrivsTestChildExitBase = 100; -- cgit v1.2.3 From ed23f547093e705ba3d6f82b2ce49592180f9a5a Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Fri, 29 Mar 2019 12:25:17 -0700 Subject: Treat ENOSPC as a state-file error during save. PiperOrigin-RevId: 241028806 Change-Id: I770bf751a2740869a93c3ab50370a727ae580470 --- pkg/sentry/kernel/kernel.go | 5 +++++ pkg/sentry/state/BUILD | 1 + pkg/sentry/state/state.go | 14 ++++++++++++-- 3 files changed, 18 insertions(+), 2 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 3533fd8f7..d9f3f4e24 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -365,6 +365,11 @@ func (ts *TaskSet) flushWritesToFiles(ctx context.Context) error { syncErr := desc.file.Fsync(ctx, 0, fs.FileMaxOffset, fs.SyncAll) if err := fs.SaveFileFsyncError(syncErr); err != nil { name, _ := desc.file.Dirent.FullName(nil /* root */) + // Wrapping this error not only allows + // for a more useful message, but is + // required to distinguish Fsync errors + // from state file errors in + // state.Save. return fmt.Errorf("%q was not sufficiently synced: %v", name, err) } } diff --git a/pkg/sentry/state/BUILD b/pkg/sentry/state/BUILD index 69385e23c..cee18f681 100644 --- a/pkg/sentry/state/BUILD +++ b/pkg/sentry/state/BUILD @@ -18,5 +18,6 @@ go_library( "//pkg/sentry/kernel", "//pkg/sentry/watchdog", "//pkg/state/statefile", + "//pkg/syserror", ], ) diff --git a/pkg/sentry/state/state.go b/pkg/sentry/state/state.go index 67db78a56..224f8b709 100644 --- a/pkg/sentry/state/state.go +++ b/pkg/sentry/state/state.go @@ -24,18 +24,20 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/watchdog" "gvisor.googlesource.com/gvisor/pkg/state/statefile" + "gvisor.googlesource.com/gvisor/pkg/syserror" ) var previousMetadata map[string]string -// ErrStateFile is returned when the state file cannot be opened. +// ErrStateFile is returned when an error is encountered writing the statefile +// (which may occur during open or close calls in addition to write). type ErrStateFile struct { err error } // Error implements error.Error(). func (e ErrStateFile) Error() string { - return fmt.Sprintf("failed to open statefile: %v", e.err) + return fmt.Sprintf("statefile error: %v", e.err) } // SaveOpts contains save-related options. @@ -76,6 +78,14 @@ func (opts SaveOpts) Save(k *kernel.Kernel, w *watchdog.Watchdog) error { } else { // Save the kernel. err = k.SaveTo(wc) + + // ENOSPC is a state file error. This error can only come from + // writing the state file, and not from fs.FileOperations.Fsync + // because we wrap those in kernel.TaskSet.flushWritesToFiles. + if err == syserror.ENOSPC { + err = ErrStateFile{err} + } + if closeErr := wc.Close(); err == nil && closeErr != nil { err = ErrStateFile{closeErr} } -- cgit v1.2.3 From e8fef3d873e4564f0979303d134478b11def8349 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Fri, 29 Mar 2019 14:47:16 -0700 Subject: Treat fsync errors during save as SaveRejection errors. PiperOrigin-RevId: 241055485 Change-Id: I70259e9fef59bdf9733b35a2cd3319359449dd45 --- pkg/sentry/kernel/kernel.go | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index d9f3f4e24..f5cbd6c23 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -365,12 +365,15 @@ func (ts *TaskSet) flushWritesToFiles(ctx context.Context) error { syncErr := desc.file.Fsync(ctx, 0, fs.FileMaxOffset, fs.SyncAll) if err := fs.SaveFileFsyncError(syncErr); err != nil { name, _ := desc.file.Dirent.FullName(nil /* root */) - // Wrapping this error not only allows - // for a more useful message, but is - // required to distinguish Fsync errors - // from state file errors in + // Wrap this error in ErrSaveRejection + // so that it will trigger a save + // error, rather than a panic. This + // also allows us to distinguish Fsync + // errors from state file errors in // state.Save. - return fmt.Errorf("%q was not sufficiently synced: %v", name, err) + return fs.ErrSaveRejection{ + Err: fmt.Errorf("%q was not sufficiently synced: %v", name, err), + } } } } -- cgit v1.2.3 From 26e8d9981fcf6d08199a9fd9c609d9715c3cf37e Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Fri, 29 Mar 2019 16:24:29 -0700 Subject: Use kernel.Task.CopyScratchBuffer in syscalls/linux where possible. PiperOrigin-RevId: 241072126 Change-Id: Ib4d9f58f550732ac4c5153d3cf159a5b1a9749da --- pkg/sentry/kernel/task.go | 9 ++++----- pkg/sentry/syscalls/linux/sys_prctl.go | 2 +- pkg/sentry/syscalls/linux/sys_socket.go | 2 +- pkg/sentry/syscalls/linux/sys_stat.go | 2 +- 4 files changed, 7 insertions(+), 8 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index f958aba26..9c365e781 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -550,17 +550,16 @@ func (t *Task) afterLoad() { t.futexWaiter = futex.NewWaiter() } -// copyScratchBufferLen is the length of the copyScratchBuffer field of the Task -// struct. -const copyScratchBufferLen = 52 +// copyScratchBufferLen is the length of Task.copyScratchBuffer. +const copyScratchBufferLen = 144 // sizeof(struct stat) // CopyScratchBuffer returns a scratch buffer to be used in CopyIn/CopyOut // functions. It must only be used within those functions and can only be used // by the task goroutine; it exists to improve performance and thus // intentionally lacks any synchronization. // -// Callers should pass a constant value as an argument, which will allow the -// compiler to inline and optimize out the if statement below. +// Callers should pass a constant value as an argument if possible, which will +// allow the compiler to inline and optimize out the if statement below. func (t *Task) CopyScratchBuffer(size int) []byte { if size > copyScratchBufferLen { return make([]byte, size) diff --git a/pkg/sentry/syscalls/linux/sys_prctl.go b/pkg/sentry/syscalls/linux/sys_prctl.go index 4938f27bd..7a29bd9b7 100644 --- a/pkg/sentry/syscalls/linux/sys_prctl.go +++ b/pkg/sentry/syscalls/linux/sys_prctl.go @@ -75,7 +75,7 @@ func Prctl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall case linux.PR_GET_NAME: addr := args[1].Pointer() - buf := make([]byte, linux.TASK_COMM_LEN) + buf := t.CopyScratchBuffer(linux.TASK_COMM_LEN) len := copy(buf, t.Name()) if len < linux.TASK_COMM_LEN { buf[len] = 0 diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go index 564357bac..49e6f4aeb 100644 --- a/pkg/sentry/syscalls/linux/sys_socket.go +++ b/pkg/sentry/syscalls/linux/sys_socket.go @@ -516,7 +516,7 @@ func SetSockOpt(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sy if optLen > maxOptLen { return 0, nil, syscall.EINVAL } - buf := make([]byte, optLen) + buf := t.CopyScratchBuffer(int(optLen)) if _, err := t.CopyIn(optValAddr, &buf); err != nil { return 0, nil, err } diff --git a/pkg/sentry/syscalls/linux/sys_stat.go b/pkg/sentry/syscalls/linux/sys_stat.go index 8d6a8f616..bdfb9b3ef 100644 --- a/pkg/sentry/syscalls/linux/sys_stat.go +++ b/pkg/sentry/syscalls/linux/sys_stat.go @@ -133,7 +133,7 @@ func stat(t *kernel.Task, d *fs.Dirent, dirPath bool, statAddr usermem.Addr) err // common syscall for many applications, and t.CopyObjectOut has // noticeable performance impact due to its many slice allocations and // use of reflection. - b := make([]byte, 0, linux.SizeOfStat) + b := t.CopyScratchBuffer(int(linux.SizeOfStat))[:0] // Dev (uint64) b = binary.AppendUint64(b, usermem.ByteOrder, uint64(d.Inode.StableAttr.DeviceID)) -- cgit v1.2.3 From a4b34e26372528ef60140acef0b7c1ab1934f82a Mon Sep 17 00:00:00 2001 From: Andrei Vagin Date: Mon, 1 Apr 2019 12:52:19 -0700 Subject: gvisor: convert ilist to ilist:generic_list ilist:generic_list works faster (cl/240185278) and the code looks cleaner without type casting. PiperOrigin-RevId: 241381175 Change-Id: I8487ab1d73637b3e9733c253c56dce9e79f0d35f --- pkg/sentry/fs/BUILD | 14 +++++++++++++- pkg/sentry/fs/inotify.go | 14 +++++--------- pkg/sentry/fs/inotify_event.go | 3 +-- pkg/sentry/kernel/pipe/BUILD | 15 ++++++++++++++- pkg/sentry/kernel/pipe/buffers.go | 6 +----- pkg/sentry/kernel/pipe/pipe.go | 8 +++----- pkg/waiter/BUILD | 27 +++++++++++++++++++++++++-- pkg/waiter/waiter.go | 12 ++++-------- 8 files changed, 66 insertions(+), 33 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 6957c1bbe..dda6a0c9f 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -14,6 +14,7 @@ go_library( "dirent_cache.go", "dirent_list.go", "dirent_state.go", + "event_list.go", "file.go", "file_operations.go", "file_overlay.go", @@ -46,7 +47,6 @@ go_library( deps = [ "//pkg/abi/linux", "//pkg/amutex", - "//pkg/ilist", "//pkg/log", "//pkg/metric", "//pkg/p9", @@ -83,6 +83,18 @@ go_template_instance( }, ) +go_template_instance( + name = "event_list", + out = "event_list.go", + package = "fs", + prefix = "event", + template = "//pkg/ilist:generic_list", + types = { + "Linker": "*Event", + "Element": "*Event", + }, +) + go_test( name = "fs_x_test", size = "small", diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 51ece5ed0..5d6a7074b 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -19,7 +19,6 @@ import ( "sync/atomic" "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/ilist" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" @@ -51,7 +50,7 @@ type Inotify struct { evMu sync.Mutex `state:"nosave"` // A list of pending events for this inotify instance. Protected by evMu. - events ilist.List + events eventList // A scratch buffer, use to serialize inotify events. Use allocate this // ahead of time and reuse performance. Protected by evMu. @@ -143,9 +142,7 @@ func (i *Inotify) Read(ctx context.Context, _ *File, dst usermem.IOSequence, _ i } var writeLen int64 - for e := i.events.Front(); e != nil; e = e.Next() { - event := e.(*Event) - + for event := i.events.Front(); event != nil; event = event.Next() { // Does the buffer have enough remaining space to hold the event we're // about to write out? if dst.NumBytes() < int64(event.sizeOf()) { @@ -160,7 +157,7 @@ func (i *Inotify) Read(ctx context.Context, _ *File, dst usermem.IOSequence, _ i // Linux always dequeues an available event as long as there's enough // buffer space to copy it out, even if the copy below fails. Emulate // this behaviour. - i.events.Remove(e) + i.events.Remove(event) // Buffer has enough space, copy event to the read buffer. n, err := event.CopyTo(ctx, i.scratch, dst) @@ -197,8 +194,7 @@ func (i *Inotify) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArg defer i.evMu.Unlock() var n uint32 for e := i.events.Front(); e != nil; e = e.Next() { - event := e.(*Event) - n += uint32(event.sizeOf()) + n += uint32(e.sizeOf()) } var buf [4]byte usermem.ByteOrder.PutUint32(buf[:], n) @@ -216,7 +212,7 @@ func (i *Inotify) queueEvent(ev *Event) { // Check if we should coalesce the event we're about to queue with the last // one currently in the queue. Events are coalesced if they are identical. if last := i.events.Back(); last != nil { - if ev.equals(last.(*Event)) { + if ev.equals(last) { // "Coalesce" the two events by simply not queuing the new one. We // don't need to raise a waiter.EventIn notification because no new // data is available for reading. diff --git a/pkg/sentry/fs/inotify_event.go b/pkg/sentry/fs/inotify_event.go index 9e3e9d816..f09928b68 100644 --- a/pkg/sentry/fs/inotify_event.go +++ b/pkg/sentry/fs/inotify_event.go @@ -18,7 +18,6 @@ import ( "bytes" "fmt" - "gvisor.googlesource.com/gvisor/pkg/ilist" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) @@ -31,7 +30,7 @@ const inotifyEventBaseSize = 16 // // +stateify savable type Event struct { - ilist.Entry + eventEntry wd int32 mask uint32 diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD index 011a3f349..6b23117d9 100644 --- a/pkg/sentry/kernel/pipe/BUILD +++ b/pkg/sentry/kernel/pipe/BUILD @@ -1,10 +1,24 @@ package(licenses = ["notice"]) +load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +go_template_instance( + name = "buffer_list", + out = "buffer_list.go", + package = "pipe", + prefix = "buffer", + template = "//pkg/ilist:generic_list", + types = { + "Element": "*Buffer", + "Linker": "*Buffer", + }, +) + go_library( name = "pipe", srcs = [ + "buffer_list.go", "buffers.go", "device.go", "node.go", @@ -18,7 +32,6 @@ go_library( deps = [ "//pkg/abi/linux", "//pkg/amutex", - "//pkg/ilist", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/device", diff --git a/pkg/sentry/kernel/pipe/buffers.go b/pkg/sentry/kernel/pipe/buffers.go index fa8045910..54e059f8b 100644 --- a/pkg/sentry/kernel/pipe/buffers.go +++ b/pkg/sentry/kernel/pipe/buffers.go @@ -14,16 +14,12 @@ package pipe -import ( - "gvisor.googlesource.com/gvisor/pkg/ilist" -) - // Buffer encapsulates a queueable byte buffer that can // easily be truncated. It is designed only for use with pipes. // // +stateify savable type Buffer struct { - ilist.Entry + bufferEntry data []byte } diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index fad077d2d..357d1162e 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -25,7 +25,6 @@ import ( "sync/atomic" "syscall" - "gvisor.googlesource.com/gvisor/pkg/ilist" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -51,7 +50,7 @@ type Pipe struct { Dirent *fs.Dirent // The buffered byte queue. - data ilist.List + data bufferList // Max size of the pipe in bytes. When this max has been reached, // writers will get EWOULDBLOCK. @@ -170,13 +169,12 @@ func (p *Pipe) read(ctx context.Context, dst usermem.IOSequence) (int64, error) return 0, syserror.ErrWouldBlock } var n int64 - for b := p.data.Front(); b != nil; b = p.data.Front() { - buffer := b.(*Buffer) + for buffer := p.data.Front(); buffer != nil; buffer = p.data.Front() { n0, err := dst.CopyOut(ctx, buffer.bytes()) n += int64(n0) p.size -= n0 if buffer.truncate(n0) == 0 { - p.data.Remove(b) + p.data.Remove(buffer) } dst = dst.DropFirst(n0) if dst.NumBytes() == 0 || err != nil { diff --git a/pkg/waiter/BUILD b/pkg/waiter/BUILD index b748246da..48ce063d7 100644 --- a/pkg/waiter/BUILD +++ b/pkg/waiter/BUILD @@ -1,13 +1,28 @@ package(licenses = ["notice"]) +load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +go_template_instance( + name = "waiter_list", + out = "waiter_list.go", + package = "waiter", + prefix = "waiter", + template = "//pkg/ilist:generic_list", + types = { + "Element": "*Entry", + "Linker": "*Entry", + }, +) + go_library( name = "waiter", - srcs = ["waiter.go"], + srcs = [ + "waiter.go", + "waiter_list.go", + ], importpath = "gvisor.googlesource.com/gvisor/pkg/waiter", visibility = ["//visibility:public"], - deps = ["//pkg/ilist"], ) go_test( @@ -18,3 +33,11 @@ go_test( ], embed = [":waiter"], ) + +filegroup( + name = "autogen", + srcs = [ + "waiter_list.go", + ], + visibility = ["//:sandbox"], +) diff --git a/pkg/waiter/waiter.go b/pkg/waiter/waiter.go index 93390b299..fd429f733 100644 --- a/pkg/waiter/waiter.go +++ b/pkg/waiter/waiter.go @@ -59,8 +59,6 @@ package waiter import ( "sync" - - "gvisor.googlesource.com/gvisor/pkg/ilist" ) // EventMask represents io events as used in the poll() syscall. @@ -127,7 +125,7 @@ type Entry struct { // The following fields are protected by the queue lock. mask EventMask - ilist.Entry + waiterEntry } type channelCallback struct{} @@ -162,7 +160,7 @@ func NewChannelEntry(c chan struct{}) (Entry, chan struct{}) { // // +stateify savable type Queue struct { - list ilist.List `state:"zerovalue"` + list waiterList `state:"zerovalue"` mu sync.RWMutex `state:"nosave"` } @@ -186,8 +184,7 @@ func (q *Queue) EventUnregister(e *Entry) { // in common with the notification mask. func (q *Queue) Notify(mask EventMask) { q.mu.RLock() - for it := q.list.Front(); it != nil; it = it.Next() { - e := it.(*Entry) + for e := q.list.Front(); e != nil; e = e.Next() { if mask&e.mask != 0 { e.Callback.Callback(e) } @@ -201,8 +198,7 @@ func (q *Queue) Events() EventMask { ret := EventMask(0) q.mu.RLock() - for it := q.list.Front(); it != nil; it = it.Next() { - e := it.(*Entry) + for e := q.list.Front(); e != nil; e = e.Next() { ret |= e.mask } q.mu.RUnlock() -- cgit v1.2.3 From 7cff746ef2bbe5351e5985bebc88efc9e0881c78 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Mon, 1 Apr 2019 15:38:08 -0700 Subject: Save/restore simple devices. We weren't saving simple devices' last allocated inode numbers, which caused inode number reuse across S/R. PiperOrigin-RevId: 241414245 Change-Id: I964289978841ef0a57d2fa48daf8eab7633c1284 --- pkg/sentry/device/BUILD | 4 +- pkg/sentry/device/device.go | 111 ++++++++++++++++++++++++++++++-------- pkg/sentry/kernel/BUILD | 2 + pkg/sentry/kernel/kernel.go | 3 ++ pkg/sentry/kernel/kernel_state.go | 11 ++++ test/syscalls/linux/stat.cc | 33 ++++++++++++ 6 files changed, 139 insertions(+), 25 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/device/BUILD b/pkg/sentry/device/BUILD index 01de708d3..4ccf0674d 100644 --- a/pkg/sentry/device/BUILD +++ b/pkg/sentry/device/BUILD @@ -1,7 +1,7 @@ -load("//tools/go_stateify:defs.bzl", "go_library", "go_test") - package(licenses = ["notice"]) +load("//tools/go_stateify:defs.bzl", "go_library", "go_test") + go_library( name = "device", srcs = ["device.go"], diff --git a/pkg/sentry/device/device.go b/pkg/sentry/device/device.go index 27e4eb258..ae4fa1d93 100644 --- a/pkg/sentry/device/device.go +++ b/pkg/sentry/device/device.go @@ -14,11 +14,6 @@ // Package device defines reserved virtual kernel devices and structures // for managing them. -// -// Saving and restoring devices is not necessary if the devices are initialized -// as package global variables. Package initialization happens in a single goroutine -// and in a deterministic order, so minor device numbers will be assigned in the -// same order as packages are loaded. package device import ( @@ -30,7 +25,83 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" ) +// Registry tracks all simple devices and related state on the system for +// save/restore. +// +// The set of devices across save/restore must remain consistent. That is, no +// devices may be created or removed on restore relative to the saved +// system. Practically, this means do not create new devices specifically as +// part of restore. +// +// +stateify savable +type Registry struct { + // lastAnonDeviceMinor is the last minor device number used for an anonymous + // device. Must be accessed atomically. + lastAnonDeviceMinor uint64 + + // mu protects the fields below. + mu sync.Mutex `state:"nosave"` + + devices map[ID]*Device +} + +// SimpleDevices is the system-wide simple device registry. This is +// saved/restored by kernel.Kernel, but defined here to allow access without +// depending on the kernel package. See kernel.Kernel.deviceRegistry. +var SimpleDevices = newRegistry() + +func newRegistry() *Registry { + return &Registry{ + devices: make(map[ID]*Device), + } +} + +// newAnonID assigns a major and minor number to an anonymous device ID. +func (r *Registry) newAnonID() ID { + return ID{ + // Anon devices always have a major number of 0. + Major: 0, + // Use the next minor number. + Minor: atomic.AddUint64(&r.lastAnonDeviceMinor, 1), + } +} + +// newAnonDevice allocates a new anonymous device with a unique minor device +// number, and registers it with r. +func (r *Registry) newAnonDevice() *Device { + r.mu.Lock() + defer r.mu.Unlock() + d := &Device{ + ID: r.newAnonID(), + } + r.devices[d.ID] = d + return d +} + +// LoadFrom initializes the internal state of all devices in r from other. The +// set of devices in both registries must match. Devices may not be created or +// destroyed across save/restore. +func (r *Registry) LoadFrom(other *Registry) { + r.mu.Lock() + defer r.mu.Unlock() + other.mu.Lock() + defer other.mu.Unlock() + if len(r.devices) != len(other.devices) { + panic(fmt.Sprintf("Devices were added or removed when restoring the registry:\nnew:\n%+v\nold:\n%+v", r.devices, other.devices)) + } + for id, otherD := range other.devices { + ourD, ok := r.devices[id] + if !ok { + panic(fmt.Sprintf("Device %+v could not be restored as it wasn't defined in the new registry", otherD)) + } + ourD.loadFrom(otherD) + } + atomic.StoreUint64(&r.lastAnonDeviceMinor, atomic.LoadUint64(&other.lastAnonDeviceMinor)) +} + // ID identifies a device. +// +// +stateify savable type ID struct { Major uint64 Minor uint64 @@ -41,18 +112,12 @@ func (i *ID) DeviceID() uint64 { return uint64(linux.MakeDeviceID(uint16(i.Major), uint32(i.Minor))) } -// nextAnonDeviceMinor is the next minor number for a new anonymous device. -// Must be accessed atomically. -var nextAnonDeviceMinor uint64 - // NewAnonDevice creates a new anonymous device. Packages that require an anonymous // device should initialize the device in a global variable in a file called device.go: // // var myDevice = device.NewAnonDevice() func NewAnonDevice() *Device { - return &Device{ - ID: newAnonID(), - } + return SimpleDevices.newAnonDevice() } // NewAnonMultiDevice creates a new multi-keyed anonymous device. Packages that require @@ -62,21 +127,13 @@ func NewAnonDevice() *Device { // var myDevice = device.NewAnonMultiDevice() func NewAnonMultiDevice() *MultiDevice { return &MultiDevice{ - ID: newAnonID(), - } -} - -// newAnonID assigns a major and minor number to an anonymous device ID. -func newAnonID() ID { - return ID{ - // Anon devices always have a major number of 0. - Major: 0, - // Use the next minor number. - Minor: atomic.AddUint64(&nextAnonDeviceMinor, 1), + ID: SimpleDevices.newAnonID(), } } // Device is a simple virtual kernel device. +// +// +stateify savable type Device struct { ID @@ -84,6 +141,14 @@ type Device struct { last uint64 } +// loadFrom initializes d from other. The IDs of both devices must match. +func (d *Device) loadFrom(other *Device) { + if d.ID != other.ID { + panic(fmt.Sprintf("Attempting to initialize a device %+v from %+v, but device IDs don't match", d, other)) + } + atomic.StoreUint64(&d.last, atomic.LoadUint64(&other.last)) +} + // NextIno generates a new inode number func (d *Device) NextIno() uint64 { return atomic.AddUint64(&d.last, 1) diff --git a/pkg/sentry/kernel/BUILD b/pkg/sentry/kernel/BUILD index 4d34bc733..99a2fd964 100644 --- a/pkg/sentry/kernel/BUILD +++ b/pkg/sentry/kernel/BUILD @@ -137,6 +137,7 @@ go_library( importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/kernel", imports = [ "gvisor.googlesource.com/gvisor/pkg/bpf", + "gvisor.googlesource.com/gvisor/pkg/sentry/device", "gvisor.googlesource.com/gvisor/pkg/tcpip", ], visibility = ["//:sandbox"], @@ -156,6 +157,7 @@ go_library( "//pkg/secio", "//pkg/sentry/arch", "//pkg/sentry/context", + "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/lock", "//pkg/sentry/fs/timerfd", diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index f5cbd6c23..f7f471aaa 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -185,6 +185,9 @@ type Kernel struct { // socketTable is used to track all sockets on the system. Protected by // extMu. socketTable map[int]map[*refs.WeakRef]struct{} + + // deviceRegistry is used to save/restore device.SimpleDevices. + deviceRegistry struct{} `state:".(*device.Registry)"` } // InitKernelArgs holds arguments to Init. diff --git a/pkg/sentry/kernel/kernel_state.go b/pkg/sentry/kernel/kernel_state.go index a0a69b498..aae6f9ad2 100644 --- a/pkg/sentry/kernel/kernel_state.go +++ b/pkg/sentry/kernel/kernel_state.go @@ -15,6 +15,7 @@ package kernel import ( + "gvisor.googlesource.com/gvisor/pkg/sentry/device" "gvisor.googlesource.com/gvisor/pkg/tcpip" ) @@ -29,3 +30,13 @@ func (k *Kernel) loadDanglingEndpoints(es []tcpip.Endpoint) { tcpip.AddDanglingEndpoint(e) } } + +// saveDeviceRegistry is invoked by stateify. +func (k *Kernel) saveDeviceRegistry() *device.Registry { + return device.SimpleDevices +} + +// loadDeviceRegistry is invoked by stateify. +func (k *Kernel) loadDeviceRegistry(r *device.Registry) { + device.SimpleDevices.LoadFrom(r) +} diff --git a/test/syscalls/linux/stat.cc b/test/syscalls/linux/stat.cc index f96da5706..553fb7e56 100644 --- a/test/syscalls/linux/stat.cc +++ b/test/syscalls/linux/stat.cc @@ -429,6 +429,39 @@ TEST_F(StatTest, LstatELOOPPath) { ASSERT_THAT(lstat(path.c_str(), &s), SyscallFailsWithErrno(ELOOP)); } +// Ensure that inode allocation for anonymous devices work correctly across +// save/restore. In particular, inode numbers should be unique across S/R. +TEST(SimpleStatTest, AnonDeviceAllocatesUniqueInodesAcrossSaveRestore) { + // Use sockets as a convenient way to create inodes on an anonymous device. + int fd; + ASSERT_THAT(fd = socket(AF_UNIX, SOCK_STREAM, 0), SyscallSucceeds()); + FileDescriptor fd1(fd); + MaybeSave(); + ASSERT_THAT(fd = socket(AF_UNIX, SOCK_STREAM, 0), SyscallSucceeds()); + FileDescriptor fd2(fd); + + struct stat st1; + struct stat st2; + ASSERT_THAT(fstat(fd1.get(), &st1), SyscallSucceeds()); + ASSERT_THAT(fstat(fd2.get(), &st2), SyscallSucceeds()); + + // The two fds should have different inode numbers. Specifically, since fd2 + // was created later, it should have a higher inode number. + EXPECT_GT(st2.st_ino, st1.st_ino); + + // Verify again after another S/R cycle. The inode numbers should remain the + // same. + MaybeSave(); + + struct stat st1_after; + struct stat st2_after; + ASSERT_THAT(fstat(fd1.get(), &st1_after), SyscallSucceeds()); + ASSERT_THAT(fstat(fd2.get(), &st2_after), SyscallSucceeds()); + + EXPECT_EQ(st1_after.st_ino, st1.st_ino); + EXPECT_EQ(st2_after.st_ino, st2.st_ino); +} + } // namespace } // namespace testing -- cgit v1.2.3 From d14a7de65865e14383e3c4e68400446189b2e5e8 Mon Sep 17 00:00:00 2001 From: Rahat Mahmood Date: Tue, 2 Apr 2019 16:45:27 -0700 Subject: Fix more data races in shm debug messages. PiperOrigin-RevId: 241630409 Change-Id: Ie0df5f5a2f20c2d32e615f16e2ba43c88f963181 --- pkg/sentry/kernel/shm/shm.go | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 2b291e4f8..349f2a26e 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -286,7 +286,7 @@ func (r *Registry) remove(s *Shm) { defer s.mu.Unlock() if s.key != linux.IPC_PRIVATE { - panic(fmt.Sprintf("Attempted to remove shm segment %d (key=%d) from the registry whose key is still associated", s.ID, s.key)) + panic(fmt.Sprintf("Attempted to remove %s from the registry whose key is still associated", s.debugLocked())) } delete(r.shms, s.ID) @@ -370,6 +370,12 @@ type Shm struct { pendingDestruction bool } +// Precondition: Caller must hold s.mu. +func (s *Shm) debugLocked() string { + return fmt.Sprintf("Shm{id: %d, key: %d, size: %d bytes, refs: %d, destroyed: %v}", + s.ID, s.key, s.size, s.ReadRefs(), s.pendingDestruction) +} + // MappedName implements memmap.MappingIdentity.MappedName. func (s *Shm) MappedName(ctx context.Context) string { s.mu.Lock() @@ -412,7 +418,7 @@ func (s *Shm) AddMapping(ctx context.Context, _ memmap.MappingSpace, _ usermem.A } else { // AddMapping is called during a syscall, so ctx should always be a task // context. - log.Warningf("Adding mapping to shm %+v but couldn't get the current pid; not updating the last attach pid", s) + log.Warningf("Adding mapping to %s but couldn't get the current pid; not updating the last attach pid", s.debugLocked()) } return nil } @@ -434,7 +440,7 @@ func (s *Shm) RemoveMapping(ctx context.Context, _ memmap.MappingSpace, _ userme if pid, ok := context.ThreadGroupIDFromContext(ctx); ok { s.lastAttachDetachPID = pid } else { - log.Debugf("Couldn't obtain pid when removing mapping to shm %+v, not updating the last detach pid.", s) + log.Debugf("Couldn't obtain pid when removing mapping to %s, not updating the last detach pid.", s.debugLocked()) } } -- cgit v1.2.3 From c4caccd54042ea80a6e8b5a8f5ce59ee87a7f424 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Tue, 2 Apr 2019 18:12:11 -0700 Subject: Set options on the correct Task in PTRACE_SEIZE. $ docker run --rm --runtime=runsc -it --cap-add=SYS_PTRACE debian bash -c "apt-get update && apt-get install strace && strace ls" ... Setting up strace (4.15-2) ... execve("/bin/ls", ["ls"], [/* 6 vars */]) = 0 brk(NULL) = 0x5646d8c1e000 uname({sysname="Linux", nodename="114ef93d2db3", ...}) = 0 ... PiperOrigin-RevId: 241643321 Change-Id: Ie4bce27a7fb147eef07bbae5895c6ef3f529e177 --- pkg/sentry/kernel/ptrace.go | 2 +- test/syscalls/linux/ptrace.cc | 40 ++++++++++++++++++++++++++++++++++++++++ 2 files changed, 41 insertions(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index e8043bf8a..8d78b2fb3 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -374,7 +374,7 @@ func (t *Task) ptraceAttach(target *Task, seize bool, opts uintptr) error { return syserror.EPERM } if seize { - if err := t.ptraceSetOptionsLocked(opts); err != nil { + if err := target.ptraceSetOptionsLocked(opts); err != nil { return syserror.EIO } } diff --git a/test/syscalls/linux/ptrace.cc b/test/syscalls/linux/ptrace.cc index 6d5c425d8..8fc0045ce 100644 --- a/test/syscalls/linux/ptrace.cc +++ b/test/syscalls/linux/ptrace.cc @@ -1120,6 +1120,46 @@ TEST(PtraceTest, Interrupt_Listen_RequireSeize) { << " status " << status; } +TEST(PtraceTest, SeizeSetOptions) { + pid_t const child_pid = fork(); + if (child_pid == 0) { + // In child process. + while (true) { + SleepSafe(absl::Seconds(1)); + } + } + + // In parent process. + ASSERT_THAT(child_pid, SyscallSucceeds()); + + // Attach to the child with PTRACE_SEIZE while setting PTRACE_O_TRACESYSGOOD. + ASSERT_THAT(ptrace(PTRACE_SEIZE, child_pid, 0, PTRACE_O_TRACESYSGOOD), + SyscallSucceeds()); + + // Stop the child with PTRACE_INTERRUPT. + ASSERT_THAT(ptrace(PTRACE_INTERRUPT, child_pid, 0, 0), SyscallSucceeds()); + int status; + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_EQ(SIGTRAP | (kPtraceEventStop << 8), status >> 8); + + // Resume the child with PTRACE_SYSCALL and wait for it to enter + // syscall-enter-stop. The stop signal status from the syscall stop should be + // SIGTRAP|0x80, reflecting PTRACE_O_TRACESYSGOOD. + ASSERT_THAT(ptrace(PTRACE_SYSCALL, child_pid, 0, 0), SyscallSucceeds()); + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFSTOPPED(status) && WSTOPSIG(status) == (SIGTRAP | 0x80)) + << " status " << status; + + // SIGKILL the child (detaching the tracer) and wait for it to exit. + ASSERT_THAT(kill(child_pid, SIGKILL), SyscallSucceeds()); + ASSERT_THAT(waitpid(child_pid, &status, 0), + SyscallSucceedsWithValue(child_pid)); + EXPECT_TRUE(WIFSIGNALED(status) && WTERMSIG(status) == SIGKILL) + << " status " << status; +} + } // namespace } // namespace testing -- cgit v1.2.3 From 4968dd1341a04e93557bdd9f4b4b83eb508e026d Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Wed, 3 Apr 2019 16:21:38 -0700 Subject: Cache ThreadGroups in PIDNamespace If there are thousands of threads, ThreadGroupsAppend becomes very expensive as it must iterate over all Tasks to find the ThreadGroup leaders. Reduce the cost by maintaining a map of ThreadGroups which can be used to grab them all directly. The one somewhat visible change is to convert PID namespace init children zapping to a group-directed SIGKILL, as Linux did in 82058d668465 "signal: Use group_send_sig_info to kill all processes in a pid namespace". In a benchmark that creates N threads which sleep for two minutes, we see approximately this much CPU time in ThreadGroupsAppend: Before: 1 thread: 0ms 1024 threads: 30ms - 9130ms 4096 threads: 50ms - 2000ms 8192 threads: 18160ms 16384 threads: 17210ms After: 1 thread: 0ms 1024 threads: 0ms 4096 threads: 0ms 8192 threads: 0ms 16384 threads: 0ms The profiling is actually extremely noisy (likely due to cache effects), as some runs show almost no samples at 1024, 4096 threads, but obviously this does not scale to lots of threads. PiperOrigin-RevId: 241828039 Change-Id: I17827c90045df4b3c49b3174f3a05bca3026a72c --- pkg/sentry/kernel/kernel.go | 10 +++++----- pkg/sentry/kernel/sessions.go | 18 +++++++++--------- pkg/sentry/kernel/task_exec.go | 2 ++ pkg/sentry/kernel/task_exit.go | 18 +++++++++++------- pkg/sentry/kernel/task_start.go | 7 +++++++ pkg/sentry/kernel/threads.go | 24 ++++++++++++++---------- 6 files changed, 48 insertions(+), 31 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index f7f471aaa..1cd2653ff 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -866,14 +866,14 @@ func (k *Kernel) SendContainerSignal(cid string, info *arch.SignalInfo) error { defer k.tasks.mu.RUnlock() var lastErr error - for t := range k.tasks.Root.tids { - if t == t.tg.leader && t.ContainerID() == cid { - t.tg.signalHandlers.mu.Lock() - defer t.tg.signalHandlers.mu.Unlock() + for tg := range k.tasks.Root.tgids { + if tg.leader.ContainerID() == cid { + tg.signalHandlers.mu.Lock() infoCopy := *info - if err := t.sendSignalLocked(&infoCopy, true /*group*/); err != nil { + if err := tg.leader.sendSignalLocked(&infoCopy, true /*group*/); err != nil { lastErr = err } + tg.signalHandlers.mu.Unlock() } } return lastErr diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index ae6daac60..65e2b73c4 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -240,14 +240,14 @@ func (pg *ProcessGroup) SendSignal(info *arch.SignalInfo) error { defer tasks.mu.RUnlock() var lastErr error - for t := range tasks.Root.tids { - if t == t.tg.leader && t.tg.ProcessGroup() == pg { - t.tg.signalHandlers.mu.Lock() - defer t.tg.signalHandlers.mu.Unlock() + for tg := range tasks.Root.tgids { + if tg.ProcessGroup() == pg { + tg.signalHandlers.mu.Lock() infoCopy := *info - if err := t.sendSignalLocked(&infoCopy, true /*group*/); err != nil { + if err := tg.leader.sendSignalLocked(&infoCopy, true /*group*/); err != nil { lastErr = err } + tg.signalHandlers.mu.Unlock() } } return lastErr @@ -268,7 +268,7 @@ func (tg *ThreadGroup) CreateSession() error { // Precondition: callers must hold TaskSet.mu for writing. func (tg *ThreadGroup) createSession() error { // Get the ID for this thread in the current namespace. - id := tg.pidns.tids[tg.leader] + id := tg.pidns.tgids[tg] // Check if this ThreadGroup already leads a Session, or // if the proposed group is already taken. @@ -337,7 +337,7 @@ func (tg *ThreadGroup) createSession() error { // Ensure a translation is added to all namespaces. for ns := tg.pidns; ns != nil; ns = ns.parent { - local := ns.tids[tg.leader] + local := ns.tgids[tg] ns.sids[s] = SessionID(local) ns.sessions[SessionID(local)] = s ns.pgids[pg] = ProcessGroupID(local) @@ -356,7 +356,7 @@ func (tg *ThreadGroup) CreateProcessGroup() error { defer tg.pidns.owner.mu.Unlock() // Get the ID for this thread in the current namespace. - id := tg.pidns.tids[tg.leader] + id := tg.pidns.tgids[tg] // Per above, check for a Session leader or existing group. for s := tg.pidns.owner.sessions.Front(); s != nil; s = s.Next() { @@ -401,7 +401,7 @@ func (tg *ThreadGroup) CreateProcessGroup() error { // Ensure this translation is added to all namespaces. for ns := tg.pidns; ns != nil; ns = ns.parent { - local := ns.tids[tg.leader] + local := ns.tgids[tg] ns.pgids[pg] = ProcessGroupID(local) ns.processGroups[ProcessGroupID(local)] = pg } diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index a9b74da8e..9fca90a1c 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -234,6 +234,8 @@ func (t *Task) promoteLocked() { ns.tids[t] = leaderTID ns.tasks[oldTID] = oldLeader ns.tasks[leaderTID] = t + // Neither the ThreadGroup nor TGID change, so no need to + // update ns.tgids. } // Inherit the old leader's start time. diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index b9c558ccb..1a0734ab6 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -329,14 +329,15 @@ func (t *Task) exitChildren() { // signal." - pid_namespaces(7) t.Debugf("Init process terminating, killing namespace") t.tg.pidns.exiting = true - for other := range t.tg.pidns.tids { - if other.tg != t.tg { - other.tg.signalHandlers.mu.Lock() - other.sendSignalLocked(&arch.SignalInfo{ - Signo: int32(linux.SIGKILL), - }, false /* group */) - other.tg.signalHandlers.mu.Unlock() + for other := range t.tg.pidns.tgids { + if other == t.tg { + continue } + other.signalHandlers.mu.Lock() + other.leader.sendSignalLocked(&arch.SignalInfo{ + Signo: int32(linux.SIGKILL), + }, true /* group */) + other.signalHandlers.mu.Unlock() } // TODO: The init process waits for all processes in the // namespace to exit before completing its own exit @@ -652,6 +653,9 @@ func (t *Task) exitNotifyLocked(fromPtraceDetach bool) { tid := ns.tids[t] delete(ns.tasks, tid) delete(ns.tids, t) + if t == t.tg.leader { + delete(ns.tgids, t.tg) + } } t.tg.exitedCPUStats.Accumulate(t.CPUStats()) t.tg.ioUsage.Accumulate(t.ioUsage) diff --git a/pkg/sentry/kernel/task_start.go b/pkg/sentry/kernel/task_start.go index c82a32c78..b7534c0a2 100644 --- a/pkg/sentry/kernel/task_start.go +++ b/pkg/sentry/kernel/task_start.go @@ -212,11 +212,18 @@ func (ts *TaskSet) assignTIDsLocked(t *Task) error { for _, a := range allocatedTIDs { delete(a.ns.tasks, a.tid) delete(a.ns.tids, t) + if t.tg.leader == nil { + delete(a.ns.tgids, t.tg) + } } return err } ns.tasks[tid] = t ns.tids[t] = tid + if t.tg.leader == nil { + // New thread group. + ns.tgids[t.tg] = tid + } allocatedTIDs = append(allocatedTIDs, allocatedTID{ns, tid}) } return nil diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index bdb907905..4af1b7dfa 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -100,10 +100,8 @@ func newTaskSet() *TaskSet { // // Preconditions: ts.mu must be locked (for reading or writing). func (ts *TaskSet) forEachThreadGroupLocked(f func(tg *ThreadGroup)) { - for t := range ts.Root.tids { - if t == t.tg.leader { - f(t.tg) - } + for tg := range ts.Root.tgids { + f(tg) } } @@ -145,6 +143,13 @@ type PIDNamespace struct { // identifiers in this namespace. tids map[*Task]ThreadID + // tgids is a mapping from thread groups visible in this namespace to + // their identifiers in this namespace. + // + // The content of tgids is equivalent to tids[tg.leader]. This exists + // primarily as an optimization to quickly find all thread groups. + tgids map[*ThreadGroup]ThreadID + // sessions is a mapping from SessionIDs in this namespace to sessions // visible in the namespace. sessions map[SessionID]*Session @@ -173,6 +178,7 @@ func newPIDNamespace(ts *TaskSet, parent *PIDNamespace, userns *auth.UserNamespa userns: userns, tasks: make(map[ThreadID]*Task), tids: make(map[*Task]ThreadID), + tgids: make(map[*ThreadGroup]ThreadID), sessions: make(map[SessionID]*Session), sids: make(map[*Session]SessionID), processGroups: make(map[ProcessGroupID]*ProcessGroup), @@ -227,7 +233,7 @@ func (ns *PIDNamespace) IDOfTask(t *Task) ThreadID { func (ns *PIDNamespace) IDOfThreadGroup(tg *ThreadGroup) ThreadID { ns.owner.mu.RLock() defer ns.owner.mu.RUnlock() - return ns.tids[tg.leader] + return ns.tgids[tg] } // Tasks returns a snapshot of the tasks in ns. @@ -250,10 +256,8 @@ func (ns *PIDNamespace) ThreadGroups() []*ThreadGroup { func (ns *PIDNamespace) ThreadGroupsAppend(tgs []*ThreadGroup) []*ThreadGroup { ns.owner.mu.RLock() defer ns.owner.mu.RUnlock() - for t := range ns.tids { - if t == t.tg.leader { - tgs = append(tgs, t.tg) - } + for tg := range ns.tgids { + tgs = append(tgs, tg) } return tgs } @@ -387,7 +391,7 @@ func (tg *ThreadGroup) MemberIDs(pidns *PIDNamespace) []ThreadID { func (tg *ThreadGroup) ID() ThreadID { tg.pidns.owner.mu.RLock() defer tg.pidns.owner.mu.RUnlock() - return tg.pidns.tids[tg.leader] + return tg.pidns.tgids[tg] } // A taskNode defines the relationship between a task and the rest of the -- cgit v1.2.3 From 9cf33960fc61309140a67587748570a36e78fc75 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Wed, 3 Apr 2019 18:05:30 -0700 Subject: Only CopyOut CPU when it changes This will save copies when preemption is not caused by a CPU migration. PiperOrigin-RevId: 241844399 Change-Id: I2ba3b64aa377846ab763425bd59b61158f576851 --- pkg/sentry/kernel/rseq.go | 2 +- pkg/sentry/kernel/task_run.go | 17 +++++++++++------ 2 files changed, 12 insertions(+), 7 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/rseq.go b/pkg/sentry/kernel/rseq.go index 46b03c700..0a954bc16 100644 --- a/pkg/sentry/kernel/rseq.go +++ b/pkg/sentry/kernel/rseq.go @@ -88,6 +88,7 @@ func (t *Task) RSEQCPUAddr() usermem.Addr { func (t *Task) SetRSEQCPUAddr(addr usermem.Addr) error { t.rseqCPUAddr = addr if addr != 0 { + t.rseqCPU = int32(hostcpu.GetCPU()) if err := t.rseqCopyOutCPU(); err != nil { t.rseqCPUAddr = 0 t.rseqCPU = -1 @@ -102,7 +103,6 @@ func (t *Task) SetRSEQCPUAddr(addr usermem.Addr) error { // Preconditions: The caller must be running on the task goroutine. t's // AddressSpace must be active. func (t *Task) rseqCopyOutCPU() error { - t.rseqCPU = int32(hostcpu.GetCPU()) buf := t.CopyScratchBuffer(4) usermem.ByteOrder.PutUint32(buf, uint32(t.rseqCPU)) _, err := t.CopyOutBytes(t.rseqCPUAddr, buf) diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 596b9aa16..359986b21 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -21,6 +21,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/hostcpu" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" @@ -169,12 +170,16 @@ func (*runApp) execute(t *Task) taskRunState { if t.rseqPreempted { t.rseqPreempted = false if t.rseqCPUAddr != 0 { - if err := t.rseqCopyOutCPU(); err != nil { - t.Warningf("Failed to copy CPU to %#x for RSEQ: %v", t.rseqCPUAddr, err) - t.forceSignal(linux.SIGSEGV, false) - t.SendSignal(sigPriv(linux.SIGSEGV)) - // Re-enter the task run loop for signal delivery. - return (*runApp)(nil) + cpu := int32(hostcpu.GetCPU()) + if t.rseqCPU != cpu { + t.rseqCPU = cpu + if err := t.rseqCopyOutCPU(); err != nil { + t.Warningf("Failed to copy CPU to %#x for RSEQ: %v", t.rseqCPUAddr, err) + t.forceSignal(linux.SIGSEGV, false) + t.SendSignal(sigPriv(linux.SIGSEGV)) + // Re-enter the task run loop for signal delivery. + return (*runApp)(nil) + } } } t.rseqInterrupt() -- cgit v1.2.3 From 75a5ccf5d98876c26305da0feff20e4a148027ec Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Thu, 4 Apr 2019 17:13:31 -0700 Subject: Remove defer from trivial ThreadID methods In particular, ns.IDOfTask and tg.ID are used for gettid and getpid, respectively, where removing defer saves ~100ns. This may be a small improvement to application logging, which may call gettid/getpid frequently. PiperOrigin-RevId: 242039616 Change-Id: I860beb62db3fe077519835e6bafa7c74cba6ca80 --- pkg/sentry/kernel/threads.go | 20 ++++++++++++-------- 1 file changed, 12 insertions(+), 8 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index 4af1b7dfa..4fd6cf4e2 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -196,8 +196,9 @@ func (ns *PIDNamespace) NewChild(userns *auth.UserNamespace) *PIDNamespace { // task has that TID, TaskWithID returns nil. func (ns *PIDNamespace) TaskWithID(tid ThreadID) *Task { ns.owner.mu.RLock() - defer ns.owner.mu.RUnlock() - return ns.tasks[tid] + t := ns.tasks[tid] + ns.owner.mu.RUnlock() + return t } // ThreadGroupWithID returns the thread group lead by the task with thread ID @@ -224,16 +225,18 @@ func (ns *PIDNamespace) ThreadGroupWithID(tid ThreadID) *ThreadGroup { // 0. func (ns *PIDNamespace) IDOfTask(t *Task) ThreadID { ns.owner.mu.RLock() - defer ns.owner.mu.RUnlock() - return ns.tids[t] + id := ns.tids[t] + ns.owner.mu.RUnlock() + return id } // IDOfThreadGroup returns the TID assigned to tg's leader in PID namespace ns. // If the task is not visible in that namespace, IDOfThreadGroup returns 0. func (ns *PIDNamespace) IDOfThreadGroup(tg *ThreadGroup) ThreadID { ns.owner.mu.RLock() - defer ns.owner.mu.RUnlock() - return ns.tgids[tg] + id := ns.tgids[tg] + ns.owner.mu.RUnlock() + return id } // Tasks returns a snapshot of the tasks in ns. @@ -390,8 +393,9 @@ func (tg *ThreadGroup) MemberIDs(pidns *PIDNamespace) []ThreadID { // is dead, ID returns 0. func (tg *ThreadGroup) ID() ThreadID { tg.pidns.owner.mu.RLock() - defer tg.pidns.owner.mu.RUnlock() - return tg.pidns.tgids[tg] + id := tg.pidns.tgids[tg] + tg.pidns.owner.mu.RUnlock() + return id } // A taskNode defines the relationship between a task and the rest of the -- cgit v1.2.3 From 9471c013483b0709479c51d470ac840621ae7d46 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Mon, 8 Apr 2019 16:31:06 -0700 Subject: Export kernel.SignalInfoPriv. Also add kernel.SignalInfoNoInfo, and use it in RLIMIT_FSIZE checks. PiperOrigin-RevId: 242562428 Change-Id: I4887c0e1c8f5fddcabfe6d4281bf76d2f2eafe90 --- pkg/sentry/fs/host/tty.go | 7 ++----- pkg/sentry/kernel/fasync/BUILD | 1 - pkg/sentry/kernel/fasync/fasync.go | 7 +------ pkg/sentry/kernel/sessions.go | 4 ++-- pkg/sentry/kernel/signal.go | 16 +++++++++++++--- pkg/sentry/kernel/task_run.go | 2 +- pkg/sentry/kernel/task_sched.go | 10 +++++----- pkg/sentry/kernel/task_signals.go | 2 +- pkg/sentry/kernel/task_syscall.go | 8 ++++---- pkg/sentry/kernel/thread_group.go | 2 +- pkg/sentry/syscalls/linux/error.go | 7 ++----- test/syscalls/linux/write.cc | 17 ++++++++++++++--- 12 files changed, 46 insertions(+), 37 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/host/tty.go b/pkg/sentry/fs/host/tty.go index 21db0086e..c5cb75df7 100644 --- a/pkg/sentry/fs/host/tty.go +++ b/pkg/sentry/fs/host/tty.go @@ -344,11 +344,8 @@ func (t *TTYFileOperations) checkChange(ctx context.Context, sig linux.Signal) e // 040b6362d58f "tty: fix leakage of -ERESTARTSYS to userland" doesn't // apply: the sentry will handle -ERESTARTSYS in // kernel.runApp.execute() even if the kernel.Task isn't interrupted. - si := arch.SignalInfo{ - Code: arch.SignalInfoKernel, - Signo: int32(sig), - } + // // Linux ignores the result of kill_pgrp(). - _ = pg.SendSignal(&si) + _ = pg.SendSignal(kernel.SignalInfoPriv(sig)) return kernel.ERESTARTSYS } diff --git a/pkg/sentry/kernel/fasync/BUILD b/pkg/sentry/kernel/fasync/BUILD index 5faf95909..59b4a49e1 100644 --- a/pkg/sentry/kernel/fasync/BUILD +++ b/pkg/sentry/kernel/fasync/BUILD @@ -9,7 +9,6 @@ go_library( visibility = ["//:sandbox"], deps = [ "//pkg/abi/linux", - "//pkg/sentry/arch", "//pkg/sentry/fs", "//pkg/sentry/kernel", "//pkg/sentry/kernel/auth", diff --git a/pkg/sentry/kernel/fasync/fasync.go b/pkg/sentry/kernel/fasync/fasync.go index aa4aac109..298d988ea 100644 --- a/pkg/sentry/kernel/fasync/fasync.go +++ b/pkg/sentry/kernel/fasync/fasync.go @@ -19,7 +19,6 @@ import ( "sync" "gvisor.googlesource.com/gvisor/pkg/abi/linux" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" @@ -72,11 +71,7 @@ func (a *FileAsync) Callback(e *waiter.Entry) { a.requester.EffectiveKUID == c.RealKUID || a.requester.RealKUID == c.SavedKUID || a.requester.RealKUID == c.RealKUID { - t.SendSignal(&arch.SignalInfo{ - Signo: int32(linux.SIGIO), - // SEND_SIG_PRIV - Code: arch.SignalInfoKernel, - }) + t.SendSignal(kernel.SignalInfoPriv(linux.SIGIO)) } a.mu.Unlock() } diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index 65e2b73c4..070c2f930 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -219,8 +219,8 @@ func (pg *ProcessGroup) handleOrphan() { return } tg.signalHandlers.mu.Lock() - tg.leader.sendSignalLocked(sigPriv(linux.SIGHUP), true /* group */) - tg.leader.sendSignalLocked(sigPriv(linux.SIGCONT), true /* group */) + tg.leader.sendSignalLocked(SignalInfoPriv(linux.SIGHUP), true /* group */) + tg.leader.sendSignalLocked(SignalInfoPriv(linux.SIGCONT), true /* group */) tg.signalHandlers.mu.Unlock() }) diff --git a/pkg/sentry/kernel/signal.go b/pkg/sentry/kernel/signal.go index b066df132..22a56c6fc 100644 --- a/pkg/sentry/kernel/signal.go +++ b/pkg/sentry/kernel/signal.go @@ -56,11 +56,21 @@ func (k *Kernel) sendExternalSignal(info *arch.SignalInfo, context string) { } } -// sigPriv returns a SignalInfo representing a signal sent by the sentry. (The -// name reflects its equivalence to Linux's SEND_SIG_PRIV.) -func sigPriv(sig linux.Signal) *arch.SignalInfo { +// SignalInfoPriv returns a SignalInfo equivalent to Linux's SEND_SIG_PRIV. +func SignalInfoPriv(sig linux.Signal) *arch.SignalInfo { return &arch.SignalInfo{ Signo: int32(sig), Code: arch.SignalInfoKernel, } } + +// SignalInfoNoInfo returns a SignalInfo equivalent to Linux's SEND_SIG_NOINFO. +func SignalInfoNoInfo(sig linux.Signal, sender, receiver *Task) *arch.SignalInfo { + info := &arch.SignalInfo{ + Signo: int32(sig), + Code: arch.SignalInfoUser, + } + info.SetPid(int32(receiver.tg.pidns.IDOfThreadGroup(sender.tg))) + info.SetUid(int32(sender.Credentials().RealKUID.In(receiver.UserNamespace()).OrOverflow())) + return info +} diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 359986b21..6b5fe7165 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -176,7 +176,7 @@ func (*runApp) execute(t *Task) taskRunState { if err := t.rseqCopyOutCPU(); err != nil { t.Warningf("Failed to copy CPU to %#x for RSEQ: %v", t.rseqCPUAddr, err) t.forceSignal(linux.SIGSEGV, false) - t.SendSignal(sigPriv(linux.SIGSEGV)) + t.SendSignal(SignalInfoPriv(linux.SIGSEGV)) // Re-enter the task run loop for signal delivery. return (*runApp)(nil) } diff --git a/pkg/sentry/kernel/task_sched.go b/pkg/sentry/kernel/task_sched.go index 3b3cdc24a..3d654bf93 100644 --- a/pkg/sentry/kernel/task_sched.go +++ b/pkg/sentry/kernel/task_sched.go @@ -394,7 +394,7 @@ func (ticker *kernelCPUClockTicker) Notify(exp uint64) { newItimerVirtSetting, exp := tg.itimerVirtSetting.At(tgVirtNow) tg.itimerVirtSetting = newItimerVirtSetting if exp != 0 { - virtReceiver.sendSignalLocked(sigPriv(linux.SIGVTALRM), true) + virtReceiver.sendSignalLocked(SignalInfoPriv(linux.SIGVTALRM), true) } } if profReceiver != nil { @@ -402,18 +402,18 @@ func (ticker *kernelCPUClockTicker) Notify(exp uint64) { newItimerProfSetting, exp := tg.itimerProfSetting.At(tgProfNow) tg.itimerProfSetting = newItimerProfSetting if exp != 0 { - profReceiver.sendSignalLocked(sigPriv(linux.SIGPROF), true) + profReceiver.sendSignalLocked(SignalInfoPriv(linux.SIGPROF), true) } // RLIMIT_CPU soft limit newRlimitCPUSoftSetting, exp := tg.rlimitCPUSoftSetting.At(tgProfNow) tg.rlimitCPUSoftSetting = newRlimitCPUSoftSetting if exp != 0 { - profReceiver.sendSignalLocked(sigPriv(linux.SIGXCPU), true) + profReceiver.sendSignalLocked(SignalInfoPriv(linux.SIGXCPU), true) } // RLIMIT_CPU hard limit rlimitCPUMax := tg.limits.Get(limits.CPU).Max if rlimitCPUMax != limits.Infinity && !tgProfNow.Before(ktime.FromSeconds(int64(rlimitCPUMax))) { - profReceiver.sendSignalLocked(sigPriv(linux.SIGKILL), true) + profReceiver.sendSignalLocked(SignalInfoPriv(linux.SIGKILL), true) } } tg.signalHandlers.mu.Unlock() @@ -471,7 +471,7 @@ func (t *Task) NotifyRlimitCPUUpdated() { tgcpu := t.tg.cpuStatsAtLocked(t.k.CPUClockNow()) tgProfNow := ktime.FromNanoseconds((tgcpu.UserTime + tgcpu.SysTime).Nanoseconds()) if !tgProfNow.Before(ktime.FromSeconds(int64(rlimitCPU.Max))) { - t.sendSignalLocked(sigPriv(linux.SIGKILL), true) + t.sendSignalLocked(SignalInfoPriv(linux.SIGKILL), true) } } t.tg.updateCPUTimersEnabledLocked() diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 6a204aa59..e177562d7 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -224,7 +224,7 @@ func (t *Task) deliverSignal(info *arch.SignalInfo, act arch.SignalAct) taskRunS // Send a forced SIGSEGV. If the signal that couldn't be delivered // was a SIGSEGV, force the handler to SIG_DFL. t.forceSignal(linux.SIGSEGV, linux.Signal(info.Signo) == linux.SIGSEGV /* unconditional */) - t.SendSignal(sigPriv(linux.SIGSEGV)) + t.SendSignal(SignalInfoPriv(linux.SIGSEGV)) } default: diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index 9e43f089a..52f5fde8d 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -250,7 +250,7 @@ type runSyscallAfterSyscallEnterStop struct{} func (*runSyscallAfterSyscallEnterStop) execute(t *Task) taskRunState { if sig := linux.Signal(t.ptraceCode); sig.IsValid() { t.tg.signalHandlers.mu.Lock() - t.sendSignalLocked(sigPriv(sig), false /* group */) + t.sendSignalLocked(SignalInfoPriv(sig), false /* group */) t.tg.signalHandlers.mu.Unlock() } if t.killed() { @@ -270,7 +270,7 @@ type runSyscallAfterSysemuStop struct{} func (*runSyscallAfterSysemuStop) execute(t *Task) taskRunState { if sig := linux.Signal(t.ptraceCode); sig.IsValid() { t.tg.signalHandlers.mu.Lock() - t.sendSignalLocked(sigPriv(sig), false /* group */) + t.sendSignalLocked(SignalInfoPriv(sig), false /* group */) t.tg.signalHandlers.mu.Unlock() } if t.killed() { @@ -335,7 +335,7 @@ func (t *Task) doVsyscall(addr usermem.Addr, sysno uintptr) taskRunState { if _, err := t.CopyIn(usermem.Addr(t.Arch().Stack()), caller); err != nil { t.Debugf("vsyscall %d: error reading return address from stack: %v", sysno, err) t.forceSignal(linux.SIGSEGV, false /* unconditional */) - t.SendSignal(sigPriv(linux.SIGSEGV)) + t.SendSignal(SignalInfoPriv(linux.SIGSEGV)) return (*runApp)(nil) } @@ -405,7 +405,7 @@ func (t *Task) doVsyscallInvoke(sysno uintptr, args arch.SyscallArguments, calle t.Debugf("vsyscall %d, caller %x: emulated syscall returned error: %v", sysno, t.Arch().Value(caller), err) if err == syserror.EFAULT { t.forceSignal(linux.SIGSEGV, false /* unconditional */) - t.SendSignal(sigPriv(linux.SIGSEGV)) + t.SendSignal(SignalInfoPriv(linux.SIGSEGV)) // A return is not emulated in this case. return (*runApp)(nil) } diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index 1b7b74319..58f3a7ec9 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -322,7 +322,7 @@ type itimerRealListener struct { // Notify implements ktime.TimerListener.Notify. func (l *itimerRealListener) Notify(exp uint64) { - l.tg.SendSignal(sigPriv(linux.SIGALRM)) + l.tg.SendSignal(SignalInfoPriv(linux.SIGALRM)) } // Destroy implements ktime.TimerListener.Destroy. diff --git a/pkg/sentry/syscalls/linux/error.go b/pkg/sentry/syscalls/linux/error.go index 9fd002955..e86bed313 100644 --- a/pkg/sentry/syscalls/linux/error.go +++ b/pkg/sentry/syscalls/linux/error.go @@ -19,9 +19,9 @@ import ( "sync" "syscall" + "gvisor.googlesource.com/gvisor/pkg/abi/linux" "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/metric" - "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" "gvisor.googlesource.com/gvisor/pkg/syserror" @@ -53,10 +53,7 @@ func handleIOError(t *kernel.Task, partialResult bool, err, intr error, op strin // // Do not consume the error and return it as EFBIG. // Simultaneously send a SIGXFSZ per setrlimit(2). - t.SendSignal(&arch.SignalInfo{ - Signo: int32(syscall.SIGXFSZ), - Code: arch.SignalInfoKernel, - }) + t.SendSignal(kernel.SignalInfoNoInfo(linux.SIGXFSZ, t, t)) return syscall.EFBIG case syserror.ErrInterrupted: // The syscall was interrupted. Return nil if it completed diff --git a/test/syscalls/linux/write.cc b/test/syscalls/linux/write.cc index ca6aafd18..432bd6066 100644 --- a/test/syscalls/linux/write.cc +++ b/test/syscalls/linux/write.cc @@ -17,6 +17,7 @@ #include #include #include +#include #include #include @@ -105,23 +106,33 @@ TEST_F(WriteTest, WriteExceedsRLimit) { EXPECT_THAT(write(fd, buf.data(), target_lim + 1), SyscallSucceedsWithValue(target_lim)); EXPECT_THAT(write(fd, buf.data(), 1), SyscallFailsWithErrno(EFBIG)); + siginfo_t info; struct timespec timelimit = {0, 0}; - EXPECT_THAT(RetryEINTR(sigtimedwait)(&filesize_mask, nullptr, &timelimit), + ASSERT_THAT(RetryEINTR(sigtimedwait)(&filesize_mask, &info, &timelimit), SyscallSucceedsWithValue(SIGXFSZ)); + EXPECT_EQ(info.si_code, SI_USER); + EXPECT_EQ(info.si_pid, getpid()); + EXPECT_EQ(info.si_uid, getuid()); EXPECT_THAT(pwrite(fd, buf.data(), target_lim + 1, 1), SyscallSucceedsWithValue(target_lim - 1)); EXPECT_THAT(pwrite(fd, buf.data(), 1, target_lim), SyscallFailsWithErrno(EFBIG)); - EXPECT_THAT(RetryEINTR(sigtimedwait)(&filesize_mask, nullptr, &timelimit), + ASSERT_THAT(RetryEINTR(sigtimedwait)(&filesize_mask, &info, &timelimit), SyscallSucceedsWithValue(SIGXFSZ)); + EXPECT_EQ(info.si_code, SI_USER); + EXPECT_EQ(info.si_pid, getpid()); + EXPECT_EQ(info.si_uid, getuid()); EXPECT_THAT(pwrite64(fd, buf.data(), target_lim + 1, 1), SyscallSucceedsWithValue(target_lim - 1)); EXPECT_THAT(pwrite64(fd, buf.data(), 1, target_lim), SyscallFailsWithErrno(EFBIG)); - EXPECT_THAT(RetryEINTR(sigtimedwait)(&filesize_mask, nullptr, &timelimit), + ASSERT_THAT(RetryEINTR(sigtimedwait)(&filesize_mask, &info, &timelimit), SyscallSucceedsWithValue(SIGXFSZ)); + EXPECT_EQ(info.si_code, SI_USER); + EXPECT_EQ(info.si_pid, getpid()); + EXPECT_EQ(info.si_uid, getuid()); ASSERT_THAT(sigprocmask(SIG_UNBLOCK, &filesize_mask, nullptr), SyscallSucceeds()); -- cgit v1.2.3 From b3b140ea4f9e1b632463cbf83c97f58464eceeac Mon Sep 17 00:00:00 2001 From: Li Qiang Date: Tue, 9 Apr 2019 14:56:04 -0700 Subject: syscalls: sendfile: limit the count to MAX_RW_COUNT From sendfile spec and also the linux kernel code, we should limit the count arg to 'MAX_RW_COUNT'. This patch export 'MAX_RW_COUNT' in kernel pkg and use it in the implementation of sendfile syscall. Signed-off-by: Li Qiang Change-Id: I1086fec0685587116984555abd22b07ac233fbd2 PiperOrigin-RevId: 242745831 --- pkg/sentry/kernel/task_usermem.go | 22 +++++++++++----------- pkg/sentry/syscalls/linux/sys_file.go | 4 ++++ 2 files changed, 15 insertions(+), 11 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_usermem.go b/pkg/sentry/kernel/task_usermem.go index c8e973bd5..cb68799d3 100644 --- a/pkg/sentry/kernel/task_usermem.go +++ b/pkg/sentry/kernel/task_usermem.go @@ -22,10 +22,10 @@ import ( "gvisor.googlesource.com/gvisor/pkg/syserror" ) -// _MAX_RW_COUNT is the maximum size in bytes of a single read or write. +// MAX_RW_COUNT is the maximum size in bytes of a single read or write. // Reads and writes that exceed this size may be silently truncated. // (Linux: include/linux/fs.h:MAX_RW_COUNT) -var _MAX_RW_COUNT = int(usermem.Addr(math.MaxInt32).RoundDown()) +var MAX_RW_COUNT = int(usermem.Addr(math.MaxInt32).RoundDown()) // Activate ensures that the task has an active address space. func (t *Task) Activate() { @@ -187,9 +187,9 @@ func (t *Task) CopyOutIovecs(addr usermem.Addr, src usermem.AddrRangeSeq) error // - If any AddrRange would include addresses outside the application address // range, CopyInIovecs returns EFAULT. // -// - The combined length of all AddrRanges is limited to _MAX_RW_COUNT. If the +// - The combined length of all AddrRanges is limited to MAX_RW_COUNT. If the // combined length of all AddrRanges would otherwise exceed this amount, ranges -// beyond _MAX_RW_COUNT are silently truncated. +// beyond MAX_RW_COUNT are silently truncated. // // Preconditions: As for usermem.IO.CopyIn. The caller must be running on the // task goroutine. t's AddressSpace must be active. @@ -228,7 +228,7 @@ func (t *Task) CopyInIovecs(addr usermem.Addr, numIovecs int) (usermem.AddrRange if numIovecs == 1 { // Special case to avoid allocating dst. - return usermem.AddrRangeSeqOf(ar).TakeFirst(_MAX_RW_COUNT), nil + return usermem.AddrRangeSeqOf(ar).TakeFirst(MAX_RW_COUNT), nil } dst = append(dst, ar) @@ -239,11 +239,11 @@ func (t *Task) CopyInIovecs(addr usermem.Addr, numIovecs int) (usermem.AddrRange return usermem.AddrRangeSeq{}, syserror.ENOSYS } - // Truncate to _MAX_RW_COUNT. + // Truncate to MAX_RW_COUNT. var total uint64 for i := range dst { dstlen := uint64(dst[i].Length()) - if rem := uint64(_MAX_RW_COUNT) - total; rem < dstlen { + if rem := uint64(MAX_RW_COUNT) - total; rem < dstlen { dst[i].End -= usermem.Addr(dstlen - rem) dstlen = rem } @@ -256,16 +256,16 @@ func (t *Task) CopyInIovecs(addr usermem.Addr, numIovecs int) (usermem.AddrRange // SingleIOSequence returns a usermem.IOSequence representing [addr, // addr+length) in t's address space. If this contains addresses outside the // application address range, it returns EFAULT. If length exceeds -// _MAX_RW_COUNT, the range is silently truncated. +// MAX_RW_COUNT, the range is silently truncated. // // SingleIOSequence is analogous to Linux's // lib/iov_iter.c:import_single_range(). (Note that the non-vectorized read and // write syscalls in Linux do not use import_single_range(). However they check // access_ok() in fs/read_write.c:vfs_read/vfs_write, and overflowing address -// ranges are truncated to _MAX_RW_COUNT by fs/read_write.c:rw_verify_area().) +// ranges are truncated to MAX_RW_COUNT by fs/read_write.c:rw_verify_area().) func (t *Task) SingleIOSequence(addr usermem.Addr, length int, opts usermem.IOOpts) (usermem.IOSequence, error) { - if length > _MAX_RW_COUNT { - length = _MAX_RW_COUNT + if length > MAX_RW_COUNT { + length = MAX_RW_COUNT } ar, ok := t.MemoryManager().CheckIORange(addr, int64(length)) if !ok { diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 5a874d935..d2d351449 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -2002,6 +2002,10 @@ func Sendfile(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc return 0, nil, syserror.EINVAL } + if count > int64(kernel.MAX_RW_COUNT) { + count = int64(kernel.MAX_RW_COUNT) + } + // Get files. outFile := t.FDMap().GetFile(outFD) if outFile == nil { -- cgit v1.2.3 From 0a0619216ec9ca96c181dd69d9bf31e7762090cb Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Wed, 10 Apr 2019 11:26:10 -0700 Subject: Start saving MountSource.DirentCache. DirentCache is already a savable type, and it ensures that it is empty at the point of Save. There is no reason not to save it along with the MountSource. This did uncover an issue where not all MountSources were properly flushed before Save. If a mount point has an open file and is then unmounted, we save the MountSource without flushing it first. This CL also fixes that by flushing all MountSources for all open FDs on Save. PiperOrigin-RevId: 242906637 Change-Id: I3acd9d52b6ce6b8c989f835a408016cb3e67018f --- pkg/sentry/fs/BUILD | 1 - pkg/sentry/fs/mount.go | 2 +- pkg/sentry/fs/mount_state.go | 25 -------------- pkg/sentry/kernel/kernel.go | 78 ++++++++++++++++++++++++++++++-------------- 4 files changed, 55 insertions(+), 51 deletions(-) delete mode 100644 pkg/sentry/fs/mount_state.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index dda6a0c9f..1742d3a65 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -32,7 +32,6 @@ go_library( "mock.go", "mount.go", "mount_overlay.go", - "mount_state.go", "mounts.go", "offset.go", "overlay.go", diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index dd6e64b4c..5cc777bef 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -119,7 +119,7 @@ type MountSource struct { // fscache keeps Dirents pinned beyond application references to them. // It must be flushed before kernel.SaveTo. - fscache *DirentCache `state:"nosave"` + fscache *DirentCache // direntRefs is the sum of references on all Dirents in this MountSource. // diff --git a/pkg/sentry/fs/mount_state.go b/pkg/sentry/fs/mount_state.go deleted file mode 100644 index 6344d5160..000000000 --- a/pkg/sentry/fs/mount_state.go +++ /dev/null @@ -1,25 +0,0 @@ -// Copyright 2018 Google LLC -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package fs - -// afterLoad is invoked by stateify. -// -// Beyond the cache, this method's existence is required to ensure that this -// object is not marked "complete" until all dependent objects are also marked -// "complete". Implementations (e.g. see gofer_state.go) reach into the -// MountSourceOperations through this object, this is necessary on restore. -func (msrc *MountSource) afterLoad() { - msrc.fscache = NewDirentCache(defaultDirentCacheSize) -} diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 1cd2653ff..a9994f23b 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -311,7 +311,9 @@ func (k *Kernel) SaveTo(w io.Writer) error { // Clear the dirent cache before saving because Dirents must be Loaded in a // particular order (parents before children), and Loading dirents from a cache // breaks that order. - k.mounts.FlushMountSourceRefs() + if err := k.flushMountSourceRefs(); err != nil { + return err + } // Ensure that all pending asynchronous work is complete: // - inode and mount release @@ -351,39 +353,67 @@ func (k *Kernel) SaveTo(w io.Writer) error { return nil } -func (ts *TaskSet) flushWritesToFiles(ctx context.Context) error { +// flushMountSourceRefs flushes the MountSources for all mounted filesystems +// and open FDs. +func (k *Kernel) flushMountSourceRefs() error { + // Flush all mount sources for currently mounted filesystems. + k.mounts.FlushMountSourceRefs() + + // There may be some open FDs whose filesystems have been unmounted. We + // must flush those as well. + return k.tasks.forEachFDPaused(func(desc descriptor) error { + desc.file.Dirent.Inode.MountSource.FlushDirentRefs() + return nil + }) +} + +// forEachFDPaused applies the given function to each open file descriptor in each +// task. +// +// Precondition: Must be called with the kernel paused. +func (ts *TaskSet) forEachFDPaused(f func(descriptor) error) error { ts.mu.RLock() defer ts.mu.RUnlock() for t := range ts.Root.tids { // We can skip locking Task.mu here since the kernel is paused. - if fdmap := t.fds; fdmap != nil { - for _, desc := range fdmap.files { - if flags := desc.file.Flags(); !flags.Write { - continue - } - if sattr := desc.file.Dirent.Inode.StableAttr; !fs.IsFile(sattr) && !fs.IsDir(sattr) { - continue - } - // Here we need all metadata synced. - syncErr := desc.file.Fsync(ctx, 0, fs.FileMaxOffset, fs.SyncAll) - if err := fs.SaveFileFsyncError(syncErr); err != nil { - name, _ := desc.file.Dirent.FullName(nil /* root */) - // Wrap this error in ErrSaveRejection - // so that it will trigger a save - // error, rather than a panic. This - // also allows us to distinguish Fsync - // errors from state file errors in - // state.Save. - return fs.ErrSaveRejection{ - Err: fmt.Errorf("%q was not sufficiently synced: %v", name, err), - } - } + if t.fds == nil { + continue + } + for _, desc := range t.fds.files { + if err := f(desc); err != nil { + return err } } } return nil } +func (ts *TaskSet) flushWritesToFiles(ctx context.Context) error { + return ts.forEachFDPaused(func(desc descriptor) error { + if flags := desc.file.Flags(); !flags.Write { + return nil + } + if sattr := desc.file.Dirent.Inode.StableAttr; !fs.IsFile(sattr) && !fs.IsDir(sattr) { + return nil + } + // Here we need all metadata synced. + syncErr := desc.file.Fsync(ctx, 0, fs.FileMaxOffset, fs.SyncAll) + if err := fs.SaveFileFsyncError(syncErr); err != nil { + name, _ := desc.file.Dirent.FullName(nil /* root */) + // Wrap this error in ErrSaveRejection + // so that it will trigger a save + // error, rather than a panic. This + // also allows us to distinguish Fsync + // errors from state file errors in + // state.Save. + return fs.ErrSaveRejection{ + Err: fmt.Errorf("%q was not sufficiently synced: %v", name, err), + } + } + return nil + }) +} + // Preconditions: The kernel must be paused. func (k *Kernel) invalidateUnsavableMappings(ctx context.Context) error { invalidated := make(map[*mm.MemoryManager]struct{}) -- cgit v1.2.3 From f7aff0aaa4320505933df838cf5b551b69d5e513 Mon Sep 17 00:00:00 2001 From: Kevin Krakauer Date: Wed, 10 Apr 2019 12:35:43 -0700 Subject: Allow threads with CAP_SYS_RESOURCE to raise hard rlimits. PiperOrigin-RevId: 242919489 Change-Id: Ie3267b3bcd8a54b54bc16a6556369a19e843376f --- pkg/sentry/kernel/fd_map_test.go | 8 ++++---- pkg/sentry/limits/limits.go | 8 ++++++-- pkg/sentry/limits/limits_test.go | 30 ++++++++++++++++++------------ pkg/sentry/limits/linux.go | 2 +- pkg/sentry/mm/mm_test.go | 2 +- pkg/sentry/syscalls/linux/sys_rlimit.go | 8 +++++++- test/syscalls/linux/rlimits.cc | 15 +++++++++++---- 7 files changed, 48 insertions(+), 25 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/fd_map_test.go b/pkg/sentry/kernel/fd_map_test.go index b49996137..9e76f0a2d 100644 --- a/pkg/sentry/kernel/fd_map_test.go +++ b/pkg/sentry/kernel/fd_map_test.go @@ -40,7 +40,7 @@ func newTestFDMap() *FDMap { func TestFDMapMany(t *testing.T) { file := filetest.NewTestFile(t) limitSet := limits.NewLimitSet() - limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}) + limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}, true /* privileged */) f := newTestFDMap() for i := 0; i < maxFD; i++ { @@ -64,7 +64,7 @@ func TestFDMapMany(t *testing.T) { func TestFDMap(t *testing.T) { file := filetest.NewTestFile(t) limitSet := limits.NewLimitSet() - limitSet.Set(limits.NumberOfFiles, limits.Limit{1, maxFD}) + limitSet.Set(limits.NumberOfFiles, limits.Limit{1, maxFD}, true /* privileged */) f := newTestFDMap() if _, err := f.NewFDFrom(0, file, FDFlags{}, limitSet); err != nil { @@ -76,7 +76,7 @@ func TestFDMap(t *testing.T) { } largeLimit := limits.Limit{maxFD, maxFD} - limitSet.Set(limits.NumberOfFiles, largeLimit) + limitSet.Set(limits.NumberOfFiles, largeLimit, true /* privileged */) if fd, err := f.NewFDFrom(0, file, FDFlags{}, limitSet); err != nil { t.Fatalf("Adding an FD to a resized map: got %v, want nil", err) @@ -117,7 +117,7 @@ func TestDescriptorFlags(t *testing.T) { file := filetest.NewTestFile(t) f := newTestFDMap() limitSet := limits.NewLimitSet() - limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}) + limitSet.Set(limits.NumberOfFiles, limits.Limit{maxFD, maxFD}, true /* privileged */) origFlags := FDFlags{CloseOnExec: true} diff --git a/pkg/sentry/limits/limits.go b/pkg/sentry/limits/limits.go index eeca01876..b0571739f 100644 --- a/pkg/sentry/limits/limits.go +++ b/pkg/sentry/limits/limits.go @@ -113,13 +113,17 @@ func (l *LimitSet) SetUnchecked(t LimitType, v Limit) { } // Set assigns value v to resource of LimitType t and returns the old value. -func (l *LimitSet) Set(t LimitType, v Limit) (Limit, error) { +// privileged should be true only when either the caller has CAP_SYS_RESOURCE +// or when creating limits for a new kernel. +func (l *LimitSet) Set(t LimitType, v Limit, privileged bool) (Limit, error) { l.mu.Lock() defer l.mu.Unlock() + // If a limit is already set, make sure the new limit doesn't // exceed the previous max limit. if _, ok := l.data[t]; ok { - if l.data[t].Max < v.Max { + // Unprivileged users can only lower their hard limits. + if l.data[t].Max < v.Max && !privileged { return Limit{}, syscall.EPERM } if v.Cur > v.Max { diff --git a/pkg/sentry/limits/limits_test.go b/pkg/sentry/limits/limits_test.go index d41f62554..945428163 100644 --- a/pkg/sentry/limits/limits_test.go +++ b/pkg/sentry/limits/limits_test.go @@ -20,18 +20,24 @@ import ( ) func TestSet(t *testing.T) { - ls := NewLimitSet() - ls.Set(1, Limit{Cur: 50, Max: 50}) - if _, err := ls.Set(1, Limit{Cur: 20, Max: 50}); err != nil { - t.Fatalf("Tried to lower Limit to valid new value: got %v, wanted nil", err) - } - if _, err := ls.Set(1, Limit{Cur: 20, Max: 60}); err != syscall.EPERM { - t.Fatalf("Tried to raise limit.Max to invalid higher value: got %v, wanted syscall.EPERM", err) - } - if _, err := ls.Set(1, Limit{Cur: 60, Max: 50}); err != syscall.EINVAL { - t.Fatalf("Tried to raise limit.Cur to invalid higher value: got %v, wanted syscall.EINVAL", err) + testCases := []struct { + limit Limit + privileged bool + expectedErr error + }{ + {limit: Limit{Cur: 50, Max: 50}, privileged: false, expectedErr: nil}, + {limit: Limit{Cur: 20, Max: 50}, privileged: false, expectedErr: nil}, + {limit: Limit{Cur: 20, Max: 60}, privileged: false, expectedErr: syscall.EPERM}, + {limit: Limit{Cur: 60, Max: 50}, privileged: false, expectedErr: syscall.EINVAL}, + {limit: Limit{Cur: 11, Max: 10}, privileged: false, expectedErr: syscall.EINVAL}, + {limit: Limit{Cur: 20, Max: 60}, privileged: true, expectedErr: nil}, } - if _, err := ls.Set(1, Limit{Cur: 11, Max: 10}); err != syscall.EINVAL { - t.Fatalf("Tried to set new limit with Cur > Max: got %v, wanted syscall.EINVAL", err) + + ls := NewLimitSet() + for _, tc := range testCases { + if _, err := ls.Set(1, tc.limit, tc.privileged); err != tc.expectedErr { + t.Fatalf("Tried to set Limit to %+v and privilege %t: got %v, wanted %v", tc.limit, tc.privileged, err, tc.expectedErr) + } } + } diff --git a/pkg/sentry/limits/linux.go b/pkg/sentry/limits/linux.go index 295f9c398..e09d0d2fb 100644 --- a/pkg/sentry/limits/linux.go +++ b/pkg/sentry/limits/linux.go @@ -95,6 +95,6 @@ func NewLinuxDistroLimitSet() (*LimitSet, error) { // 1,048,576 ought to be enough for anyone. l := ls.Get(ProcessCount) l.Cur = 1 << 20 - ls.Set(ProcessCount, l) + ls.Set(ProcessCount, l, true /* privileged */) return ls, nil } diff --git a/pkg/sentry/mm/mm_test.go b/pkg/sentry/mm/mm_test.go index e12cb3bd1..ae4fba478 100644 --- a/pkg/sentry/mm/mm_test.go +++ b/pkg/sentry/mm/mm_test.go @@ -70,7 +70,7 @@ func TestUsageASUpdates(t *testing.T) { func TestBrkDataLimitUpdates(t *testing.T) { limitSet := limits.NewLimitSet() - limitSet.Set(limits.Data, limits.Limit{}) // zero RLIMIT_DATA + limitSet.Set(limits.Data, limits.Limit{}, true /* privileged */) // zero RLIMIT_DATA ctx := contexttest.WithLimitSet(contexttest.Context(t), limitSet) mm := testMemoryManager(ctx) diff --git a/pkg/sentry/syscalls/linux/sys_rlimit.go b/pkg/sentry/syscalls/linux/sys_rlimit.go index b0b216045..443334693 100644 --- a/pkg/sentry/syscalls/linux/sys_rlimit.go +++ b/pkg/sentry/syscalls/linux/sys_rlimit.go @@ -106,7 +106,13 @@ func prlimit64(t *kernel.Task, resource limits.LimitType, newLim *limits.Limit) if _, ok := setableLimits[resource]; !ok { return limits.Limit{}, syserror.EPERM } - oldLim, err := t.ThreadGroup().Limits().Set(resource, *newLim) + + // "A privileged process (under Linux: one with the CAP_SYS_RESOURCE + // capability in the initial user namespace) may make arbitrary changes + // to either limit value." + privileged := t.HasCapabilityIn(linux.CAP_SYS_RESOURCE, t.Kernel().RootUserNamespace()) + + oldLim, err := t.ThreadGroup().Limits().Set(resource, *newLim, privileged) if err != nil { return limits.Limit{}, err } diff --git a/test/syscalls/linux/rlimits.cc b/test/syscalls/linux/rlimits.cc index dc31bad9a..5a6174d99 100644 --- a/test/syscalls/linux/rlimits.cc +++ b/test/syscalls/linux/rlimits.cc @@ -25,15 +25,12 @@ namespace { TEST(RlimitTest, SetRlimitHigher) { SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(HaveCapability(CAP_SYS_RESOURCE))); - SKIP_IF(!IsRunningOnGvisor()); struct rlimit rl = {}; EXPECT_THAT(getrlimit(RLIMIT_NOFILE, &rl), SyscallSucceeds()); - // TODO: Even with CAP_SYS_RESOURCE, gVisor does not allow - // setting a higher rlimit. rl.rlim_max++; - EXPECT_THAT(setrlimit(RLIMIT_NOFILE, &rl), SyscallFailsWithErrno(EPERM)); + EXPECT_THAT(setrlimit(RLIMIT_NOFILE, &rl), SyscallSucceeds()); } TEST(RlimitTest, UnprivilegedSetRlimit) { @@ -56,6 +53,16 @@ TEST(RlimitTest, UnprivilegedSetRlimit) { EXPECT_THAT(setrlimit(RLIMIT_NOFILE, &rl), SyscallFailsWithErrno(EPERM)); } +TEST(RlimitTest, SetSoftRlimitAboveHard) { + SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(HaveCapability(CAP_SYS_RESOURCE))); + + struct rlimit rl = {}; + EXPECT_THAT(getrlimit(RLIMIT_NOFILE, &rl), SyscallSucceeds()); + + rl.rlim_cur = rl.rlim_max + 1; + EXPECT_THAT(setrlimit(RLIMIT_NOFILE, &rl), SyscallFailsWithErrno(EINVAL)); +} + } // namespace } // namespace testing -- cgit v1.2.3 From d93d19fd4eefdfd868919a73c9498e7da7eb9258 Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Wed, 10 Apr 2019 16:35:22 -0700 Subject: Fix uses of RootFromContext. RootFromContext can return a dirent with reference taken, or nil. We must call DecRef if (and only if) a real dirent is returned. PiperOrigin-RevId: 242965515 Change-Id: Ie2b7b4cb19ee09b6ccf788b71f3fd7efcdf35a11 --- pkg/sentry/fs/copy_up.go | 10 +++++++--- pkg/sentry/fs/file.go | 6 +++++- pkg/sentry/fs/file_overlay.go | 10 ++++++++-- pkg/sentry/fs/fsutil/file.go | 4 +++- pkg/sentry/fs/gofer/file.go | 4 +++- pkg/sentry/fs/host/file.go | 4 +++- pkg/sentry/fs/proc/fds.go | 4 +++- pkg/sentry/fs/proc/proc.go | 4 +++- pkg/sentry/fs/proc/task.go | 4 +++- pkg/sentry/fs/ramfs/dir.go | 4 +++- pkg/sentry/fs/tty/dir.go | 4 +++- pkg/sentry/kernel/kernel.go | 4 ++++ 12 files changed, 48 insertions(+), 14 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/copy_up.go b/pkg/sentry/fs/copy_up.go index 6d4ebaaa4..ba69e718d 100644 --- a/pkg/sentry/fs/copy_up.go +++ b/pkg/sentry/fs/copy_up.go @@ -188,11 +188,15 @@ func copyUpLocked(ctx context.Context, parent *Dirent, next *Dirent) error { var childUpperInode *Inode parentUpper := parent.Inode.overlay.upper + root := RootFromContext(ctx) + if root != nil { + defer root.DecRef() + } // Create the file in the upper filesystem and get an Inode for it. switch next.Inode.StableAttr.Type { case RegularFile: - childFile, err := parentUpper.Create(ctx, RootFromContext(ctx), next.name, FileFlags{Read: true, Write: true}, attrs.Perms) + childFile, err := parentUpper.Create(ctx, root, next.name, FileFlags{Read: true, Write: true}, attrs.Perms) if err != nil { log.Warningf("copy up failed to create file: %v", err) return syserror.EIO @@ -201,7 +205,7 @@ func copyUpLocked(ctx context.Context, parent *Dirent, next *Dirent) error { childUpperInode = childFile.Dirent.Inode case Directory: - if err := parentUpper.CreateDirectory(ctx, RootFromContext(ctx), next.name, attrs.Perms); err != nil { + if err := parentUpper.CreateDirectory(ctx, root, next.name, attrs.Perms); err != nil { log.Warningf("copy up failed to create directory: %v", err) return syserror.EIO } @@ -221,7 +225,7 @@ func copyUpLocked(ctx context.Context, parent *Dirent, next *Dirent) error { log.Warningf("copy up failed to read symlink value: %v", err) return syserror.EIO } - if err := parentUpper.CreateLink(ctx, RootFromContext(ctx), link, next.name); err != nil { + if err := parentUpper.CreateLink(ctx, root, link, next.name); err != nil { log.Warningf("copy up failed to create symlink: %v", err) return syserror.EIO } diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index d66813103..01c18647c 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -410,7 +410,11 @@ func (f *File) ConfigureMMap(ctx context.Context, opts *memmap.MMapOpts) error { // MappedName implements memmap.MappingIdentity.MappedName. func (f *File) MappedName(ctx context.Context) string { - name, _ := f.Dirent.FullName(RootFromContext(ctx)) + root := RootFromContext(ctx) + if root != nil { + defer root.DecRef() + } + name, _ := f.Dirent.FullName(root) return name } diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index cd231bdef..4efe85832 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -169,7 +169,9 @@ func (f *overlayFileOperations) Seek(ctx context.Context, file *File, whence See // Readdir implements FileOperations.Readdir. func (f *overlayFileOperations) Readdir(ctx context.Context, file *File, serializer DentrySerializer) (int64, error) { root := RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } dirCtx := &DirCtx{ Serializer: serializer, DirCursor: &f.dirCursor, @@ -440,7 +442,11 @@ func (omi *overlayMappingIdentity) InodeID() uint64 { // MappedName implements MappingIdentity.MappedName. func (omi *overlayMappingIdentity) MappedName(ctx context.Context) string { - name, _ := omi.overlayFile.Dirent.FullName(RootFromContext(ctx)) + root := RootFromContext(ctx) + if root != nil { + defer root.DecRef() + } + name, _ := omi.overlayFile.Dirent.FullName(root) return name } diff --git a/pkg/sentry/fs/fsutil/file.go b/pkg/sentry/fs/fsutil/file.go index 32f8133fb..ce329b37a 100644 --- a/pkg/sentry/fs/fsutil/file.go +++ b/pkg/sentry/fs/fsutil/file.go @@ -277,7 +277,9 @@ func (sdfo *StaticDirFileOperations) IterateDir(ctx context.Context, dirCtx *fs. // Readdir implements fs.FileOperations.Readdir. func (sdfo *StaticDirFileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { root := fs.RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } dirCtx := &fs.DirCtx{ Serializer: serializer, DirCursor: &sdfo.dirCursor, diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index 631cc80ae..e49ae2201 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -120,7 +120,9 @@ func (f *fileOperations) Release() { // Readdir implements fs.FileOperations.Readdir. func (f *fileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { root := fs.RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } dirCtx := &fs.DirCtx{ Serializer: serializer, diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index 4e84d1d6c..d67a0795f 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -166,7 +166,9 @@ func (f *fileOperations) Readiness(mask waiter.EventMask) waiter.EventMask { // Readdir implements fs.FileOperations.Readdir. func (f *fileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { root := fs.RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } dirCtx := &fs.DirCtx{ Serializer: serializer, DirCursor: &f.dirCursor, diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index b8a0a5eff..3c471bad9 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -119,7 +119,9 @@ func (f *fd) GetFile(context.Context, *fs.Dirent, fs.FileFlags) (*fs.File, error // Readlink returns the current target. func (f *fd) Readlink(ctx context.Context, _ *fs.Inode) (string, error) { root := fs.RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } n, _ := f.Dirent.FullName(root) return n, nil } diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index 88018e707..c9e659533 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -210,7 +210,9 @@ func (rpf *rootProcFile) Readdir(ctx context.Context, file *fs.File, ser fs.Dent // Add dot and dotdot. root := fs.RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } dot, dotdot := file.Dirent.GetDotAttrs(root) names = append(names, ".", "..") m["."] = dot diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 5a90c5578..4b1f84942 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -162,7 +162,9 @@ func (f *subtasksFile) Readdir(ctx context.Context, file *fs.File, ser fs.Dentry if offset == 0 { // Serialize "." and "..". root := fs.RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } dot, dotdot := file.Dirent.GetDotAttrs(root) if err := dirCtx.DirEmit(".", dot); err != nil { return offset, err diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index 05d716afb..a3b33c0f8 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -417,7 +417,9 @@ func (dfo *dirFileOperations) IterateDir(ctx context.Context, dirCtx *fs.DirCtx, // Readdir implements FileOperations.Readdir. func (dfo *dirFileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { root := fs.RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } dirCtx := &fs.DirCtx{ Serializer: serializer, DirCursor: &dfo.dirCursor, diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index 485cdb456..11bf736d6 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -315,7 +315,9 @@ func (df *dirFileOperations) IterateDir(ctx context.Context, dirCtx *fs.DirCtx, // Readdir implements FileOperations.Readdir. func (df *dirFileOperations) Readdir(ctx context.Context, file *fs.File, serializer fs.DentrySerializer) (int64, error) { root := fs.RootFromContext(ctx) - defer root.DecRef() + if root != nil { + defer root.DecRef() + } dirCtx := &fs.DirCtx{ Serializer: serializer, DirCursor: &df.dirCursor, diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index a9994f23b..b8953657c 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -676,6 +676,10 @@ func (k *Kernel) CreateProcess(args CreateProcessArgs) (*ThreadGroup, ThreadID, root := args.Root if root == nil { root = fs.RootFromContext(ctx) + // Is the root STILL nil? + if root == nil { + return nil, 0, fmt.Errorf("CreateProcessArgs.Root was not provided, and failed to get root from context") + } } defer root.DecRef() args.Root = nil -- cgit v1.2.3 From cc48969bb72e3efdc22746c5e7463b79b1942c2b Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Wed, 10 Apr 2019 17:59:02 -0700 Subject: Internal change PiperOrigin-RevId: 242978508 Change-Id: I0ea59ac5ba1dd499e87c53f2e24709371048679b --- pkg/sentry/arch/arch_x86.go | 4 ++-- pkg/sentry/arch/signal_amd64.go | 11 ++++++----- pkg/sentry/kernel/syscalls.go | 4 ++-- pkg/sentry/loader/loader.go | 6 +++--- pkg/sentry/loader/vdso.go | 5 +++-- pkg/sentry/mm/syscalls.go | 10 +++++----- pkg/sentry/strace/strace.go | 4 ++-- pkg/sentry/syscalls/linux/error.go | 4 ++-- test/syscalls/linux/32bit.cc | 4 ++-- test/syscalls/linux/exec_binary.cc | 10 +++++----- vdso/vdso.lds | 4 ++-- 11 files changed, 34 insertions(+), 32 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/arch/arch_x86.go b/pkg/sentry/arch/arch_x86.go index e50a76083..c8bf0e7f2 100644 --- a/pkg/sentry/arch/arch_x86.go +++ b/pkg/sentry/arch/arch_x86.go @@ -306,8 +306,8 @@ func (s *State) ptraceGetRegs() syscall.PtraceRegs { // FS/GS_TLS_SEL when fs_base/gs_base is a 64-bit value. (We do the // same in PtraceSetRegs.) // - // TODO: Remove this fixup since newer Linux doesn't have - // this behavior anymore. + // TODO: Remove this fixup since newer Linux + // doesn't have this behavior anymore. if regs.Fs == 0 && regs.Fs_base <= 0xffffffff { regs.Fs = _FS_TLS_SEL } diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index f7f054b0b..c9de36897 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -392,15 +392,16 @@ func (c *context64) SignalSetup(st *Stack, act *SignalAct, info *SignalInfo, alt Sigset: sigset, } - // TODO: Set SignalContext64.Err, Trapno, and Cr2 based on - // the fault that caused the signal. For now, leave Err and Trapno - // unset and assume CR2 == info.Addr() for SIGSEGVs and SIGBUSes. + // TODO: Set SignalContext64.Err, Trapno, and Cr2 + // based on the fault that caused the signal. For now, leave Err and + // Trapno unset and assume CR2 == info.Addr() for SIGSEGVs and + // SIGBUSes. if linux.Signal(info.Signo) == linux.SIGSEGV || linux.Signal(info.Signo) == linux.SIGBUS { uc.MContext.Cr2 = info.Addr() } - // "... the value (%rsp+8) is always a multiple of 16 (...) when control is - // transferred to the function entry point." - AMD64 ABI + // "... the value (%rsp+8) is always a multiple of 16 (...) when + // control is transferred to the function entry point." - AMD64 ABI ucSize := binary.Size(uc) if ucSize < 0 { // This can only happen if we've screwed up the definition of diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go index 19b711e9c..7eb99718d 100644 --- a/pkg/sentry/kernel/syscalls.go +++ b/pkg/sentry/kernel/syscalls.go @@ -165,8 +165,8 @@ type Stracer interface { // // The returned private data is passed to SyscallExit. // - // TODO: remove kernel imports from the strace package so - // that the type can be used directly. + // TODO: remove kernel imports from the strace + // package so that the type can be used directly. SyscallEnter(t *Task, sysno uintptr, args arch.SyscallArguments, flags uint32) interface{} // SyscallExit is called on syscall exit. diff --git a/pkg/sentry/loader/loader.go b/pkg/sentry/loader/loader.go index deb8892f6..80ad59dde 100644 --- a/pkg/sentry/loader/loader.go +++ b/pkg/sentry/loader/loader.go @@ -70,9 +70,9 @@ func openPath(ctx context.Context, mm *fs.MountNamespace, root, wd *fs.Dirent, m defer d.DecRef() perms := fs.PermMask{ - // TODO: Linux requires only execute permission, - // not read. However, our backing filesystems may prevent us - // from reading the file without read permission. + // TODO: Linux requires only execute + // permission, not read. However, our backing filesystems may + // prevent us from reading the file without read permission. // // Additionally, a task with a non-readable executable has // additional constraints on access via ptrace and procfs. diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index 273f6b5b9..fabf0cbe4 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -261,8 +261,9 @@ func PrepareVDSO(mfp pgalloc.MemoryFileProvider) (*VDSO, error) { return &VDSO{ ParamPage: mm.NewSpecialMappable("[vvar]", mfp, paramPage), - // TODO: Don't advertise the VDSO, as some applications may - // not be able to handle multiple [vdso] hints. + // TODO: Don't advertise the VDSO, as + // some applications may not be able to handle multiple [vdso] + // hints. vdso: mm.NewSpecialMappable("", mfp, vdso), phdrs: info.phdrs, }, nil diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index f8f095fed..cc7eb76d2 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -698,11 +698,11 @@ func (mm *MemoryManager) Brk(ctx context.Context, addr usermem.Addr) (usermem.Ad return mm.brk.End, syserror.EINVAL } - // TODO: This enforces RLIMIT_DATA, but is slightly more - // permissive than the usual data limit. In particular, this only - // limits the size of the heap; a true RLIMIT_DATA limits the size of - // heap + data + bss. The segment sizes need to be plumbed from the - // loader package to fully enforce RLIMIT_DATA. + // TODO: This enforces RLIMIT_DATA, but is + // slightly more permissive than the usual data limit. In particular, + // this only limits the size of the heap; a true RLIMIT_DATA limits the + // size of heap + data + bss. The segment sizes need to be plumbed from + // the loader package to fully enforce RLIMIT_DATA. if uint64(addr-mm.brk.Start) > limits.FromContext(ctx).Get(limits.Data).Cur { mm.mappingMu.Unlock() return mm.brk.End, syserror.ENOMEM diff --git a/pkg/sentry/strace/strace.go b/pkg/sentry/strace/strace.go index 6c93d7de7..a7e9df268 100644 --- a/pkg/sentry/strace/strace.go +++ b/pkg/sentry/strace/strace.go @@ -686,8 +686,8 @@ func (s SyscallMap) Name(sysno uintptr) string { // N.B. This is not in an init function because we can't be sure all syscall // tables are registered with the kernel when init runs. // -// TODO: remove kernel package dependencies from this package and -// have the kernel package self-initialize all syscall tables. +// TODO: remove kernel package dependencies from this +// package and have the kernel package self-initialize all syscall tables. func Initialize() { for _, table := range kernel.SyscallTables() { // Is this known? diff --git a/pkg/sentry/syscalls/linux/error.go b/pkg/sentry/syscalls/linux/error.go index e86bed313..8759e5e32 100644 --- a/pkg/sentry/syscalls/linux/error.go +++ b/pkg/sentry/syscalls/linux/error.go @@ -89,8 +89,8 @@ func handleIOError(t *kernel.Task, partialResult bool, err, intr error, op strin // side is gone. The partial write is returned. EPIPE will be // returned on the next call. // - // TODO: In some cases SIGPIPE should also be sent - // to the application. + // TODO: In some cases SIGPIPE should + // also be sent to the application. return nil case syserror.ErrWouldBlock: // Syscall would block, but completed a partial read/write. diff --git a/test/syscalls/linux/32bit.cc b/test/syscalls/linux/32bit.cc index b8d5f0355..230648c9b 100644 --- a/test/syscalls/linux/32bit.cc +++ b/test/syscalls/linux/32bit.cc @@ -84,8 +84,8 @@ TEST(Syscall32Bit, Int80) { // disabled). return; case Platform::kPtrace: - // TODO: The ptrace platform does not have a consistent story - // here. + // TODO: The ptrace platform does not have a + // consistent story here. return; case Platform::kNative: break; diff --git a/test/syscalls/linux/exec_binary.cc b/test/syscalls/linux/exec_binary.cc index cfc898699..cb5c7ae51 100644 --- a/test/syscalls/linux/exec_binary.cc +++ b/test/syscalls/linux/exec_binary.cc @@ -285,9 +285,9 @@ ElfBinary<64> StandardElf() { elf.header.e_phoff = sizeof(elf.header); elf.header.e_phentsize = sizeof(decltype(elf)::ElfPhdr); - // TODO: Always include a PT_GNU_STACK segment to disable - // executable stacks. With this omitted the stack (and all PROT_READ) mappings - // should be executable, but gVisor doesn't support that. + // TODO: Always include a PT_GNU_STACK segment to + // disable executable stacks. With this omitted the stack (and all PROT_READ) + // mappings should be executable, but gVisor doesn't support that. decltype(elf)::ElfPhdr phdr = {}; phdr.p_type = PT_GNU_STACK; phdr.p_flags = PF_R | PF_W; @@ -1005,8 +1005,8 @@ TEST(ElfTest, NoExecute) { // Execute, but no read permissions on the binary works just fine. TEST(ElfTest, NoRead) { - // TODO: gVisor's backing filesystem may prevent the sentry from - // reading the executable. + // TODO: gVisor's backing filesystem may prevent the + // sentry from reading the executable. SKIP_IF(IsRunningOnGvisor()); ElfBinary<64> elf = StandardElf(); diff --git a/vdso/vdso.lds b/vdso/vdso.lds index 97bb6d0c1..166779931 100644 --- a/vdso/vdso.lds +++ b/vdso/vdso.lds @@ -56,8 +56,8 @@ SECTIONS { .altinstr_replacement : { *(.altinstr_replacement) } /* - * TODO: Remove this alignment? Then the VDSO would fit in a - * single page. + * TODO: Remove this alignment? Then the VDSO would fit + * in a single page. */ . = ALIGN(0x1000); .text : { *(.text*) } :text =0x90909090 -- cgit v1.2.3 From 4209edafb6a9eeff8741a4360100557179b47b35 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Thu, 11 Apr 2019 00:41:42 -0700 Subject: Use open fids when fstat()ing gofer files. PiperOrigin-RevId: 243018347 Change-Id: I1e5b80607c1df0747482abea61db7fcf24536d37 --- pkg/sentry/fs/ashmem/area.go | 9 ++++---- pkg/sentry/fs/binder/binder.go | 7 +++--- pkg/sentry/fs/dev/full.go | 19 +++++++-------- pkg/sentry/fs/dev/null.go | 40 +++++++++++++++++--------------- pkg/sentry/fs/dev/random.go | 19 +++++++-------- pkg/sentry/fs/fdpipe/pipe.go | 15 ++++++------ pkg/sentry/fs/file.go | 12 ++++++++++ pkg/sentry/fs/file_operations.go | 6 +++++ pkg/sentry/fs/file_overlay.go | 26 +++++++++++++++++++++ pkg/sentry/fs/filetest/filetest.go | 17 +++++++------- pkg/sentry/fs/fsutil/file.go | 35 ++++++++++++++++++---------- pkg/sentry/fs/fsutil/inode.go | 19 +++++++-------- pkg/sentry/fs/gofer/file.go | 16 +++++++++++++ pkg/sentry/fs/host/file.go | 5 ++-- pkg/sentry/fs/inotify.go | 5 ++++ pkg/sentry/fs/proc/exec_args.go | 19 +++++++-------- pkg/sentry/fs/proc/fds.go | 21 +++++++++-------- pkg/sentry/fs/proc/proc.go | 3 ++- pkg/sentry/fs/proc/rpcinet_proc.go | 17 +++++++------- pkg/sentry/fs/proc/seqfile/seqfile.go | 17 +++++++------- pkg/sentry/fs/proc/sys.go | 19 +++++++-------- pkg/sentry/fs/proc/sys_net.go | 34 ++++++++++++++------------- pkg/sentry/fs/proc/task.go | 41 ++++++++++++++++++--------------- pkg/sentry/fs/proc/uid_gid_map.go | 17 +++++++------- pkg/sentry/fs/proc/uptime.go | 19 +++++++-------- pkg/sentry/fs/ramfs/dir.go | 3 ++- pkg/sentry/fs/ramfs/socket.go | 21 +++++++++-------- pkg/sentry/fs/ramfs/symlink.go | 21 +++++++++-------- pkg/sentry/fs/timerfd/timerfd.go | 13 ++++++----- pkg/sentry/fs/tmpfs/file_regular.go | 15 ++++++------ pkg/sentry/fs/tty/dir.go | 15 ++++++------ pkg/sentry/fs/tty/master.go | 11 +++++---- pkg/sentry/fs/tty/slave.go | 11 +++++---- pkg/sentry/kernel/epoll/epoll.go | 13 ++++++----- pkg/sentry/kernel/eventfd/eventfd.go | 15 ++++++------ pkg/sentry/kernel/pipe/reader_writer.go | 11 +++++---- pkg/sentry/loader/vdso.go | 17 +++++++------- pkg/sentry/socket/epsocket/epsocket.go | 11 +++++---- pkg/sentry/socket/hostinet/socket.go | 11 +++++---- pkg/sentry/socket/netlink/socket.go | 11 +++++---- pkg/sentry/socket/rpcinet/socket.go | 11 +++++---- pkg/sentry/socket/unix/unix.go | 11 +++++---- pkg/sentry/syscalls/linux/sys_stat.go | 37 +++++++++++++++++++---------- 43 files changed, 422 insertions(+), 293 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index 710b5185f..651cbc164 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -42,10 +42,11 @@ const ( // // +stateify savable type Area struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` ad *Device diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index 188353961..a41b5dcae 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -85,9 +85,10 @@ func (bd *Device) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) // // +stateify savable type Proc struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` bd *Device task *kernel.Task diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index 0cb513004..82da9aae9 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -59,15 +59,16 @@ func (f *fullDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.Fi // +stateify savable type fullFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - readZeros `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` + readZeros `state:"nosave"` } var _ fs.FileOperations = (*fullFileOperations)(nil) diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 83f43c203..5d306d352 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -60,16 +60,17 @@ func (n *nullDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.Fi // +stateify savable type nullFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRead `state:"nosave"` - fsutil.FileNoopWrite `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRead `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` } var _ fs.FileOperations = (*nullFileOperations)(nil) @@ -100,15 +101,16 @@ func (zd *zeroDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.F // +stateify savable type zeroFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoopWrite `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - readZeros `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` + readZeros `state:"nosave"` } var _ fs.FileOperations = (*zeroFileOperations)(nil) diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index b9b78db7a..ffd5cf6c3 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -57,15 +57,16 @@ func (*randomDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.Fi // +stateify savable type randomFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoopWrite `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` } var _ fs.FileOperations = (*randomFileOperations)(nil) diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index b4d11cb45..98483ab68 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -37,13 +37,14 @@ import ( // // +stateify savable type pipeOperations struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - waiter.Queue `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.Queue `state:"nosave"` // flags are the flags used to open the pipe. flags fs.FileFlags `state:".(fs.FileFlags)"` diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 01c18647c..2c2126f17 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -408,6 +408,18 @@ func (f *File) ConfigureMMap(ctx context.Context, opts *memmap.MMapOpts) error { return f.FileOperations.ConfigureMMap(ctx, f, opts) } +// UnstableAttr calls f.FileOperations.UnstableAttr with f as the File. +// +// Returns syserror.ErrInterrupted if interrupted. +func (f *File) UnstableAttr(ctx context.Context) (UnstableAttr, error) { + if !f.mu.Lock(ctx) { + return UnstableAttr{}, syserror.ErrInterrupted + } + defer f.mu.Unlock() + + return f.FileOperations.UnstableAttr(ctx, f) +} + // MappedName implements memmap.MappingIdentity.MappedName. func (f *File) MappedName(ctx context.Context) string { root := RootFromContext(ctx) diff --git a/pkg/sentry/fs/file_operations.go b/pkg/sentry/fs/file_operations.go index 81c6e2b5d..e0fa5135f 100644 --- a/pkg/sentry/fs/file_operations.go +++ b/pkg/sentry/fs/file_operations.go @@ -96,6 +96,12 @@ type FileOperations interface { // memmap.Mappable. ConfigureMMap(ctx context.Context, file *File, opts *memmap.MMapOpts) error + // UnstableAttr returns the "unstable" attributes of the inode represented + // by the file. Most implementations can embed + // fsutil.FileUseInodeUnstableAttr, which delegates to + // InodeOperations.UnstableAttr. + UnstableAttr(ctx context.Context, file *File) (UnstableAttr, error) + // Ioctl implements the ioctl(2) linux syscall. // // io provides access to the virtual memory space to which pointers in args diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index 4efe85832..e1f02f0f4 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -339,6 +339,32 @@ func (*overlayFileOperations) ConfigureMMap(ctx context.Context, file *File, opt return nil } +// UnstableAttr implements fs.FileOperations.UnstableAttr. +func (f *overlayFileOperations) UnstableAttr(ctx context.Context, file *File) (UnstableAttr, error) { + // Hot path. Avoid defers. + f.upperMu.Lock() + if f.upper != nil { + attr, err := f.upper.UnstableAttr(ctx) + f.upperMu.Unlock() + return attr, err + } + f.upperMu.Unlock() + + // It's possible that copy-up has occurred, but we haven't opened a upper + // file yet. If this is the case, just use the upper inode's UnstableAttr + // rather than opening a file. + o := file.Dirent.Inode.overlay + o.copyMu.RLock() + if o.upper != nil { + attr, err := o.upper.UnstableAttr(ctx) + o.copyMu.RUnlock() + return attr, err + } + o.copyMu.RUnlock() + + return f.lower.UnstableAttr(ctx) +} + // Ioctl implements fs.FileOperations.Ioctl and always returns ENOTTY. func (*overlayFileOperations) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { return 0, syserror.ENOTTY diff --git a/pkg/sentry/fs/filetest/filetest.go b/pkg/sentry/fs/filetest/filetest.go index 40d84d9f2..388a1ce36 100644 --- a/pkg/sentry/fs/filetest/filetest.go +++ b/pkg/sentry/fs/filetest/filetest.go @@ -31,14 +31,15 @@ import ( // TestFileOperations is an implementation of the File interface. It provides all // required methods. type TestFileOperations struct { - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } // NewTestFile creates and initializes a new test file. diff --git a/pkg/sentry/fs/fsutil/file.go b/pkg/sentry/fs/fsutil/file.go index ce329b37a..df34dc788 100644 --- a/pkg/sentry/fs/fsutil/file.go +++ b/pkg/sentry/fs/fsutil/file.go @@ -224,7 +224,7 @@ func (FileNoIoctl) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallAr } // DirFileOperations implements most of fs.FileOperations for directories, -// except for Readdir which the embedding type must implement. +// except for Readdir and UnstableAttr which the embedding type must implement. type DirFileOperations struct { waiter.AlwaysReady FileGenericSeek @@ -250,7 +250,8 @@ func (*DirFileOperations) Write(context.Context, *fs.File, usermem.IOSequence, i // // +stateify savable type StaticDirFileOperations struct { - DirFileOperations + DirFileOperations `state:"nosave"` + FileUseInodeUnstableAttr `state:"nosave"` // dentryMap is a SortedDentryMap used to implement Readdir. dentryMap *fs.SortedDentryMap @@ -291,16 +292,17 @@ func (sdfo *StaticDirFileOperations) Readdir(ctx context.Context, file *fs.File, // // +stateify savable type NoReadWriteFile struct { - waiter.AlwaysReady `state:"nosave"` - FileGenericSeek `state:"nosave"` - FileNoIoctl `state:"nosave"` - FileNoMMap `state:"nosave"` - FileNoopFsync `state:"nosave"` - FileNoopFlush `state:"nosave"` - FileNoopRelease `state:"nosave"` - FileNoRead `state:"nosave"` - FileNoWrite `state:"nosave"` - FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + FileGenericSeek `state:"nosave"` + FileNoIoctl `state:"nosave"` + FileNoMMap `state:"nosave"` + FileNoopFsync `state:"nosave"` + FileNoopFlush `state:"nosave"` + FileNoopRelease `state:"nosave"` + FileNoRead `state:"nosave"` + FileNoWrite `state:"nosave"` + FileNotDirReaddir `state:"nosave"` + FileUseInodeUnstableAttr `state:"nosave"` } var _ fs.FileOperations = (*NoReadWriteFile)(nil) @@ -365,3 +367,12 @@ type FileNoopRead struct{} func (FileNoopRead) Read(context.Context, *fs.File, usermem.IOSequence, int64) (int64, error) { return 0, nil } + +// FileUseInodeUnstableAttr implements fs.FileOperations.UnstableAttr by calling +// InodeOperations.UnstableAttr. +type FileUseInodeUnstableAttr struct{} + +// UnstableAttr implements fs.FileOperations.UnstableAttr. +func (FileUseInodeUnstableAttr) UnstableAttr(ctx context.Context, file *fs.File) (fs.UnstableAttr, error) { + return file.Dirent.Inode.UnstableAttr(ctx) +} diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index 37490e5b2..468171a9b 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -238,15 +238,16 @@ func (i *InodeSimpleExtendedAttributes) Listxattr(_ *fs.Inode) (map[string]struc // // +stateify savable type staticFile struct { - waiter.AlwaysReady `state:"nosave"` - FileGenericSeek `state:"nosave"` - FileNoIoctl `state:"nosave"` - FileNoMMap `state:"nosave"` - FileNoopFsync `state:"nosave"` - FileNoopFlush `state:"nosave"` - FileNoopRelease `state:"nosave"` - FileNoopWrite `state:"nosave"` - FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + FileGenericSeek `state:"nosave"` + FileNoIoctl `state:"nosave"` + FileNoMMap `state:"nosave"` + FileNoopFsync `state:"nosave"` + FileNoopFlush `state:"nosave"` + FileNoopRelease `state:"nosave"` + FileNoopWrite `state:"nosave"` + FileNotDirReaddir `state:"nosave"` + FileUseInodeUnstableAttr `state:"nosave"` FileStaticContentReader } diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index e49ae2201..80d1e08a6 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -310,6 +310,22 @@ func (f *fileOperations) ConfigureMMap(ctx context.Context, file *fs.File, opts return f.inodeOperations.configureMMap(file, opts) } +// UnstableAttr implements fs.FileOperations.UnstableAttr. +func (f *fileOperations) UnstableAttr(ctx context.Context, file *fs.File) (fs.UnstableAttr, error) { + s := f.inodeOperations.session() + if s.cachePolicy.cacheUAttrs(file.Dirent.Inode) { + return f.inodeOperations.cachingInodeOps.UnstableAttr(ctx, file.Dirent.Inode) + } + // Use f.handles.File, which represents 9P fids that have been opened, + // instead of inodeFileState.file, which represents 9P fids that have not. + // This may be significantly more efficient in some implementations. + _, valid, pattr, err := getattr(ctx, f.handles.File) + if err != nil { + return fs.UnstableAttr{}, err + } + return unstable(ctx, valid, pattr, s.mounter, s.client), nil +} + // Seek implements fs.FileOperations.Seek. func (f *fileOperations) Seek(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64) (int64, error) { return fsutil.SeekWithDirCursor(ctx, file, whence, offset, &f.dirCursor) diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index d67a0795f..2a8f285ff 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -36,8 +36,9 @@ import ( // // +stateify savable type fileOperations struct { - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // iops are the Inode operations for this file. iops *inodeOperations `state:"wait"` diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 5d6a7074b..59fa662f3 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -186,6 +186,11 @@ func (*Inotify) ConfigureMMap(context.Context, *File, *memmap.MMapOpts) error { return syserror.ENODEV } +// UnstableAttr implements FileOperations.UnstableAttr. +func (i *Inotify) UnstableAttr(ctx context.Context, file *File) (UnstableAttr, error) { + return file.Dirent.Inode.UnstableAttr(ctx) +} + // Ioctl implements fs.FileOperations.Ioctl. func (i *Inotify) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { switch args[1].Int() { diff --git a/pkg/sentry/fs/proc/exec_args.go b/pkg/sentry/fs/proc/exec_args.go index 9daad5d2b..fc21dfbbd 100644 --- a/pkg/sentry/fs/proc/exec_args.go +++ b/pkg/sentry/fs/proc/exec_args.go @@ -77,15 +77,16 @@ func (i *execArgInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs. // +stateify savable type execArgFile struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopWrite `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // arg is the type of exec argument this file contains. arg execArgType diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index 3c471bad9..939ebaba1 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -91,7 +91,7 @@ func readDescriptors(t *kernel.Task, c *fs.DirCtx, offset int64, toDentAttr func // fd implements fs.InodeOperations for a file in /proc/TID/fd/. type fd struct { ramfs.Symlink - *fs.File + file *fs.File } var _ fs.InodeOperations = (*fd)(nil) @@ -103,7 +103,7 @@ func newFd(t *kernel.Task, f *fs.File, msrc *fs.MountSource) *fs.Inode { fd := &fd{ // RootOwner overridden by taskOwnedInodeOps.UnstableAttrs(). Symlink: *ramfs.NewSymlink(t, fs.RootOwner, ""), - File: f, + file: f, } return newProcInode(fd, msrc, fs.Symlink, t) } @@ -112,8 +112,8 @@ func newFd(t *kernel.Task, f *fs.File, msrc *fs.MountSource) *fs.Inode { // arguments are ignored. func (f *fd) GetFile(context.Context, *fs.Dirent, fs.FileFlags) (*fs.File, error) { // Take a reference on the fs.File. - f.File.IncRef() - return f.File, nil + f.file.IncRef() + return f.file, nil } // Readlink returns the current target. @@ -122,14 +122,14 @@ func (f *fd) Readlink(ctx context.Context, _ *fs.Inode) (string, error) { if root != nil { defer root.DecRef() } - n, _ := f.Dirent.FullName(root) + n, _ := f.file.Dirent.FullName(root) return n, nil } // Getlink implements fs.InodeOperations.Getlink. func (f *fd) Getlink(context.Context, *fs.Inode) (*fs.Dirent, error) { - f.Dirent.IncRef() - return f.Dirent, nil + f.file.Dirent.IncRef() + return f.file.Dirent, nil } // Truncate is ignored. @@ -139,12 +139,12 @@ func (f *fd) Truncate(context.Context, *fs.Inode, int64) error { func (f *fd) Release(ctx context.Context) { f.Symlink.Release(ctx) - f.File.DecRef() + f.file.DecRef() } // Close releases the reference on the file. func (f *fd) Close() error { - f.DecRef() + f.file.DecRef() return nil } @@ -212,7 +212,8 @@ func (f *fdDir) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFla // +stateify savable type fdDirFile struct { - fsutil.DirFileOperations `state:"nosave"` + fsutil.DirFileOperations `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` isInfoFile bool diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index c9e659533..64e1e1998 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -191,7 +191,8 @@ func (p *proc) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlag // // +stateify savable type rootProcFile struct { - fsutil.DirFileOperations `state:"nosave"` + fsutil.DirFileOperations `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` iops *proc } diff --git a/pkg/sentry/fs/proc/rpcinet_proc.go b/pkg/sentry/fs/proc/rpcinet_proc.go index 65faa21f2..81f64a28b 100644 --- a/pkg/sentry/fs/proc/rpcinet_proc.go +++ b/pkg/sentry/fs/proc/rpcinet_proc.go @@ -60,14 +60,15 @@ func (i *rpcInetInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs. // rpcInetFile implements fs.FileOperations as RPCs. type rpcInetFile struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` inode *rpcInetInode } diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index 16fc6789e..0a0eb45e2 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -183,14 +183,15 @@ func (s *SeqFile) updateSourceLocked(ctx context.Context, record int) { // // +stateify savable type seqFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` seqFile *SeqFile } diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index ee6b9f262..a7bc9198e 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -134,15 +134,16 @@ var _ fs.InodeOperations = (*hostname)(nil) // +stateify savable type hostnameFile struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoSeek `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSeek `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` } // Read implements fs.FileOperations.Read. diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index 42e9bc47f..728a46a74 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -85,14 +85,15 @@ func (m *tcpMemInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.F // +stateify savable type tcpMemFile struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` tcpMemInode *tcpMemInode } @@ -197,14 +198,15 @@ func (s *tcpSack) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileF // +stateify savable type tcpSackFile struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` tcpSack *tcpSack diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 4b1f84942..0edcdfce2 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -142,7 +142,8 @@ func (s *subtasks) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.File // +stateify savable type subtasksFile struct { - fsutil.DirFileOperations `state:"nosave"` + fsutil.DirFileOperations `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` t *kernel.Task pidns *kernel.PIDNamespace @@ -669,15 +670,16 @@ func (c *comm) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlag // +stateify savable type commFile struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` t *kernel.Task } @@ -724,15 +726,16 @@ func (a *auxvec) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFl // +stateify savable type auxvecFile struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` t *kernel.Task } diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index 0c68bbfc9..d433632cf 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -81,14 +81,15 @@ func (imio *idMapInodeOperations) GetFile(ctx context.Context, dirent *fs.Dirent // +stateify savable type idMapFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` iops *idMapInodeOperations } diff --git a/pkg/sentry/fs/proc/uptime.go b/pkg/sentry/fs/proc/uptime.go index 40d0fd1fd..d7ae26fcf 100644 --- a/pkg/sentry/fs/proc/uptime.go +++ b/pkg/sentry/fs/proc/uptime.go @@ -54,15 +54,16 @@ func (u *uptime) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFl // +stateify savable type uptimeFile struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` startTime ktime.Time } diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index a3b33c0f8..011cf3a16 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -388,7 +388,8 @@ func (*Dir) Rename(ctx context.Context, oldParent *fs.Inode, oldName string, new // // +stateify savable type dirFileOperations struct { - fsutil.DirFileOperations `state:"nosave"` + fsutil.DirFileOperations `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // dirCursor contains the name of the last directory entry that was // serialized. diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index 2c1295897..5bcb6c364 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -67,16 +67,17 @@ func (s *Socket) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFl // +stateify savable type socketFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoRead `state:"nosave"` - fsutil.FileNoSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoRead `state:"nosave"` + fsutil.FileNoSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` } var _ fs.FileOperations = (*socketFileOperations)(nil) diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index 47dae380b..35dabdad2 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -88,16 +88,17 @@ func (s *Symlink) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileF // +stateify savable type symlinkFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoRead `state:"nosave"` - fsutil.FileNoSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoRead `state:"nosave"` + fsutil.FileNoSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` } var _ fs.FileOperations = (*symlinkFileOperations)(nil) diff --git a/pkg/sentry/fs/timerfd/timerfd.go b/pkg/sentry/fs/timerfd/timerfd.go index b26466b9d..ef9a08854 100644 --- a/pkg/sentry/fs/timerfd/timerfd.go +++ b/pkg/sentry/fs/timerfd/timerfd.go @@ -33,12 +33,13 @@ import ( // // +stateify savable type TimerOperations struct { - fsutil.FileZeroSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileZeroSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` events waiter.Queue `state:"zerovalue"` timer *ktime.Timer diff --git a/pkg/sentry/fs/tmpfs/file_regular.go b/pkg/sentry/fs/tmpfs/file_regular.go index be6298130..d0c9b8bea 100644 --- a/pkg/sentry/fs/tmpfs/file_regular.go +++ b/pkg/sentry/fs/tmpfs/file_regular.go @@ -28,13 +28,14 @@ import ( // // +stateify savable type regularFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // iops is the InodeOperations of a regular tmpfs file. It is // guaranteed to be the same as file.Dirent.Inode.InodeOperations, diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index 11bf736d6..33b4c6438 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -285,13 +285,14 @@ func (d *dirInodeOperations) masterClose(t *Terminal) { // // +stateify savable type dirFileOperations struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // di is the inode operations. di *dirInodeOperations diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index b5e13ab36..7c256abb0 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -95,11 +95,12 @@ func (mi *masterInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flag // // +stateify savable type masterFileOperations struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // d is the containing dir. d *dirInodeOperations diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index 6dbce90b4..e8368bcdd 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -84,11 +84,12 @@ func (si *slaveInodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags // // +stateify savable type slaveFileOperations struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // si is the inode operations. si *slaveInodeOperations diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index 61c0fb7c5..befefb11c 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -98,12 +98,13 @@ func (p *pollEntry) WeakRefGone() { // // +stateify savable type EventPoll struct { - fsutil.FilePipeSeek `state:"zerovalue"` - fsutil.FileNotDirReaddir `state:"zerovalue"` - fsutil.FileNoFsync `state:"zerovalue"` - fsutil.FileNoopFlush `state:"zerovalue"` - fsutil.FileNoMMap `state:"zerovalue"` - fsutil.FileNoIoctl `state:"zerovalue"` + fsutil.FilePipeSeek `state:"zerovalue"` + fsutil.FileNotDirReaddir `state:"zerovalue"` + fsutil.FileNoFsync `state:"zerovalue"` + fsutil.FileNoopFlush `state:"zerovalue"` + fsutil.FileNoMMap `state:"zerovalue"` + fsutil.FileNoIoctl `state:"zerovalue"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // Wait queue is used to notify interested parties when the event poll // object itself becomes readable or writable. diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index 2d43c986d..b448ad813 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -38,13 +38,14 @@ import ( // // +stateify savable type EventOperations struct { - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` // Mutex that protects accesses to the fields of this event. mu sync.Mutex `state:"nosave"` diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index 028175530..1090432d7 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -35,11 +35,12 @@ import ( // // +stateify savable type ReaderWriter struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` *Pipe } diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index fabf0cbe4..18b7e90d8 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -52,14 +52,15 @@ func (f *fileContext) Value(key interface{}) interface{} { // byteReader implements fs.FileOperations for reading from a []byte source. type byteReader struct { - waiter.AlwaysReady `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FilePipeSeek `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` data []byte } diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index 5bcafad98..78f43178f 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -210,11 +210,12 @@ type commonEndpoint interface { // // +stateify savable type SocketOperations struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` socket.SendReceiveTimeout *waiter.Queue diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index a0a8a3220..be63823d8 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -46,11 +46,12 @@ const ( // socketOperations implements fs.FileOperations and socket.Socket for a socket // implemented using a host socket. type socketOperations struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` socket.SendReceiveTimeout family int // Read-only. diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index 7223773ad..2503a67c5 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -65,11 +65,12 @@ var netlinkSocketDevice = device.NewAnonDevice() // // +stateify savable type Socket struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` socket.SendReceiveTimeout // ports provides netlink port allocation. diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index 548a22f32..896b5b7ce 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -45,11 +45,12 @@ import ( // socketOperations implements fs.FileOperations and socket.Socket for a socket // implemented using a host socket. type socketOperations struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` socket.SendReceiveTimeout family int // Read-only. diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index c857a0f33..92411c901 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -45,11 +45,12 @@ import ( // // +stateify savable type SocketOperations struct { - fsutil.FilePipeSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` + fsutil.FilePipeSeek `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileUseInodeUnstableAttr `state:"nosave"` refs.AtomicRefCount socket.SendReceiveTimeout diff --git a/pkg/sentry/syscalls/linux/sys_stat.go b/pkg/sentry/syscalls/linux/sys_stat.go index bdfb9b3ef..02634b2dd 100644 --- a/pkg/sentry/syscalls/linux/sys_stat.go +++ b/pkg/sentry/syscalls/linux/sys_stat.go @@ -60,7 +60,7 @@ func Fstatat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca } defer file.DecRef() - return 0, nil, stat(t, file.Dirent, false, statAddr) + return 0, nil, fstat(t, file, statAddr) } return 0, nil, fileOpOn(t, fd, path, flags&linux.AT_SYMLINK_NOFOLLOW == 0, func(root *fs.Dirent, d *fs.Dirent) error { @@ -98,7 +98,7 @@ func Fstat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall } defer file.DecRef() - return 0, nil, stat(t, file.Dirent, false /* dirPath */, statAddr) + return 0, nil, fstat(t, file, statAddr) } // stat implements stat from the given *fs.Dirent. @@ -110,9 +110,26 @@ func stat(t *kernel.Task, d *fs.Dirent, dirPath bool, statAddr usermem.Addr) err if err != nil { return err } + return copyOutStat(t, statAddr, d.Inode.StableAttr, uattr) +} + +// fstat implements fstat for the given *fs.File. +func fstat(t *kernel.Task, f *fs.File, statAddr usermem.Addr) error { + uattr, err := f.UnstableAttr(t) + if err != nil { + return err + } + return copyOutStat(t, statAddr, f.Dirent.Inode.StableAttr, uattr) +} +// copyOutStat copies the attributes (sattr, uattr) to the struct stat at +// address dst in t's address space. It encodes the stat struct to bytes +// manually, as stat() is a very common syscall for many applications, and +// t.CopyObjectOut has noticeable performance impact due to its many slice +// allocations and use of reflection. +func copyOutStat(t *kernel.Task, dst usermem.Addr, sattr fs.StableAttr, uattr fs.UnstableAttr) error { var mode uint32 - switch d.Inode.StableAttr.Type { + switch sattr.Type { case fs.RegularFile, fs.SpecialFile: mode |= linux.ModeRegular case fs.Symlink: @@ -129,16 +146,12 @@ func stat(t *kernel.Task, d *fs.Dirent, dirPath bool, statAddr usermem.Addr) err mode |= linux.ModeSocket } - // We encode the stat struct to bytes manually, as stat() is a very - // common syscall for many applications, and t.CopyObjectOut has - // noticeable performance impact due to its many slice allocations and - // use of reflection. b := t.CopyScratchBuffer(int(linux.SizeOfStat))[:0] // Dev (uint64) - b = binary.AppendUint64(b, usermem.ByteOrder, uint64(d.Inode.StableAttr.DeviceID)) + b = binary.AppendUint64(b, usermem.ByteOrder, uint64(sattr.DeviceID)) // Ino (uint64) - b = binary.AppendUint64(b, usermem.ByteOrder, uint64(d.Inode.StableAttr.InodeID)) + b = binary.AppendUint64(b, usermem.ByteOrder, uint64(sattr.InodeID)) // Nlink (uint64) b = binary.AppendUint64(b, usermem.ByteOrder, uattr.Links) // Mode (uint32) @@ -150,11 +163,11 @@ func stat(t *kernel.Task, d *fs.Dirent, dirPath bool, statAddr usermem.Addr) err // Padding (uint32) b = binary.AppendUint32(b, usermem.ByteOrder, 0) // Rdev (uint64) - b = binary.AppendUint64(b, usermem.ByteOrder, uint64(linux.MakeDeviceID(d.Inode.StableAttr.DeviceFileMajor, d.Inode.StableAttr.DeviceFileMinor))) + b = binary.AppendUint64(b, usermem.ByteOrder, uint64(linux.MakeDeviceID(sattr.DeviceFileMajor, sattr.DeviceFileMinor))) // Size (uint64) b = binary.AppendUint64(b, usermem.ByteOrder, uint64(uattr.Size)) // Blksize (uint64) - b = binary.AppendUint64(b, usermem.ByteOrder, uint64(d.Inode.StableAttr.BlockSize)) + b = binary.AppendUint64(b, usermem.ByteOrder, uint64(sattr.BlockSize)) // Blocks (uint64) b = binary.AppendUint64(b, usermem.ByteOrder, uint64(uattr.Usage/512)) @@ -173,7 +186,7 @@ func stat(t *kernel.Task, d *fs.Dirent, dirPath bool, statAddr usermem.Addr) err b = binary.AppendUint64(b, usermem.ByteOrder, uint64(ctime.Sec)) b = binary.AppendUint64(b, usermem.ByteOrder, uint64(ctime.Nsec)) - _, err = t.CopyOutBytes(statAddr, b) + _, err := t.CopyOutBytes(dst, b) return err } -- cgit v1.2.3 From c8cee7108f1a1b37e89961c6dd69ccab97952c86 Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Wed, 17 Apr 2019 12:56:23 -0700 Subject: Use FD limit and file size limit from host FD limit and file size limit is read from the host, instead of using hard-coded defaults, given that they effect the sandbox process. Also limit the direct cache to use no more than half if the available FDs. PiperOrigin-RevId: 244050323 Change-Id: I787ad0fdf07c49d589e51aebfeae477324fe26e6 --- pkg/sentry/fs/BUILD | 1 + pkg/sentry/fs/context.go | 12 +++++ pkg/sentry/fs/dirent_cache.go | 43 ++++++++++++++--- pkg/sentry/fs/dirent_cache_limiter.go | 55 +++++++++++++++++++++ pkg/sentry/fs/dirent_cache_test.go | 90 +++++++++++++++++++++++++++++++++++ pkg/sentry/fs/gofer/session.go | 9 ++++ pkg/sentry/fs/mount.go | 20 ++++++-- pkg/sentry/fs/mount_overlay.go | 11 ++++- pkg/sentry/kernel/kernel.go | 9 ++++ pkg/sentry/kernel/task.go | 2 + runsc/boot/fs.go | 19 +++++++- runsc/boot/limits.go | 77 +++++++++++++++++++++++++++++- runsc/boot/loader.go | 4 ++ test/syscalls/linux/poll.cc | 9 ++++ 14 files changed, 347 insertions(+), 14 deletions(-) create mode 100644 pkg/sentry/fs/dirent_cache_limiter.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 1742d3a65..1fd9e30f6 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -12,6 +12,7 @@ go_library( "dentry.go", "dirent.go", "dirent_cache.go", + "dirent_cache_limiter.go", "dirent_list.go", "dirent_state.go", "event_list.go", diff --git a/pkg/sentry/fs/context.go b/pkg/sentry/fs/context.go index c0e6075e4..4869428a8 100644 --- a/pkg/sentry/fs/context.go +++ b/pkg/sentry/fs/context.go @@ -26,6 +26,9 @@ type contextID int const ( // CtxRoot is a Context.Value key for a Dirent. CtxRoot contextID = iota + + // CtxDirentCacheLimiter is a Context.Value key for DirentCacheLimiter. + CtxDirentCacheLimiter ) // ContextCanAccessFile determines whether `file` can be accessed in the requested way @@ -100,3 +103,12 @@ func RootFromContext(ctx context.Context) *Dirent { } return nil } + +// DirentCacheLimiterFromContext returns the DirentCacheLimiter used by ctx, or +// nil if ctx does not have a dirent cache limiter. +func DirentCacheLimiterFromContext(ctx context.Context) *DirentCacheLimiter { + if v := ctx.Value(CtxDirentCacheLimiter); v != nil { + return v.(*DirentCacheLimiter) + } + return nil +} diff --git a/pkg/sentry/fs/dirent_cache.go b/pkg/sentry/fs/dirent_cache.go index 502b0a09b..d26a06971 100644 --- a/pkg/sentry/fs/dirent_cache.go +++ b/pkg/sentry/fs/dirent_cache.go @@ -32,6 +32,10 @@ type DirentCache struct { // when cache is nil. maxSize uint64 + // limit restricts the number of entries in the cache amoung multiple caches. + // It may be nil if there are no global limit for this cache. + limit *DirentCacheLimiter + // mu protects currentSize and direntList. mu sync.Mutex `state:"nosave"` @@ -45,8 +49,7 @@ type DirentCache struct { list direntList `state:"zerovalue"` } -// NewDirentCache returns a new DirentCache with the given maxSize. If maxSize -// is 0, nil is returned. +// NewDirentCache returns a new DirentCache with the given maxSize. func NewDirentCache(maxSize uint64) *DirentCache { return &DirentCache{ maxSize: maxSize, @@ -71,15 +74,24 @@ func (c *DirentCache) Add(d *Dirent) { return } + // First check against the global limit. + for c.limit != nil && !c.limit.tryInc() { + if c.currentSize == 0 { + // If the global limit is reached, but there is nothing more to drop from + // this cache, there is not much else to do. + c.mu.Unlock() + return + } + c.remove(c.list.Back()) + } + // d is not in cache. Add it and take a reference. c.list.PushFront(d) d.IncRef() c.currentSize++ - // Remove the oldest until we are under the size limit. - for c.maxSize > 0 && c.currentSize > c.maxSize { - c.remove(c.list.Back()) - } + c.maybeShrink() + c.mu.Unlock() } @@ -92,6 +104,9 @@ func (c *DirentCache) remove(d *Dirent) { d.SetNext(nil) d.DecRef() c.currentSize-- + if c.limit != nil { + c.limit.dec() + } } // Remove removes the element from the cache and decrements its refCount. It @@ -142,3 +157,19 @@ func (c *DirentCache) Invalidate() { } c.mu.Unlock() } + +// setMaxSize sets cache max size. If current size is larger than max size, the +// cache shrinks to acommodate the new max. +func (c *DirentCache) setMaxSize(max uint64) { + c.mu.Lock() + c.maxSize = max + c.maybeShrink() + c.mu.Unlock() +} + +// shrink removes the oldest element until the list is under the size limit. +func (c *DirentCache) maybeShrink() { + for c.maxSize > 0 && c.currentSize > c.maxSize { + c.remove(c.list.Back()) + } +} diff --git a/pkg/sentry/fs/dirent_cache_limiter.go b/pkg/sentry/fs/dirent_cache_limiter.go new file mode 100644 index 000000000..024c7b2d5 --- /dev/null +++ b/pkg/sentry/fs/dirent_cache_limiter.go @@ -0,0 +1,55 @@ +// Copyright 2018 Google LLC +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package fs + +import ( + "fmt" + "sync" +) + +// DirentCacheLimiter acts as a global limit for all dirent caches in the +// process. +// +// +stateify savable +type DirentCacheLimiter struct { + mu sync.Mutex `state:"nosave"` + max uint64 + count uint64 `state:"zerovalue"` +} + +// NewDirentCacheLimiter creates a new DirentCacheLimiter. +func NewDirentCacheLimiter(max uint64) *DirentCacheLimiter { + return &DirentCacheLimiter{max: max} +} + +func (d *DirentCacheLimiter) tryInc() bool { + d.mu.Lock() + if d.count >= d.max { + d.mu.Unlock() + return false + } + d.count++ + d.mu.Unlock() + return true +} + +func (d *DirentCacheLimiter) dec() { + d.mu.Lock() + if d.count == 0 { + panic(fmt.Sprintf("underflowing DirentCacheLimiter count: %+v", d)) + } + d.count-- + d.mu.Unlock() +} diff --git a/pkg/sentry/fs/dirent_cache_test.go b/pkg/sentry/fs/dirent_cache_test.go index 5d0e9d91c..93e8d415f 100644 --- a/pkg/sentry/fs/dirent_cache_test.go +++ b/pkg/sentry/fs/dirent_cache_test.go @@ -120,6 +120,96 @@ func TestDirentCache(t *testing.T) { } } +func TestDirentCacheLimiter(t *testing.T) { + const ( + globalMaxSize = 5 + maxSize = 3 + ) + + limit := NewDirentCacheLimiter(globalMaxSize) + c1 := NewDirentCache(maxSize) + c1.limit = limit + c2 := NewDirentCache(maxSize) + c2.limit = limit + + // Create a Dirent d. + d := NewNegativeDirent("") + + // Add d to the cache. + c1.Add(d) + if got, want := c1.Size(), uint64(1); got != want { + t.Errorf("c1.Size() got %v, want %v", got, want) + } + + // Add maxSize-1 more elements. d should be oldest element. + for i := 0; i < maxSize-1; i++ { + c1.Add(NewNegativeDirent("")) + } + if got, want := c1.Size(), uint64(maxSize); got != want { + t.Errorf("c1.Size() got %v, want %v", got, want) + } + + // Check that d is still there. + if got, want := c1.contains(d), true; got != want { + t.Errorf("c1.contains(d) got %v want %v", got, want) + } + + // Fill up the other cache, it will start dropping old entries from the cache + // when the global limit is reached. + for i := 0; i < maxSize; i++ { + c2.Add(NewNegativeDirent("")) + } + + // Check is what's remaining from global max. + if got, want := c2.Size(), globalMaxSize-maxSize; int(got) != want { + t.Errorf("c2.Size() got %v, want %v", got, want) + } + + // Check that d was not dropped. + if got, want := c1.contains(d), true; got != want { + t.Errorf("c1.contains(d) got %v want %v", got, want) + } + + // Add an entry that will eventually be dropped. Check is done later... + drop := NewNegativeDirent("") + c1.Add(drop) + + // Check that d is bumped to front even when global limit is reached. + c1.Add(d) + if got, want := c1.contains(d), true; got != want { + t.Errorf("c1.contains(d) got %v want %v", got, want) + } + + // Add 2 more element and check that: + // - d is still in the list: to verify that d was bumped + // - d2/d3 are in the list: older entries are dropped when global limit is + // reached. + // - drop is not in the list: indeed older elements are dropped. + d2 := NewNegativeDirent("") + c1.Add(d2) + d3 := NewNegativeDirent("") + c1.Add(d3) + if got, want := c1.contains(d), true; got != want { + t.Errorf("c1.contains(d) got %v want %v", got, want) + } + if got, want := c1.contains(d2), true; got != want { + t.Errorf("c1.contains(d2) got %v want %v", got, want) + } + if got, want := c1.contains(d3), true; got != want { + t.Errorf("c1.contains(d3) got %v want %v", got, want) + } + if got, want := c1.contains(drop), false; got != want { + t.Errorf("c1.contains(drop) got %v want %v", got, want) + } + + // Drop all entries from one cache. The other will be allowed to grow. + c1.Invalidate() + c2.Add(NewNegativeDirent("")) + if got, want := c2.Size(), uint64(maxSize); got != want { + t.Errorf("c2.Size() got %v, want %v", got, want) + } +} + // TestNilDirentCache tests that a nil cache supports all cache operations, but // treats them as noop. func TestNilDirentCache(t *testing.T) { diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index d626b86f5..ed5147c65 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -28,6 +28,10 @@ import ( "gvisor.googlesource.com/gvisor/pkg/unet" ) +// DefaultDirentCacheSize is the default dirent cache size for 9P mounts. It can +// be adjusted independentely from the other dirent caches. +var DefaultDirentCacheSize uint64 = fs.DefaultDirentCacheSize + // +stateify savable type endpointMaps struct { // mu protexts the direntMap, the keyMap, and the pathMap below. @@ -249,6 +253,11 @@ func Root(ctx context.Context, dev string, filesystem fs.Filesystem, superBlockF // Construct the MountSource with the session and superBlockFlags. m := fs.NewMountSource(s, filesystem, superBlockFlags) + // Given that gofer files can consume host FDs, restrict the number + // of files that can be held by the cache. + m.SetDirentCacheMaxSize(DefaultDirentCacheSize) + m.SetDirentCacheLimiter(fs.DirentCacheLimiterFromContext(ctx)) + // Send the Tversion request. s.client, err = p9.NewClient(conn, s.msize, s.version) if err != nil { diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index 5cc777bef..1e245ae5f 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -151,9 +151,9 @@ type MountSource struct { children map[*MountSource]struct{} } -// defaultDirentCacheSize is the number of Dirents that the VFS can hold an extra -// reference on. -const defaultDirentCacheSize uint64 = 1000 +// DefaultDirentCacheSize is the number of Dirents that the VFS can hold an +// extra reference on. +const DefaultDirentCacheSize uint64 = 1000 // NewMountSource returns a new MountSource. Filesystem may be nil if there is no // filesystem backing the mount. @@ -162,7 +162,7 @@ func NewMountSource(mops MountSourceOperations, filesystem Filesystem, flags Mou MountSourceOperations: mops, Flags: flags, Filesystem: filesystem, - fscache: NewDirentCache(defaultDirentCacheSize), + fscache: NewDirentCache(DefaultDirentCacheSize), children: make(map[*MountSource]struct{}), } } @@ -246,6 +246,18 @@ func (msrc *MountSource) FlushDirentRefs() { msrc.fscache.Invalidate() } +// SetDirentCacheMaxSize sets the max size to the dirent cache associated with +// this mount source. +func (msrc *MountSource) SetDirentCacheMaxSize(max uint64) { + msrc.fscache.setMaxSize(max) +} + +// SetDirentCacheLimiter sets the limiter objcet to the dirent cache associated +// with this mount source. +func (msrc *MountSource) SetDirentCacheLimiter(l *DirentCacheLimiter) { + msrc.fscache.limit = l +} + // NewCachingMountSource returns a generic mount that will cache dirents // aggressively. func NewCachingMountSource(filesystem Filesystem, flags MountSourceFlags) *MountSource { diff --git a/pkg/sentry/fs/mount_overlay.go b/pkg/sentry/fs/mount_overlay.go index 4c89673b5..fb60a1aec 100644 --- a/pkg/sentry/fs/mount_overlay.go +++ b/pkg/sentry/fs/mount_overlay.go @@ -31,10 +31,19 @@ type overlayMountSourceOperations struct { func newOverlayMountSource(upper, lower *MountSource, flags MountSourceFlags) *MountSource { upper.IncRef() lower.IncRef() - return NewMountSource(&overlayMountSourceOperations{ + msrc := NewMountSource(&overlayMountSourceOperations{ upper: upper, lower: lower, }, &overlayFilesystem{}, flags) + + // Use the minimum number to keep resource usage under limits. + size := lower.fscache.maxSize + if size > upper.fscache.maxSize { + size = upper.fscache.maxSize + } + msrc.fscache.setMaxSize(size) + + return msrc } // Revalidate implements MountSourceOperations.Revalidate for an overlay by diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index b8953657c..290c4a53c 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -188,6 +188,11 @@ type Kernel struct { // deviceRegistry is used to save/restore device.SimpleDevices. deviceRegistry struct{} `state:".(*device.Registry)"` + + // DirentCacheLimiter controls the number of total dirent entries can be in + // caches. Not all caches use it, only the caches that use host resources use + // the limiter. It may be nil if disabled. + DirentCacheLimiter *fs.DirentCacheLimiter } // InitKernelArgs holds arguments to Init. @@ -626,6 +631,8 @@ func (ctx *createProcessContext) Value(key interface{}) interface{} { return ctx.k.mounts.Root() } return nil + case fs.CtxDirentCacheLimiter: + return ctx.k.DirentCacheLimiter case ktime.CtxRealtimeClock: return ctx.k.RealtimeClock() case limits.CtxLimits: @@ -1170,6 +1177,8 @@ func (ctx supervisorContext) Value(key interface{}) interface{} { return auth.NewRootCredentials(ctx.k.rootUserNamespace) case fs.CtxRoot: return ctx.k.mounts.Root() + case fs.CtxDirentCacheLimiter: + return ctx.k.DirentCacheLimiter case ktime.CtxRealtimeClock: return ctx.k.RealtimeClock() case limits.CtxLimits: diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index 9c365e781..ed2175c37 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -601,6 +601,8 @@ func (t *Task) Value(key interface{}) interface{} { return int32(t.ThreadGroup().ID()) case fs.CtxRoot: return t.fsc.RootDirectory() + case fs.CtxDirentCacheLimiter: + return t.k.DirentCacheLimiter case inet.CtxStack: return t.NetworkContext() case ktime.CtxRealtimeClock: diff --git a/runsc/boot/fs.go b/runsc/boot/fs.go index 8dfb6dce6..761142d98 100644 --- a/runsc/boot/fs.go +++ b/runsc/boot/fs.go @@ -20,10 +20,10 @@ import ( "path/filepath" "strconv" "strings" + "syscall" // Include filesystem types that OCI spec might mount. _ "gvisor.googlesource.com/gvisor/pkg/sentry/fs/dev" - _ "gvisor.googlesource.com/gvisor/pkg/sentry/fs/gofer" _ "gvisor.googlesource.com/gvisor/pkg/sentry/fs/host" _ "gvisor.googlesource.com/gvisor/pkg/sentry/fs/proc" _ "gvisor.googlesource.com/gvisor/pkg/sentry/fs/sys" @@ -38,6 +38,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs/gofer" "gvisor.googlesource.com/gvisor/pkg/sentry/fs/ramfs" "gvisor.googlesource.com/gvisor/pkg/syserror" "gvisor.googlesource.com/gvisor/runsc/specutils" @@ -81,6 +82,22 @@ func (f *fdDispenser) empty() bool { return len(f.fds) == 0 } +func adjustDirentCache(k *kernel.Kernel) error { + var hl syscall.Rlimit + if err := syscall.Getrlimit(syscall.RLIMIT_NOFILE, &hl); err != nil { + return fmt.Errorf("getting RLIMIT_NOFILE: %v", err) + } + if int64(hl.Cur) != syscall.RLIM_INFINITY { + newSize := hl.Cur / 2 + if newSize < gofer.DefaultDirentCacheSize { + log.Infof("Setting gofer dirent cache size to %d", newSize) + gofer.DefaultDirentCacheSize = newSize + k.DirentCacheLimiter = fs.NewDirentCacheLimiter(newSize) + } + } + return nil +} + // setupRootContainerFS creates a mount namespace containing the root filesystem // and all mounts. 'rootCtx' is used to walk directories to find mount points. // 'setMountNS' is called after namespace is created. It must set the mount NS diff --git a/runsc/boot/limits.go b/runsc/boot/limits.go index e3e716bf9..32e62cdf7 100644 --- a/runsc/boot/limits.go +++ b/runsc/boot/limits.go @@ -16,8 +16,11 @@ package boot import ( "fmt" + "sync" + "syscall" specs "github.com/opencontainers/runtime-spec/specs-go" + "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" ) @@ -41,10 +44,43 @@ var fromLinuxResource = map[string]limits.LimitType{ "RLIMIT_STACK": limits.Stack, } -func createLimitSet(spec *specs.Spec) (*limits.LimitSet, error) { +func findName(lt limits.LimitType) string { + for k, v := range fromLinuxResource { + if v == lt { + return k + } + } + return "unknown" +} + +var defaults defs + +type defs struct { + mu sync.Mutex + set *limits.LimitSet + err error +} + +func (d *defs) get() (*limits.LimitSet, error) { + d.mu.Lock() + defer d.mu.Unlock() + + if d.err != nil { + return nil, d.err + } + if d.set == nil { + if err := d.initDefaults(); err != nil { + d.err = err + return nil, err + } + } + return d.set, nil +} + +func (d *defs) initDefaults() error { ls, err := limits.NewLinuxLimitSet() if err != nil { - return nil, err + return err } // Set default limits based on what containers get by default, ex: @@ -66,6 +102,43 @@ func createLimitSet(spec *specs.Spec) (*limits.LimitSet, error) { ls.SetUnchecked(limits.SignalsPending, limits.Limit{Cur: 0, Max: 0}) ls.SetUnchecked(limits.Stack, limits.Limit{Cur: 8388608, Max: limits.Infinity}) + // Read host limits that directly affect the sandbox and adjust the defaults + // based on them. + for _, res := range []int{syscall.RLIMIT_FSIZE, syscall.RLIMIT_NOFILE} { + var hl syscall.Rlimit + if err := syscall.Getrlimit(res, &hl); err != nil { + return err + } + + lt, ok := limits.FromLinuxResource[res] + if !ok { + return fmt.Errorf("unknown rlimit type %v", res) + } + hostLimit := limits.Limit{ + Cur: limits.FromLinux(hl.Cur), + Max: limits.FromLinux(hl.Max), + } + + defaultLimit := ls.Get(lt) + if hostLimit.Cur != limits.Infinity && hostLimit.Cur < defaultLimit.Cur { + log.Warningf("Host limit is lower than recommended, resource: %q, host: %d, recommended: %d", findName(lt), hostLimit.Cur, defaultLimit.Cur) + } + if hostLimit.Cur != defaultLimit.Cur || hostLimit.Max != defaultLimit.Max { + log.Infof("Setting limit from host, resource: %q {soft: %d, hard: %d}", findName(lt), hostLimit.Cur, hostLimit.Max) + ls.SetUnchecked(lt, hostLimit) + } + } + + d.set = ls + return nil +} + +func createLimitSet(spec *specs.Spec) (*limits.LimitSet, error) { + ls, err := defaults.get() + if err != nil { + return nil, err + } + // Then apply overwrites on top of defaults. for _, rl := range spec.Process.Rlimits { lt, ok := fromLinuxResource[rl.Type] diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 56cb137f0..88a834aa5 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -274,6 +274,10 @@ func New(args Args) (*Loader, error) { return nil, fmt.Errorf("initializing kernel: %v", err) } + if err := adjustDirentCache(k); err != nil { + return nil, err + } + // Turn on packet logging if enabled. if args.Conf.LogPackets { log.Infof("Packet logging enabled") diff --git a/test/syscalls/linux/poll.cc b/test/syscalls/linux/poll.cc index 7a6a39444..67a86cc22 100644 --- a/test/syscalls/linux/poll.cc +++ b/test/syscalls/linux/poll.cc @@ -255,7 +255,16 @@ TEST_F(PollTest, Nfds) { // Stash value of RLIMIT_NOFILES. struct rlimit rlim; TEST_PCHECK(getrlimit(RLIMIT_NOFILE, &rlim) == 0); + + // gVisor caps the number of FDs that epoll can use beyond RLIMIT_NOFILE. + constexpr rlim_t gVisorMax = 1048576; + if (rlim.rlim_cur > gVisorMax) { + rlim.rlim_cur = gVisorMax; + TEST_PCHECK(setrlimit(RLIMIT_NOFILE, &rlim) == 0); + } + rlim_t max_fds = rlim.rlim_cur; + LOG(INFO) << "Using limit: " << max_fds; // Create an eventfd. Since its value is initially zero, it is writable. FileDescriptor efd = ASSERT_NO_ERRNO_AND_VALUE(NewEventFD()); -- cgit v1.2.3 From b52cbd60280342f25411561702e97fe650fdaa9c Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Wed, 17 Apr 2019 13:42:16 -0700 Subject: Don't allow sigtimedwait to catch unblockable signals The existing logic attempting to do this is incorrect. Unary ^ has higher precedence than &^, so mask always has UnblockableSignals cleared, allowing dequeueSignalLocked to dequeue unblockable signals (which allows userspace to ignore them). Switch the logic so that unblockable signals are always masked. PiperOrigin-RevId: 244058487 Change-Id: Ib19630ac04068a1fbfb9dc4a8eab1ccbdb21edc3 --- pkg/sentry/kernel/task_signals.go | 2 +- test/syscalls/linux/BUILD | 1 + test/syscalls/linux/sigtimedwait.cc | 76 +++++++++++++++++++++++++++++++++++++ 3 files changed, 78 insertions(+), 1 deletion(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index e177562d7..3a8e61900 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -307,7 +307,7 @@ func (t *Task) SignalReturn(rt bool) (*SyscallControl, error) { func (t *Task) Sigtimedwait(set linux.SignalSet, timeout time.Duration) (*arch.SignalInfo, error) { // set is the set of signals we're interested in; invert it to get the set // of signals to block. - mask := ^set &^ UnblockableSignals + mask := ^(set &^ UnblockableSignals) t.tg.signalHandlers.mu.Lock() defer t.tg.signalHandlers.mu.Unlock() diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index 38faba267..d99733fc9 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -1825,6 +1825,7 @@ cc_binary( srcs = ["sigtimedwait.cc"], linkstatic = 1, deps = [ + "//test/util:file_descriptor", "//test/util:logging", "//test/util:signal_util", "//test/util:test_util", diff --git a/test/syscalls/linux/sigtimedwait.cc b/test/syscalls/linux/sigtimedwait.cc index 3a350fc28..1df9c013f 100644 --- a/test/syscalls/linux/sigtimedwait.cc +++ b/test/syscalls/linux/sigtimedwait.cc @@ -18,6 +18,7 @@ #include "gtest/gtest.h" #include "absl/time/clock.h" #include "absl/time/time.h" +#include "test/util/file_descriptor.h" #include "test/util/logging.h" #include "test/util/signal_util.h" #include "test/util/test_util.h" @@ -163,6 +164,81 @@ TEST(SigtimedwaitTest, ChildExitGeneratedSIGCHLDWithHandler) { EXPECT_TRUE(WIFEXITED(status) && WEXITSTATUS(status) == 0) << status; } +// sigtimedwait cannot catch SIGKILL. +TEST(SigtimedwaitTest, SIGKILLUncaught) { + // This is a regression test for sigtimedwait dequeuing SIGKILLs, thus + // preventing the task from exiting. + // + // The explanation below is specific to behavior in gVisor. The Linux behavior + // here is irrelevant because without a bug that prevents delivery of SIGKILL, + // none of this behavior is visible (in Linux or gVisor). + // + // SIGKILL is rather intrusive. Simply sending the SIGKILL marks + // ThreadGroup.exitStatus as exiting with SIGKILL, before the SIGKILL is even + // delivered. + // + // As a result, we cannot simply exit the child with a different exit code if + // it survives and expect to see that code in waitpid because: + // 1. PrepareGroupExit will override Task.exitStatus with + // ThreadGroup.exitStatus. + // 2. waitpid(2) will always return ThreadGroup.exitStatus rather than + // Task.exitStatus. + // + // We could use exit(2) to set Task.exitStatus without override, and a SIGCHLD + // handler to receive Task.exitStatus in the parent, but with that much + // test complexity, it is cleaner to simply use a pipe to notify the parent + // that we survived. + constexpr auto kSigtimedwaitSetupTime = absl::Seconds(2); + + int pipe_fds[2]; + ASSERT_THAT(pipe(pipe_fds), SyscallSucceeds()); + FileDescriptor rfd(pipe_fds[0]); + FileDescriptor wfd(pipe_fds[1]); + + pid_t pid = fork(); + if (pid == 0) { + rfd.reset(); + + sigset_t mask; + sigemptyset(&mask); + sigaddset(&mask, SIGKILL); + RetryEINTR(sigtimedwait)(&mask, nullptr, nullptr); + + // Survived. + char c = 'a'; + TEST_PCHECK(WriteFd(wfd.get(), &c, 1) == 1); + _exit(1); + } + ASSERT_THAT(pid, SyscallSucceeds()); + + wfd.reset(); + + // Wait for child to block in sigtimedwait, then kill it. + absl::SleepFor(kSigtimedwaitSetupTime); + + // Sending SIGKILL will attempt to enqueue the signal twice: once in the + // normal signal sending path, and once to all Tasks in the ThreadGroup when + // applying SIGKILL side-effects. + // + // If we use kill(2), the former will be on the ThreadGroup signal queue and + // the latter will be on the Task signal queue. sigtimedwait can only dequeue + // one signal, so the other would kill the Task, masking bugs. + // + // If we use tkill(2), the former will be on the Task signal queue and the + // latter will be dropped as a duplicate. Then sigtimedwait can theoretically + // dequeue the single SIGKILL. + EXPECT_THAT(syscall(SYS_tkill, pid, SIGKILL), SyscallSucceeds()); + + int status; + EXPECT_THAT(RetryEINTR(waitpid)(pid, &status, 0), + SyscallSucceedsWithValue(pid)); + EXPECT_TRUE(WIFSIGNALED(status) && WTERMSIG(status) == SIGKILL) << status; + + // Child shouldn't have survived. + char c; + EXPECT_THAT(ReadFd(rfd.get(), &c, 1), SyscallSucceedsWithValue(0)); +} + TEST(SigtimedwaitTest, IgnoredUnmaskedSignal) { constexpr int kSigno = SIGUSR1; constexpr auto kSigtimedwaitSetupTime = absl::Seconds(2); -- cgit v1.2.3 From f86c35a51ff92718e36ff6075339300be11e09b3 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Mon, 22 Apr 2019 18:17:25 -0700 Subject: Clean up state error handling PiperOrigin-RevId: 244773836 Change-Id: I32223f79d2314fe1ac4ddfc63004fc22ff634adf --- pkg/sentry/control/state.go | 2 +- pkg/sentry/kernel/kernel.go | 23 ++++++++++++----------- pkg/state/state.go | 20 +++++++++++++++----- 3 files changed, 28 insertions(+), 17 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/control/state.go b/pkg/sentry/control/state.go index 0a480c84a..b6bbf69fa 100644 --- a/pkg/sentry/control/state.go +++ b/pkg/sentry/control/state.go @@ -64,7 +64,7 @@ func (s *State) Save(o *SaveOpts, _ *struct{}) error { log.Infof("Save succeeded: exiting...") } else { log.Warningf("Save failed: exiting...") - s.Kernel.SetExitError(err) + s.Kernel.SetSaveError(err) } s.Kernel.Kill(kernel.ExitStatus{}) }, diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 290c4a53c..ee6334509 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -175,9 +175,9 @@ type Kernel struct { // netlinkPorts manages allocation of netlink socket port IDs. netlinkPorts *port.Manager - // exitErr is the error causing the sandbox to exit, if any. It is - // protected by extMu. - exitErr error `state:"nosave"` + // saveErr is the error causing the sandbox to exit during save, if + // any. It is protected by extMu. + saveErr error `state:"nosave"` // danglingEndpoints is used to save / restore tcpip.DanglingEndpoints. danglingEndpoints struct{} `state:".([]tcpip.Endpoint)"` @@ -1029,20 +1029,21 @@ func (k *Kernel) NetlinkPorts() *port.Manager { return k.netlinkPorts } -// ExitError returns the sandbox error that caused the kernel to exit. -func (k *Kernel) ExitError() error { +// SaveError returns the sandbox error that caused the kernel to exit during +// save. +func (k *Kernel) SaveError() error { k.extMu.Lock() defer k.extMu.Unlock() - return k.exitErr + return k.saveErr } -// SetExitError sets the sandbox error that caused the kernel to exit, if one is -// not already set. -func (k *Kernel) SetExitError(err error) { +// SetSaveError sets the sandbox error that caused the kernel to exit during +// save, if one is not already set. +func (k *Kernel) SetSaveError(err error) { k.extMu.Lock() defer k.extMu.Unlock() - if k.exitErr == nil { - k.exitErr = err + if k.saveErr == nil { + k.saveErr = err } } diff --git a/pkg/state/state.go b/pkg/state/state.go index 4b141777e..4486f83a7 100644 --- a/pkg/state/state.go +++ b/pkg/state/state.go @@ -60,8 +60,8 @@ import ( // ErrState is returned when an error is encountered during encode/decode. type ErrState struct { - // Err is the underlying error. - Err error + // err is the underlying error. + err error // path is the visit path from root to the current object. path string @@ -72,7 +72,17 @@ type ErrState struct { // Error returns a sensible description of the state error. func (e *ErrState) Error() string { - return fmt.Sprintf("%v:\nstate path: %s\n%s", e.Err, e.path, e.trace) + return fmt.Sprintf("%v:\nstate path: %s\n%s", e.err, e.path, e.trace) +} + +// UnwrapErrState returns the underlying error in ErrState. +// +// If err is not *ErrState, err is returned directly. +func UnwrapErrState(err error) error { + if e, ok := err.(*ErrState); ok { + return e.err + } + return err } // Save saves the given object state. @@ -318,9 +328,9 @@ func (sr *recoverable) safely(fn func()) (err error) { if r := recover(); r != nil { es := new(ErrState) if e, ok := r.(error); ok { - es.Err = e + es.err = e } else { - es.Err = fmt.Errorf("%v", r) + es.err = fmt.Errorf("%v", r) } es.path = sr.path() -- cgit v1.2.3 From f17cfa4d53742923b5c91b149b82a05bcda3ea20 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Thu, 25 Apr 2019 17:45:56 -0700 Subject: Perform explicit CPUID and FP state compatibility checks on restore PiperOrigin-RevId: 245341004 Change-Id: Ic4d581039d034a8ae944b43e45e84eb2c3973657 --- pkg/cpuid/cpuid.go | 34 ++++++++++++++++++++++ pkg/sentry/arch/arch_state_x86.go | 59 ++++++++++++++++++++++++++++++--------- pkg/sentry/kernel/kernel.go | 30 ++++++++++++++++++++ 3 files changed, 110 insertions(+), 13 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/cpuid/cpuid.go b/pkg/cpuid/cpuid.go index 64e2e68f1..61441150e 100644 --- a/pkg/cpuid/cpuid.go +++ b/pkg/cpuid/cpuid.go @@ -446,6 +446,20 @@ const ( extendedFeatures // Returns some extended feature bits in edx and ecx. ) +// These are the extended floating point state features. They are used to +// enumerate floating point features in XCR0, XSTATE_BV, etc. +const ( + XSAVEFeatureX87 = 1 << 0 + XSAVEFeatureSSE = 1 << 1 + XSAVEFeatureAVX = 1 << 2 + XSAVEFeatureBNDREGS = 1 << 3 + XSAVEFeatureBNDCSR = 1 << 4 + XSAVEFeatureAVX512op = 1 << 5 + XSAVEFeatureAVX512zmm0 = 1 << 6 + XSAVEFeatureAVX512zmm16 = 1 << 7 + XSAVEFeaturePKRU = 1 << 9 +) + var cpuFreqMHz float64 // x86FeaturesFromString includes features from x86FeatureStrings and @@ -561,6 +575,26 @@ func (fs *FeatureSet) Intel() bool { return fs.VendorID == "GenuineIntel" } +// ErrIncompatible is returned by FeatureSet.HostCompatible if fs is not a +// subset of the host feature set. +type ErrIncompatible struct { + message string +} + +// Error implements error. +func (e ErrIncompatible) Error() string { + return e.message +} + +// CheckHostCompatible returns nil if fs is a subset of the host feature set. +func (fs *FeatureSet) CheckHostCompatible() error { + hfs := HostFeatureSet() + if diff := fs.Subtract(hfs); diff != nil { + return ErrIncompatible{fmt.Sprintf("CPU feature set %v incompatible with host feature set %v (missing: %v)", fs.FlagsString(false), hfs.FlagsString(false), diff)} + } + return nil +} + // Helper to convert 3 regs into 12-byte vendor ID. func vendorIDFromRegs(bx, cx, dx uint32) string { bytes := make([]byte, 0, 12) diff --git a/pkg/sentry/arch/arch_state_x86.go b/pkg/sentry/arch/arch_state_x86.go index 604bd08a6..01949049d 100644 --- a/pkg/sentry/arch/arch_state_x86.go +++ b/pkg/sentry/arch/arch_state_x86.go @@ -15,14 +15,31 @@ package arch import ( - "sync" + "fmt" "syscall" - "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/cpuid" + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" ) -// warnOnce is used to warn about truncated state only once. -var warnOnce sync.Once +// ErrFloatingPoint indicates a failed restore due to unusable floating point +// state. +type ErrFloatingPoint struct { + // supported is the supported floating point state. + supported uint64 + + // saved is the saved floating point state. + saved uint64 +} + +// Error returns a sensible description of the restore error. +func (e ErrFloatingPoint) Error() string { + return fmt.Sprintf("floating point state contains unsupported features; supported: %#x saved: %#x", e.supported, e.saved) +} + +// XSTATE_BV does not exist if FXSAVE is used, but FXSAVE implicitly saves x87 +// and SSE state, so this is the equivalent XSTATE_BV value. +const fxsaveBV uint64 = cpuid.XSAVEFeatureX87 | cpuid.XSAVEFeatureSSE // afterLoad is invoked by stateify. func (s *State) afterLoad() { @@ -33,7 +50,8 @@ func (s *State) afterLoad() { // state that may be saved by the new CPU. Even if extraneous new state // is saved, the state we care about is guaranteed to be a subset of // new state. Later optimizations can use less space when using a - // smaller state component bitmap. Intel SDM section 13 has more info. + // smaller state component bitmap. Intel SDM Volume 1 Chapter 13 has + // more info. s.x86FPState = newX86FPState() // x86FPState always contains all the FP state supported by the host. @@ -41,15 +59,30 @@ func (s *State) afterLoad() { // which we cannot restore. // // The x86 FP state areas are backwards compatible, so we can simply - // truncate the additional floating point state. Applications should - // not depend on the truncated state because it should relate only to - // features that were not exposed in the app FeatureSet. + // truncate the additional floating point state. + // + // Applications should not depend on the truncated state because it + // should relate only to features that were not exposed in the app + // FeatureSet. However, because we do not *prevent* them from using + // this state, we must verify here that there is no in-use state + // (according to XSTATE_BV) which we do not support. if len(s.x86FPState) < len(old) { - warnOnce.Do(func() { - // This will occur on every instance of state, don't - // bother warning more than once. - log.Infof("dropping %d bytes of floating point state; the application should not depend on this state", len(old)-len(s.x86FPState)) - }) + // What do we support? + supportedBV := fxsaveBV + if fs := cpuid.HostFeatureSet(); fs.UseXsave() { + supportedBV = fs.ValidXCR0Mask() + } + + // What was in use? + savedBV := fxsaveBV + if len(old) >= xstateBVOffset+8 { + savedBV = usermem.ByteOrder.Uint64(old[xstateBVOffset:]) + } + + // Supported features must be a superset of saved features. + if savedBV&^supportedBV != 0 { + panic(ErrFloatingPoint{supported: supportedBV, saved: savedBV}) + } } // Copy to the new, aligned location. diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index ee6334509..a1b2d7161 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -337,6 +337,17 @@ func (k *Kernel) SaveTo(w io.Writer) error { return fmt.Errorf("failed to invalidate unsavable mappings: %v", err) } + // Save the CPUID FeatureSet before the rest of the kernel so we can + // verify its compatibility on restore before attempting to restore the + // entire kernel, which may fail on an incompatible machine. + // + // N.B. This will also be saved along with the full kernel save below. + cpuidStart := time.Now() + if err := state.Save(w, k.FeatureSet(), nil); err != nil { + return err + } + log.Infof("CPUID save took [%s].", time.Since(cpuidStart)) + // Save the kernel state. kernelStart := time.Now() var stats state.Stats @@ -469,6 +480,25 @@ func (k *Kernel) LoadFrom(r io.Reader, net inet.Stack) error { initAppCores := k.applicationCores + // Load the pre-saved CPUID FeatureSet. + // + // N.B. This was also saved along with the full kernel below, so we + // don't need to explicitly install it in the Kernel. + cpuidStart := time.Now() + var features cpuid.FeatureSet + if err := state.Load(r, &features, nil); err != nil { + return err + } + log.Infof("CPUID load took [%s].", time.Since(cpuidStart)) + + // Verify that the FeatureSet is usable on this host. We do this before + // Kernel load so that the explicit CPUID mismatch error has priority + // over floating point state restore errors that may occur on load on + // an incompatible machine. + if err := features.CheckHostCompatible(); err != nil { + return err + } + // Load the kernel state. kernelStart := time.Now() var stats state.Stats -- cgit v1.2.3 From f4ce43e1f426148d99c28c1b0e5c43ddda17a8cb Mon Sep 17 00:00:00 2001 From: Nicolas Lacasse Date: Mon, 29 Apr 2019 14:03:04 -0700 Subject: Allow and document bug ids in gVisor codebase. PiperOrigin-RevId: 245818639 Change-Id: I03703ef0fb9b6675955637b9fe2776204c545789 --- CONTRIBUTING.md | 7 +++ pkg/cpuid/cpuid_test.go | 2 +- pkg/dhcp/client.go | 2 +- pkg/log/glog.go | 2 +- pkg/metric/metric.go | 4 +- pkg/segment/set.go | 2 +- pkg/segment/test/set_functions.go | 2 +- pkg/sentry/arch/arch.go | 2 +- pkg/sentry/arch/arch_amd64.go | 4 +- pkg/sentry/arch/arch_x86.go | 2 +- pkg/sentry/arch/signal_amd64.go | 6 +-- pkg/sentry/arch/stack.go | 6 +-- pkg/sentry/context/context.go | 2 +- pkg/sentry/control/proc.go | 2 +- pkg/sentry/fs/README.md | 2 +- pkg/sentry/fs/ashmem/area.go | 4 +- pkg/sentry/fs/binder/binder.go | 22 ++++---- pkg/sentry/fs/dentry.go | 2 +- pkg/sentry/fs/dirent.go | 8 +-- pkg/sentry/fs/file.go | 2 +- pkg/sentry/fs/file_overlay.go | 4 +- pkg/sentry/fs/fsutil/file.go | 8 +-- pkg/sentry/fs/fsutil/inode_cached.go | 4 +- pkg/sentry/fs/gofer/cache_policy.go | 4 +- pkg/sentry/fs/gofer/file.go | 2 +- pkg/sentry/fs/gofer/file_state.go | 2 +- pkg/sentry/fs/gofer/handles.go | 2 +- pkg/sentry/fs/gofer/inode.go | 6 +-- pkg/sentry/fs/gofer/inode_state.go | 2 +- pkg/sentry/fs/gofer/session.go | 2 +- pkg/sentry/fs/gofer/session_state.go | 2 +- pkg/sentry/fs/host/fs.go | 4 +- pkg/sentry/fs/host/inode.go | 10 ++-- pkg/sentry/fs/inode.go | 6 +-- pkg/sentry/fs/inode_operations.go | 2 +- pkg/sentry/fs/inode_overlay.go | 6 +-- pkg/sentry/fs/mount.go | 4 +- pkg/sentry/fs/mount_test.go | 2 +- pkg/sentry/fs/proc/README.md | 12 ++--- pkg/sentry/fs/proc/fds.go | 2 +- pkg/sentry/fs/proc/loadavg.go | 2 +- pkg/sentry/fs/proc/meminfo.go | 6 +-- pkg/sentry/fs/proc/mounts.go | 2 +- pkg/sentry/fs/proc/net.go | 2 +- pkg/sentry/fs/proc/stat.go | 12 ++--- pkg/sentry/fs/proc/sys_net.go | 2 +- pkg/sentry/fs/proc/task.go | 8 +-- pkg/sentry/fs/proc/version.go | 2 +- pkg/sentry/fs/ramfs/dir.go | 2 +- pkg/sentry/fs/tmpfs/fs.go | 2 +- pkg/sentry/fs/tmpfs/inode_file.go | 2 +- pkg/sentry/fs/tmpfs/tmpfs.go | 2 +- pkg/sentry/fs/tty/dir.go | 6 +-- pkg/sentry/fs/tty/fs.go | 2 +- pkg/sentry/fs/tty/master.go | 6 +-- pkg/sentry/fs/tty/slave.go | 6 +-- pkg/sentry/kernel/auth/credentials.go | 2 +- pkg/sentry/kernel/auth/user_namespace.go | 2 +- pkg/sentry/kernel/pending_signals.go | 2 +- pkg/sentry/kernel/ptrace.go | 4 +- pkg/sentry/kernel/rseq.go | 2 +- pkg/sentry/kernel/sched/cpuset.go | 2 +- pkg/sentry/kernel/semaphore/semaphore.go | 6 +-- pkg/sentry/kernel/shm/shm.go | 2 +- pkg/sentry/kernel/syscalls.go | 2 +- pkg/sentry/kernel/task_context.go | 2 +- pkg/sentry/kernel/task_exec.go | 2 +- pkg/sentry/kernel/task_exit.go | 4 +- pkg/sentry/kernel/task_identity.go | 2 +- pkg/sentry/kernel/task_run.go | 2 +- pkg/sentry/kernel/task_signals.go | 4 +- pkg/sentry/kernel/task_stop.go | 2 +- pkg/sentry/loader/loader.go | 2 +- pkg/sentry/loader/vdso.go | 6 +-- pkg/sentry/memmap/memmap.go | 2 +- pkg/sentry/mm/aio_context.go | 2 +- pkg/sentry/mm/procfs.go | 10 ++-- pkg/sentry/mm/special_mappable.go | 2 +- pkg/sentry/mm/syscalls.go | 6 +-- pkg/sentry/mm/vma.go | 2 +- pkg/sentry/platform/kvm/kvm_amd64_unsafe.go | 2 +- pkg/sentry/platform/platform.go | 2 +- pkg/sentry/platform/ptrace/subprocess.go | 2 +- pkg/sentry/platform/ring0/x86.go | 4 +- pkg/sentry/sighandling/sighandling.go | 2 +- pkg/sentry/sighandling/sighandling_unsafe.go | 2 +- pkg/sentry/socket/epsocket/epsocket.go | 32 ++++++------ pkg/sentry/socket/epsocket/save_restore.go | 2 +- pkg/sentry/socket/epsocket/stack.go | 2 +- pkg/sentry/socket/hostinet/socket.go | 2 +- pkg/sentry/socket/netlink/route/protocol.go | 8 +-- pkg/sentry/socket/netlink/socket.go | 10 ++-- pkg/sentry/socket/rpcinet/conn/conn.go | 2 +- pkg/sentry/socket/rpcinet/notifier/notifier.go | 4 +- pkg/sentry/socket/rpcinet/socket.go | 6 +-- pkg/sentry/socket/rpcinet/syscall_rpc.proto | 2 +- pkg/sentry/strace/strace.go | 2 +- pkg/sentry/syscalls/linux/error.go | 2 +- pkg/sentry/syscalls/linux/linux64.go | 60 +++++++++++----------- pkg/sentry/syscalls/linux/sys_aio.go | 2 +- pkg/sentry/syscalls/linux/sys_file.go | 4 +- pkg/sentry/syscalls/linux/sys_mmap.go | 4 +- pkg/sentry/syscalls/linux/sys_read.go | 2 +- pkg/sentry/syscalls/linux/sys_socket.go | 4 +- pkg/sentry/syscalls/linux/sys_thread.go | 2 +- pkg/sentry/syscalls/linux/sys_write.go | 4 +- pkg/sentry/time/calibrated_clock.go | 6 +-- pkg/sentry/time/parameters.go | 2 +- pkg/sentry/usermem/usermem.go | 4 +- pkg/sentry/watchdog/watchdog.go | 2 +- pkg/syserr/syserr.go | 10 ++-- pkg/tcpip/network/ipv4/icmp.go | 2 +- pkg/tcpip/network/ipv6/icmp.go | 4 +- pkg/tcpip/stack/nic.go | 6 +-- pkg/tcpip/stack/stack.go | 4 +- pkg/tcpip/stack/stack_global_state.go | 2 +- pkg/tcpip/stack/transport_test.go | 2 +- pkg/tcpip/tcpip.go | 2 +- pkg/tcpip/transport/raw/raw.go | 2 +- pkg/tcpip/transport/tcp/BUILD | 2 +- pkg/unet/unet.go | 2 +- pkg/unet/unet_test.go | 2 +- runsc/boot/controller.go | 4 +- runsc/boot/fs.go | 6 +-- runsc/boot/loader.go | 2 +- runsc/cmd/checkpoint.go | 2 +- runsc/container/container.go | 2 +- runsc/container/container_test.go | 4 +- runsc/sandbox/sandbox.go | 6 +-- runsc/specutils/specutils.go | 4 +- test/syscalls/BUILD | 6 +-- test/syscalls/build_defs.bzl | 4 +- test/syscalls/linux/32bit.cc | 14 ++--- test/syscalls/linux/aio.cc | 2 +- test/syscalls/linux/chmod.cc | 2 +- test/syscalls/linux/epoll.cc | 2 +- test/syscalls/linux/exec_binary.cc | 12 ++--- test/syscalls/linux/file_base.h | 4 +- test/syscalls/linux/ioctl.cc | 4 +- test/syscalls/linux/ip_socket_test_util.cc | 2 +- test/syscalls/linux/lseek.cc | 2 +- test/syscalls/linux/mkdir.cc | 2 +- test/syscalls/linux/mmap.cc | 18 +++---- test/syscalls/linux/open.cc | 2 +- test/syscalls/linux/partial_bad_buffer.cc | 18 +++---- test/syscalls/linux/pipe.cc | 6 +-- test/syscalls/linux/proc.cc | 32 ++++++------ test/syscalls/linux/proc_pid_smaps.cc | 2 +- test/syscalls/linux/ptrace.cc | 2 +- test/syscalls/linux/pwrite64.cc | 2 +- test/syscalls/linux/readv_socket.cc | 2 +- test/syscalls/linux/rtsignal.cc | 2 +- test/syscalls/linux/socket_inet_loopback.cc | 10 ++-- .../socket_ipv4_udp_unbound_external_networking.cc | 4 +- test/syscalls/linux/socket_netlink_route.cc | 4 +- test/syscalls/linux/socket_stream_blocking.cc | 2 +- test/syscalls/linux/socket_test_util.cc | 2 +- test/syscalls/linux/socket_unix.cc | 16 +++--- test/syscalls/linux/socket_unix_dgram.cc | 2 +- .../linux/socket_unix_dgram_non_blocking.cc | 2 +- test/syscalls/linux/socket_unix_non_stream.cc | 10 ++-- .../linux/socket_unix_unbound_seqpacket.cc | 2 +- test/syscalls/linux/socket_unix_unbound_stream.cc | 4 +- test/syscalls/linux/stat.cc | 2 +- test/syscalls/linux/stat_times.cc | 8 +-- test/syscalls/linux/tcp_socket.cc | 2 +- test/syscalls/linux/tkill.cc | 2 +- test/syscalls/linux/udp_bind.cc | 4 +- test/syscalls/linux/uidgid.cc | 2 +- test/syscalls/linux/utimes.cc | 4 +- test/syscalls/linux/wait.cc | 2 +- test/syscalls/linux/write.cc | 2 +- third_party/gvsync/downgradable_rwmutex_unsafe.go | 2 +- vdso/cycle_clock.h | 2 +- vdso/vdso_amd64.lds | 2 +- vdso/vdso_arm64.lds | 2 +- 176 files changed, 403 insertions(+), 396 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index d6dafc595..238dd6665 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -126,6 +126,13 @@ change. When approved, the change will be submitted by a team member and automatically merged into the repository. +### Bug IDs + +Some TODOs and NOTEs sprinkled throughout the code have associated IDs of the +form b/1234. These correspond to bugs in our internal bug tracker. Eventually +these bugs will be moved to the GitHub Issues, but until then they can simply be +ignored. + ### The small print Contributions made by corporations are covered by a different agreement than the diff --git a/pkg/cpuid/cpuid_test.go b/pkg/cpuid/cpuid_test.go index 35e7b8e50..64ade1cbe 100644 --- a/pkg/cpuid/cpuid_test.go +++ b/pkg/cpuid/cpuid_test.go @@ -78,7 +78,7 @@ func TestTakeFeatureIntersection(t *testing.T) { } } -// TODO: Run this test on a very old platform, and make sure more +// TODO(b/73346484): Run this test on a very old platform, and make sure more // bits are enabled than just FPU and PAE. This test currently may not detect // if HostFeatureSet gives back junk bits. func TestHostFeatureSet(t *testing.T) { diff --git a/pkg/dhcp/client.go b/pkg/dhcp/client.go index 354205e63..2ba79be32 100644 --- a/pkg/dhcp/client.go +++ b/pkg/dhcp/client.go @@ -120,7 +120,7 @@ func (c *Client) Config() Config { // If the server sets a lease limit a timer is set to automatically // renew it. func (c *Client) Request(ctx context.Context, requestedAddr tcpip.Address) (cfg Config, reterr error) { - // TODO: remove calls to {Add,Remove}Address when they're no + // TODO(b/127321246): remove calls to {Add,Remove}Address when they're no // longer required to send and receive broadcast. if err := c.stack.AddAddressWithOptions(c.nicid, ipv4.ProtocolNumber, tcpipHeader.IPv4Any, stack.NeverPrimaryEndpoint); err != nil && err != tcpip.ErrDuplicateAddress { return Config{}, fmt.Errorf("dhcp: AddAddressWithOptions(): %s", err) diff --git a/pkg/log/glog.go b/pkg/log/glog.go index fbb58501b..24d5390d7 100644 --- a/pkg/log/glog.go +++ b/pkg/log/glog.go @@ -144,7 +144,7 @@ func (g GoogleEmitter) Emit(level Level, timestamp time.Time, format string, arg b.writeAll(pid) b.write(' ') - // FIXME: The caller, fabricated. This really sucks, but it + // FIXME(b/73383460): The caller, fabricated. This really sucks, but it // is unacceptable to put runtime.Callers() in the hot path. b.writeAll(caller) b.write(']') diff --git a/pkg/metric/metric.go b/pkg/metric/metric.go index 02af75974..e5eb95f89 100644 --- a/pkg/metric/metric.go +++ b/pkg/metric/metric.go @@ -44,8 +44,8 @@ var ( // // Metrics are not saved across save/restore and thus reset to zero on restore. // -// TODO: Support non-cumulative metrics. -// TODO: Support metric fields. +// TODO(b/67298402): Support non-cumulative metrics. +// TODO(b/67298427): Support metric fields. // type Uint64Metric struct { // value is the actual value of the metric. It must be accessed diff --git a/pkg/segment/set.go b/pkg/segment/set.go index a9a3b8875..74a916ea3 100644 --- a/pkg/segment/set.go +++ b/pkg/segment/set.go @@ -1270,7 +1270,7 @@ func segmentAfterPosition(n *node, i int) Iterator { } func zeroValueSlice(slice []Value) { - // TODO: check if Go is actually smart enough to optimize a + // TODO(jamieliu): check if Go is actually smart enough to optimize a // ClearValue that assigns nil to a memset here for i := range slice { Functions{}.ClearValue(&slice[i]) diff --git a/pkg/segment/test/set_functions.go b/pkg/segment/test/set_functions.go index 05ba5fbb9..41f649011 100644 --- a/pkg/segment/test/set_functions.go +++ b/pkg/segment/test/set_functions.go @@ -15,7 +15,7 @@ package segment // Basic numeric constants that we define because the math package doesn't. -// TODO: These should be Math.MaxInt64/MinInt64? +// TODO(nlacasse): These should be Math.MaxInt64/MinInt64? const ( maxInt = int(^uint(0) >> 1) minInt = -maxInt - 1 diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go index 4cd7a9af5..16d8eb2b2 100644 --- a/pkg/sentry/arch/arch.go +++ b/pkg/sentry/arch/arch.go @@ -53,7 +53,7 @@ type FloatingPointData byte // Context provides architecture-dependent information for a specific thread. // -// NOTE: Currently we use uintptr here to refer to a generic native +// NOTE(b/34169503): Currently we use uintptr here to refer to a generic native // register value. While this will work for the foreseeable future, it isn't // strictly correct. We may want to create some abstraction that makes this // more clear or enables us to store values of arbitrary widths. This is diff --git a/pkg/sentry/arch/arch_amd64.go b/pkg/sentry/arch/arch_amd64.go index 2507774f7..7ec2f2c84 100644 --- a/pkg/sentry/arch/arch_amd64.go +++ b/pkg/sentry/arch/arch_amd64.go @@ -305,7 +305,7 @@ func (c *context64) PtracePeekUser(addr uintptr) (interface{}, error) { buf := binary.Marshal(nil, usermem.ByteOrder, c.ptraceGetRegs()) return c.Native(uintptr(usermem.ByteOrder.Uint64(buf[addr:]))), nil } - // TODO: debug registers + // TODO(b/34088053): debug registers return c.Native(0), nil } @@ -320,6 +320,6 @@ func (c *context64) PtracePokeUser(addr, data uintptr) error { _, err := c.PtraceSetRegs(bytes.NewBuffer(buf)) return err } - // TODO: debug registers + // TODO(b/34088053): debug registers return nil } diff --git a/pkg/sentry/arch/arch_x86.go b/pkg/sentry/arch/arch_x86.go index c8bf0e7f2..4305fe2cb 100644 --- a/pkg/sentry/arch/arch_x86.go +++ b/pkg/sentry/arch/arch_x86.go @@ -306,7 +306,7 @@ func (s *State) ptraceGetRegs() syscall.PtraceRegs { // FS/GS_TLS_SEL when fs_base/gs_base is a 64-bit value. (We do the // same in PtraceSetRegs.) // - // TODO: Remove this fixup since newer Linux + // TODO(gvisor.dev/issue/168): Remove this fixup since newer Linux // doesn't have this behavior anymore. if regs.Fs == 0 && regs.Fs_base <= 0xffffffff { regs.Fs = _FS_TLS_SEL diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index c9de36897..7f76eba27 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -319,7 +319,7 @@ func (c *context64) NewSignalStack() NativeSignalStack { // From Linux 'arch/x86/include/uapi/asm/sigcontext.h' the following is the // size of the magic cookie at the end of the xsave frame. // -// NOTE: Currently we don't actually populate the fpstate +// NOTE(b/33003106#comment11): Currently we don't actually populate the fpstate // on the signal stack. const _FP_XSTATE_MAGIC2_SIZE = 4 @@ -392,7 +392,7 @@ func (c *context64) SignalSetup(st *Stack, act *SignalAct, info *SignalInfo, alt Sigset: sigset, } - // TODO: Set SignalContext64.Err, Trapno, and Cr2 + // TODO(gvisor.dev/issue/159): Set SignalContext64.Err, Trapno, and Cr2 // based on the fault that caused the signal. For now, leave Err and // Trapno unset and assume CR2 == info.Addr() for SIGSEGVs and // SIGBUSes. @@ -505,7 +505,7 @@ func (c *context64) SignalRestore(st *Stack, rt bool) (linux.SignalSet, SignalSt l := len(c.sigFPState) if l > 0 { c.x86FPState = c.sigFPState[l-1] - // NOTE: State save requires that any slice + // NOTE(cl/133042258): State save requires that any slice // elements from '[len:cap]' to be zero value. c.sigFPState[l-1] = nil c.sigFPState = c.sigFPState[0 : l-1] diff --git a/pkg/sentry/arch/stack.go b/pkg/sentry/arch/stack.go index f2cfb0426..2e33ccdf5 100644 --- a/pkg/sentry/arch/stack.go +++ b/pkg/sentry/arch/stack.go @@ -97,7 +97,7 @@ func (s *Stack) Push(vals ...interface{}) (usermem.Addr, error) { if c < 0 { return 0, fmt.Errorf("bad binary.Size for %T", v) } - // TODO: Use a real context.Context. + // TODO(b/38173783): Use a real context.Context. n, err := usermem.CopyObjectOut(context.Background(), s.IO, s.Bottom-usermem.Addr(c), norm, usermem.IOOpts{}) if err != nil || c != n { return 0, err @@ -121,11 +121,11 @@ func (s *Stack) Pop(vals ...interface{}) (usermem.Addr, error) { var err error if isVaddr { value := s.Arch.Native(uintptr(0)) - // TODO: Use a real context.Context. + // TODO(b/38173783): Use a real context.Context. n, err = usermem.CopyObjectIn(context.Background(), s.IO, s.Bottom, value, usermem.IOOpts{}) *vaddr = usermem.Addr(s.Arch.Value(value)) } else { - // TODO: Use a real context.Context. + // TODO(b/38173783): Use a real context.Context. n, err = usermem.CopyObjectIn(context.Background(), s.IO, s.Bottom, v, usermem.IOOpts{}) } if err != nil { diff --git a/pkg/sentry/context/context.go b/pkg/sentry/context/context.go index 7ed6a5e8a..eefc3e1b4 100644 --- a/pkg/sentry/context/context.go +++ b/pkg/sentry/context/context.go @@ -114,7 +114,7 @@ var bgContext = &logContext{Logger: log.Log()} // Background returns an empty context using the default logger. // // Users should be wary of using a Background context. Please tag any use with -// FIXME and a note to remove this use. +// FIXME(b/38173783) and a note to remove this use. // // Generally, one should use the Task as their context when available, or avoid // having to use a context in places where a Task is unavailable. diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go index e848def14..aca2267a7 100644 --- a/pkg/sentry/control/proc.go +++ b/pkg/sentry/control/proc.go @@ -261,7 +261,7 @@ func (proc *Proc) Ps(args *PsArgs, out *string) error { } // Process contains information about a single process in a Sandbox. -// TODO: Implement TTY field. +// TODO(b/117881927): Implement TTY field. type Process struct { UID auth.KUID `json:"uid"` PID kernel.ThreadID `json:"pid"` diff --git a/pkg/sentry/fs/README.md b/pkg/sentry/fs/README.md index a88a0cd3a..f53ed3eaa 100644 --- a/pkg/sentry/fs/README.md +++ b/pkg/sentry/fs/README.md @@ -59,7 +59,7 @@ two categories: The first is always necessary to save and restore. An application may never have any open file descriptors, but across save and restore it should see a coherent -view of any mount namespace. NOTE: Currently only one "initial" +view of any mount namespace. NOTE(b/63601033): Currently only one "initial" mount namespace is supported. The second is so that system calls across save and restore are coherent with diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index 651cbc164..1f61c5711 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -240,7 +240,7 @@ func (a *Area) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArgume return 0, syserror.EINVAL } - // TODO: If personality flag + // TODO(b/30946773,gvisor.dev/issue/153): If personality flag // READ_IMPLIES_EXEC is set, set PROT_EXEC if PORT_READ is set. a.perms = perms @@ -290,7 +290,7 @@ func (a *Area) pinOperation(pin linux.AshmemPin, op uint32) (uintptr, error) { return linux.AshmemNotPurged, nil case linux.AshmemUnpinIoctl: - // TODO: Implement purge on unpin. + // TODO(b/30946773): Implement purge on unpin. a.pb.UnpinRange(r) return 0, nil diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index a41b5dcae..d9f1559de 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -69,7 +69,7 @@ func NewDevice(ctx context.Context, owner fs.FileOwner, fp fs.FilePermissions) * // GetFile implements fs.InodeOperations.GetFile. // -// TODO: Add functionality to GetFile: Additional fields will be +// TODO(b/30946773): Add functionality to GetFile: Additional fields will be // needed in the Device structure, initialize them here. Also, Device will need // to keep track of the created Procs in order to implement BINDER_READ_WRITE // ioctl. @@ -133,7 +133,7 @@ func (bp *Proc) Write(ctx context.Context, file *fs.File, src usermem.IOSequence // Flush implements fs.FileOperations.Flush. // -// TODO: Implement. +// TODO(b/30946773): Implement. func (bp *Proc) Flush(ctx context.Context, file *fs.File) error { return nil } @@ -149,7 +149,7 @@ func (bp *Proc) ConfigureMMap(ctx context.Context, file *fs.File, opts *memmap.M } opts.MaxPerms.Write = false - // TODO: Binder sets VM_DONTCOPY, preventing the created vma + // TODO(b/30946773): Binder sets VM_DONTCOPY, preventing the created vma // from being copied across fork(), but we don't support this yet. As // a result, MMs containing a Binder mapping cannot be forked (MM.Fork will // fail when AddMapping returns EBUSY). @@ -159,7 +159,7 @@ func (bp *Proc) ConfigureMMap(ctx context.Context, file *fs.File, opts *memmap.M // Ioctl implements fs.FileOperations.Ioctl. // -// TODO: Implement. +// TODO(b/30946773): Implement. func (bp *Proc) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { // Switch on ioctl request. switch uint32(args[1].Int()) { @@ -173,22 +173,22 @@ func (bp *Proc) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArgum }) return 0, err case linux.BinderWriteReadIoctl: - // TODO: Implement. + // TODO(b/30946773): Implement. fallthrough case linux.BinderSetIdleTimeoutIoctl: - // TODO: Implement. + // TODO(b/30946773): Implement. fallthrough case linux.BinderSetMaxThreadsIoctl: - // TODO: Implement. + // TODO(b/30946773): Implement. fallthrough case linux.BinderSetIdlePriorityIoctl: - // TODO: Implement. + // TODO(b/30946773): Implement. fallthrough case linux.BinderSetContextMgrIoctl: - // TODO: Implement. + // TODO(b/30946773): Implement. fallthrough case linux.BinderThreadExitIoctl: - // TODO: Implement. + // TODO(b/30946773): Implement. return 0, syserror.ENOSYS default: // Ioctls irrelevant to Binder. @@ -228,7 +228,7 @@ func (bp *Proc) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, // Translate implements memmap.Mappable.Translate. func (bp *Proc) Translate(ctx context.Context, required, optional memmap.MappableRange, at usermem.AccessType) ([]memmap.Translation, error) { - // TODO: In addition to the page initially allocated and mapped + // TODO(b/30946773): In addition to the page initially allocated and mapped // in AddMapping (Linux: binder_mmap), Binder allocates and maps pages for // each transaction (Linux: binder_ioctl => binder_ioctl_write_read => // binder_thread_write => binder_transaction => binder_alloc_buf => diff --git a/pkg/sentry/fs/dentry.go b/pkg/sentry/fs/dentry.go index 4879df4d6..29fb155a4 100644 --- a/pkg/sentry/fs/dentry.go +++ b/pkg/sentry/fs/dentry.go @@ -83,7 +83,7 @@ type DirCtx struct { attrs map[string]DentAttr // DirCursor is the directory cursor. - // TODO: Once Handles are removed this can just live in the + // TODO(b/67778717): Once Handles are removed this can just live in the // respective FileOperations implementations and not need to get // plumbed everywhere. DirCursor *string diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go index 4bcdf530a..54fc11fe1 100644 --- a/pkg/sentry/fs/dirent.go +++ b/pkg/sentry/fs/dirent.go @@ -318,7 +318,7 @@ func (d *Dirent) SyncAll(ctx context.Context) { // There is nothing to sync for a read-only filesystem. if !d.Inode.MountSource.Flags.ReadOnly { - // FIXME: This should be a mount traversal, not a + // FIXME(b/34856369): This should be a mount traversal, not a // Dirent traversal, because some Inodes that need to be synced // may no longer be reachable by name (after sys_unlink). // @@ -1506,7 +1506,7 @@ func Rename(ctx context.Context, root *Dirent, oldParent *Dirent, oldName string } // Are we frozen? - // TODO: Is this the right errno? + // TODO(jamieliu): Is this the right errno? if oldParent.frozen && !oldParent.Inode.IsVirtual() { return syscall.ENOENT } @@ -1565,7 +1565,7 @@ func Rename(ctx context.Context, root *Dirent, oldParent *Dirent, oldName string } else { // Check constraints on the dirent being replaced. - // NOTE: We don't want to keep replaced alive + // NOTE(b/111808347): We don't want to keep replaced alive // across the Rename, so must call DecRef manually (no defer). // Check that we can delete replaced. @@ -1606,7 +1606,7 @@ func Rename(ctx context.Context, root *Dirent, oldParent *Dirent, oldName string // Allow the file system to drop extra references on replaced. replaced.dropExtendedReference() - // NOTE: Keeping a dirent + // NOTE(b/31798319,b/31867149,b/31867671): Keeping a dirent // open across renames is currently broken for multiple // reasons, so we flush all references on the replaced node and // its children. diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 2c2126f17..5d5026661 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -65,7 +65,7 @@ const FileMaxOffset = math.MaxInt64 // under a single abortable mutex which also synchronizes lseek(2), read(2), // and write(2). // -// FIXME: Split synchronization from cancellation. +// FIXME(b/38451980): Split synchronization from cancellation. // // +stateify savable type File struct { diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index e1f02f0f4..6e680f0a4 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -160,7 +160,7 @@ func (f *overlayFileOperations) Seek(ctx context.Context, file *File, whence See // If this was a seek on a directory, we must update the cursor. if seekDir && whence == SeekSet && offset == 0 { // Currently only seeking to 0 on a directory is supported. - // FIXME: Lift directory seeking limitations. + // FIXME(b/33075855): Lift directory seeking limitations. f.dirCursor = "" } return n, nil @@ -329,7 +329,7 @@ func (*overlayFileOperations) ConfigureMMap(ctx context.Context, file *File, opt if !o.isMappableLocked() { return syserror.ENODEV } - // FIXME: This is a copy/paste of fsutil.GenericConfigureMMap, + // FIXME(jamieliu): This is a copy/paste of fsutil.GenericConfigureMMap, // which we can't use because the overlay implementation is in package fs, // so depending on fs/fsutil would create a circular dependency. Move // overlay to fs/overlay. diff --git a/pkg/sentry/fs/fsutil/file.go b/pkg/sentry/fs/fsutil/file.go index df34dc788..42afdd11c 100644 --- a/pkg/sentry/fs/fsutil/file.go +++ b/pkg/sentry/fs/fsutil/file.go @@ -36,7 +36,7 @@ func (FileNoopRelease) Release() {} // // Currently only seeking to 0 on a directory is supported. // -// FIXME: Lift directory seeking limitations. +// FIXME(b/33075855): Lift directory seeking limitations. func SeekWithDirCursor(ctx context.Context, file *fs.File, whence fs.SeekWhence, offset int64, dirCursor *string) (int64, error) { inode := file.Dirent.Inode current := file.Offset() @@ -50,7 +50,7 @@ func SeekWithDirCursor(ctx context.Context, file *fs.File, whence fs.SeekWhence, if fs.IsCharDevice(inode.StableAttr) { // Ignore seek requests. // - // FIXME: This preserves existing + // FIXME(b/34716638): This preserves existing // behavior but is not universally correct. return 0, nil } @@ -104,7 +104,7 @@ func SeekWithDirCursor(ctx context.Context, file *fs.File, whence fs.SeekWhence, return current, syserror.EINVAL } return sz + offset, nil - // FIXME: This is not universally correct. + // FIXME(b/34778850): This is not universally correct. // Remove SpecialDirectory. case fs.SpecialDirectory: if offset != 0 { @@ -112,7 +112,7 @@ func SeekWithDirCursor(ctx context.Context, file *fs.File, whence fs.SeekWhence, } // SEEK_END to 0 moves the directory "cursor" to the end. // - // FIXME: The ensures that after the seek, + // FIXME(b/35442290): The ensures that after the seek, // reading on the directory will get EOF. But it is not // correct in general because the directory can grow in // size; attempting to read those new entries will be diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index b690cfe93..ba33b9912 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -479,7 +479,7 @@ func (c *CachingInodeOperations) Read(ctx context.Context, file *fs.File, dst us // common: getting a return value of 0 from a read syscall is the only way // to detect EOF. // - // TODO: Separate out c.attr.Size and use atomics instead of + // TODO(jamieliu): Separate out c.attr.Size and use atomics instead of // c.dataMu. c.dataMu.RLock() size := c.attr.Size @@ -776,7 +776,7 @@ func (c *CachingInodeOperations) Translate(ctx context.Context, required, option var translatedEnd uint64 for seg := c.cache.FindSegment(required.Start); seg.Ok() && seg.Start() < required.End; seg, _ = seg.NextNonEmpty() { segMR := seg.Range().Intersect(optional) - // TODO: Make Translations writable even if writability is + // TODO(jamieliu): Make Translations writable even if writability is // not required if already kept-dirty by another writable translation. perms := usermem.AccessType{ Read: true, diff --git a/pkg/sentry/fs/gofer/cache_policy.go b/pkg/sentry/fs/gofer/cache_policy.go index d7fbb71b7..51c573aef 100644 --- a/pkg/sentry/fs/gofer/cache_policy.go +++ b/pkg/sentry/fs/gofer/cache_policy.go @@ -136,7 +136,7 @@ func (cp cachePolicy) revalidate(ctx context.Context, name string, parent, child // Walk from parent to child again. // - // TODO: If we have a directory FD in the parent + // TODO(b/112031682): If we have a directory FD in the parent // inodeOperations, then we can use fstatat(2) to get the inode // attributes instead of making this RPC. qids, _, mask, attr, err := parentIops.fileState.file.walkGetAttr(ctx, []string{name}) @@ -171,7 +171,7 @@ func (cp cachePolicy) keep(d *fs.Dirent) bool { return false } sattr := d.Inode.StableAttr - // NOTE: Only cache files, directories, and symlinks. + // NOTE(b/31979197): Only cache files, directories, and symlinks. return fs.IsFile(sattr) || fs.IsDir(sattr) || fs.IsSymlink(sattr) } diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index 80d1e08a6..35caa42cd 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -297,7 +297,7 @@ func (f *fileOperations) Flush(ctx context.Context, file *fs.File) error { // We do this because some p9 server implementations of Flush are // over-zealous. // - // FIXME: weaken these implementations and remove this check. + // FIXME(edahlgren): weaken these implementations and remove this check. if !file.Flags().Write { return nil } diff --git a/pkg/sentry/fs/gofer/file_state.go b/pkg/sentry/fs/gofer/file_state.go index f770ca4ea..d0c64003c 100644 --- a/pkg/sentry/fs/gofer/file_state.go +++ b/pkg/sentry/fs/gofer/file_state.go @@ -28,7 +28,7 @@ func (f *fileOperations) afterLoad() { // Manually load the open handles. var err error - // TODO: Context is not plumbed to save/restore. + // TODO(b/38173783): Context is not plumbed to save/restore. f.handles, err = f.inodeOperations.fileState.getHandles(context.Background(), f.flags) if err != nil { return fmt.Errorf("failed to re-open handle: %v", err) diff --git a/pkg/sentry/fs/gofer/handles.go b/pkg/sentry/fs/gofer/handles.go index f32e99ce0..0b33e80c3 100644 --- a/pkg/sentry/fs/gofer/handles.go +++ b/pkg/sentry/fs/gofer/handles.go @@ -49,7 +49,7 @@ func (h *handles) DecRef() { log.Warningf("error closing host file: %v", err) } } - // FIXME: Context is not plumbed here. + // FIXME(b/38173783): Context is not plumbed here. if err := h.File.close(context.Background()); err != nil { log.Warningf("error closing p9 file: %v", err) } diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go index 29af1010c..1181a24cc 100644 --- a/pkg/sentry/fs/gofer/inode.go +++ b/pkg/sentry/fs/gofer/inode.go @@ -570,13 +570,13 @@ func init() { } // AddLink implements InodeOperations.AddLink, but is currently a noop. -// FIXME: Remove this from InodeOperations altogether. +// FIXME(b/63117438): Remove this from InodeOperations altogether. func (*inodeOperations) AddLink() {} // DropLink implements InodeOperations.DropLink, but is currently a noop. -// FIXME: Remove this from InodeOperations altogether. +// FIXME(b/63117438): Remove this from InodeOperations altogether. func (*inodeOperations) DropLink() {} // NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange. -// FIXME: Remove this from InodeOperations altogether. +// FIXME(b/63117438): Remove this from InodeOperations altogether. func (i *inodeOperations) NotifyStatusChange(ctx context.Context) {} diff --git a/pkg/sentry/fs/gofer/inode_state.go b/pkg/sentry/fs/gofer/inode_state.go index ad4d3df58..44d76ba9f 100644 --- a/pkg/sentry/fs/gofer/inode_state.go +++ b/pkg/sentry/fs/gofer/inode_state.go @@ -123,7 +123,7 @@ func (i *inodeFileState) afterLoad() { // beforeSave. return fmt.Errorf("failed to find path for inode number %d. Device %s contains %s", i.sattr.InodeID, i.s.connID, fs.InodeMappings(i.s.inodeMappings)) } - // TODO: Context is not plumbed to save/restore. + // TODO(b/38173783): Context is not plumbed to save/restore. ctx := &dummyClockContext{context.Background()} _, i.file, err = i.s.attach.walk(ctx, splitAbsolutePath(name)) diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index ed5147c65..4ed688ce5 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -134,7 +134,7 @@ type session struct { // socket files. This allows unix domain sockets to be used with paths that // belong to a gofer. // - // TODO: there are few possible races with someone stat'ing the + // TODO(b/77154739): there are few possible races with someone stat'ing the // file and another deleting it concurrently, where the file will not be // reported as socket file. endpoints *endpointMaps `state:"wait"` diff --git a/pkg/sentry/fs/gofer/session_state.go b/pkg/sentry/fs/gofer/session_state.go index 0ad5d63b5..b1f299be5 100644 --- a/pkg/sentry/fs/gofer/session_state.go +++ b/pkg/sentry/fs/gofer/session_state.go @@ -104,7 +104,7 @@ func (s *session) afterLoad() { // If private unix sockets are enabled, create and fill the session's endpoint // maps. if opts.privateunixsocket { - // TODO: Context is not plumbed to save/restore. + // TODO(b/38173783): Context is not plumbed to save/restore. ctx := &dummyClockContext{context.Background()} if err = s.restoreEndpointMaps(ctx); err != nil { diff --git a/pkg/sentry/fs/host/fs.go b/pkg/sentry/fs/host/fs.go index 800649211..de349a41a 100644 --- a/pkg/sentry/fs/host/fs.go +++ b/pkg/sentry/fs/host/fs.go @@ -87,7 +87,7 @@ func (f *Filesystem) Mount(ctx context.Context, _ string, flags fs.MountSourceFl options := fs.GenericMountSourceOptions(data) // Grab the whitelist if one was specified. - // TODO: require another option "testonly" in order to allow + // TODO(edahlgren/mpratt/hzy): require another option "testonly" in order to allow // no whitelist. if wl, ok := options[whitelistKey]; ok { f.paths = strings.Split(wl, "|") @@ -320,7 +320,7 @@ func (m *superOperations) SaveInodeMapping(inode *fs.Inode, path string) { // Keep implements fs.MountSourceOperations.Keep. // -// TODO: It is possible to change the permissions on a +// TODO(b/72455313,b/77596690): It is possible to change the permissions on a // host file while it is in the dirent cache (say from RO to RW), but it is not // possible to re-open the file with more relaxed permissions, since the host // FD is already open and stored in the inode. diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index 2030edcb4..69c648f67 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -95,7 +95,7 @@ type inodeFileState struct { // ReadToBlocksAt implements fsutil.CachedFileObject.ReadToBlocksAt. func (i *inodeFileState) ReadToBlocksAt(ctx context.Context, dsts safemem.BlockSeq, offset uint64) (uint64, error) { - // TODO: Using safemem.FromIOReader here is wasteful for two + // TODO(jamieliu): Using safemem.FromIOReader here is wasteful for two // reasons: // // - Using preadv instead of iterated preads saves on host system calls. @@ -325,7 +325,7 @@ func (i *inodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags fs.Fi // canMap returns true if this fs.Inode can be memory mapped. func canMap(inode *fs.Inode) bool { - // FIXME: Some obscure character devices can be mapped. + // FIXME(b/38213152): Some obscure character devices can be mapped. return fs.IsFile(inode.StableAttr) } @@ -428,15 +428,15 @@ func (i *inodeOperations) StatFS(context.Context) (fs.Info, error) { } // AddLink implements fs.InodeOperations.AddLink. -// FIXME: Remove this from InodeOperations altogether. +// FIXME(b/63117438): Remove this from InodeOperations altogether. func (i *inodeOperations) AddLink() {} // DropLink implements fs.InodeOperations.DropLink. -// FIXME: Remove this from InodeOperations altogether. +// FIXME(b/63117438): Remove this from InodeOperations altogether. func (i *inodeOperations) DropLink() {} // NotifyStatusChange implements fs.InodeOperations.NotifyStatusChange. -// FIXME: Remove this from InodeOperations altogether. +// FIXME(b/63117438): Remove this from InodeOperations altogether. func (i *inodeOperations) NotifyStatusChange(ctx context.Context) {} // readdirAll returns all of the directory entries in i. diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index d82f9740e..fe411a766 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -93,10 +93,10 @@ func (i *Inode) DecRef() { // destroy releases the Inode and releases the msrc reference taken. func (i *Inode) destroy() { - // FIXME: Context is not plumbed here. + // FIXME(b/38173783): Context is not plumbed here. ctx := context.Background() if err := i.WriteOut(ctx); err != nil { - // FIXME: Mark as warning again once noatime is + // FIXME(b/65209558): Mark as warning again once noatime is // properly supported. log.Debugf("Inode %+v, failed to sync all metadata: %v", i.StableAttr, err) } @@ -359,7 +359,7 @@ func (i *Inode) Getlink(ctx context.Context) (*Dirent, error) { // AddLink calls i.InodeOperations.AddLink. func (i *Inode) AddLink() { if i.overlay != nil { - // FIXME: Remove this from InodeOperations altogether. + // FIXME(b/63117438): Remove this from InodeOperations altogether. // // This interface is only used by ramfs to update metadata of // children. These filesystems should _never_ have overlay diff --git a/pkg/sentry/fs/inode_operations.go b/pkg/sentry/fs/inode_operations.go index ceacc7659..ff8b75f31 100644 --- a/pkg/sentry/fs/inode_operations.go +++ b/pkg/sentry/fs/inode_operations.go @@ -118,7 +118,7 @@ type InodeOperations interface { // // The caller must ensure that this operation is permitted. // - // TODO: merge Remove and RemoveDirectory, Remove + // TODO(b/67778723): merge Remove and RemoveDirectory, Remove // just needs a type flag. Remove(ctx context.Context, dir *Inode, name string) error diff --git a/pkg/sentry/fs/inode_overlay.go b/pkg/sentry/fs/inode_overlay.go index 254646176..bda3e1861 100644 --- a/pkg/sentry/fs/inode_overlay.go +++ b/pkg/sentry/fs/inode_overlay.go @@ -142,7 +142,7 @@ func overlayLookup(ctx context.Context, parent *overlayEntry, inode *Inode, name } else { // If we have something from the upper, we can only use it if the types // match. - // NOTE: Allow SpecialDirectories and Directories to merge. + // NOTE(b/112312863): Allow SpecialDirectories and Directories to merge. // This is needed to allow submounts in /proc and /sys. if upperInode.StableAttr.Type == child.Inode.StableAttr.Type || (IsDir(upperInode.StableAttr) && IsDir(child.Inode.StableAttr)) { @@ -226,7 +226,7 @@ func overlayCreate(ctx context.Context, o *overlayEntry, parent *Dirent, name st return nil, err } - // NOTE: Replace the Dirent with a transient Dirent, since + // NOTE(b/71766861): Replace the Dirent with a transient Dirent, since // we are about to create the real Dirent: an overlay Dirent. // // This ensures the *fs.File returned from overlayCreate is in the same @@ -338,7 +338,7 @@ func overlayRename(ctx context.Context, o *overlayEntry, oldParent *Dirent, rena // directory will appear empty in the upper fs, which will then // allow the rename to proceed when it should return ENOTEMPTY. // - // NOTE: Ideally, we'd just pass in the replaced + // NOTE(b/111808347): Ideally, we'd just pass in the replaced // Dirent from Rename, but we must drop the reference on // replaced before we make the rename call, so Rename can't // pass the Dirent to the Inode without significantly diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index 1e245ae5f..4d1693204 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -42,7 +42,7 @@ type DirentOperations interface { // MountSourceOperations contains filesystem specific operations. type MountSourceOperations interface { - // TODO: Add: + // TODO(b/67778729): Add: // BlockSize() int64 // FS() Filesystem @@ -101,7 +101,7 @@ func (i InodeMappings) String() string { // amalgamation implies that a mount source cannot be shared by multiple mounts // (e.g. cannot be mounted at different locations). // -// TODO: Move mount-specific information out of MountSource. +// TODO(b/63601033): Move mount-specific information out of MountSource. // // +stateify savable type MountSource struct { diff --git a/pkg/sentry/fs/mount_test.go b/pkg/sentry/fs/mount_test.go index 269d6b9da..d7605b2c9 100644 --- a/pkg/sentry/fs/mount_test.go +++ b/pkg/sentry/fs/mount_test.go @@ -33,7 +33,7 @@ func cacheReallyContains(cache *DirentCache, d *Dirent) bool { } // TestMountSourceOnlyCachedOnce tests that a Dirent that is mounted over only ends -// up in a single Dirent Cache. NOTE: Having a dirent in multiple +// up in a single Dirent Cache. NOTE(b/63848693): Having a dirent in multiple // caches causes major consistency issues. func TestMountSourceOnlyCachedOnce(t *testing.T) { ctx := contexttest.Context(t) diff --git a/pkg/sentry/fs/proc/README.md b/pkg/sentry/fs/proc/README.md index 3cc5f197c..5d4ec6c7b 100644 --- a/pkg/sentry/fs/proc/README.md +++ b/pkg/sentry/fs/proc/README.md @@ -91,7 +91,7 @@ CPU.IO utilization in last 10 minutes | Always zero Num currently running processes | Always zero Total num processes | Always zero -TODO: Populate the columns with accurate statistics. +TODO(b/62345059): Populate the columns with accurate statistics. ### meminfo @@ -128,12 +128,12 @@ Field name | Notes Buffers | Always zero, no block devices SwapCache | Always zero, no swap Inactive(anon) | Always zero, see SwapCache -Unevictable | Always zero TODO -Mlocked | Always zero TODO +Unevictable | Always zero TODO(b/31823263) +Mlocked | Always zero TODO(b/31823263) SwapTotal | Always zero, no swap SwapFree | Always zero, no swap -Dirty | Always zero TODO -Writeback | Always zero TODO +Dirty | Always zero TODO(b/31823263) +Writeback | Always zero TODO(b/31823263) MemAvailable | Uses the same value as MemFree since there is no swap. Slab | Missing SReclaimable | Missing @@ -185,7 +185,7 @@ softirq 0 0 0 0 0 0 0 0 0 0 0 All fields except for `btime` are always zero. -TODO: Populate with accurate fields. +TODO(b/37226836): Populate with accurate fields. ### sys diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index 25da06f5d..f2329e623 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -258,7 +258,7 @@ func newFdInfoDir(t *kernel.Task, msrc *fs.MountSource) *fs.Inode { // Lookup loads an fd in /proc/TID/fdinfo into a Dirent. func (fdid *fdInfoDir) Lookup(ctx context.Context, dir *fs.Inode, p string) (*fs.Dirent, error) { inode, err := walkDescriptors(fdid.t, p, func(file *fs.File, fdFlags kernel.FDFlags) *fs.Inode { - // TODO: Using a static inode here means that the + // TODO(b/121266871): Using a static inode here means that the // data can be out-of-date if, for instance, the flags on the // FD change before we read this file. We should switch to // generating the data on Read(). Also, we should include pos, diff --git a/pkg/sentry/fs/proc/loadavg.go b/pkg/sentry/fs/proc/loadavg.go index 78f3a1dc0..3ee0e570a 100644 --- a/pkg/sentry/fs/proc/loadavg.go +++ b/pkg/sentry/fs/proc/loadavg.go @@ -40,7 +40,7 @@ func (d *loadavgData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) var buf bytes.Buffer - // TODO: Include real data in fields. + // TODO(b/62345059): Include real data in fields. // Column 1-3: CPU and IO utilization of the last 1, 5, and 10 minute periods. // Column 4-5: currently running processes and the total number of processes. // Column 6: the last process ID used. diff --git a/pkg/sentry/fs/proc/meminfo.go b/pkg/sentry/fs/proc/meminfo.go index 620e93ce3..75cbf3e77 100644 --- a/pkg/sentry/fs/proc/meminfo.go +++ b/pkg/sentry/fs/proc/meminfo.go @@ -58,7 +58,7 @@ func (d *meminfoData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) fmt.Fprintf(&buf, "MemTotal: %8d kB\n", totalSize/1024) memFree := (totalSize - totalUsage) / 1024 // We use MemFree as MemAvailable because we don't swap. - // TODO: When reclaim is implemented the value of MemAvailable + // TODO(rahat): When reclaim is implemented the value of MemAvailable // should change. fmt.Fprintf(&buf, "MemFree: %8d kB\n", memFree) fmt.Fprintf(&buf, "MemAvailable: %8d kB\n", memFree) @@ -72,8 +72,8 @@ func (d *meminfoData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) fmt.Fprintf(&buf, "Inactive(anon): 0 kB\n") fmt.Fprintf(&buf, "Active(file): %8d kB\n", activeFile/1024) fmt.Fprintf(&buf, "Inactive(file): %8d kB\n", inactiveFile/1024) - fmt.Fprintf(&buf, "Unevictable: 0 kB\n") // TODO - fmt.Fprintf(&buf, "Mlocked: 0 kB\n") // TODO + fmt.Fprintf(&buf, "Unevictable: 0 kB\n") // TODO(b/31823263) + fmt.Fprintf(&buf, "Mlocked: 0 kB\n") // TODO(b/31823263) fmt.Fprintf(&buf, "SwapTotal: 0 kB\n") fmt.Fprintf(&buf, "SwapFree: 0 kB\n") fmt.Fprintf(&buf, "Dirty: 0 kB\n") diff --git a/pkg/sentry/fs/proc/mounts.go b/pkg/sentry/fs/proc/mounts.go index 1e62af8c6..fe62b167b 100644 --- a/pkg/sentry/fs/proc/mounts.go +++ b/pkg/sentry/fs/proc/mounts.go @@ -114,7 +114,7 @@ func (mif *mountInfoFile) ReadSeqFileData(ctx context.Context, handle seqfile.Se // (4) Root: the pathname of the directory in the filesystem // which forms the root of this mount. // - // NOTE: This will always be "/" until we implement + // NOTE(b/78135857): This will always be "/" until we implement // bind mounts. fmt.Fprintf(&buf, "/ ") diff --git a/pkg/sentry/fs/proc/net.go b/pkg/sentry/fs/proc/net.go index 55a958f9e..d24b2d370 100644 --- a/pkg/sentry/fs/proc/net.go +++ b/pkg/sentry/fs/proc/net.go @@ -154,7 +154,7 @@ func (n *netDev) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([]se contents[1] = " face |bytes packets errs drop fifo frame compressed multicast|bytes packets errs drop fifo colls carrier compressed\n" for _, i := range interfaces { - // TODO: Collect stats from each inet.Stack + // TODO(b/71872867): Collect stats from each inet.Stack // implementation (hostinet, epsocket, and rpcinet). // Implements the same format as diff --git a/pkg/sentry/fs/proc/stat.go b/pkg/sentry/fs/proc/stat.go index f2bbef375..18bd8e9b6 100644 --- a/pkg/sentry/fs/proc/stat.go +++ b/pkg/sentry/fs/proc/stat.go @@ -83,7 +83,7 @@ func (s *statData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([] var buf bytes.Buffer - // TODO: We currently export only zero CPU stats. We could + // TODO(b/37226836): We currently export only zero CPU stats. We could // at least provide some aggregate stats. var cpu cpuStats fmt.Fprintf(&buf, "cpu %s\n", cpu) @@ -100,7 +100,7 @@ func (s *statData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([] const numInterrupts = 256 // The Kernel doesn't handle real interrupts, so report all zeroes. - // TODO: We could count page faults as #PF. + // TODO(b/37226836): We could count page faults as #PF. fmt.Fprintf(&buf, "intr 0") // total for i := 0; i < numInterrupts; i++ { fmt.Fprintf(&buf, " 0") @@ -108,22 +108,22 @@ func (s *statData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) ([] fmt.Fprintf(&buf, "\n") // Total number of context switches. - // TODO: Count this. + // TODO(b/37226836): Count this. fmt.Fprintf(&buf, "ctxt 0\n") // CLOCK_REALTIME timestamp from boot, in seconds. fmt.Fprintf(&buf, "btime %d\n", s.k.Timekeeper().BootTime().Seconds()) // Total number of clones. - // TODO: Count this. + // TODO(b/37226836): Count this. fmt.Fprintf(&buf, "processes 0\n") // Number of runnable tasks. - // TODO: Count this. + // TODO(b/37226836): Count this. fmt.Fprintf(&buf, "procs_running 0\n") // Number of tasks waiting on IO. - // TODO: Count this. + // TODO(b/37226836): Count this. fmt.Fprintf(&buf, "procs_blocked 0\n") // Number of each softirq handled. diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index 728a46a74..0ce77f04f 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -39,7 +39,7 @@ const ( // tcpMemInode is used to read/write the size of netstack tcp buffers. // -// TODO: If we have multiple proc mounts, concurrent writes can +// TODO(b/121381035): If we have multiple proc mounts, concurrent writes can // leave netstack and the proc files in an inconsistent state. Since we set the // buffer size from these proc files on restore, we may also race and end up in // an inconsistent state on restore. diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 0edcdfce2..9f65a8337 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -77,7 +77,7 @@ func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace "fd": newFdDir(t, msrc), "fdinfo": newFdInfoDir(t, msrc), "gid_map": newGIDMap(t, msrc), - // FIXME: create the correct io file for threads. + // FIXME(b/123511468): create the correct io file for threads. "io": newIO(t, msrc), "maps": newMaps(t, msrc), "mountinfo": seqfile.NewSeqFileInode(t, &mountInfoFile{t: t}, msrc), @@ -93,7 +93,7 @@ func newTaskDir(t *kernel.Task, msrc *fs.MountSource, pidns *kernel.PIDNamespace contents["task"] = newSubtasks(t, msrc, pidns) } - // TODO: Set EUID/EGID based on dumpability. + // TODO(b/31916171): Set EUID/EGID based on dumpability. d := &taskDir{ Dir: *ramfs.NewDir(t, contents, fs.RootOwner, fs.FilePermsFromMode(0555)), t: t, @@ -245,7 +245,7 @@ func (e *exe) executable() (d *fs.Dirent, err error) { e.t.WithMuLocked(func(t *kernel.Task) { mm := t.MemoryManager() if mm == nil { - // TODO: Check shouldn't allow Readlink once the + // TODO(b/34851096): Check shouldn't allow Readlink once the // Task is zombied. err = syserror.EACCES return @@ -297,7 +297,7 @@ type namespaceSymlink struct { } func newNamespaceSymlink(t *kernel.Task, msrc *fs.MountSource, name string) *fs.Inode { - // TODO: Namespace symlinks should contain the namespace name and the + // TODO(rahat): Namespace symlinks should contain the namespace name and the // inode number for the namespace instance, so for example user:[123456]. We // currently fake the inode number by sticking the symlink inode in its // place. diff --git a/pkg/sentry/fs/proc/version.go b/pkg/sentry/fs/proc/version.go index b6d49d5e9..58e0c793c 100644 --- a/pkg/sentry/fs/proc/version.go +++ b/pkg/sentry/fs/proc/version.go @@ -65,7 +65,7 @@ func (v *versionData) ReadSeqFileData(ctx context.Context, h seqfile.SeqHandle) // Since we don't really want to expose build information to // applications, those fields are omitted. // - // FIXME: Using Version from the init task SyscallTable + // FIXME(mpratt): Using Version from the init task SyscallTable // disregards the different version a task may have (e.g., in a uts // namespace). ver := init.Leader().SyscallTable().Version diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index 159fd2981..c0400b67d 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -358,7 +358,7 @@ func (d *Dir) CreateDirectory(ctx context.Context, dir *fs.Inode, name string, p _, err := d.createInodeOperationsCommon(ctx, name, func() (*fs.Inode, error) { return d.NewDir(ctx, dir, perms) }) - // TODO: Support updating status times, as those should be + // TODO(nlacasse): Support updating status times, as those should be // updated by links. return err } diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go index d0c93028f..8e44421b6 100644 --- a/pkg/sentry/fs/tmpfs/fs.go +++ b/pkg/sentry/fs/tmpfs/fs.go @@ -34,7 +34,7 @@ const ( // GID for the root directory. rootGIDKey = "gid" - // TODO: support a tmpfs size limit. + // TODO(edahlgren/mpratt): support a tmpfs size limit. // size = "size" // Permissions that exceed modeMask will be rejected. diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 7c80d711b..4450e1363 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -309,7 +309,7 @@ func (f *fileInodeOperations) read(ctx context.Context, file *fs.File, dst userm // common: getting a return value of 0 from a read syscall is the only way // to detect EOF. // - // TODO: Separate out f.attr.Size and use atomics instead of + // TODO(jamieliu): Separate out f.attr.Size and use atomics instead of // f.dataMu. f.dataMu.RLock() size := f.attr.Size diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 555692505..5bb4922cb 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -32,7 +32,7 @@ import ( var fsInfo = fs.Info{ Type: linux.TMPFS_MAGIC, - // TODO: allow configuring a tmpfs size and enforce it. + // TODO(b/29637826): allow configuring a tmpfs size and enforce it. TotalBlocks: 0, FreeBlocks: 0, } diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index 33b4c6438..f8713471a 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -66,7 +66,7 @@ type dirInodeOperations struct { // msrc is the super block this directory is on. // - // TODO: Plumb this through instead of storing it here. + // TODO(chrisko): Plumb this through instead of storing it here. msrc *fs.MountSource // mu protects the fields below. @@ -89,7 +89,7 @@ type dirInodeOperations struct { // next is the next pty index to use. // - // TODO: reuse indices when ptys are closed. + // TODO(b/29356795): reuse indices when ptys are closed. next uint32 } @@ -118,7 +118,7 @@ func newDir(ctx context.Context, m *fs.MountSource) *fs.Inode { // N.B. Linux always uses inode id 1 for the directory. See // fs/devpts/inode.c:devpts_fill_super. // - // TODO: Since ptsDevice must be shared between + // TODO(b/75267214): Since ptsDevice must be shared between // different mounts, we must not assign fixed numbers. InodeID: ptsDevice.NextIno(), BlockSize: usermem.PageSize, diff --git a/pkg/sentry/fs/tty/fs.go b/pkg/sentry/fs/tty/fs.go index 43e0e2a04..a53448c47 100644 --- a/pkg/sentry/fs/tty/fs.go +++ b/pkg/sentry/fs/tty/fs.go @@ -43,7 +43,7 @@ func (*filesystem) Name() string { // AllowUserMount allows users to mount(2) this file system. func (*filesystem) AllowUserMount() bool { - // TODO: Users may mount this once the terminals are in a + // TODO(b/29356795): Users may mount this once the terminals are in a // usable state. return false } diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index 7c256abb0..e2686a074 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -51,7 +51,7 @@ func newMasterInode(ctx context.Context, d *dirInodeOperations, owner fs.FileOwn // N.B. Linux always uses inode id 2 for ptmx. See // fs/devpts/inode.c:mknod_ptmx. // - // TODO: Since ptsDevice must be shared between + // TODO(b/75267214): Since ptsDevice must be shared between // different mounts, we must not assign fixed numbers. InodeID: ptsDevice.NextIno(), Type: fs.CharacterDevice, @@ -157,7 +157,7 @@ func (mf *masterFileOperations) Ioctl(ctx context.Context, io usermem.IO, args a // of the slave end. return mf.t.ld.setTermios(ctx, io, args) case linux.TCSETSW: - // TODO: This should drain the output queue first. + // TODO(b/29356795): This should drain the output queue first. return mf.t.ld.setTermios(ctx, io, args) case linux.TIOCGPTN: _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), uint32(mf.t.n), usermem.IOOpts{ @@ -165,7 +165,7 @@ func (mf *masterFileOperations) Ioctl(ctx context.Context, io usermem.IO, args a }) return 0, err case linux.TIOCSPTLCK: - // TODO: Implement pty locking. For now just pretend we do. + // TODO(b/29356795): Implement pty locking. For now just pretend we do. return 0, nil case linux.TIOCGWINSZ: return 0, mf.t.ld.windowSize(ctx, io, args) diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index e8368bcdd..ed080ca0f 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -56,7 +56,7 @@ func newSlaveInode(ctx context.Context, d *dirInodeOperations, t *Terminal, owne // N.B. Linux always uses inode id = tty index + 3. See // fs/devpts/inode.c:devpts_pty_new. // - // TODO: Since ptsDevice must be shared between + // TODO(b/75267214): Since ptsDevice must be shared between // different mounts, we must not assign fixed numbers. InodeID: ptsDevice.NextIno(), Type: fs.CharacterDevice, @@ -137,7 +137,7 @@ func (sf *slaveFileOperations) Ioctl(ctx context.Context, io usermem.IO, args ar case linux.TCSETS: return sf.si.t.ld.setTermios(ctx, io, args) case linux.TCSETSW: - // TODO: This should drain the output queue first. + // TODO(b/29356795): This should drain the output queue first. return sf.si.t.ld.setTermios(ctx, io, args) case linux.TIOCGPTN: _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), uint32(sf.si.t.n), usermem.IOOpts{ @@ -151,7 +151,7 @@ func (sf *slaveFileOperations) Ioctl(ctx context.Context, io usermem.IO, args ar case linux.TIOCSCTTY: // Make the given terminal the controlling terminal of the // calling process. - // TODO: Implement once we have support for job + // TODO(b/129283598): Implement once we have support for job // control. return 0, nil default: diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go index a843b9aab..2055da196 100644 --- a/pkg/sentry/kernel/auth/credentials.go +++ b/pkg/sentry/kernel/auth/credentials.go @@ -125,7 +125,7 @@ func NewUserCredentials(kuid KUID, kgid KGID, extraKGIDs []KGID, capabilities *T creds.EffectiveCaps = capabilities.EffectiveCaps creds.BoundingCaps = capabilities.BoundingCaps creds.InheritableCaps = capabilities.InheritableCaps - // TODO: Support ambient capabilities. + // TODO(nlacasse): Support ambient capabilities. } else { // If no capabilities are specified, grant capabilities consistent with // setresuid + setresgid from NewRootCredentials to the given uid and diff --git a/pkg/sentry/kernel/auth/user_namespace.go b/pkg/sentry/kernel/auth/user_namespace.go index 30957bb9a..159940a69 100644 --- a/pkg/sentry/kernel/auth/user_namespace.go +++ b/pkg/sentry/kernel/auth/user_namespace.go @@ -49,7 +49,7 @@ type UserNamespace struct { gidMapFromParent idMapSet gidMapToParent idMapSet - // TODO: Support disabling setgroups(2). + // TODO(b/27454212): Support disabling setgroups(2). } // NewRootUserNamespace returns a UserNamespace that is appropriate for a diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go index 373e11772..deff6def9 100644 --- a/pkg/sentry/kernel/pending_signals.go +++ b/pkg/sentry/kernel/pending_signals.go @@ -30,7 +30,7 @@ const ( // rtSignalCap is the maximum number of instances of a given realtime // signal that may be pending. // - // TODO: In Linux, the minimum signal queue size is + // TODO(igudger): In Linux, the minimum signal queue size is // RLIMIT_SIGPENDING, which is by default max_threads/2. rtSignalCap = 32 ) diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index 8d78b2fb3..15f2e2964 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -162,7 +162,7 @@ func (t *Task) CanTrace(target *Task, attach bool) bool { if cgid := callerCreds.RealKGID; cgid != targetCreds.RealKGID || cgid != targetCreds.EffectiveKGID || cgid != targetCreds.SavedKGID { return false } - // TODO: dumpability check + // TODO(b/31916171): dumpability check if callerCreds.UserNamespace != targetCreds.UserNamespace { return false } @@ -396,7 +396,7 @@ func (t *Task) ptraceAttach(target *Task, seize bool, opts uintptr) error { if target.stop == (*groupStop)(nil) { target.trapStopPending = true target.endInternalStopLocked() - // TODO: Linux blocks ptrace_attach() until the task has + // TODO(jamieliu): Linux blocks ptrace_attach() until the task has // entered the ptrace-stop (or exited) via JOBCTL_TRAPPING. } target.tg.signalHandlers.mu.Unlock() diff --git a/pkg/sentry/kernel/rseq.go b/pkg/sentry/kernel/rseq.go index 0a954bc16..6d3314e81 100644 --- a/pkg/sentry/kernel/rseq.go +++ b/pkg/sentry/kernel/rseq.go @@ -66,7 +66,7 @@ func (t *Task) SetRSEQCriticalRegion(rscr RSEQCriticalRegion) error { if rscr.CriticalSection.Contains(rscr.Restart) { return syserror.EINVAL } - // TODO: check that rscr.CriticalSection and rscr.Restart are in + // TODO(jamieliu): check that rscr.CriticalSection and rscr.Restart are in // the application address range, for consistency with Linux t.tg.rscr.Store(&rscr) return nil diff --git a/pkg/sentry/kernel/sched/cpuset.go b/pkg/sentry/kernel/sched/cpuset.go index 69aee9127..41ac1067d 100644 --- a/pkg/sentry/kernel/sched/cpuset.go +++ b/pkg/sentry/kernel/sched/cpuset.go @@ -29,7 +29,7 @@ type CPUSet []byte // CPUSetSize returns the size in bytes of a CPUSet that can contain num cpus. func CPUSetSize(num uint) uint { - // NOTE: Applications may expect that the size of a CPUSet in + // NOTE(b/68859821): Applications may expect that the size of a CPUSet in // bytes is always a multiple of sizeof(unsigned long), since this is true // in Linux. Thus we always round up. bytes := (num + bitsPerByte - 1) / bitsPerByte diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index 29a2eb804..2b7c1a9bc 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -302,7 +302,7 @@ func (s *Set) SetVal(ctx context.Context, num int32, val int16, creds *auth.Cred return syserror.ERANGE } - // TODO: Clear undo entries in all processes + // TODO(b/29354920): Clear undo entries in all processes sem.value = val sem.pid = pid s.changeTime = ktime.NowFromContext(ctx) @@ -336,7 +336,7 @@ func (s *Set) SetValAll(ctx context.Context, vals []uint16, creds *auth.Credenti for i, val := range vals { sem := &s.sems[i] - // TODO: Clear undo entries in all processes + // TODO(b/29354920): Clear undo entries in all processes sem.value = int16(val) sem.pid = pid sem.wakeWaiters() @@ -481,7 +481,7 @@ func (s *Set) executeOps(ctx context.Context, ops []linux.Sembuf, pid int32) (ch } // All operations succeeded, apply them. - // TODO: handle undo operations. + // TODO(b/29354920): handle undo operations. for i, v := range tmpVals { s.sems[i].value = v s.sems[i].wakeWaiters() diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index 349f2a26e..d4812a065 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -427,7 +427,7 @@ func (s *Shm) AddMapping(ctx context.Context, _ memmap.MappingSpace, _ usermem.A func (s *Shm) RemoveMapping(ctx context.Context, _ memmap.MappingSpace, _ usermem.AddrRange, _ uint64, _ bool) { s.mu.Lock() defer s.mu.Unlock() - // TODO: RemoveMapping may be called during task exit, when ctx + // TODO(b/38173783): RemoveMapping may be called during task exit, when ctx // is context.Background. Gracefully handle missing clocks. Failing to // update the detach time in these cases is ok, since no one can observe the // omission. diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go index 7eb99718d..293b21249 100644 --- a/pkg/sentry/kernel/syscalls.go +++ b/pkg/sentry/kernel/syscalls.go @@ -165,7 +165,7 @@ type Stracer interface { // // The returned private data is passed to SyscallExit. // - // TODO: remove kernel imports from the strace + // TODO(gvisor.dev/issue/155): remove kernel imports from the strace // package so that the type can be used directly. SyscallEnter(t *Task, sysno uintptr, args arch.SyscallArguments, flags uint32) interface{} diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index 1b4d4cf2f..ac38dd157 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -60,7 +60,7 @@ func (tc *TaskContext) release() { // Nil out pointers so that if the task is saved after release, it doesn't // follow the pointers to possibly now-invalid objects. if tc.MemoryManager != nil { - // TODO + // TODO(b/38173783) tc.MemoryManager.DecUsers(context.Background()) tc.MemoryManager = nil } diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index 9fca90a1c..b49f902a5 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -208,7 +208,7 @@ func (r *runSyscallAfterExecStop) execute(t *Task) taskRunState { t.tc = *r.tc t.mu.Unlock() t.unstopVforkParent() - // NOTE: All locks must be dropped prior to calling Activate. + // NOTE(b/30316266): All locks must be dropped prior to calling Activate. t.MemoryManager().Activate() t.ptraceExec(oldTID) diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index 1a0734ab6..a07956208 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -339,7 +339,7 @@ func (t *Task) exitChildren() { }, true /* group */) other.signalHandlers.mu.Unlock() } - // TODO: The init process waits for all processes in the + // TODO(b/37722272): The init process waits for all processes in the // namespace to exit before completing its own exit // (kernel/pid_namespace.c:zap_pid_ns_processes()). Stop until all // other tasks in the namespace are dead, except possibly for this @@ -692,7 +692,7 @@ func (t *Task) exitNotificationSignal(sig linux.Signal, receiver *Task) *arch.Si info.Code = arch.CLD_EXITED info.SetStatus(int32(t.exitStatus.Code)) } - // TODO: Set utime, stime. + // TODO(b/72102453): Set utime, stime. return info } diff --git a/pkg/sentry/kernel/task_identity.go b/pkg/sentry/kernel/task_identity.go index e105eba13..6c9608f8d 100644 --- a/pkg/sentry/kernel/task_identity.go +++ b/pkg/sentry/kernel/task_identity.go @@ -421,7 +421,7 @@ func (t *Task) SetKeepCaps(k bool) { // updateCredsForExec updates t.creds to reflect an execve(). // -// NOTE: We currently do not implement privileged executables +// NOTE(b/30815691): We currently do not implement privileged executables // (set-user/group-ID bits and file capabilities). This allows us to make a lot // of simplifying assumptions: // diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 6b5fe7165..7115aa967 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -110,7 +110,7 @@ func (t *Task) doStop() { return } t.Deactivate() - // NOTE: t.Activate() must be called without any locks held, so + // NOTE(b/30316266): t.Activate() must be called without any locks held, so // this defer must precede the defer for unlocking the signal mutex. defer t.Activate() t.accountTaskGoroutineEnter(TaskGoroutineStopped) diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 3a8e61900..7f2e0df72 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -509,7 +509,7 @@ func (t *Task) canReceiveSignalLocked(sig linux.Signal) bool { if t.stop != nil { return false } - // - TODO: No special case for when t is also the sending task, + // - TODO(b/38173783): No special case for when t is also the sending task, // because the identity of the sender is unknown. // - Do not choose tasks that have already been interrupted, as they may be // busy handling another signal. @@ -895,7 +895,7 @@ func (t *Task) signalStop(target *Task, code int32, status int32) { sigchld.SetPid(int32(t.tg.pidns.tids[target])) sigchld.SetUid(int32(target.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow())) sigchld.SetStatus(status) - // TODO: Set utime, stime. + // TODO(b/72102453): Set utime, stime. t.sendSignalLocked(sigchld, true /* group */) } } diff --git a/pkg/sentry/kernel/task_stop.go b/pkg/sentry/kernel/task_stop.go index 36846484c..1302cadc1 100644 --- a/pkg/sentry/kernel/task_stop.go +++ b/pkg/sentry/kernel/task_stop.go @@ -69,7 +69,7 @@ import ( // A TaskStop is a condition visible to the task control flow graph that // prevents a task goroutine from running or exiting, i.e. an internal stop. // -// NOTE: Most TaskStops don't contain any data; they're +// NOTE(b/30793614): Most TaskStops don't contain any data; they're // distinguished by their type. The obvious way to implement such a TaskStop // is: // diff --git a/pkg/sentry/loader/loader.go b/pkg/sentry/loader/loader.go index 80ad59dde..79051befa 100644 --- a/pkg/sentry/loader/loader.go +++ b/pkg/sentry/loader/loader.go @@ -70,7 +70,7 @@ func openPath(ctx context.Context, mm *fs.MountNamespace, root, wd *fs.Dirent, m defer d.DecRef() perms := fs.PermMask{ - // TODO: Linux requires only execute + // TODO(gvisor.dev/issue/160): Linux requires only execute // permission, not read. However, our backing filesystems may // prevent us from reading the file without read permission. // diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index 18b7e90d8..8c196df84 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -194,7 +194,7 @@ func validateVDSO(ctx context.Context, f *fs.File, size uint64) (elfInfo, error) // VDSO describes a VDSO. // -// NOTE: to support multiple architectures or operating systems, this +// NOTE(mpratt): to support multiple architectures or operating systems, this // would need to contain a VDSO for each. // // +stateify savable @@ -262,7 +262,7 @@ func PrepareVDSO(mfp pgalloc.MemoryFileProvider) (*VDSO, error) { return &VDSO{ ParamPage: mm.NewSpecialMappable("[vvar]", mfp, paramPage), - // TODO: Don't advertise the VDSO, as + // TODO(gvisor.dev/issue/157): Don't advertise the VDSO, as // some applications may not be able to handle multiple [vdso] // hints. vdso: mm.NewSpecialMappable("", mfp, vdso), @@ -279,7 +279,7 @@ func PrepareVDSO(mfp pgalloc.MemoryFileProvider) (*VDSO, error) { // kernel simply directly maps the entire file into process memory, with very // little real ELF parsing. // -// NOTE: This means that userspace can, and unfortunately does, +// NOTE(b/25323870): This means that userspace can, and unfortunately does, // depend on parts of the ELF that would normally not be mapped. To maintain // compatibility with such binaries, we load the VDSO much like Linux. // diff --git a/pkg/sentry/memmap/memmap.go b/pkg/sentry/memmap/memmap.go index 1ef1f0dd8..3f6f7ebd0 100644 --- a/pkg/sentry/memmap/memmap.go +++ b/pkg/sentry/memmap/memmap.go @@ -356,6 +356,6 @@ type MMapOpts struct { // Hint is the name used for the mapping in /proc/[pid]/maps. If Hint is // empty, MappingIdentity.MappedName() will be used instead. // - // TODO: Replace entirely with MappingIdentity? + // TODO(jamieliu): Replace entirely with MappingIdentity? Hint string } diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index f7ff06de0..7075792e0 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -331,7 +331,7 @@ func (mm *MemoryManager) NewAIOContext(ctx context.Context, events uint32) (uint Length: aioRingBufferSize, MappingIdentity: m, Mappable: m, - // TODO: Linux does "do_mmap_pgoff(..., PROT_READ | + // TODO(fvoznika): Linux does "do_mmap_pgoff(..., PROT_READ | // PROT_WRITE, ...)" in fs/aio.c:aio_setup_ring(); why do we make this // mapping read-only? Perms: usermem.Read, diff --git a/pkg/sentry/mm/procfs.go b/pkg/sentry/mm/procfs.go index 0c4b8895d..7cdbf6e25 100644 --- a/pkg/sentry/mm/procfs.go +++ b/pkg/sentry/mm/procfs.go @@ -69,7 +69,7 @@ func (mm *MemoryManager) ReadMapsSeqFileData(ctx context.Context, handle seqfile start = *handle.(*usermem.Addr) } for vseg := mm.vmas.LowerBoundSegment(start); vseg.Ok(); vseg = vseg.NextSegment() { - // FIXME: If we use a usermem.Addr for the handle, we get + // FIXME(b/30793614): If we use a usermem.Addr for the handle, we get // "panic: autosave error: type usermem.Addr is not registered". vmaAddr := vseg.End() data = append(data, seqfile.SeqData{ @@ -88,7 +88,7 @@ func (mm *MemoryManager) ReadMapsSeqFileData(ctx context.Context, handle seqfile // // Artifically adjust the seqfile handle so we only output vsyscall entry once. if start != vsyscallEnd { - // FIXME: Can't get a pointer to constant vsyscallEnd. + // FIXME(b/30793614): Can't get a pointer to constant vsyscallEnd. vmaAddr := vsyscallEnd data = append(data, seqfile.SeqData{ Buf: []byte(vsyscallMapsEntry), @@ -134,7 +134,7 @@ func (mm *MemoryManager) appendVMAMapsEntryLocked(ctx context.Context, vseg vmaI if vma.hint != "" { s = vma.hint } else if vma.id != nil { - // FIXME: We are holding mm.mappingMu here, which is + // FIXME(jamieliu): We are holding mm.mappingMu here, which is // consistent with Linux's holding mmap_sem in // fs/proc/task_mmu.c:show_map_vma() => fs/seq_file.c:seq_file_path(). // However, it's not clear that fs.File.MappedName() is actually @@ -162,7 +162,7 @@ func (mm *MemoryManager) ReadSmapsSeqFileData(ctx context.Context, handle seqfil start = *handle.(*usermem.Addr) } for vseg := mm.vmas.LowerBoundSegment(start); vseg.Ok(); vseg = vseg.NextSegment() { - // FIXME: If we use a usermem.Addr for the handle, we get + // FIXME(b/30793614): If we use a usermem.Addr for the handle, we get // "panic: autosave error: type usermem.Addr is not registered". vmaAddr := vseg.End() data = append(data, seqfile.SeqData{ @@ -174,7 +174,7 @@ func (mm *MemoryManager) ReadSmapsSeqFileData(ctx context.Context, handle seqfil // We always emulate vsyscall, so advertise it here. See // ReadMapsSeqFileData for additional commentary. if start != vsyscallEnd { - // FIXME: Can't get a pointer to constant vsyscallEnd. + // FIXME(b/30793614): Can't get a pointer to constant vsyscallEnd. vmaAddr := vsyscallEnd data = append(data, seqfile.SeqData{ Buf: []byte(vsyscallSmapsEntry), diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index cfbf7a104..3b5161998 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -136,7 +136,7 @@ func (m *SpecialMappable) Length() uint64 { // NewSharedAnonMappable returns a SpecialMappable that implements the // semantics of mmap(MAP_SHARED|MAP_ANONYMOUS) and mappings of /dev/zero. // -// TODO: The use of SpecialMappable is a lazy code reuse hack. Linux +// TODO(jamieliu): The use of SpecialMappable is a lazy code reuse hack. Linux // uses an ephemeral file created by mm/shmem.c:shmem_zero_setup(); we should // do the same to get non-zero device and inode IDs. func NewSharedAnonMappable(length uint64, mfp pgalloc.MemoryFileProvider) (*SpecialMappable, error) { diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index cc7eb76d2..7b675b9b5 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -137,7 +137,7 @@ func (mm *MemoryManager) MMap(ctx context.Context, opts memmap.MMapOpts) (userme return 0, err } - // TODO: In Linux, VM_LOCKONFAULT (which may be set on the new + // TODO(jamieliu): In Linux, VM_LOCKONFAULT (which may be set on the new // vma by mlockall(MCL_FUTURE|MCL_ONFAULT) => mm_struct::def_flags) appears // to effectively disable MAP_POPULATE by unsetting FOLL_POPULATE in // mm/util.c:vm_mmap_pgoff() => mm/gup.c:__mm_populate() => @@ -148,7 +148,7 @@ func (mm *MemoryManager) MMap(ctx context.Context, opts memmap.MMapOpts) (userme mm.populateVMAAndUnlock(ctx, vseg, ar, true) case opts.Mappable == nil && length <= privateAllocUnit: - // NOTE: Get pmas and map eagerly in the hope + // NOTE(b/63077076, b/63360184): Get pmas and map eagerly in the hope // that doing so will save on future page faults. We only do this for // anonymous mappings, since otherwise the cost of // memmap.Mappable.Translate is unknown; and only for small mappings, @@ -698,7 +698,7 @@ func (mm *MemoryManager) Brk(ctx context.Context, addr usermem.Addr) (usermem.Ad return mm.brk.End, syserror.EINVAL } - // TODO: This enforces RLIMIT_DATA, but is + // TODO(gvisor.dev/issue/156): This enforces RLIMIT_DATA, but is // slightly more permissive than the usual data limit. In particular, // this only limits the size of the heap; a true RLIMIT_DATA limits the // size of heap + data + bss. The segment sizes need to be plumbed from diff --git a/pkg/sentry/mm/vma.go b/pkg/sentry/mm/vma.go index e9c9a80ea..931995254 100644 --- a/pkg/sentry/mm/vma.go +++ b/pkg/sentry/mm/vma.go @@ -274,7 +274,7 @@ func (mm *MemoryManager) getVMAsLocked(ctx context.Context, ar usermem.AddrRange // Loop invariants: vgap = vseg.PrevGap(); addr < vseg.End(). vma := vseg.ValuePtr() if addr < vseg.Start() { - // TODO: Implement vma.growsDown here. + // TODO(jamieliu): Implement vma.growsDown here. return vbegin, vgap, syserror.EFAULT } diff --git a/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go b/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go index c0a0af92d..d0f6bb225 100644 --- a/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go +++ b/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go @@ -62,7 +62,7 @@ func updateSystemValues(fd int) error { // Calculate whether guestPCID is supported. // - // FIXME: These should go through the much more pleasant + // FIXME(ascannell): These should go through the much more pleasant // cpuid package interfaces, once a way to accept raw kvm CPUID entries // is plumbed (or some rough equivalent). for i := 0; i < int(cpuidSupported.nr); i++ { diff --git a/pkg/sentry/platform/platform.go b/pkg/sentry/platform/platform.go index d1c9458ea..0e48417b9 100644 --- a/pkg/sentry/platform/platform.go +++ b/pkg/sentry/platform/platform.go @@ -181,7 +181,7 @@ var ( // this signal both to Contexts and to the sentry itself, under the assumption // that they originate from races with Context.Interrupt(). // -// NOTE: The Go runtime only guarantees that a small subset +// NOTE(b/23420492): The Go runtime only guarantees that a small subset // of signals will be always be unblocked on all threads, one of which // is SIGCHLD. const SignalInterrupt = linux.SIGCHLD diff --git a/pkg/sentry/platform/ptrace/subprocess.go b/pkg/sentry/platform/ptrace/subprocess.go index 82f125073..2a5d699ec 100644 --- a/pkg/sentry/platform/ptrace/subprocess.go +++ b/pkg/sentry/platform/ptrace/subprocess.go @@ -79,7 +79,7 @@ func (tp *threadPool) lookupOrCreate(currentTID int32, newThread func() *thread) // Before creating a new thread, see if we can find a thread // whose system tid has disappeared. // - // TODO: Other parts of this package depend on + // TODO(b/77216482): Other parts of this package depend on // threads never exiting. for origTID, t := range tp.threads { // Signal zero is an easy existence check. diff --git a/pkg/sentry/platform/ring0/x86.go b/pkg/sentry/platform/ring0/x86.go index 7c88010d8..4c6daec22 100644 --- a/pkg/sentry/platform/ring0/x86.go +++ b/pkg/sentry/platform/ring0/x86.go @@ -116,7 +116,7 @@ const ( // // Note that sign-extension semantics apply to the highest order bit. // -// FIXME: This should use the cpuid passed to Init. +// FIXME(b/69382326): This should use the cpuid passed to Init. func VirtualAddressBits() uint32 { ax, _, _, _ := cpuid.HostID(0x80000008, 0) return (ax >> 8) & 0xff @@ -124,7 +124,7 @@ func VirtualAddressBits() uint32 { // PhysicalAddressBits returns the number of bits available for physical addresses. // -// FIXME: This should use the cpuid passed to Init. +// FIXME(b/69382326): This should use the cpuid passed to Init. func PhysicalAddressBits() uint32 { ax, _, _, _ := cpuid.HostID(0x80000008, 0) return ax & 0xff diff --git a/pkg/sentry/sighandling/sighandling.go b/pkg/sentry/sighandling/sighandling.go index 6b5d5f993..571245ce5 100644 --- a/pkg/sentry/sighandling/sighandling.go +++ b/pkg/sentry/sighandling/sighandling.go @@ -86,7 +86,7 @@ func handleSignals(sigchans []chan os.Signal, handler func(linux.Signal), start, // // Otherwise ignore the signal. // - // TODO: Drop in Go 1.12, which uses tgkill + // TODO(b/114489875): Drop in Go 1.12, which uses tgkill // in runtime.raise. switch signal { case linux.SIGHUP, linux.SIGINT, linux.SIGTERM: diff --git a/pkg/sentry/sighandling/sighandling_unsafe.go b/pkg/sentry/sighandling/sighandling_unsafe.go index 5913d47a8..db6e71487 100644 --- a/pkg/sentry/sighandling/sighandling_unsafe.go +++ b/pkg/sentry/sighandling/sighandling_unsafe.go @@ -23,7 +23,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/abi/linux" ) -// TODO: Move to pkg/abi/linux along with definitions in +// TODO(b/34161764): Move to pkg/abi/linux along with definitions in // pkg/sentry/arch. type sigaction struct { handler uintptr diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index 23138d874..768fa0dfa 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -608,7 +608,7 @@ func (s *SocketOperations) Shutdown(t *kernel.Task, how int) *syserr.Error { // GetSockOpt implements the linux syscall getsockopt(2) for sockets backed by // tcpip.Endpoint. func (s *SocketOperations) GetSockOpt(t *kernel.Task, level, name, outLen int) (interface{}, *syserr.Error) { - // TODO: Unlike other socket options, SO_TIMESTAMP is + // TODO(b/78348848): Unlike other socket options, SO_TIMESTAMP is // implemented specifically for epsocket.SocketOperations rather than // commonEndpoint. commonEndpoint should be extended to support socket // options where the implementation is not shared, as unix sockets need @@ -658,7 +658,7 @@ func GetSockOpt(t *kernel.Task, s socket.Socket, ep commonEndpoint, family int, // getSockOptSocket implements GetSockOpt when level is SOL_SOCKET. func getSockOptSocket(t *kernel.Task, s socket.Socket, ep commonEndpoint, family int, skType transport.SockType, name, outLen int) (interface{}, *syserr.Error) { - // TODO: Stop rejecting short optLen values in getsockopt. + // TODO(b/124056281): Stop rejecting short optLen values in getsockopt. switch name { case linux.SO_TYPE: if outLen < sizeOfInt32 { @@ -789,7 +789,7 @@ func getSockOptSocket(t *kernel.Task, s socket.Socket, ep commonEndpoint, family return linux.Linger{}, nil case linux.SO_SNDTIMEO: - // TODO: Linux allows shorter lengths for partial results. + // TODO(igudger): Linux allows shorter lengths for partial results. if outLen < linux.SizeOfTimeval { return nil, syserr.ErrInvalidArgument } @@ -797,7 +797,7 @@ func getSockOptSocket(t *kernel.Task, s socket.Socket, ep commonEndpoint, family return linux.NsecToTimeval(s.SendTimeout()), nil case linux.SO_RCVTIMEO: - // TODO: Linux allows shorter lengths for partial results. + // TODO(igudger): Linux allows shorter lengths for partial results. if outLen < linux.SizeOfTimeval { return nil, syserr.ErrInvalidArgument } @@ -894,7 +894,7 @@ func getSockOptTCP(t *kernel.Task, ep commonEndpoint, name, outLen int) (interfa return nil, syserr.TranslateNetstackError(err) } - // TODO: Translate fields once they are added to + // TODO(b/64800844): Translate fields once they are added to // tcpip.TCPInfoOption. info := linux.TCPInfo{} @@ -995,7 +995,7 @@ func getSockOptIP(t *kernel.Task, ep commonEndpoint, name, outLen int) (interfac // SetSockOpt implements the linux syscall setsockopt(2) for sockets backed by // tcpip.Endpoint. func (s *SocketOperations) SetSockOpt(t *kernel.Task, level int, name int, optVal []byte) *syserr.Error { - // TODO: Unlike other socket options, SO_TIMESTAMP is + // TODO(b/78348848): Unlike other socket options, SO_TIMESTAMP is // implemented specifically for epsocket.SocketOperations rather than // commonEndpoint. commonEndpoint should be extended to support socket // options where the implementation is not shared, as unix sockets need @@ -1338,7 +1338,7 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.AddMembershipOption{ NIC: tcpip.NICID(req.InterfaceIndex), - // TODO: Change AddMembership to use the standard + // TODO(igudger): Change AddMembership to use the standard // any address representation. InterfaceAddr: tcpip.Address(req.InterfaceAddr[:]), MulticastAddr: tcpip.Address(req.MulticastAddr[:]), @@ -1352,7 +1352,7 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.RemoveMembershipOption{ NIC: tcpip.NICID(req.InterfaceIndex), - // TODO: Change DropMembership to use the standard + // TODO(igudger): Change DropMembership to use the standard // any address representation. InterfaceAddr: tcpip.Address(req.InterfaceAddr[:]), MulticastAddr: tcpip.Address(req.MulticastAddr[:]), @@ -1380,7 +1380,7 @@ func setSockOptIP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) *s )) case linux.MCAST_JOIN_GROUP: - // FIXME: Implement MCAST_JOIN_GROUP. + // FIXME(b/124219304): Implement MCAST_JOIN_GROUP. t.Kernel().EmitUnimplementedEvent(t) return syserr.ErrInvalidArgument @@ -1695,7 +1695,7 @@ func (s *SocketOperations) coalescingRead(ctx context.Context, dst usermem.IOSeq // nonBlockingRead issues a non-blocking read. // -// TODO: Support timestamps for stream sockets. +// TODO(b/78348848): Support timestamps for stream sockets. func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSequence, peek, trunc, senderRequested bool) (int, int, interface{}, uint32, socket.ControlMessages, *syserr.Error) { isPacket := s.isPacketBased() @@ -1762,7 +1762,7 @@ func (s *SocketOperations) nonBlockingRead(ctx context.Context, dst usermem.IOSe dst = dst.DropFirst(n) num, err := dst.CopyOutFrom(ctx, safemem.FromVecReaderFunc{func(dsts [][]byte) (int64, error) { n, _, err := s.Endpoint.Peek(dsts) - // TODO: Handle peek timestamp. + // TODO(b/78348848): Handle peek timestamp. if err != nil { return int64(n), syserr.TranslateNetstackError(err).ToError() } @@ -1963,7 +1963,7 @@ func (s *SocketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to [] func (s *SocketOperations) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { // SIOCGSTAMP is implemented by epsocket rather than all commonEndpoint // sockets. - // TODO: Add a commonEndpoint method to support SIOCGSTAMP. + // TODO(b/78348848): Add a commonEndpoint method to support SIOCGSTAMP. if int(args[1].Int()) == syscall.SIOCGSTAMP { s.readMu.Lock() defer s.readMu.Unlock() @@ -2153,19 +2153,19 @@ func interfaceIoctl(ctx context.Context, io usermem.IO, arg int, ifr *linux.IFRe case syscall.SIOCGIFMAP: // Gets the hardware parameters of the device. - // TODO: Implement. + // TODO(b/71872867): Implement. case syscall.SIOCGIFTXQLEN: // Gets the transmit queue length of the device. - // TODO: Implement. + // TODO(b/71872867): Implement. case syscall.SIOCGIFDSTADDR: // Gets the destination address of a point-to-point device. - // TODO: Implement. + // TODO(b/71872867): Implement. case syscall.SIOCGIFBRDADDR: // Gets the broadcast address of a device. - // TODO: Implement. + // TODO(b/71872867): Implement. case syscall.SIOCGIFNETMASK: // Gets the network mask of a device. diff --git a/pkg/sentry/socket/epsocket/save_restore.go b/pkg/sentry/socket/epsocket/save_restore.go index 34d9a7cf0..f19afb6c0 100644 --- a/pkg/sentry/socket/epsocket/save_restore.go +++ b/pkg/sentry/socket/epsocket/save_restore.go @@ -20,7 +20,7 @@ import ( // afterLoad is invoked by stateify. func (s *Stack) afterLoad() { - s.Stack = stack.StackFromEnv // FIXME + s.Stack = stack.StackFromEnv // FIXME(b/36201077) if s.Stack == nil { panic("can't restore without netstack/tcpip/stack.Stack") } diff --git a/pkg/sentry/socket/epsocket/stack.go b/pkg/sentry/socket/epsocket/stack.go index c0081c819..37c48f4bc 100644 --- a/pkg/sentry/socket/epsocket/stack.go +++ b/pkg/sentry/socket/epsocket/stack.go @@ -77,7 +77,7 @@ func (s *Stack) InterfaceAddrs() map[int32][]inet.InterfaceAddr { Family: family, PrefixLen: uint8(len(a.Address) * 8), Addr: []byte(a.Address), - // TODO: Other fields. + // TODO(b/68878065): Other fields. }) } nicAddrs[int32(id)] = addrs diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index c4848b313..49349074f 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -348,7 +348,7 @@ func (s *socketOperations) SetSockOpt(t *kernel.Task, level int, name int, opt [ func (s *socketOperations) RecvMsg(t *kernel.Task, dst usermem.IOSequence, flags int, haveDeadline bool, deadline ktime.Time, senderRequested bool, controlDataLen uint64) (int, int, interface{}, uint32, socket.ControlMessages, *syserr.Error) { // Whitelist flags. // - // FIXME: We can't support MSG_ERRQUEUE because it uses ancillary + // FIXME(jamieliu): We can't support MSG_ERRQUEUE because it uses ancillary // messages that netstack/tcpip/transport/unix doesn't understand. Kill the // Socket interface's dependence on netstack. if flags&^(syscall.MSG_DONTWAIT|syscall.MSG_PEEK|syscall.MSG_TRUNC) != 0 { diff --git a/pkg/sentry/socket/netlink/route/protocol.go b/pkg/sentry/socket/netlink/route/protocol.go index 7e70b09b2..e414b829b 100644 --- a/pkg/sentry/socket/netlink/route/protocol.go +++ b/pkg/sentry/socket/netlink/route/protocol.go @@ -110,7 +110,7 @@ func (p *Protocol) dumpLinks(ctx context.Context, hdr linux.NetlinkMessageHeader m.PutAttr(linux.IFLA_ADDRESS, mac) m.PutAttr(linux.IFLA_BROADCAST, brd) - // TODO: There are many more attributes. + // TODO(b/68878065): There are many more attributes. } return nil @@ -122,7 +122,7 @@ func (p *Protocol) dumpAddrs(ctx context.Context, hdr linux.NetlinkMessageHeader // netlink header and 1 byte protocol family common to all // NETLINK_ROUTE requests. // - // TODO: Filter output by passed protocol family. + // TODO(b/68878065): Filter output by passed protocol family. // The RTM_GETADDR dump response is a set of RTM_NEWADDR messages each // containing an InterfaceAddrMessage followed by a set of netlink @@ -151,7 +151,7 @@ func (p *Protocol) dumpAddrs(ctx context.Context, hdr linux.NetlinkMessageHeader m.PutAttr(linux.IFA_ADDRESS, []byte(a.Addr)) - // TODO: There are many more attributes. + // TODO(b/68878065): There are many more attributes. } } @@ -175,7 +175,7 @@ func (p *Protocol) ProcessMessage(ctx context.Context, hdr linux.NetlinkMessageH } } - // TODO: Only the dump variant of the types below are + // TODO(b/68878065): Only the dump variant of the types below are // supported. if hdr.Flags&linux.NLM_F_DUMP != linux.NLM_F_DUMP { return syserr.ErrNotSupported diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index 0fe9b39b6..a34f9d3ca 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -168,7 +168,7 @@ func (s *Socket) EventUnregister(e *waiter.Entry) { // Ioctl implements fs.FileOperations.Ioctl. func (s *Socket) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallArguments) (uintptr, error) { - // TODO: no ioctls supported. + // TODO(b/68878065): no ioctls supported. return 0, syserror.ENOTTY } @@ -319,7 +319,7 @@ func (s *Socket) GetSockOpt(t *kernel.Task, level int, name int, outLen int) (in t.Kernel().EmitUnimplementedEvent(t) } } - // TODO: other sockopts are not supported. + // TODO(b/68878065): other sockopts are not supported. return nil, syserr.ErrProtocolNotAvailable } @@ -369,7 +369,7 @@ func (s *Socket) SetSockOpt(t *kernel.Task, level int, name int, opt []byte) *sy } } - // TODO: other sockopts are not supported. + // TODO(b/68878065): other sockopts are not supported. return syserr.ErrProtocolNotAvailable } @@ -389,7 +389,7 @@ func (s *Socket) GetSockName(t *kernel.Task) (interface{}, uint32, *syserr.Error func (s *Socket) GetPeerName(t *kernel.Task) (interface{}, uint32, *syserr.Error) { sa := linux.SockAddrNetlink{ Family: linux.AF_NETLINK, - // TODO: Support non-kernel peers. For now the peer + // TODO(b/68878065): Support non-kernel peers. For now the peer // must be the kernel. PortID: 0, } @@ -540,7 +540,7 @@ func (s *Socket) processMessages(ctx context.Context, buf []byte) *syserr.Error continue } - // TODO: ACKs not supported yet. + // TODO(b/68877377): ACKs not supported yet. if hdr.Flags&linux.NLM_F_ACK == linux.NLM_F_ACK { return syserr.ErrNotSupported } diff --git a/pkg/sentry/socket/rpcinet/conn/conn.go b/pkg/sentry/socket/rpcinet/conn/conn.go index 9c749b888..64106c4b5 100644 --- a/pkg/sentry/socket/rpcinet/conn/conn.go +++ b/pkg/sentry/socket/rpcinet/conn/conn.go @@ -50,7 +50,7 @@ type RPCConnection struct { // NewRPCConnection initializes a RPC connection to a socket gofer. func NewRPCConnection(s *unet.Socket) *RPCConnection { conn := &RPCConnection{socket: s, requests: map[uint64]request{}} - go func() { // S/R-FIXME + go func() { // S/R-FIXME(b/77962828) var nums [16]byte for { for n := 0; n < len(nums); { diff --git a/pkg/sentry/socket/rpcinet/notifier/notifier.go b/pkg/sentry/socket/rpcinet/notifier/notifier.go index d9bda78b0..f06d12231 100644 --- a/pkg/sentry/socket/rpcinet/notifier/notifier.go +++ b/pkg/sentry/socket/rpcinet/notifier/notifier.go @@ -64,7 +64,7 @@ func NewRPCNotifier(cn *conn.RPCConnection) (*Notifier, error) { fdMap: make(map[uint32]*fdInfo), } - go w.waitAndNotify() // S/R-FIXME + go w.waitAndNotify() // S/R-FIXME(b/77962828) return w, nil } @@ -166,7 +166,7 @@ func (n *Notifier) waitAndNotify() error { res := n.rpcConn.Request(id).Result.(*pb.SyscallResponse_EpollWait).EpollWait.Result if e, ok := res.(*pb.EpollWaitResponse_ErrorNumber); ok { err := syscall.Errno(e.ErrorNumber) - // NOTE: I don't think epoll_wait can return EAGAIN but I'm being + // NOTE(magi): I don't think epoll_wait can return EAGAIN but I'm being // conseratively careful here since exiting the notification thread // would be really bad. if err == syscall.EINTR || err == syscall.EAGAIN { diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index 3418a6d75..cf8f69efb 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -288,7 +288,7 @@ func (s *socketOperations) Accept(t *kernel.Task, peerRequested bool, flags int, if blocking && se == syserr.ErrTryAgain { // Register for notifications. e, ch := waiter.NewChannelEntry(nil) - // FIXME: This waiter.EventHUp is a partial + // FIXME(b/119878986): This waiter.EventHUp is a partial // measure, need to figure out how to translate linux events to // internal events. s.EventRegister(&e, waiter.EventIn|waiter.EventHUp) @@ -370,7 +370,7 @@ func (s *socketOperations) Shutdown(t *kernel.Task, how int) *syserr.Error { // We save the shutdown state because of strange differences on linux // related to recvs on blocking vs. non-blocking sockets after a SHUT_RD. // We need to emulate that behavior on the blocking side. - // TODO: There is a possible race that can exist with loopback, + // TODO(b/120096741): There is a possible race that can exist with loopback, // where data could possibly be lost. s.setShutdownFlags(how) @@ -771,7 +771,7 @@ func (s *socketOperations) SendMsg(t *kernel.Task, src usermem.IOSequence, to [] return 0, syserr.FromError(err) } - // TODO: this needs to change to map directly to a SendMsg syscall + // TODO(bgeffon): this needs to change to map directly to a SendMsg syscall // in the RPC. totalWritten := 0 n, err := rpcSendMsg(t, &pb.SyscallRequest_Sendmsg{&pb.SendmsgRequest{ diff --git a/pkg/sentry/socket/rpcinet/syscall_rpc.proto b/pkg/sentry/socket/rpcinet/syscall_rpc.proto index c056e4c9d..9586f5923 100644 --- a/pkg/sentry/socket/rpcinet/syscall_rpc.proto +++ b/pkg/sentry/socket/rpcinet/syscall_rpc.proto @@ -3,7 +3,7 @@ syntax = "proto3"; // package syscall_rpc is a set of networking related system calls that can be // forwarded to a socket gofer. // -// TODO: Document individual RPCs. +// TODO(b/77963526): Document individual RPCs. package syscall_rpc; message SendmsgRequest { diff --git a/pkg/sentry/strace/strace.go b/pkg/sentry/strace/strace.go index a6d870b44..434a200d9 100644 --- a/pkg/sentry/strace/strace.go +++ b/pkg/sentry/strace/strace.go @@ -722,7 +722,7 @@ func (s SyscallMap) Name(sysno uintptr) string { // N.B. This is not in an init function because we can't be sure all syscall // tables are registered with the kernel when init runs. // -// TODO: remove kernel package dependencies from this +// TODO(gvisor.dev/issue/155): remove kernel package dependencies from this // package and have the kernel package self-initialize all syscall tables. func Initialize() { for _, table := range kernel.SyscallTables() { diff --git a/pkg/sentry/syscalls/linux/error.go b/pkg/sentry/syscalls/linux/error.go index 8759e5e32..304a12dde 100644 --- a/pkg/sentry/syscalls/linux/error.go +++ b/pkg/sentry/syscalls/linux/error.go @@ -89,7 +89,7 @@ func handleIOError(t *kernel.Task, partialResult bool, err, intr error, op strin // side is gone. The partial write is returned. EPIPE will be // returned on the next call. // - // TODO: In some cases SIGPIPE should + // TODO(gvisor.dev/issue/161): In some cases SIGPIPE should // also be sent to the application. return nil case syserror.ErrWouldBlock: diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index be793ca11..b9b4ccbd1 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -143,10 +143,10 @@ var AMD64 = &kernel.SyscallTable{ 65: Semop, 66: Semctl, 67: Shmdt, - // 68: @Syscall(Msgget), TODO - // 69: @Syscall(Msgsnd), TODO - // 70: @Syscall(Msgrcv), TODO - // 71: @Syscall(Msgctl), TODO + // 68: @Syscall(Msgget), TODO(b/29354921) + // 69: @Syscall(Msgsnd), TODO(b/29354921) + // 70: @Syscall(Msgrcv), TODO(b/29354921) + // 71: @Syscall(Msgctl), TODO(b/29354921) 72: Fcntl, 73: Flock, 74: Fsync, @@ -197,8 +197,8 @@ var AMD64 = &kernel.SyscallTable{ 119: Setresgid, 120: Getresgid, 121: Getpgid, - // 122: @Syscall(Setfsuid), TODO - // 123: @Syscall(Setfsgid), TODO + // 122: @Syscall(Setfsuid), TODO(b/112851702) + // 123: @Syscall(Setfsgid), TODO(b/112851702) 124: Getsid, 125: Capget, 126: Capset, @@ -217,7 +217,7 @@ var AMD64 = &kernel.SyscallTable{ 136: syscalls.ErrorWithEvent(syscall.ENOSYS), 137: Statfs, 138: Fstatfs, - // 139: @Syscall(Sysfs), TODO + // 139: @Syscall(Sysfs), TODO(gvisor.dev/issue/165) 140: Getpriority, 141: Setpriority, // @Syscall(SchedSetparam, returns:EPERM or ENOSYS, note:Returns EPERM if the process does not have cap_sys_nice; ENOSYS otherwise) @@ -291,7 +291,7 @@ var AMD64 = &kernel.SyscallTable{ // @Syscall(Security, note:Not implemented in Linux) 185: syscalls.Error(syscall.ENOSYS), 186: Gettid, - 187: nil, // @Syscall(Readahead), TODO + 187: nil, // @Syscall(Readahead), TODO(b/29351341) // @Syscall(Setxattr, returns:ENOTSUP, note:Requires filesystem support) 188: syscalls.ErrorWithEvent(syscall.ENOTSUP), // @Syscall(Lsetxattr, returns:ENOTSUP, note:Requires filesystem support) @@ -342,7 +342,7 @@ var AMD64 = &kernel.SyscallTable{ 217: Getdents64, 218: SetTidAddress, 219: RestartSyscall, - // 220: @Syscall(Semtimedop), TODO + // 220: @Syscall(Semtimedop), TODO(b/29354920) 221: Fadvise64, 222: TimerCreate, 223: TimerSettime, @@ -360,16 +360,16 @@ var AMD64 = &kernel.SyscallTable{ 235: Utimes, // @Syscall(Vserver, note:Not implemented by Linux) 236: syscalls.Error(syscall.ENOSYS), // Vserver, not implemented by Linux - // @Syscall(Mbind, returns:EPERM or ENOSYS, note:Returns EPERM if the process does not have cap_sys_nice; ENOSYS otherwise), TODO + // @Syscall(Mbind, returns:EPERM or ENOSYS, note:Returns EPERM if the process does not have cap_sys_nice; ENOSYS otherwise), TODO(b/117792295) 237: syscalls.CapError(linux.CAP_SYS_NICE), // may require cap_sys_nice 238: SetMempolicy, 239: GetMempolicy, - // 240: @Syscall(MqOpen), TODO - // 241: @Syscall(MqUnlink), TODO - // 242: @Syscall(MqTimedsend), TODO - // 243: @Syscall(MqTimedreceive), TODO - // 244: @Syscall(MqNotify), TODO - // 245: @Syscall(MqGetsetattr), TODO + // 240: @Syscall(MqOpen), TODO(b/29354921) + // 241: @Syscall(MqUnlink), TODO(b/29354921) + // 242: @Syscall(MqTimedsend), TODO(b/29354921) + // 243: @Syscall(MqTimedreceive), TODO(b/29354921) + // 244: @Syscall(MqNotify), TODO(b/29354921) + // 245: @Syscall(MqGetsetattr), TODO(b/29354921) 246: syscalls.CapError(linux.CAP_SYS_BOOT), // kexec_load, requires cap_sys_boot 247: Waitid, // @Syscall(AddKey, returns:EACCES, note:Not available to user) @@ -407,22 +407,22 @@ var AMD64 = &kernel.SyscallTable{ 273: syscalls.Error(syscall.ENOSYS), // @Syscall(GetRobustList, note:Obsolete) 274: syscalls.Error(syscall.ENOSYS), - // 275: @Syscall(Splice), TODO - // 276: @Syscall(Tee), TODO + // 275: @Syscall(Splice), TODO(b/29354098) + // 276: @Syscall(Tee), TODO(b/29354098) 277: SyncFileRange, - // 278: @Syscall(Vmsplice), TODO + // 278: @Syscall(Vmsplice), TODO(b/29354098) // @Syscall(MovePages, returns:EPERM or ENOSYS, note:Returns EPERM if the process does not have cap_sys_nice; ENOSYS otherwise) 279: syscalls.CapError(linux.CAP_SYS_NICE), // requires cap_sys_nice (mostly) 280: Utimensat, 281: EpollPwait, - // 282: @Syscall(Signalfd), TODO + // 282: @Syscall(Signalfd), TODO(b/19846426) 283: TimerfdCreate, 284: Eventfd, 285: Fallocate, 286: TimerfdSettime, 287: TimerfdGettime, 288: Accept4, - // 289: @Syscall(Signalfd4), TODO + // 289: @Syscall(Signalfd4), TODO(b/19846426) 290: Eventfd2, 291: EpollCreate1, 292: Dup3, @@ -447,17 +447,17 @@ var AMD64 = &kernel.SyscallTable{ 305: syscalls.CapError(linux.CAP_SYS_TIME), // requires cap_sys_time 306: Syncfs, 307: SendMMsg, - // 308: @Syscall(Setns), TODO + // 308: @Syscall(Setns), TODO(b/29354995) 309: Getcpu, - // 310: @Syscall(ProcessVmReadv), TODO may require cap_sys_ptrace - // 311: @Syscall(ProcessVmWritev), TODO may require cap_sys_ptrace + // 310: @Syscall(ProcessVmReadv), TODO(gvisor.dev/issue/158) may require cap_sys_ptrace + // 311: @Syscall(ProcessVmWritev), TODO(gvisor.dev/issue/158) may require cap_sys_ptrace // @Syscall(Kcmp, returns:EPERM or ENOSYS, note:Requires cap_sys_ptrace) 312: syscalls.CapError(linux.CAP_SYS_PTRACE), // @Syscall(FinitModule, returns:EPERM or ENOSYS, note:Returns EPERM if the process does not have cap_sys_module; ENOSYS otherwise) 313: syscalls.CapError(linux.CAP_SYS_MODULE), - // 314: @Syscall(SchedSetattr), TODO, we have no scheduler - // 315: @Syscall(SchedGetattr), TODO, we have no scheduler - // 316: @Syscall(Renameat2), TODO + // 314: @Syscall(SchedSetattr), TODO(b/118902272), we have no scheduler + // 315: @Syscall(SchedGetattr), TODO(b/118902272), we have no scheduler + // 316: @Syscall(Renameat2), TODO(b/118902772) 317: Seccomp, 318: GetRandom, 319: MemfdCreate, @@ -465,9 +465,9 @@ var AMD64 = &kernel.SyscallTable{ 320: syscalls.CapError(linux.CAP_SYS_BOOT), // @Syscall(Bpf, returns:EPERM or ENOSYS, note:Returns EPERM if the process does not have cap_sys_boot; ENOSYS otherwise) 321: syscalls.CapError(linux.CAP_SYS_ADMIN), // requires cap_sys_admin for all commands - // 322: @Syscall(Execveat), TODO - // 323: @Syscall(Userfaultfd), TODO - // 324: @Syscall(Membarrier), TODO + // 322: @Syscall(Execveat), TODO(b/118901836) + // 323: @Syscall(Userfaultfd), TODO(b/118906345) + // 324: @Syscall(Membarrier), TODO(b/118904897) 325: Mlock2, // Syscalls after 325 are "backports" from versions of Linux after 4.4. // 326: @Syscall(CopyFileRange), diff --git a/pkg/sentry/syscalls/linux/sys_aio.go b/pkg/sentry/syscalls/linux/sys_aio.go index 355071131..61c2647bf 100644 --- a/pkg/sentry/syscalls/linux/sys_aio.go +++ b/pkg/sentry/syscalls/linux/sys_aio.go @@ -120,7 +120,7 @@ func IoDestroy(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys // Does not exist. return 0, nil, syserror.EINVAL } - // FIXME: Linux blocks until all AIO to the destroyed context is + // FIXME(fvoznika): Linux blocks until all AIO to the destroyed context is // done. return 0, nil, nil } diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 50151f7b6..967464c85 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -259,7 +259,7 @@ func mknodAt(t *kernel.Task, dirFD kdefs.FD, addr usermem.Addr, mode linux.FileM case linux.ModeCharacterDevice: fallthrough case linux.ModeBlockDevice: - // TODO: We don't support creating block or character + // TODO(b/72101894): We don't support creating block or character // devices at the moment. // // When we start supporting block and character devices, we'll @@ -1532,7 +1532,7 @@ func chown(t *kernel.Task, d *fs.Dirent, uid auth.UID, gid auth.GID) error { owner.GID = kgid } - // FIXME: This is racy; the inode's owner may have changed in + // FIXME(b/62949101): This is racy; the inode's owner may have changed in // the meantime. (Linux holds i_mutex while calling // fs/attr.c:notify_change() => inode_operations::setattr => // inode_change_ok().) diff --git a/pkg/sentry/syscalls/linux/sys_mmap.go b/pkg/sentry/syscalls/linux/sys_mmap.go index 8732861e0..805b251b1 100644 --- a/pkg/sentry/syscalls/linux/sys_mmap.go +++ b/pkg/sentry/syscalls/linux/sys_mmap.go @@ -185,7 +185,7 @@ func Madvise(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca case linux.MADV_MERGEABLE, linux.MADV_UNMERGEABLE: fallthrough case linux.MADV_DONTDUMP, linux.MADV_DODUMP: - // TODO: Core dumping isn't implemented, so these are + // TODO(b/72045799): Core dumping isn't implemented, so these are // no-ops. fallthrough case linux.MADV_NORMAL, linux.MADV_RANDOM, linux.MADV_SEQUENTIAL, linux.MADV_WILLNEED: @@ -223,7 +223,7 @@ func GetMempolicy(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel. nodeFlag := flags&linux.MPOL_F_NODE != 0 addrFlag := flags&linux.MPOL_F_ADDR != 0 - // TODO: Once sysfs is implemented, report a single numa node in + // TODO(rahat): Once sysfs is implemented, report a single numa node in // /sys/devices/system/node. if nodemask != 0 && maxnode < 1 { return 0, nil, syserror.EINVAL diff --git a/pkg/sentry/syscalls/linux/sys_read.go b/pkg/sentry/syscalls/linux/sys_read.go index 8105e9b43..50c7d7a74 100644 --- a/pkg/sentry/syscalls/linux/sys_read.go +++ b/pkg/sentry/syscalls/linux/sys_read.go @@ -192,7 +192,7 @@ func Preadv(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal } // Preadv2 implements linux syscall preadv2(2). -// TODO: Implement RWF_HIPRI functionality. +// TODO(b/120162627): Implement RWF_HIPRI functionality. func Preadv2(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { // While the syscall is // preadv2(int fd, struct iovec* iov, int iov_cnt, off_t offset, int flags) diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go index 30ccc3f66..c8748958a 100644 --- a/pkg/sentry/syscalls/linux/sys_socket.go +++ b/pkg/sentry/syscalls/linux/sys_socket.go @@ -317,7 +317,7 @@ func accept(t *kernel.Task, fd kdefs.FD, addr usermem.Addr, addrLen usermem.Addr return 0, syserror.ConvertIntr(e.ToError(), kernel.ERESTARTSYS) } if peerRequested { - // NOTE: Linux does not give you an error if it can't + // NOTE(magi): Linux does not give you an error if it can't // write the data back out so neither do we. if err := writeAddress(t, peer, peerLen, addr, addrLen); err == syscall.EINVAL { return 0, err @@ -735,7 +735,7 @@ func recvSingleMsg(t *kernel.Task, s socket.Socket, msgPtr usermem.Addr, flags i return 0, err } - // FIXME: Pretend we have an empty error queue. + // FIXME(b/63594852): Pretend we have an empty error queue. if flags&linux.MSG_ERRQUEUE != 0 { return 0, syscall.EAGAIN } diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index 61cafefb9..ddcb5b789 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -350,7 +350,7 @@ func Waitid(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal } si.SetPid(int32(wr.TID)) si.SetUid(int32(wr.UID)) - // TODO: convert kernel.ExitStatus to functions and make + // TODO(b/73541790): convert kernel.ExitStatus to functions and make // WaitResult.Status a linux.WaitStatus s := syscall.WaitStatus(wr.Status) switch { diff --git a/pkg/sentry/syscalls/linux/sys_write.go b/pkg/sentry/syscalls/linux/sys_write.go index a5ad7efb2..e405608c4 100644 --- a/pkg/sentry/syscalls/linux/sys_write.go +++ b/pkg/sentry/syscalls/linux/sys_write.go @@ -192,8 +192,8 @@ func Pwritev(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysca } // Pwritev2 implements linux syscall pwritev2(2). -// TODO: Implement RWF_HIPRI functionality. -// TODO: Implement O_SYNC and D_SYNC functionality. +// TODO(b/120162627): Implement RWF_HIPRI functionality. +// TODO(b/120161091): Implement O_SYNC and D_SYNC functionality. func Pwritev2(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { // While the syscall is // pwritev2(int fd, struct iovec* iov, int iov_cnt, off_t offset, int flags) diff --git a/pkg/sentry/time/calibrated_clock.go b/pkg/sentry/time/calibrated_clock.go index c8cf4eca4..a98bcd7de 100644 --- a/pkg/sentry/time/calibrated_clock.go +++ b/pkg/sentry/time/calibrated_clock.go @@ -37,7 +37,7 @@ var fallbackMetric = metric.MustCreateNewUint64Metric("/time/fallback", false /* // clock. type CalibratedClock struct { // mu protects the fields below. - // TODO: consider a sequence counter for read locking. + // TODO(mpratt): consider a sequence counter for read locking. mu sync.RWMutex // ref sample the reference clock that this clock is calibrated @@ -140,7 +140,7 @@ func (c *CalibratedClock) updateParams(actual Parameters) { // N.B. logErrorAdjustment will have already logged the error // at warning level. // - // TODO: We could allow Realtime clock jumps here. + // TODO(mpratt): We could allow Realtime clock jumps here. c.resetLocked("Extreme clock error.") return } @@ -229,7 +229,7 @@ func (c *CalibratedClock) GetTime() (int64, error) { // CalibratedClocks contains calibrated monotonic and realtime clocks. // -// TODO: We know that Linux runs the monotonic and realtime clocks at +// TODO(mpratt): We know that Linux runs the monotonic and realtime clocks at // the same rate, so rather than tracking both individually, we could do one // calibration for both clocks. type CalibratedClocks struct { diff --git a/pkg/sentry/time/parameters.go b/pkg/sentry/time/parameters.go index f3ad58454..8568b1193 100644 --- a/pkg/sentry/time/parameters.go +++ b/pkg/sentry/time/parameters.go @@ -43,7 +43,7 @@ const ( // These statements assume that the host clock does not change. Actual // error will depend upon host clock changes. // - // TODO: make error correction more robust to delayed + // TODO(b/68779214): make error correction more robust to delayed // updates. ApproxUpdateInterval = 1 * time.Second diff --git a/pkg/sentry/usermem/usermem.go b/pkg/sentry/usermem/usermem.go index 99766a803..4c7d5014a 100644 --- a/pkg/sentry/usermem/usermem.go +++ b/pkg/sentry/usermem/usermem.go @@ -28,7 +28,7 @@ import ( // IO provides access to the contents of a virtual memory space. // -// FIXME: Implementations of IO cannot expect ctx to contain any +// FIXME(b/38173783): Implementations of IO cannot expect ctx to contain any // meaningful data. type IO interface { // CopyOut copies len(src) bytes from src to the memory mapped at addr. It @@ -85,7 +85,7 @@ type IO interface { // order. CopyInTo(ctx context.Context, ars AddrRangeSeq, dst safemem.Writer, opts IOOpts) (int64, error) - // TODO: The requirement that CopyOutFrom/CopyInTo call src/dst + // TODO(jamieliu): The requirement that CopyOutFrom/CopyInTo call src/dst // at most once, which is unnecessary in most cases, forces implementations // to gather safemem.Blocks into a single slice to pass to src/dst. Add // CopyOutFromIter/CopyInToIter, which relaxes this restriction, to avoid diff --git a/pkg/sentry/watchdog/watchdog.go b/pkg/sentry/watchdog/watchdog.go index c49b537a5..b4f1e3a4f 100644 --- a/pkg/sentry/watchdog/watchdog.go +++ b/pkg/sentry/watchdog/watchdog.go @@ -236,7 +236,7 @@ func (w *Watchdog) runTurn() { if !ok { // New stuck task detected. // - // TODO: Tasks blocked doing IO may be considered stuck in kernel. + // TODO(b/65849403): Tasks blocked doing IO may be considered stuck in kernel. tc = &offender{lastUpdateTime: lastUpdateTime} stuckTasks.Increment() newTaskFound = true diff --git a/pkg/syserr/syserr.go b/pkg/syserr/syserr.go index dad83e80c..232634dd4 100644 --- a/pkg/syserr/syserr.go +++ b/pkg/syserr/syserr.go @@ -49,7 +49,7 @@ func New(message string, linuxTranslation *linux.Errno) *Error { return err } - // TODO: Remove this. + // TODO(b/34162363): Remove this. errno := linuxTranslation.Number() if errno <= 0 || errno >= len(linuxBackwardsTranslations) { panic(fmt.Sprint("invalid errno: ", errno)) @@ -106,12 +106,12 @@ type linuxBackwardsTranslation struct { ok bool } -// TODO: Remove this. +// TODO(b/34162363): Remove this. var linuxBackwardsTranslations [maxErrno]linuxBackwardsTranslation // ToError translates an Error to a corresponding error value. // -// TODO: Remove this. +// TODO(b/34162363): Remove this. func (e *Error) ToError() error { if e == nil { return nil @@ -138,7 +138,7 @@ func (e *Error) ToLinux() *linux.Errno { return e.errno } -// TODO: Remove or replace most of these errors. +// TODO(b/34162363): Remove or replace most of these errors. // // Some of the errors should be replaced with package specific errors and // others should be removed entirely. @@ -278,7 +278,7 @@ var ( // FromError converts a generic error to an *Error. // -// TODO: Remove this function. +// TODO(b/34162363): Remove this function. func FromError(err error) *Error { if err == nil { return nil diff --git a/pkg/tcpip/network/ipv4/icmp.go b/pkg/tcpip/network/ipv4/icmp.go index ed9a4eee5..1c3acda4b 100644 --- a/pkg/tcpip/network/ipv4/icmp.go +++ b/pkg/tcpip/network/ipv4/icmp.go @@ -64,7 +64,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, vv buffer.V } h := header.ICMPv4(v) - // TODO: Meaningfully handle all ICMP types. + // TODO(b/112892170): Meaningfully handle all ICMP types. switch h.Type() { case header.ICMPv4Echo: received.Echo.Increment() diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go index 3210e6fc7..be28be36d 100644 --- a/pkg/tcpip/network/ipv6/icmp.go +++ b/pkg/tcpip/network/ipv6/icmp.go @@ -73,7 +73,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, vv buffer.V } h := header.ICMPv6(v) - // TODO: Meaningfully handle all ICMP types. + // TODO(b/112892170): Meaningfully handle all ICMP types. switch h.Type() { case header.ICMPv6PacketTooBig: received.PacketTooBig.Increment() @@ -247,7 +247,7 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack. DstAddr: r.RemoteAddress, }) - // TODO: count this in ICMP stats. + // TODO(stijlist): count this in ICMP stats. return linkEP.WritePacket(r, nil /* gso */, hdr, buffer.VectorisedView{}, ProtocolNumber) } diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go index 8b6c17a90..c18571b0f 100644 --- a/pkg/tcpip/stack/nic.go +++ b/pkg/tcpip/stack/nic.go @@ -176,7 +176,7 @@ func (n *NIC) primaryEndpoint(protocol tcpip.NetworkProtocolNumber) *referencedN for e := list.Front(); e != nil; e = e.Next() { r := e.(*referencedNetworkEndpoint) - // TODO: allow broadcast address when SO_BROADCAST is set. + // TODO(crawshaw): allow broadcast address when SO_BROADCAST is set. switch r.ep.ID().LocalAddress { case header.IPv4Broadcast, header.IPv4Any: continue @@ -476,7 +476,7 @@ func (n *NIC) DeliverNetworkPacket(linkEP LinkEndpoint, remote, _ tcpip.LinkAddr n.mu.RUnlock() if ok && ref.tryIncRef() { r.RemoteAddress = src - // TODO: Update the source NIC as well. + // TODO(b/123449044): Update the source NIC as well. ref.ep.HandlePacket(&r, vv) ref.decRef() } else { @@ -485,7 +485,7 @@ func (n *NIC) DeliverNetworkPacket(linkEP LinkEndpoint, remote, _ tcpip.LinkAddr hdr := buffer.NewPrependableFromView(vv.First()) vv.RemoveFirst() - // TODO: use route.WritePacket. + // TODO(b/128629022): use route.WritePacket. if err := n.linkEP.WritePacket(&r, nil /* gso */, hdr, vv, protocol); err != nil { r.Stats().IP.OutgoingPacketErrors.Increment() } else { diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index 8f7b6f781..cb9ffe9c2 100644 --- a/pkg/tcpip/stack/stack.go +++ b/pkg/tcpip/stack/stack.go @@ -476,7 +476,7 @@ func (s *Stack) Stats() tcpip.Stats { // SetForwarding enables or disables the packet forwarding between NICs. func (s *Stack) SetForwarding(enable bool) { - // TODO: Expose via /proc/sys/net/ipv4/ip_forward. + // TODO(igudger, bgeffon): Expose via /proc/sys/net/ipv4/ip_forward. s.mu.Lock() s.forwarding = enable s.mu.Unlock() @@ -484,7 +484,7 @@ func (s *Stack) SetForwarding(enable bool) { // Forwarding returns if the packet forwarding between NICs is enabled. func (s *Stack) Forwarding() bool { - // TODO: Expose via /proc/sys/net/ipv4/ip_forward. + // TODO(igudger, bgeffon): Expose via /proc/sys/net/ipv4/ip_forward. s.mu.RLock() defer s.mu.RUnlock() return s.forwarding diff --git a/pkg/tcpip/stack/stack_global_state.go b/pkg/tcpip/stack/stack_global_state.go index f2c6c9a8d..3d7e4b719 100644 --- a/pkg/tcpip/stack/stack_global_state.go +++ b/pkg/tcpip/stack/stack_global_state.go @@ -15,5 +15,5 @@ package stack // StackFromEnv is the global stack created in restore run. -// FIXME +// FIXME(b/36201077) var StackFromEnv *Stack diff --git a/pkg/tcpip/stack/transport_test.go b/pkg/tcpip/stack/transport_test.go index 0c2589083..2df974bf2 100644 --- a/pkg/tcpip/stack/transport_test.go +++ b/pkg/tcpip/stack/transport_test.go @@ -453,7 +453,7 @@ func TestTransportForwarding(t *testing.T) { s := stack.New([]string{"fakeNet"}, []string{"fakeTrans"}, stack.Options{}) s.SetForwarding(true) - // TODO: Change this to a channel NIC. + // TODO(b/123449044): Change this to a channel NIC. id1 := loopback.New() if err := s.CreateNIC(1, id1); err != nil { t.Fatalf("CreateNIC #1 failed: %v", err) diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index 80cd6b4e5..b09137f08 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -444,7 +444,7 @@ type PasscredOption int // TCPInfoOption is used by GetSockOpt to expose TCP statistics. // -// TODO: Add and populate stat fields. +// TODO(b/64800844): Add and populate stat fields. type TCPInfoOption struct { RTT time.Duration RTTVar time.Duration diff --git a/pkg/tcpip/transport/raw/raw.go b/pkg/tcpip/transport/raw/raw.go index 8dada2e4f..f0f60ce91 100644 --- a/pkg/tcpip/transport/raw/raw.go +++ b/pkg/tcpip/transport/raw/raw.go @@ -100,7 +100,7 @@ type endpoint struct { } // NewEndpoint returns a raw endpoint for the given protocols. -// TODO: IP_HDRINCL, IPPROTO_RAW, and AF_PACKET. +// TODO(b/129292371): IP_HDRINCL, IPPROTO_RAW, and AF_PACKET. func NewEndpoint(stack *stack.Stack, netProto tcpip.NetworkProtocolNumber, transProto tcpip.TransportProtocolNumber, waiterQueue *waiter.Queue) (tcpip.Endpoint, *tcpip.Error) { if netProto != header.IPv4ProtocolNumber { return nil, tcpip.ErrUnknownProtocol diff --git a/pkg/tcpip/transport/tcp/BUILD b/pkg/tcpip/transport/tcp/BUILD index e5c05f8c0..d44d63e95 100644 --- a/pkg/tcpip/transport/tcp/BUILD +++ b/pkg/tcpip/transport/tcp/BUILD @@ -73,7 +73,7 @@ go_test( "tcp_test.go", "tcp_timestamp_test.go", ], - # FIXME + # FIXME(b/68809571) tags = ["flaky"], deps = [ ":tcp", diff --git a/pkg/unet/unet.go b/pkg/unet/unet.go index deeea078d..114fb8c5b 100644 --- a/pkg/unet/unet.go +++ b/pkg/unet/unet.go @@ -211,7 +211,7 @@ func SocketPair(packet bool) (*Socket, *Socket, error) { // variable between our two sockets. We only use SocketPair in tests // anyway. // - // NOTE: This is purely due to the fact that the raw + // NOTE(b/27107811): This is purely due to the fact that the raw // syscall does not serve as a boundary for the sanitizer. var race int32 a, err := NewSocket(fds[0]) diff --git a/pkg/unet/unet_test.go b/pkg/unet/unet_test.go index ecc670925..db5485539 100644 --- a/pkg/unet/unet_test.go +++ b/pkg/unet/unet_test.go @@ -40,7 +40,7 @@ func randomFilename() (string, error) { return "", err } - // NOTE: We try to use relative path if possible. This is + // NOTE(b/26918832): We try to use relative path if possible. This is // to help conforming to the unix path length limit. if rel, err := filepath.Rel(cwd, file); err == nil { return rel, nil diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index 2488981f9..712c50ee9 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -231,7 +231,7 @@ func (cm *containerManager) Start(args *StartArgs, _ *struct{}) error { } // Prevent CIDs containing ".." from confusing the sentry when creating // /containers/ directory. - // TODO: Once we have multiple independent roots, this + // TODO(b/129293409): Once we have multiple independent roots, this // check won't be necessary. if path.Clean(args.CID) != args.CID { return fmt.Errorf("container ID shouldn't contain directory traversals such as \"..\": %q", args.CID) @@ -352,7 +352,7 @@ func (cm *containerManager) Restore(o *RestoreOpts, _ *struct{}) error { return fmt.Errorf("creating network: %v", err) } if eps, ok := networkStack.(*epsocket.Stack); ok { - stack.StackFromEnv = eps.Stack // FIXME + stack.StackFromEnv = eps.Stack // FIXME(b/36201077) } info, err := o.FilePayload.Files[0].Stat() if err != nil { diff --git a/runsc/boot/fs.go b/runsc/boot/fs.go index 761142d98..07061b9b3 100644 --- a/runsc/boot/fs.go +++ b/runsc/boot/fs.go @@ -274,7 +274,7 @@ func getMountNameAndOptions(conf *Config, m specs.Mount, fds *fdDispenser) (stri useOverlay = conf.Overlay && !mountFlags(m.Options).ReadOnly default: - // TODO: Support all the mount types and make this a + // TODO(nlacasse): Support all the mount types and make this a // fatal error. Most applications will "just work" without // them, so this is a warning for now. // we do not support. @@ -425,7 +425,7 @@ func addRestoreMount(conf *Config, renv *fs.RestoreEnvironment, m specs.Mount, f if err != nil { return err } - // TODO: Fix this when we support all the mount types and + // TODO(nlacasse): Fix this when we support all the mount types and // make this a fatal error. if fsName == "" { return nil @@ -475,7 +475,7 @@ func createRestoreEnvironment(spec *specs.Spec, conf *Config, fds *fdDispenser) } } - // TODO: handle '/tmp' properly (see mountTmp()). + // TODO(b/67958150): handle '/tmp' properly (see mountTmp()). if !tmpMounted { tmpMount := specs.Mount{ Type: tmpfs, diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 48ecb2626..75ec19c32 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -577,7 +577,7 @@ func (l *Loader) startContainer(k *kernel.Kernel, spec *specs.Spec, conf *Config // sentry currently supports only 1 mount namespace, which is tied to a // single user namespace. Thus we must run in the same user namespace // to access mounts. - // TODO: Create a new mount namespace for the container. + // TODO(b/63601033): Create a new mount namespace for the container. creds := auth.NewUserCredentials( auth.KUID(spec.Process.User.UID), auth.KGID(spec.Process.User.GID), diff --git a/runsc/cmd/checkpoint.go b/runsc/cmd/checkpoint.go index d8f748aa0..f722df055 100644 --- a/runsc/cmd/checkpoint.go +++ b/runsc/cmd/checkpoint.go @@ -105,7 +105,7 @@ func (c *Checkpoint) Execute(_ context.Context, f *flag.FlagSet, args ...interfa return subcommands.ExitSuccess } - // TODO: Make it possible to restore into same container. + // TODO(b/110843694): Make it possible to restore into same container. // For now, we can fake it by destroying the container and making a // new container with the same ID. This hack does not work with docker // which uses the container pid to ensure that the restore-container is diff --git a/runsc/container/container.go b/runsc/container/container.go index 1bed1a97e..a30c217f7 100644 --- a/runsc/container/container.go +++ b/runsc/container/container.go @@ -529,7 +529,7 @@ func (c *Container) WaitPID(pid int32, clearStatus bool) (syscall.WaitStatus, er // SignalContainer sends the signal to the container. If all is true and signal // is SIGKILL, then waits for all processes to exit before returning. // SignalContainer returns an error if the container is already stopped. -// TODO: Distinguish different error types. +// TODO(b/113680494): Distinguish different error types. func (c *Container) SignalContainer(sig syscall.Signal, all bool) error { log.Debugf("Signal container %q: %v", c.ID, sig) // Signaling container in Stopped state is allowed. When all=false, diff --git a/runsc/container/container_test.go b/runsc/container/container_test.go index 9fe584aa3..603c4d929 100644 --- a/runsc/container/container_test.go +++ b/runsc/container/container_test.go @@ -242,10 +242,10 @@ func configs(opts ...configOption) []*boot.Config { case overlay: c.Overlay = true case kvm: - // TODO: KVM tests are flaky. Disable until fixed. + // TODO(b/112165693): KVM tests are flaky. Disable until fixed. continue - // TODO: KVM doesn't work with --race. + // TODO(b/68787993): KVM doesn't work with --race. if testutil.RaceEnabled { continue } diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 92495c69e..48a0dafe2 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -267,7 +267,7 @@ func (s *Sandbox) Event(cid string) (*boot.Event, error) { defer conn.Close() var e boot.Event - // TODO: Pass in the container id (cid) here. The sandbox + // TODO(b/129292330): Pass in the container id (cid) here. The sandbox // should return events only for that container. if err := conn.Call(boot.ContainerEvent, nil, &e); err != nil { return nil, fmt.Errorf("retrieving event data from sandbox: %v", err) @@ -457,7 +457,7 @@ func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bund } if conf.Platform == boot.PlatformPtrace { - // TODO: Also set a new PID namespace so that we limit + // TODO(b/75837838): Also set a new PID namespace so that we limit // access to other host processes. log.Infof("Sandbox will be started in the current PID namespace") } else { @@ -520,7 +520,7 @@ func (s *Sandbox) createSandboxProcess(spec *specs.Spec, conf *boot.Config, bund // root for itself, so it has to have the CAP_SYS_ADMIN // capability. // - // FIXME: The current implementations of + // FIXME(b/122554829): The current implementations of // os/exec doesn't allow to set ambient capabilities if // a process is started in a new user namespace. As a // workaround, we start the sandbox process with the 0 diff --git a/runsc/specutils/specutils.go b/runsc/specutils/specutils.go index 32f81b8d4..ac85bec71 100644 --- a/runsc/specutils/specutils.go +++ b/runsc/specutils/specutils.go @@ -90,7 +90,7 @@ func ValidateSpec(spec *specs.Spec) error { log.Warningf("AppArmor profile %q is being ignored", spec.Process.ApparmorProfile) } - // TODO: Apply seccomp to application inside sandbox. + // TODO(b/72226747): Apply seccomp to application inside sandbox. if spec.Linux != nil && spec.Linux.Seccomp != nil { log.Warningf("Seccomp spec is being ignored") } @@ -220,7 +220,7 @@ func Capabilities(enableRaw bool, specCaps *specs.LinuxCapabilities) (*auth.Task if caps.PermittedCaps, err = capsFromNames(specCaps.Permitted, skipSet); err != nil { return nil, err } - // TODO: Support ambient capabilities. + // TODO(nlacasse): Support ambient capabilities. } return &caps, nil } diff --git a/test/syscalls/BUILD b/test/syscalls/BUILD index 94e0f24e0..d35f59433 100644 --- a/test/syscalls/BUILD +++ b/test/syscalls/BUILD @@ -277,7 +277,7 @@ syscall_test(test = "//test/syscalls/linux:sendfile_test") syscall_test(test = "//test/syscalls/linux:sigaction_test") -# TODO: Enable once the test passes in runsc. +# TODO(b/119826902): Enable once the test passes in runsc. # syscall_test(test = "//test/syscalls/linux:sigaltstack_test") syscall_test(test = "//test/syscalls/linux:sigiret_test") @@ -414,7 +414,7 @@ syscall_test( ) syscall_test( - # NOTE: Large sendmsg may stall a long time. + # NOTE(b/116636318): Large sendmsg may stall a long time. size = "enormous", test = "//test/syscalls/linux:socket_unix_dgram_local_test", ) @@ -437,7 +437,7 @@ syscall_test( ) syscall_test( - # NOTE: Large sendmsg may stall a long time. + # NOTE(b/116636318): Large sendmsg may stall a long time. size = "enormous", test = "//test/syscalls/linux:socket_unix_seqpacket_local_test", ) diff --git a/test/syscalls/build_defs.bzl b/test/syscalls/build_defs.bzl index 610b030b2..cd74a769d 100644 --- a/test/syscalls/build_defs.bzl +++ b/test/syscalls/build_defs.bzl @@ -78,10 +78,10 @@ def _syscall_test( tags += [full_platform, "file_" + file_access] # Add tag to prevent the tests from running in a Bazel sandbox. - # TODO: Make the tests run without this tag. + # TODO(b/120560048): Make the tests run without this tag. tags.append("no-sandbox") - # TODO: KVM tests are tagged "manual" to until the platform is + # TODO(b/112165693): KVM tests are tagged "manual" to until the platform is # more stable. if platform == "kvm": tags += ["manual"] diff --git a/test/syscalls/linux/32bit.cc b/test/syscalls/linux/32bit.cc index 230648c9b..78baf548e 100644 --- a/test/syscalls/linux/32bit.cc +++ b/test/syscalls/linux/32bit.cc @@ -80,11 +80,11 @@ constexpr int kExitCode = 42; TEST(Syscall32Bit, Int80) { switch (GvisorPlatform()) { case Platform::kKVM: - // TODO: 32-bit segments are broken (but not explictly + // TODO(b/111805002): 32-bit segments are broken (but not explictly // disabled). return; case Platform::kPtrace: - // TODO: The ptrace platform does not have a + // TODO(gvisor.dev/issue/167): The ptrace platform does not have a // consistent story here. return; case Platform::kNative: @@ -99,10 +99,10 @@ TEST(Syscall32Bit, Int80) { TEST(Syscall32Bit, Sysenter) { switch (GvisorPlatform()) { case Platform::kKVM: - // TODO: See above. + // TODO(b/111805002): See above. return; case Platform::kPtrace: - // TODO: See above. + // TODO(gvisor.dev/issue/167): See above. return; case Platform::kNative: break; @@ -123,10 +123,10 @@ TEST(Syscall32Bit, Sysenter) { TEST(Syscall32Bit, Syscall) { switch (GvisorPlatform()) { case Platform::kKVM: - // TODO: See above. + // TODO(b/111805002): See above. return; case Platform::kPtrace: - // TODO: See above. + // TODO(gvisor.dev/issue/167): See above. return; case Platform::kNative: break; @@ -207,7 +207,7 @@ void FarCall32() { TEST(Call32Bit, Disallowed) { switch (GvisorPlatform()) { case Platform::kKVM: - // TODO: See above. + // TODO(b/111805002): See above. return; case Platform::kPtrace: // The ptrace platform cannot prevent switching to compatibility mode. diff --git a/test/syscalls/linux/aio.cc b/test/syscalls/linux/aio.cc index 06643ccb8..b96aab9b9 100644 --- a/test/syscalls/linux/aio.cc +++ b/test/syscalls/linux/aio.cc @@ -103,7 +103,7 @@ TEST_F(AIOTest, BasicWrite) { // aio implementation uses aio_ring. gVisor doesn't and returns all zeroes. // Linux implements aio_ring, so skip the zeroes check. // - // TODO: Remove when gVisor implements aio_ring. + // TODO(b/65486370): Remove when gVisor implements aio_ring. auto ring = reinterpret_cast(ctx_); auto magic = IsRunningOnGvisor() ? 0 : AIO_RING_MAGIC; EXPECT_EQ(ring->magic, magic); diff --git a/test/syscalls/linux/chmod.cc b/test/syscalls/linux/chmod.cc index 2f2ff3b7d..2f42fe326 100644 --- a/test/syscalls/linux/chmod.cc +++ b/test/syscalls/linux/chmod.cc @@ -235,7 +235,7 @@ TEST(ChmodTest, FchmodFileToNoPermissionsSucceeds_NoRandomSave) { // Verify that we can get a RW FD after chmod, even if a RO fd is left open. TEST(ChmodTest, ChmodWritableWithOpenFD) { - // FIXME: broken on hostfs. + // FIXME(b/72455313): broken on hostfs. if (IsRunningOnGvisor()) { return; } diff --git a/test/syscalls/linux/epoll.cc b/test/syscalls/linux/epoll.cc index 7b1d83ad8..b4a3bfcba 100644 --- a/test/syscalls/linux/epoll.cc +++ b/test/syscalls/linux/epoll.cc @@ -56,7 +56,7 @@ TEST(EpollTest, AllWritable) { struct epoll_event result[kFDsPerEpoll]; ASSERT_THAT(RetryEINTR(epoll_wait)(epollfd.get(), result, kFDsPerEpoll, -1), SyscallSucceedsWithValue(kFDsPerEpoll)); - // TODO: Why do some tests check epoll_event::data, and others + // TODO(edahlgren): Why do some tests check epoll_event::data, and others // don't? Does Linux actually guarantee that, in any of these test cases, // epoll_wait will necessarily write out the epoll_events in the order that // they were registered? diff --git a/test/syscalls/linux/exec_binary.cc b/test/syscalls/linux/exec_binary.cc index 187696ed9..c10d85398 100644 --- a/test/syscalls/linux/exec_binary.cc +++ b/test/syscalls/linux/exec_binary.cc @@ -285,7 +285,7 @@ ElfBinary<64> StandardElf() { elf.header.e_phoff = sizeof(elf.header); elf.header.e_phentsize = sizeof(decltype(elf)::ElfPhdr); - // TODO: Always include a PT_GNU_STACK segment to + // TODO(gvisor.dev/issue/153): Always include a PT_GNU_STACK segment to // disable executable stacks. With this omitted the stack (and all PROT_READ) // mappings should be executable, but gVisor doesn't support that. decltype(elf)::ElfPhdr phdr = {}; @@ -403,7 +403,7 @@ TEST(ElfTest, DataSegment) { // Linux will allow PT_LOAD segments to overlap. TEST(ElfTest, DirectlyOverlappingSegments) { - // NOTE: see PIEOutOfOrderSegments. + // NOTE(b/37289926): see PIEOutOfOrderSegments. SKIP_IF(IsRunningOnGvisor()); ElfBinary<64> elf = StandardElf(); @@ -439,7 +439,7 @@ TEST(ElfTest, DirectlyOverlappingSegments) { // Linux allows out-of-order PT_LOAD segments. TEST(ElfTest, OutOfOrderSegments) { - // NOTE: see PIEOutOfOrderSegments. + // NOTE(b/37289926): see PIEOutOfOrderSegments. SKIP_IF(IsRunningOnGvisor()); ElfBinary<64> elf = StandardElf(); @@ -670,7 +670,7 @@ TEST(ElfTest, PIENonZeroStart) { } TEST(ElfTest, PIEOutOfOrderSegments) { - // TODO: This triggers a bug in Linux where it computes the size + // TODO(b/37289926): This triggers a bug in Linux where it computes the size // of the binary as 0x20000 - 0x40000 = 0xfffffffffffe0000, which obviously // fails to map. // @@ -1005,7 +1005,7 @@ TEST(ElfTest, NoExecute) { // Execute, but no read permissions on the binary works just fine. TEST(ElfTest, NoRead) { - // TODO: gVisor's backing filesystem may prevent the + // TODO(gvisor.dev/issue/160): gVisor's backing filesystem may prevent the // sentry from reading the executable. SKIP_IF(IsRunningOnGvisor()); @@ -1024,7 +1024,7 @@ TEST(ElfTest, NoRead) { ASSERT_NO_ERRNO(WaitStopped(child)); - // TODO: A task with a non-readable executable is marked + // TODO(gvisor.dev/issue/160): A task with a non-readable executable is marked // non-dumpable, preventing access to proc files. gVisor does not implement // this behavior. } diff --git a/test/syscalls/linux/file_base.h b/test/syscalls/linux/file_base.h index 19c9a5053..43f568111 100644 --- a/test/syscalls/linux/file_base.h +++ b/test/syscalls/linux/file_base.h @@ -52,7 +52,7 @@ class FileTest : public ::testing::Test { test_file_fd_ = ASSERT_NO_ERRNO_AND_VALUE( Open(test_file_name_, O_CREAT | O_RDWR, S_IRUSR | S_IWUSR)); - // FIXME: enable when mknod syscall is supported. + // FIXME(edahlgren): enable when mknod syscall is supported. // test_fifo_name_ = NewTempAbsPath(); // ASSERT_THAT(mknod(test_fifo_name_.c_str()), S_IFIFO|0644, 0, // SyscallSucceeds()); @@ -97,7 +97,7 @@ class FileTest : public ::testing::Test { UnlinkFile(); ClosePipes(); - // FIXME: enable when mknod syscall is supported. + // FIXME(edahlgren): enable when mknod syscall is supported. // close(test_fifo_[0]); // close(test_fifo_[1]); // unlink(test_fifo_name_.c_str()); diff --git a/test/syscalls/linux/ioctl.cc b/test/syscalls/linux/ioctl.cc index de29047e0..c7741a177 100644 --- a/test/syscalls/linux/ioctl.cc +++ b/test/syscalls/linux/ioctl.cc @@ -158,7 +158,7 @@ TEST_F(IoctlTest, FIOASYNCNoTarget) { } TEST_F(IoctlTest, FIOASYNCSelfTarget) { - // FIXME: gVisor erroneously sends SIGIO on close(2), which would + // FIXME(b/120624367): gVisor erroneously sends SIGIO on close(2), which would // kill the test when pair goes out of scope. Temporarily ignore SIGIO so that // that the close signal is ignored. struct sigaction sa; @@ -195,7 +195,7 @@ TEST_F(IoctlTest, FIOASYNCSelfTarget) { // Equivalent to FIOASYNCSelfTarget except that FIOSETOWN is called before // FIOASYNC. TEST_F(IoctlTest, FIOASYNCSelfTarget2) { - // FIXME: gVisor erroneously sends SIGIO on close(2), which would + // FIXME(b/120624367): gVisor erroneously sends SIGIO on close(2), which would // kill the test when pair goes out of scope. Temporarily ignore SIGIO so that // that the close signal is ignored. struct sigaction sa; diff --git a/test/syscalls/linux/ip_socket_test_util.cc b/test/syscalls/linux/ip_socket_test_util.cc index 4ad787cc0..0a149c2e5 100644 --- a/test/syscalls/linux/ip_socket_test_util.cc +++ b/test/syscalls/linux/ip_socket_test_util.cc @@ -24,7 +24,7 @@ namespace gvisor { namespace testing { PosixErrorOr InterfaceIndex(std::string name) { - // TODO: Consider using netlink. + // TODO(igudger): Consider using netlink. ifreq req = {}; memcpy(req.ifr_name, name.c_str(), name.size()); ASSIGN_OR_RETURN_ERRNO(auto sock, Socket(AF_INET, SOCK_DGRAM, 0)); diff --git a/test/syscalls/linux/lseek.cc b/test/syscalls/linux/lseek.cc index fb6a1546e..6a4f1423c 100644 --- a/test/syscalls/linux/lseek.cc +++ b/test/syscalls/linux/lseek.cc @@ -194,7 +194,7 @@ TEST(LseekTest, EtcPasswdDup) { ASSERT_THAT(lseek(fd3.get(), 0, SEEK_CUR), SyscallSucceedsWithValue(1000)); } -// TODO: Add tests where we have donated in sockets. +// TODO(magi): Add tests where we have donated in sockets. } // namespace diff --git a/test/syscalls/linux/mkdir.cc b/test/syscalls/linux/mkdir.cc index 84db45eb3..50807b68f 100644 --- a/test/syscalls/linux/mkdir.cc +++ b/test/syscalls/linux/mkdir.cc @@ -36,7 +36,7 @@ class MkdirTest : public ::testing::Test { // TearDown unlinks created files. void TearDown() override { - // FIXME: We don't currently implement rmdir. + // FIXME(edahlgren): We don't currently implement rmdir. // We do this unconditionally because there's no harm in trying. rmdir(dirname_.c_str()); } diff --git a/test/syscalls/linux/mmap.cc b/test/syscalls/linux/mmap.cc index b500e79a4..a4fb9d1e0 100644 --- a/test/syscalls/linux/mmap.cc +++ b/test/syscalls/linux/mmap.cc @@ -816,7 +816,7 @@ class MMapFileTest : public MMapTest { // MAP_POPULATE allowed. // There isn't a good way to verify it actually did anything. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, MapPopulate) { ASSERT_THAT( Map(0, kPageSize, PROT_READ, MAP_PRIVATE | MAP_POPULATE, fd_.get(), 0), @@ -825,7 +825,7 @@ TEST_F(MMapFileTest, MapPopulate) { // MAP_POPULATE on a short file. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, MapPopulateShort) { ASSERT_THAT(Map(0, 2 * kPageSize, PROT_READ, MAP_PRIVATE | MAP_POPULATE, fd_.get(), 0), @@ -923,7 +923,7 @@ TEST_F(MMapFileTest, WriteSharedOnReadOnlyFd) { // MAP_SHARED PROT_READ not allowed on write-only FDs. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, ReadSharedOnWriteOnlyFd) { const FileDescriptor fd = ASSERT_NO_ERRNO_AND_VALUE(Open(filename_, O_WRONLY)); @@ -936,7 +936,7 @@ TEST_F(MMapFileTest, ReadSharedOnWriteOnlyFd) { // MAP_SHARED PROT_WRITE not allowed on write-only FDs. // The FD must always be readable. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, WriteSharedOnWriteOnlyFd) { const FileDescriptor fd = ASSERT_NO_ERRNO_AND_VALUE(Open(filename_, O_WRONLY)); @@ -1371,7 +1371,7 @@ TEST_F(MMapFileTest, WritePrivate) { // SIGBUS raised when writing past end of file to a private mapping. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, SigBusDeathWritePrivate) { SetupGvisorDeathTest(); @@ -1390,7 +1390,7 @@ TEST_F(MMapFileTest, SigBusDeathWritePrivate) { // SIGBUS raised when reading past end of file on a shared mapping. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, SigBusDeathReadShared) { SetupGvisorDeathTest(); @@ -1410,7 +1410,7 @@ TEST_F(MMapFileTest, SigBusDeathReadShared) { // SIGBUS raised when reading past end of file on a shared mapping. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, SigBusDeathWriteShared) { SetupGvisorDeathTest(); @@ -1459,7 +1459,7 @@ TEST_F(MMapFileTest, NoSigBusOnPageContainingEOFWritePrivate) { // Tests that SIGBUS is not raised when reading from a file-mapped page // containing EOF, *after* the EOF for a shared mapping. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, NoSigBusOnPageContainingEOFReadShared) { uintptr_t addr; ASSERT_THAT(addr = Map(0, 2 * kPageSize, PROT_READ, MAP_SHARED, fd_.get(), 0), @@ -1476,7 +1476,7 @@ TEST_F(MMapFileTest, NoSigBusOnPageContainingEOFReadShared) { // Tests that SIGBUS is not raised when writing to a file-mapped page containing // EOF, *after* the EOF for a shared mapping. // -// FIXME: Parameterize. +// FIXME(b/37222275): Parameterize. TEST_F(MMapFileTest, NoSigBusOnPageContainingEOFWriteShared) { uintptr_t addr; ASSERT_THAT(addr = Map(0, 2 * kPageSize, PROT_READ | PROT_WRITE, MAP_SHARED, diff --git a/test/syscalls/linux/open.cc b/test/syscalls/linux/open.cc index cdc226300..22e4666c2 100644 --- a/test/syscalls/linux/open.cc +++ b/test/syscalls/linux/open.cc @@ -279,7 +279,7 @@ TEST_F(OpenTest, Null) { ASSERT_THAT(open(&c, O_RDONLY), SyscallFailsWithErrno(ENOENT)); } -// NOTE: While the man pages specify that this behavior should be +// NOTE(b/119785738): While the man pages specify that this behavior should be // undefined, Linux truncates the file on opening read only if we have write // permission, so we will too. TEST_F(OpenTest, CanTruncateReadOnly) { diff --git a/test/syscalls/linux/partial_bad_buffer.cc b/test/syscalls/linux/partial_bad_buffer.cc index 073a6b8c1..71288ebc4 100644 --- a/test/syscalls/linux/partial_bad_buffer.cc +++ b/test/syscalls/linux/partial_bad_buffer.cc @@ -158,7 +158,7 @@ TEST_F(PartialBadBufferTest, PreadvSmall) { } TEST_F(PartialBadBufferTest, WriteBig) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); @@ -168,7 +168,7 @@ TEST_F(PartialBadBufferTest, WriteBig) { } TEST_F(PartialBadBufferTest, WriteSmall) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); @@ -178,7 +178,7 @@ TEST_F(PartialBadBufferTest, WriteSmall) { } TEST_F(PartialBadBufferTest, PwriteBig) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); @@ -188,7 +188,7 @@ TEST_F(PartialBadBufferTest, PwriteBig) { } TEST_F(PartialBadBufferTest, PwriteSmall) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); @@ -198,7 +198,7 @@ TEST_F(PartialBadBufferTest, PwriteSmall) { } TEST_F(PartialBadBufferTest, WritevBig) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); @@ -211,7 +211,7 @@ TEST_F(PartialBadBufferTest, WritevBig) { } TEST_F(PartialBadBufferTest, WritevSmall) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); @@ -224,7 +224,7 @@ TEST_F(PartialBadBufferTest, WritevSmall) { } TEST_F(PartialBadBufferTest, PwritevBig) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); @@ -238,7 +238,7 @@ TEST_F(PartialBadBufferTest, PwritevBig) { } TEST_F(PartialBadBufferTest, PwritevSmall) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); @@ -279,7 +279,7 @@ TEST_F(PartialBadBufferTest, GetdentsOneEntry) { // Verify that when write returns EFAULT the kernel hasn't silently written // the initial valid bytes. TEST_F(PartialBadBufferTest, WriteEfaultIsntPartial) { - // FIXME: The sentry write syscalls will return immediately + // FIXME(b/24788078): The sentry write syscalls will return immediately // if Access returns an error, but Access may not return an error // and the sentry will instead perform a partial write. SKIP_IF(IsRunningOnGvisor()); diff --git a/test/syscalls/linux/pipe.cc b/test/syscalls/linux/pipe.cc index c49ec9f09..abd10b11b 100644 --- a/test/syscalls/linux/pipe.cc +++ b/test/syscalls/linux/pipe.cc @@ -36,7 +36,7 @@ namespace { // Buffer size of a pipe. // -// TODO: Get this from F_GETPIPE_SZ. +// TODO(b/35762278): Get this from F_GETPIPE_SZ. constexpr int kPipeSize = 65536; class PipeTest : public ::testing::Test { @@ -316,7 +316,7 @@ TEST_F(PipeTest, BlockWriteClosed) { // Blocking write returns EPIPE when read end is closed even if something has // been written. // -// FIXME: Pipe writes blocking early allows S/R to interrupt the +// FIXME(b/35924046): Pipe writes blocking early allows S/R to interrupt the // write(2) call before the buffer is full. Then the next call will will return // non-zero instead of EPIPE. TEST_F(PipeTest, BlockPartialWriteClosed_NoRandomSave) { @@ -329,7 +329,7 @@ TEST_F(PipeTest, BlockPartialWriteClosed_NoRandomSave) { // Write more than fits in the buffer. Blocks then returns partial write // when the other end is closed. The next call returns EPIPE. if (IsRunningOnGvisor()) { - // FIXME: Pipe writes block early on gVisor, resulting in a + // FIXME(b/35924046): Pipe writes block early on gVisor, resulting in a // shorter than expected partial write. ASSERT_THAT(write(wfd, buf.data(), buf.size()), SyscallSucceedsWithValue(::testing::Gt(0))); diff --git a/test/syscalls/linux/proc.cc b/test/syscalls/linux/proc.cc index 3ec31ae8b..7ba274226 100644 --- a/test/syscalls/linux/proc.cc +++ b/test/syscalls/linux/proc.cc @@ -61,7 +61,7 @@ #include "test/util/thread_util.h" #include "test/util/timer_util.h" -// NOTE: No, this isn't really a syscall but this is a really simple +// NOTE(magi): No, this isn't really a syscall but this is a really simple // way to get it tested on both gVisor, PTrace and Linux. using ::testing::AllOf; @@ -489,7 +489,7 @@ TEST(ProcSelfMaps, Map1) { } TEST(ProcSelfMaps, Map2) { - // NOTE: The permissions must be different or the pages will get merged. + // NOTE(magi): The permissions must be different or the pages will get merged. Mapping map1 = ASSERT_NO_ERRNO_AND_VALUE( MmapAnon(kPageSize, PROT_READ | PROT_EXEC, MAP_PRIVATE)); Mapping map2 = @@ -564,7 +564,7 @@ TEST(ProcSelfMaps, MapUnmap) { } TEST(ProcSelfMaps, Mprotect) { - // FIXME: Linux's mprotect() sometimes fails to merge VMAs in this + // FIXME(jamieliu): Linux's mprotect() sometimes fails to merge VMAs in this // case. SKIP_IF(!IsRunningOnGvisor()); @@ -977,7 +977,7 @@ void MapPopulateRSS(int prot, uint64_t* before, uint64_t* after) { *after = ASSERT_NO_ERRNO_AND_VALUE(CurrentRSS()); } -// TODO: Test for PROT_READ + MAP_POPULATE anonymous mappings. Their +// TODO(b/73896574): Test for PROT_READ + MAP_POPULATE anonymous mappings. Their // semantics are more subtle: // // Small pages -> Zero page mapped, not counted in RSS @@ -1140,7 +1140,7 @@ TEST(ProcPidStatusTest, ValuesAreTabDelimited) { // Threads properly counts running threads. // -// TODO: Test zombied threads while the thread group leader is still +// TODO(mpratt): Test zombied threads while the thread group leader is still // running with generalized fork and clone children from the wait test. TEST(ProcPidStatusTest, Threads) { char buf[4096] = {}; @@ -1274,7 +1274,7 @@ TEST(ProcPidSymlink, SubprocessRunning) { SyscallSucceedsWithValue(sizeof(buf))); } -// FIXME: Inconsistent behavior between gVisor and linux +// FIXME(gvisor.dev/issue/164): Inconsistent behavior between gVisor and linux // on proc files. TEST(ProcPidSymlink, SubprocessZombied) { ASSERT_NO_ERRNO(SetCapability(CAP_DAC_OVERRIDE, false)); @@ -1298,13 +1298,13 @@ TEST(ProcPidSymlink, SubprocessZombied) { SyscallFailsWithErrno(want)); } - // FIXME: Inconsistent behavior between gVisor and linux + // FIXME(gvisor.dev/issue/164): Inconsistent behavior between gVisor and linux // on proc files. // 4.17 & gVisor: Syscall succeeds and returns 1 // EXPECT_THAT(ReadlinkWhileZombied("ns/pid", buf, sizeof(buf)), // SyscallFailsWithErrno(EACCES)); - // FIXME: Inconsistent behavior between gVisor and linux + // FIXME(gvisor.dev/issue/164): Inconsistent behavior between gVisor and linux // on proc files. // 4.17 & gVisor: Syscall succeeds and returns 1. // EXPECT_THAT(ReadlinkWhileZombied("ns/user", buf, sizeof(buf)), @@ -1313,7 +1313,7 @@ TEST(ProcPidSymlink, SubprocessZombied) { // Test whether /proc/PID/ symlinks can be read for an exited process. TEST(ProcPidSymlink, SubprocessExited) { - // FIXME: These all succeed on gVisor. + // FIXME(gvisor.dev/issue/164): These all succeed on gVisor. SKIP_IF(IsRunningOnGvisor()); char buf[1]; @@ -1404,7 +1404,7 @@ TEST(ProcPidFile, SubprocessZombie) { EXPECT_THAT(ReadWhileZombied("uid_map", buf, sizeof(buf)), SyscallSucceedsWithValue(sizeof(buf))); - // FIXME: Inconsistent behavior between gVisor and linux + // FIXME(gvisor.dev/issue/164): Inconsistent behavior between gVisor and linux // on proc files. // gVisor & 4.17: Succeeds and returns 1. // EXPECT_THAT(ReadWhileZombied("io", buf, sizeof(buf)), @@ -1415,7 +1415,7 @@ TEST(ProcPidFile, SubprocessZombie) { TEST(ProcPidFile, SubprocessExited) { char buf[1]; - // FIXME: Inconsistent behavior between kernels + // FIXME(gvisor.dev/issue/164): Inconsistent behavior between kernels // gVisor: Fails with ESRCH. // 4.17: Succeeds and returns 1. // EXPECT_THAT(ReadWhileExited("auxv", buf, sizeof(buf)), @@ -1425,7 +1425,7 @@ TEST(ProcPidFile, SubprocessExited) { SyscallFailsWithErrno(ESRCH)); if (!IsRunningOnGvisor()) { - // FIXME: Succeeds on gVisor. + // FIXME(gvisor.dev/issue/164): Succeeds on gVisor. EXPECT_THAT(ReadWhileExited("comm", buf, sizeof(buf)), SyscallFailsWithErrno(ESRCH)); } @@ -1434,25 +1434,25 @@ TEST(ProcPidFile, SubprocessExited) { SyscallSucceedsWithValue(sizeof(buf))); if (!IsRunningOnGvisor()) { - // FIXME: Succeeds on gVisor. + // FIXME(gvisor.dev/issue/164): Succeeds on gVisor. EXPECT_THAT(ReadWhileExited("io", buf, sizeof(buf)), SyscallFailsWithErrno(ESRCH)); } if (!IsRunningOnGvisor()) { - // FIXME: Returns EOF on gVisor. + // FIXME(gvisor.dev/issue/164): Returns EOF on gVisor. EXPECT_THAT(ReadWhileExited("maps", buf, sizeof(buf)), SyscallFailsWithErrno(ESRCH)); } if (!IsRunningOnGvisor()) { - // FIXME: Succeeds on gVisor. + // FIXME(gvisor.dev/issue/164): Succeeds on gVisor. EXPECT_THAT(ReadWhileExited("stat", buf, sizeof(buf)), SyscallFailsWithErrno(ESRCH)); } if (!IsRunningOnGvisor()) { - // FIXME: Succeeds on gVisor. + // FIXME(gvisor.dev/issue/164): Succeeds on gVisor. EXPECT_THAT(ReadWhileExited("status", buf, sizeof(buf)), SyscallFailsWithErrno(ESRCH)); } diff --git a/test/syscalls/linux/proc_pid_smaps.cc b/test/syscalls/linux/proc_pid_smaps.cc index 5f9c42ce5..cf5c462f3 100644 --- a/test/syscalls/linux/proc_pid_smaps.cc +++ b/test/syscalls/linux/proc_pid_smaps.cc @@ -82,7 +82,7 @@ struct ProcPidSmapsEntry { // Given the value part of a /proc/[pid]/smaps field containing a value in kB // (for example, " 4 kB", returns the value in kB (in this example, 4). PosixErrorOr SmapsValueKb(absl::string_view value) { - // TODO: let us use RE2 or + // TODO(jamieliu): let us use RE2 or std::pair parts = absl::StrSplit(value, ' ', absl::SkipEmpty()); if (parts.second != "kB") { diff --git a/test/syscalls/linux/ptrace.cc b/test/syscalls/linux/ptrace.cc index 1c9d7d4f4..e0c56f1fc 100644 --- a/test/syscalls/linux/ptrace.cc +++ b/test/syscalls/linux/ptrace.cc @@ -823,7 +823,7 @@ TEST(PtraceTest, TEST(PtraceTest, Int3) { switch (GvisorPlatform()) { case Platform::kKVM: - // TODO: int3 isn't handled properly. + // TODO(b/124248694): int3 isn't handled properly. return; default: break; diff --git a/test/syscalls/linux/pwrite64.cc b/test/syscalls/linux/pwrite64.cc index 60ae6de1f..485b1e48d 100644 --- a/test/syscalls/linux/pwrite64.cc +++ b/test/syscalls/linux/pwrite64.cc @@ -30,7 +30,7 @@ namespace { // This test is currently very rudimentary. // -// TODO: +// TODO(edahlgren): // * bad buffer states (EFAULT). // * bad fds (wrong permission, wrong type of file, EBADF). // * check offset is not incremented. diff --git a/test/syscalls/linux/readv_socket.cc b/test/syscalls/linux/readv_socket.cc index 2c129b7e8..cf22c395e 100644 --- a/test/syscalls/linux/readv_socket.cc +++ b/test/syscalls/linux/readv_socket.cc @@ -41,7 +41,7 @@ class ReadvSocketTest : public SocketTest { ASSERT_THAT(write(test_unix_seqpacket_socket_[1], kReadvTestData, kReadvTestDataSize), SyscallSucceedsWithValue(kReadvTestDataSize)); - // FIXME: Enable when possible. + // FIXME(b/69821513): Enable when possible. // ASSERT_THAT(write(test_tcp_socket_[1], kReadvTestData, // kReadvTestDataSize), // SyscallSucceedsWithValue(kReadvTestDataSize)); diff --git a/test/syscalls/linux/rtsignal.cc b/test/syscalls/linux/rtsignal.cc index 1f2fed7cc..ff948f9d5 100644 --- a/test/syscalls/linux/rtsignal.cc +++ b/test/syscalls/linux/rtsignal.cc @@ -75,7 +75,7 @@ class RtSignalTest : public ::testing::Test { static int rt_sigqueueinfo(pid_t tgid, int sig, siginfo_t* uinfo) { int ret; do { - // NOTE: rt_sigqueueinfo(2) could return EAGAIN for RT signals. + // NOTE(b/25434735): rt_sigqueueinfo(2) could return EAGAIN for RT signals. ret = syscall(SYS_rt_sigqueueinfo, tgid, sig, uinfo); } while (ret == -1 && errno == EAGAIN); return ret; diff --git a/test/syscalls/linux/socket_inet_loopback.cc b/test/syscalls/linux/socket_inet_loopback.cc index cdc5c0ce8..14d7827c2 100644 --- a/test/syscalls/linux/socket_inet_loopback.cc +++ b/test/syscalls/linux/socket_inet_loopback.cc @@ -221,7 +221,7 @@ TEST_P(SocketInetReusePortTest, TcpPortReuseMultiThread) { std::atomic connects_received = ATOMIC_VAR_INIT(0); std::unique_ptr listen_thread[kThreadCount]; int accept_counts[kThreadCount] = {}; - // TODO: figure how to not disable S/R for the whole test. + // TODO(avagin): figure how to not disable S/R for the whole test. // We need to take into account that this test executes a lot of system // calls from many threads. DisableSave ds; @@ -325,7 +325,7 @@ TEST_P(SocketInetReusePortTest, UdpPortReuseMultiThread) { std::atomic packets_received = ATOMIC_VAR_INIT(0); std::unique_ptr receiver_thread[kThreadCount]; int packets_per_socket[kThreadCount] = {}; - // TODO: figure how to not disable S/R for the whole test. + // TODO(avagin): figure how to not disable S/R for the whole test. DisableSave ds; // Too expensive. for (int i = 0; i < kThreadCount; i++) { @@ -642,7 +642,7 @@ TEST_P(SocketMultiProtocolInetLoopbackTest, V6OnlyV6AnyReservesV6) { TEST_P(SocketMultiProtocolInetLoopbackTest, V6EphemeralPortReserved) { auto const& param = GetParam(); - // FIXME + // FIXME(b/114268588) SKIP_IF(IsRunningOnGvisor() && param.type == SOCK_STREAM); for (int i = 0; true; i++) { @@ -743,7 +743,7 @@ TEST_P(SocketMultiProtocolInetLoopbackTest, V6EphemeralPortReserved) { TEST_P(SocketMultiProtocolInetLoopbackTest, V4MappedEphemeralPortReserved) { auto const& param = GetParam(); - // FIXME + // FIXME(b/114268588) SKIP_IF(IsRunningOnGvisor() && param.type == SOCK_STREAM); for (int i = 0; true; i++) { @@ -867,7 +867,7 @@ TEST_P(SocketMultiProtocolInetLoopbackTest, V4MappedEphemeralPortReserved) { TEST_P(SocketMultiProtocolInetLoopbackTest, V4EphemeralPortReserved) { auto const& param = GetParam(); - // FIXME + // FIXME(b/114268588) SKIP_IF(IsRunningOnGvisor() && param.type == SOCK_STREAM); for (int i = 0; true; i++) { diff --git a/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc b/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc index 8b4fc57b6..9dd9e1bd6 100644 --- a/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc +++ b/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc @@ -244,7 +244,7 @@ TestAddress V4Multicast() { // set interface or group membership. TEST_P(IPv4UDPUnboundExternalNetworkingSocketTest, TestSendMulticastSelfNoGroup) { - // FIXME: A group membership is not required for external + // FIXME(b/125485338): A group membership is not required for external // multicast on gVisor. SKIP_IF(IsRunningOnGvisor()); @@ -371,7 +371,7 @@ TEST_P(IPv4UDPUnboundExternalNetworkingSocketTest, // Check that multicast packets won't be delivered to another socket with no // set interface or group membership. TEST_P(IPv4UDPUnboundExternalNetworkingSocketTest, TestSendMulticastNoGroup) { - // FIXME: A group membership is not required for external + // FIXME(b/125485338): A group membership is not required for external // multicast on gVisor. SKIP_IF(IsRunningOnGvisor()); diff --git a/test/syscalls/linux/socket_netlink_route.cc b/test/syscalls/linux/socket_netlink_route.cc index 8d2e7d333..ed4ae1c71 100644 --- a/test/syscalls/linux/socket_netlink_route.cc +++ b/test/syscalls/linux/socket_netlink_route.cc @@ -180,7 +180,7 @@ void CheckGetLinkResponse(const struct nlmsghdr* hdr, int seq, int port) { // RTM_NEWLINK contains at least the header and ifinfomsg. EXPECT_GE(hdr->nlmsg_len, NLMSG_SPACE(sizeof(struct ifinfomsg))); - // TODO: Check ifinfomsg contents and following attrs. + // TODO(mpratt): Check ifinfomsg contents and following attrs. } TEST(NetlinkRouteTest, GetLinkDump) { @@ -370,7 +370,7 @@ TEST(NetlinkRouteTest, GetAddrDump) { // RTM_NEWADDR contains at least the header and ifaddrmsg. EXPECT_GE(hdr->nlmsg_len, sizeof(*hdr) + sizeof(struct ifaddrmsg)); - // TODO: Check ifaddrmsg contents and following attrs. + // TODO(mpratt): Check ifaddrmsg contents and following attrs. })); } diff --git a/test/syscalls/linux/socket_stream_blocking.cc b/test/syscalls/linux/socket_stream_blocking.cc index 8b3f6a647..f0f86c01c 100644 --- a/test/syscalls/linux/socket_stream_blocking.cc +++ b/test/syscalls/linux/socket_stream_blocking.cc @@ -33,7 +33,7 @@ namespace gvisor { namespace testing { TEST_P(BlockingStreamSocketPairTest, BlockPartialWriteClosed) { - // FIXME: gVisor doesn't support SO_SNDBUF on UDS, nor does it + // FIXME(b/35921550): gVisor doesn't support SO_SNDBUF on UDS, nor does it // enforce any limit; it will write arbitrary amounts of data without // blocking. SKIP_IF(IsRunningOnGvisor()); diff --git a/test/syscalls/linux/socket_test_util.cc b/test/syscalls/linux/socket_test_util.cc index 035087566..0be23e541 100644 --- a/test/syscalls/linux/socket_test_util.cc +++ b/test/syscalls/linux/socket_test_util.cc @@ -353,7 +353,7 @@ PosixErrorOr> CreateTCPAcceptBindSocketPair( } MaybeSave(); // Successful accept. - // FIXME + // FIXME(b/110484944) if (connect_result == -1) { absl::SleepFor(absl::Seconds(1)); } diff --git a/test/syscalls/linux/socket_unix.cc b/test/syscalls/linux/socket_unix.cc index 7332b768e..fafb23ad1 100644 --- a/test/syscalls/linux/socket_unix.cc +++ b/test/syscalls/linux/socket_unix.cc @@ -186,7 +186,7 @@ TEST_P(UnixSocketPairTest, BasicFDPassNoSpace) { // BasicFDPassNoSpaceMsgCtrunc sends an FD, but does not provide any space to // receive it. It then verifies that the MSG_CTRUNC flag is set in the msghdr. TEST_P(UnixSocketPairTest, BasicFDPassNoSpaceMsgCtrunc) { - // FIXME: Support MSG_CTRUNC. + // FIXME(gvisor.dev/issue/206): Support MSG_CTRUNC. SKIP_IF(IsRunningOnGvisor()); auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); @@ -224,7 +224,7 @@ TEST_P(UnixSocketPairTest, BasicFDPassNoSpaceMsgCtrunc) { // accomidate the FD, but msg_control is set to NULL. In this case, msg_control // should override msg_controllen. TEST_P(UnixSocketPairTest, BasicFDPassNullControlMsgCtrunc) { - // FIXME: Fix handling of NULL msg_control. + // FIXME(gvisor.dev/issue/207): Fix handling of NULL msg_control. SKIP_IF(IsRunningOnGvisor()); auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); @@ -259,7 +259,7 @@ TEST_P(UnixSocketPairTest, BasicFDPassNullControlMsgCtrunc) { // space to receive it. It then verifies that the MSG_CTRUNC flag is set in the // msghdr. TEST_P(UnixSocketPairTest, BasicFDPassNotEnoughSpaceMsgCtrunc) { - // FIXME: Support MSG_CTRUNC. + // FIXME(gvisor.dev/issue/206): Support MSG_CTRUNC. SKIP_IF(IsRunningOnGvisor()); auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); @@ -296,7 +296,7 @@ TEST_P(UnixSocketPairTest, BasicFDPassNotEnoughSpaceMsgCtrunc) { // space to receive two of them. It then verifies that the MSG_CTRUNC flag is // set in the msghdr. TEST_P(UnixSocketPairTest, BasicThreeFDPassTruncationMsgCtrunc) { - // FIXME: Support MSG_CTRUNC. + // FIXME(gvisor.dev/issue/206): Support MSG_CTRUNC. SKIP_IF(IsRunningOnGvisor()); auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); @@ -408,7 +408,7 @@ TEST_P(UnixSocketPairTest, BasicFDPassUnalignedRecvNoMsgTrunc) { // provides enough space to receive one of them. It then verifies that the // MSG_CTRUNC flag is set in the msghdr. TEST_P(UnixSocketPairTest, BasicTwoFDPassUnalignedRecvTruncationMsgTrunc) { - // FIXME: Support MSG_CTRUNC. + // FIXME(gvisor.dev/issue/206): Support MSG_CTRUNC. SKIP_IF(IsRunningOnGvisor()); auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); @@ -1010,7 +1010,7 @@ TEST_P(UnixSocketPairTest, CredPassNoMsgCtrunc) { // the data without providing space for any credentials and verifies that // MSG_CTRUNC is set in the msghdr. TEST_P(UnixSocketPairTest, CredPassNoSpaceMsgCtrunc) { - // FIXME: Support MSG_CTRUNC. + // FIXME(gvisor.dev/issue/206): Support MSG_CTRUNC. SKIP_IF(IsRunningOnGvisor()); auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); @@ -1061,7 +1061,7 @@ TEST_P(UnixSocketPairTest, CredPassNoSpaceMsgCtrunc) { // the data while providing enough space for only the first field of the // credentials and verifies that MSG_CTRUNC is set in the msghdr. TEST_P(UnixSocketPairTest, CredPassTruncatedMsgCtrunc) { - // FIXME: Support MSG_CTRUNC. + // FIXME(gvisor.dev/issue/206): Support MSG_CTRUNC. SKIP_IF(IsRunningOnGvisor()); auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); @@ -1615,7 +1615,7 @@ TEST_P(UnixSocketPairTest, SocketShutdown) { } TEST_P(UnixSocketPairTest, SocketReopenFromProcfs) { - // TODO: We should be returning ENXIO and NOT EIO. + // TODO(b/122310852): We should be returning ENXIO and NOT EIO. SKIP_IF(IsRunningOnGvisor()); auto sockets = ASSERT_NO_ERRNO_AND_VALUE(NewSocketPair()); diff --git a/test/syscalls/linux/socket_unix_dgram.cc b/test/syscalls/linux/socket_unix_dgram.cc index c17d3990f..5dd5e6d77 100644 --- a/test/syscalls/linux/socket_unix_dgram.cc +++ b/test/syscalls/linux/socket_unix_dgram.cc @@ -28,7 +28,7 @@ namespace testing { namespace { TEST_P(DgramUnixSocketPairTest, WriteOneSideClosed) { - // FIXME: gVisor datagram sockets return EPIPE instead of + // FIXME(b/35925052): gVisor datagram sockets return EPIPE instead of // ECONNREFUSED. SKIP_IF(IsRunningOnGvisor()); diff --git a/test/syscalls/linux/socket_unix_dgram_non_blocking.cc b/test/syscalls/linux/socket_unix_dgram_non_blocking.cc index 460eb8320..3becb513d 100644 --- a/test/syscalls/linux/socket_unix_dgram_non_blocking.cc +++ b/test/syscalls/linux/socket_unix_dgram_non_blocking.cc @@ -31,7 +31,7 @@ using NonBlockingDgramUnixSocketPairTest = SocketPairTest; TEST_P(NonBlockingDgramUnixSocketPairTest, ReadOneSideClosed) { if (IsRunningOnGvisor()) { - // FIXME: gVisor datagram sockets return 0 instead of + // FIXME(b/70803293): gVisor datagram sockets return 0 instead of // EAGAIN. return; } diff --git a/test/syscalls/linux/socket_unix_non_stream.cc b/test/syscalls/linux/socket_unix_non_stream.cc index 8e0cbee4c..a565978f9 100644 --- a/test/syscalls/linux/socket_unix_non_stream.cc +++ b/test/syscalls/linux/socket_unix_non_stream.cc @@ -47,7 +47,7 @@ TEST_P(UnixNonStreamSocketPairTest, RecvMsgTooLarge) { const int ret = RetryEINTR(write)(sockets->second_fd(), write_buf.data(), write_buf.size()); if (ret < 0 && errno == ENOBUFS) { - // NOTE: Linux may stall the write for a long time and + // NOTE(b/116636318): Linux may stall the write for a long time and // ultimately return ENOBUFS. Allow this error, since a retry will likely // result in the same error. return; @@ -136,7 +136,7 @@ TEST_P(UnixNonStreamSocketPairTest, FragmentedSendMsg) { // N.B. At minimum, the socketpair gofer should provide a socket that is // already the correct size. // - // TODO: When internal UDS support SO_SNDBUF, we can assert that + // TODO(b/35921550): When internal UDS support SO_SNDBUF, we can assert that // we always get the right SO_SNDBUF on gVisor. GTEST_SKIP() << "SO_SNDBUF = " << actual_sndbuf << ", want " << sndbuf; } @@ -156,7 +156,7 @@ TEST_P(UnixNonStreamSocketPairTest, FragmentedSendMsg) { msg.msg_iov = &iov; msg.msg_iovlen = 1; - // NOTE: Linux has poor behavior in the presence of + // NOTE(b/116636318,b/115833655): Linux has poor behavior in the presence of // physical memory fragmentation. As a result, this may stall for a long time // and ultimately return ENOBUFS. Allow this error, since it means that we // made it to the host kernel and started the sendmsg. @@ -192,7 +192,7 @@ TEST_P(UnixNonStreamSocketPairTest, FragmentedRecvMsg) { // N.B. At minimum, the socketpair gofer should provide a socket that is // already the correct size. // - // TODO: When internal UDS support SO_SNDBUF, we can assert that + // TODO(b/35921550): When internal UDS support SO_SNDBUF, we can assert that // we always get the right SO_SNDBUF on gVisor. GTEST_SKIP() << "SO_SNDBUF = " << actual_sndbuf << ", want " << sndbuf; } @@ -201,7 +201,7 @@ TEST_P(UnixNonStreamSocketPairTest, FragmentedRecvMsg) { const int ret = RetryEINTR(write)(sockets->first_fd(), write_buf.data(), write_buf.size()); if (ret < 0 && errno == ENOBUFS) { - // NOTE: Linux may stall the write for a long time and + // NOTE(b/116636318): Linux may stall the write for a long time and // ultimately return ENOBUFS. Allow this error, since a retry will likely // result in the same error. return; diff --git a/test/syscalls/linux/socket_unix_unbound_seqpacket.cc b/test/syscalls/linux/socket_unix_unbound_seqpacket.cc index 270d7203f..21209b244 100644 --- a/test/syscalls/linux/socket_unix_unbound_seqpacket.cc +++ b/test/syscalls/linux/socket_unix_unbound_seqpacket.cc @@ -42,7 +42,7 @@ TEST_P(UnboundUnixSeqpacketSocketPairTest, SendtoWithoutConnect) { } TEST_P(UnboundUnixSeqpacketSocketPairTest, SendtoWithoutConnectIgnoresAddr) { - // FIXME: gVisor tries to find /foo/bar and thus returns ENOENT. + // FIXME(b/68223466): gVisor tries to find /foo/bar and thus returns ENOENT. if (IsRunningOnGvisor()) { return; } diff --git a/test/syscalls/linux/socket_unix_unbound_stream.cc b/test/syscalls/linux/socket_unix_unbound_stream.cc index 4db5b4be1..b95f9569e 100644 --- a/test/syscalls/linux/socket_unix_unbound_stream.cc +++ b/test/syscalls/linux/socket_unix_unbound_stream.cc @@ -269,7 +269,7 @@ TEST_P(UnixStreamSocketPairTest, SinglePeek) { // 9f389e35674f5b086edd70ed524ca0f287259725 which changes this behavior. We // used to target 3.11 compatibility, so disable this test on newer kernels. // - // NOTE: Bring this up to Linux 4.4 compatibility. + // NOTE(b/118902768): Bring this up to Linux 4.4 compatibility. auto version = ASSERT_NO_ERRNO_AND_VALUE(GetKernelVersion()); SKIP_IF(version.major > 4 || (version.major == 4 && version.minor >= 3)); } @@ -686,7 +686,7 @@ TEST_P(UnboundUnixStreamSocketPairTest, SendtoWithoutConnect) { } TEST_P(UnboundUnixStreamSocketPairTest, SendtoWithoutConnectIgnoresAddr) { - // FIXME: gVisor tries to find /foo/bar and thus returns ENOENT. + // FIXME(b/68223466): gVisor tries to find /foo/bar and thus returns ENOENT. if (IsRunningOnGvisor()) { return; } diff --git a/test/syscalls/linux/stat.cc b/test/syscalls/linux/stat.cc index 48a2059de..746318d09 100644 --- a/test/syscalls/linux/stat.cc +++ b/test/syscalls/linux/stat.cc @@ -416,7 +416,7 @@ TEST_F(StatTest, ZeroLinksOpenFdRegularFileChild_NoRandomSave) { EXPECT_EQ(st_child_before.st_gid, st_child_fd.st_gid); EXPECT_EQ(st_child_before.st_size, st_child_fd.st_size); - // TODO: This isn't ideal but since fstatfs(2) will always return + // TODO(b/34861058): This isn't ideal but since fstatfs(2) will always return // OVERLAYFS_SUPER_MAGIC we have no way to know if this fs is backed by a // gofer which doesn't support links. EXPECT_TRUE(st_child_fd.st_nlink == 0 || st_child_fd.st_nlink == 1); diff --git a/test/syscalls/linux/stat_times.cc b/test/syscalls/linux/stat_times.cc index 442957c65..8346e9a8e 100644 --- a/test/syscalls/linux/stat_times.cc +++ b/test/syscalls/linux/stat_times.cc @@ -68,7 +68,7 @@ TEST_F(StatTimesTest, FileCreationTimes) { TEST_F(StatTimesTest, FileCtimeChanges) { auto file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); - MaybeSave(); // FIXME: ctime is inconsistent. + MaybeSave(); // FIXME(b/69865927): ctime is inconsistent. absl::Time atime, mtime, ctime; std::tie(atime, mtime, ctime) = GetTime(file); @@ -150,7 +150,7 @@ TEST_F(StatTimesTest, FileAtimeChanges) { const auto file = ASSERT_NO_ERRNO_AND_VALUE( TempPath::CreateFileWith(GetAbsoluteTestTmpdir(), contents, 0666)); - MaybeSave(); // FIXME: ctime is inconsistent. + MaybeSave(); // FIXME(b/69865927): ctime is inconsistent. absl::Time atime, mtime, ctime; std::tie(atime, mtime, ctime) = GetTime(file); @@ -184,7 +184,7 @@ TEST_F(StatTimesTest, DirAtimeChanges) { const auto file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFileIn(dir.path())); - MaybeSave(); // FIXME: ctime is inconsistent. + MaybeSave(); // FIXME(b/69865927): ctime is inconsistent. absl::Time atime, mtime, ctime; std::tie(atime, mtime, ctime) = GetTime(dir); @@ -193,7 +193,7 @@ TEST_F(StatTimesTest, DirAtimeChanges) { const absl::Time before = absl::Now() - absl::Seconds(1); - // NOTE: Keep an fd open. This ensures that the inode backing the + // NOTE(b/37756234): Keep an fd open. This ensures that the inode backing the // directory won't be destroyed before the final GetTime to avoid writing out // timestamps and causing side effects. const auto fd = ASSERT_NO_ERRNO_AND_VALUE(Open(dir.path(), O_RDONLY, 0)); diff --git a/test/syscalls/linux/tcp_socket.cc b/test/syscalls/linux/tcp_socket.cc index 1057f5892..33620a874 100644 --- a/test/syscalls/linux/tcp_socket.cc +++ b/test/syscalls/linux/tcp_socket.cc @@ -191,7 +191,7 @@ TEST_P(TcpSocketTest, SenderAddressIgnoredOnPeek) { TEST_P(TcpSocketTest, SendtoAddressIgnored) { struct sockaddr_storage addr; memset(&addr, 0, sizeof(addr)); - addr.ss_family = GetParam(); // FIXME + addr.ss_family = GetParam(); // FIXME(b/63803955) char data = '\0'; EXPECT_THAT( diff --git a/test/syscalls/linux/tkill.cc b/test/syscalls/linux/tkill.cc index 9842ccc9b..3e8ce5327 100644 --- a/test/syscalls/linux/tkill.cc +++ b/test/syscalls/linux/tkill.cc @@ -32,7 +32,7 @@ namespace { static int tkill(pid_t tid, int sig) { int ret; do { - // NOTE: tkill(2) could return EAGAIN for RT signals. + // NOTE(b/25434735): tkill(2) could return EAGAIN for RT signals. ret = syscall(SYS_tkill, tid, sig); } while (ret == -1 && errno == EAGAIN); return ret; diff --git a/test/syscalls/linux/udp_bind.cc b/test/syscalls/linux/udp_bind.cc index 902be47d3..547eb2a6c 100644 --- a/test/syscalls/linux/udp_bind.cc +++ b/test/syscalls/linux/udp_bind.cc @@ -286,7 +286,7 @@ INSTANTIATE_TEST_SUITE_P( []() { SendtoTestParam param = {}; param.description = "connected IPv6 sendto IPv4 mapped IPv6"; - // TODO: Determine if this inconsistent behavior is worth + // TODO(igudger): Determine if this inconsistent behavior is worth // implementing. param.skip_on_gvisor = true; param.send_domain = AF_INET6; @@ -299,7 +299,7 @@ INSTANTIATE_TEST_SUITE_P( []() { SendtoTestParam param = {}; param.description = "connected IPv6 sendto IPv4"; - // TODO: Determine if this inconsistent behavior is worth + // TODO(igudger): Determine if this inconsistent behavior is worth // implementing. param.skip_on_gvisor = true; param.send_domain = AF_INET6; diff --git a/test/syscalls/linux/uidgid.cc b/test/syscalls/linux/uidgid.cc index c0c1f2960..d78a09b1e 100644 --- a/test/syscalls/linux/uidgid.cc +++ b/test/syscalls/linux/uidgid.cc @@ -169,7 +169,7 @@ TEST(UidGidRootTest, SetgidNotFromThreadGroupLeader) { SKIP_IF(!ASSERT_NO_ERRNO_AND_VALUE(IsRoot())); const gid_t gid = FLAGS_scratch_gid1; - // NOTE: Do setgid in a separate thread so that we can test if + // NOTE(b/64676707): Do setgid in a separate thread so that we can test if // info.si_pid is set correctly. ScopedThread([gid] { ASSERT_THAT(setgid(gid), SyscallSucceeds()); }); EXPECT_NO_ERRNO(CheckGIDs(gid, gid, gid)); diff --git a/test/syscalls/linux/utimes.cc b/test/syscalls/linux/utimes.cc index d95ee74ec..bf776cd93 100644 --- a/test/syscalls/linux/utimes.cc +++ b/test/syscalls/linux/utimes.cc @@ -33,7 +33,7 @@ namespace testing { namespace { -// TODO: utimes(nullptr) does not pick the "now" time in the +// TODO(b/36516566): utimes(nullptr) does not pick the "now" time in the // application's time domain, so when asserting that times are within a window, // we expand the window to allow for differences between the time domains. constexpr absl::Duration kClockSlack = absl::Milliseconds(100); @@ -235,7 +235,7 @@ void TestUtimensat(int dirFd, std::string const& path) { EXPECT_LE(mtime3, after); if (!IsRunningOnGvisor()) { - // FIXME: Gofers set atime and mtime to different "now" times. + // FIXME(b/36516566): Gofers set atime and mtime to different "now" times. EXPECT_EQ(atime3, mtime3); } } diff --git a/test/syscalls/linux/wait.cc b/test/syscalls/linux/wait.cc index cfab8a976..fcd606bec 100644 --- a/test/syscalls/linux/wait.cc +++ b/test/syscalls/linux/wait.cc @@ -40,7 +40,7 @@ using ::testing::UnorderedElementsAre; // These unit tests focus on the wait4(2) system call, but include a basic // checks for the i386 waitpid(2) syscall, which is a subset of wait4(2). // -// NOTE: Some functionality is not tested as +// NOTE(b/22640830,b/27680907,b/29049891): Some functionality is not tested as // it is not currently supported by gVisor: // * UID in waitid(2) siginfo. // * Process groups. diff --git a/test/syscalls/linux/write.cc b/test/syscalls/linux/write.cc index 432bd6066..7f80b2fa8 100644 --- a/test/syscalls/linux/write.cc +++ b/test/syscalls/linux/write.cc @@ -33,7 +33,7 @@ namespace testing { namespace { // This test is currently very rudimentary. // -// TODO: +// TODO(edahlgren): // * bad buffer states (EFAULT). // * bad fds (wrong permission, wrong type of file, EBADF). // * check offset is incremented. diff --git a/third_party/gvsync/downgradable_rwmutex_unsafe.go b/third_party/gvsync/downgradable_rwmutex_unsafe.go index a63a0d084..131f0a2ba 100644 --- a/third_party/gvsync/downgradable_rwmutex_unsafe.go +++ b/third_party/gvsync/downgradable_rwmutex_unsafe.go @@ -49,7 +49,7 @@ func (rw *DowngradableRWMutex) RLock() { // RUnlock undoes a single RLock call. func (rw *DowngradableRWMutex) RUnlock() { if RaceEnabled { - // TODO: Why does this need to be ReleaseMerge instead of + // TODO(jamieliu): Why does this need to be ReleaseMerge instead of // Release? IIUC this establishes Unlock happens-before RUnlock, which // seems unnecessary. RaceReleaseMerge(unsafe.Pointer(&rw.writerSem)) diff --git a/vdso/cycle_clock.h b/vdso/cycle_clock.h index 26d6690c0..309e07a3f 100644 --- a/vdso/cycle_clock.h +++ b/vdso/cycle_clock.h @@ -23,7 +23,7 @@ namespace vdso { #if __x86_64__ -// TODO: The appropriate barrier instruction to use with rdtsc on +// TODO(b/74613497): The appropriate barrier instruction to use with rdtsc on // x86_64 depends on the vendor. Intel processors can use lfence but AMD may // need mfence, depending on MSR_F10H_DECFG_LFENCE_SERIALIZE_BIT. diff --git a/vdso/vdso_amd64.lds b/vdso/vdso_amd64.lds index 166779931..e2615ae9e 100644 --- a/vdso/vdso_amd64.lds +++ b/vdso/vdso_amd64.lds @@ -56,7 +56,7 @@ SECTIONS { .altinstr_replacement : { *(.altinstr_replacement) } /* - * TODO: Remove this alignment? Then the VDSO would fit + * TODO(gvisor.dev/issue/157): Remove this alignment? Then the VDSO would fit * in a single page. */ . = ALIGN(0x1000); diff --git a/vdso/vdso_arm64.lds b/vdso/vdso_arm64.lds index 19f8efa01..469185468 100644 --- a/vdso/vdso_arm64.lds +++ b/vdso/vdso_arm64.lds @@ -59,7 +59,7 @@ SECTIONS { .altinstr_replacement : { *(.altinstr_replacement) } /* - * TODO: Remove this alignment? Then the VDSO would fit + * TODO(gvisor.dev/issue/157): Remove this alignment? Then the VDSO would fit * in a single page. */ . = ALIGN(0x1000); -- cgit v1.2.3 From 4d52a5520101a88424fb63dd99412a1db33fbd06 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Mon, 29 Apr 2019 14:25:05 -0700 Subject: Change copyright notice to "The gVisor Authors" Based on the guidelines at https://opensource.google.com/docs/releasing/authors/. 1. $ rg -l "Google LLC" | xargs sed -i 's/Google LLC.*/The gVisor Authors./' 2. Manual fixup of "Google Inc" references. 3. Add AUTHORS file. Authors may request to be added to this file. 4. Point netstack AUTHORS to gVisor AUTHORS. Drop CONTRIBUTORS. Fixes #209 PiperOrigin-RevId: 245823212 Change-Id: I64530b24ad021a7d683137459cafc510f5ee1de9 --- AUTHORS | 8 ++++++++ kokoro/run_build.sh | 2 +- kokoro/run_tests.sh | 2 +- pkg/abi/abi.go | 2 +- pkg/abi/abi_linux.go | 2 +- pkg/abi/flag.go | 2 +- pkg/abi/linux/aio.go | 2 +- pkg/abi/linux/ashmem.go | 2 +- pkg/abi/linux/audit.go | 2 +- pkg/abi/linux/binder.go | 2 +- pkg/abi/linux/bpf.go | 2 +- pkg/abi/linux/capability.go | 2 +- pkg/abi/linux/dev.go | 2 +- pkg/abi/linux/elf.go | 2 +- pkg/abi/linux/errors.go | 2 +- pkg/abi/linux/eventfd.go | 2 +- pkg/abi/linux/exec.go | 2 +- pkg/abi/linux/fcntl.go | 2 +- pkg/abi/linux/file.go | 2 +- pkg/abi/linux/fs.go | 2 +- pkg/abi/linux/futex.go | 2 +- pkg/abi/linux/inotify.go | 2 +- pkg/abi/linux/ioctl.go | 2 +- pkg/abi/linux/ip.go | 2 +- pkg/abi/linux/ipc.go | 2 +- pkg/abi/linux/limits.go | 2 +- pkg/abi/linux/linux.go | 2 +- pkg/abi/linux/mm.go | 2 +- pkg/abi/linux/netdevice.go | 2 +- pkg/abi/linux/netlink.go | 2 +- pkg/abi/linux/netlink_route.go | 2 +- pkg/abi/linux/poll.go | 2 +- pkg/abi/linux/prctl.go | 2 +- pkg/abi/linux/ptrace.go | 2 +- pkg/abi/linux/rusage.go | 2 +- pkg/abi/linux/sched.go | 2 +- pkg/abi/linux/seccomp.go | 2 +- pkg/abi/linux/sem.go | 2 +- pkg/abi/linux/shm.go | 2 +- pkg/abi/linux/signal.go | 2 +- pkg/abi/linux/socket.go | 2 +- pkg/abi/linux/tcp.go | 2 +- pkg/abi/linux/time.go | 2 +- pkg/abi/linux/timer.go | 2 +- pkg/abi/linux/tty.go | 2 +- pkg/abi/linux/uio.go | 2 +- pkg/abi/linux/utsname.go | 2 +- pkg/amutex/amutex.go | 2 +- pkg/amutex/amutex_test.go | 2 +- pkg/atomicbitops/atomic_bitops.go | 2 +- pkg/atomicbitops/atomic_bitops_amd64.s | 2 +- pkg/atomicbitops/atomic_bitops_common.go | 2 +- pkg/atomicbitops/atomic_bitops_test.go | 2 +- pkg/binary/binary.go | 2 +- pkg/binary/binary_test.go | 2 +- pkg/bits/bits.go | 2 +- pkg/bits/bits_template.go | 2 +- pkg/bits/uint64_arch_amd64.go | 2 +- pkg/bits/uint64_arch_amd64_asm.s | 2 +- pkg/bits/uint64_arch_generic.go | 2 +- pkg/bits/uint64_test.go | 2 +- pkg/bpf/bpf.go | 2 +- pkg/bpf/decoder.go | 2 +- pkg/bpf/decoder_test.go | 2 +- pkg/bpf/input_bytes.go | 2 +- pkg/bpf/interpreter.go | 2 +- pkg/bpf/interpreter_test.go | 2 +- pkg/bpf/program_builder.go | 2 +- pkg/bpf/program_builder_test.go | 2 +- pkg/compressio/compressio.go | 2 +- pkg/compressio/compressio_test.go | 2 +- pkg/control/client/client.go | 2 +- pkg/control/server/server.go | 2 +- pkg/cpuid/cpu_amd64.s | 2 +- pkg/cpuid/cpuid.go | 2 +- pkg/cpuid/cpuid_parse_test.go | 2 +- pkg/cpuid/cpuid_test.go | 2 +- pkg/dhcp/client.go | 2 +- pkg/dhcp/dhcp.go | 2 +- pkg/dhcp/dhcp_string.go | 2 +- pkg/dhcp/dhcp_test.go | 2 +- pkg/dhcp/server.go | 2 +- pkg/eventchannel/event.go | 2 +- pkg/eventchannel/event.proto | 2 +- pkg/fd/fd.go | 2 +- pkg/fd/fd_test.go | 2 +- pkg/fdnotifier/fdnotifier.go | 2 +- pkg/fdnotifier/poll_unsafe.go | 2 +- pkg/gate/gate.go | 2 +- pkg/gate/gate_test.go | 2 +- pkg/ilist/list.go | 2 +- pkg/ilist/list_test.go | 2 +- pkg/linewriter/linewriter.go | 2 +- pkg/linewriter/linewriter_test.go | 2 +- pkg/log/glog.go | 2 +- pkg/log/glog_unsafe.go | 2 +- pkg/log/json.go | 2 +- pkg/log/json_k8s.go | 2 +- pkg/log/json_test.go | 2 +- pkg/log/log.go | 2 +- pkg/log/log_test.go | 2 +- pkg/metric/metric.go | 2 +- pkg/metric/metric.proto | 2 +- pkg/metric/metric_test.go | 2 +- pkg/p9/buffer.go | 2 +- pkg/p9/buffer_test.go | 2 +- pkg/p9/client.go | 2 +- pkg/p9/client_file.go | 2 +- pkg/p9/client_test.go | 2 +- pkg/p9/file.go | 2 +- pkg/p9/handlers.go | 2 +- pkg/p9/local_server/local_server.go | 2 +- pkg/p9/messages.go | 2 +- pkg/p9/messages_test.go | 2 +- pkg/p9/p9.go | 2 +- pkg/p9/p9_test.go | 2 +- pkg/p9/p9test/client_test.go | 2 +- pkg/p9/p9test/p9test.go | 2 +- pkg/p9/path_tree.go | 2 +- pkg/p9/pool.go | 2 +- pkg/p9/pool_test.go | 2 +- pkg/p9/server.go | 2 +- pkg/p9/transport.go | 2 +- pkg/p9/transport_test.go | 2 +- pkg/p9/version.go | 2 +- pkg/p9/version_test.go | 2 +- pkg/rand/rand.go | 2 +- pkg/rand/rand_linux.go | 2 +- pkg/refs/refcounter.go | 2 +- pkg/refs/refcounter_state.go | 2 +- pkg/refs/refcounter_test.go | 2 +- pkg/seccomp/seccomp.go | 2 +- pkg/seccomp/seccomp_rules.go | 2 +- pkg/seccomp/seccomp_test.go | 2 +- pkg/seccomp/seccomp_test_victim.go | 2 +- pkg/seccomp/seccomp_unsafe.go | 2 +- pkg/secio/full_reader.go | 2 +- pkg/secio/secio.go | 2 +- pkg/secio/secio_test.go | 2 +- pkg/segment/range.go | 2 +- pkg/segment/set.go | 2 +- pkg/segment/set_state.go | 2 +- pkg/segment/test/segment_test.go | 2 +- pkg/segment/test/set_functions.go | 2 +- pkg/sentry/arch/aligned.go | 2 +- pkg/sentry/arch/arch.go | 2 +- pkg/sentry/arch/arch_amd64.go | 2 +- pkg/sentry/arch/arch_amd64.s | 2 +- pkg/sentry/arch/arch_state_x86.go | 2 +- pkg/sentry/arch/arch_x86.go | 2 +- pkg/sentry/arch/auxv.go | 2 +- pkg/sentry/arch/registers.proto | 2 +- pkg/sentry/arch/signal_act.go | 2 +- pkg/sentry/arch/signal_amd64.go | 2 +- pkg/sentry/arch/signal_info.go | 2 +- pkg/sentry/arch/signal_stack.go | 2 +- pkg/sentry/arch/stack.go | 2 +- pkg/sentry/arch/syscalls_amd64.go | 2 +- pkg/sentry/context/context.go | 2 +- pkg/sentry/context/contexttest/contexttest.go | 2 +- pkg/sentry/control/control.go | 2 +- pkg/sentry/control/pprof.go | 2 +- pkg/sentry/control/proc.go | 2 +- pkg/sentry/control/proc_test.go | 2 +- pkg/sentry/control/state.go | 2 +- pkg/sentry/device/device.go | 2 +- pkg/sentry/device/device_test.go | 2 +- pkg/sentry/fs/anon/anon.go | 2 +- pkg/sentry/fs/anon/device.go | 2 +- pkg/sentry/fs/ashmem/area.go | 2 +- pkg/sentry/fs/ashmem/device.go | 2 +- pkg/sentry/fs/ashmem/pin_board.go | 2 +- pkg/sentry/fs/ashmem/pin_board_test.go | 2 +- pkg/sentry/fs/attr.go | 2 +- pkg/sentry/fs/binder/binder.go | 2 +- pkg/sentry/fs/context.go | 2 +- pkg/sentry/fs/copy_up.go | 2 +- pkg/sentry/fs/copy_up_test.go | 2 +- pkg/sentry/fs/dentry.go | 2 +- pkg/sentry/fs/dev/dev.go | 2 +- pkg/sentry/fs/dev/device.go | 2 +- pkg/sentry/fs/dev/fs.go | 2 +- pkg/sentry/fs/dev/full.go | 2 +- pkg/sentry/fs/dev/null.go | 2 +- pkg/sentry/fs/dev/random.go | 2 +- pkg/sentry/fs/dirent.go | 2 +- pkg/sentry/fs/dirent_cache.go | 2 +- pkg/sentry/fs/dirent_cache_limiter.go | 2 +- pkg/sentry/fs/dirent_cache_test.go | 2 +- pkg/sentry/fs/dirent_refs_test.go | 2 +- pkg/sentry/fs/dirent_state.go | 2 +- pkg/sentry/fs/fdpipe/pipe.go | 2 +- pkg/sentry/fs/fdpipe/pipe_opener.go | 2 +- pkg/sentry/fs/fdpipe/pipe_opener_test.go | 2 +- pkg/sentry/fs/fdpipe/pipe_state.go | 2 +- pkg/sentry/fs/fdpipe/pipe_test.go | 2 +- pkg/sentry/fs/file.go | 2 +- pkg/sentry/fs/file_operations.go | 2 +- pkg/sentry/fs/file_overlay.go | 2 +- pkg/sentry/fs/file_overlay_test.go | 2 +- pkg/sentry/fs/file_state.go | 2 +- pkg/sentry/fs/file_test.go | 2 +- pkg/sentry/fs/filesystems.go | 2 +- pkg/sentry/fs/filetest/filetest.go | 2 +- pkg/sentry/fs/flags.go | 2 +- pkg/sentry/fs/fs.go | 2 +- pkg/sentry/fs/fsutil/dirty_set.go | 2 +- pkg/sentry/fs/fsutil/dirty_set_test.go | 2 +- pkg/sentry/fs/fsutil/file.go | 2 +- pkg/sentry/fs/fsutil/file_range_set.go | 2 +- pkg/sentry/fs/fsutil/frame_ref_set.go | 2 +- pkg/sentry/fs/fsutil/fsutil.go | 2 +- pkg/sentry/fs/fsutil/host_file_mapper.go | 2 +- pkg/sentry/fs/fsutil/host_file_mapper_state.go | 2 +- pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go | 2 +- pkg/sentry/fs/fsutil/host_mappable.go | 2 +- pkg/sentry/fs/fsutil/inode.go | 2 +- pkg/sentry/fs/fsutil/inode_cached.go | 2 +- pkg/sentry/fs/fsutil/inode_cached_test.go | 2 +- pkg/sentry/fs/gofer/attr.go | 2 +- pkg/sentry/fs/gofer/cache_policy.go | 2 +- pkg/sentry/fs/gofer/context_file.go | 2 +- pkg/sentry/fs/gofer/device.go | 2 +- pkg/sentry/fs/gofer/file.go | 2 +- pkg/sentry/fs/gofer/file_state.go | 2 +- pkg/sentry/fs/gofer/fs.go | 2 +- pkg/sentry/fs/gofer/gofer_test.go | 2 +- pkg/sentry/fs/gofer/handles.go | 2 +- pkg/sentry/fs/gofer/inode.go | 2 +- pkg/sentry/fs/gofer/inode_state.go | 2 +- pkg/sentry/fs/gofer/path.go | 2 +- pkg/sentry/fs/gofer/session.go | 2 +- pkg/sentry/fs/gofer/session_state.go | 2 +- pkg/sentry/fs/gofer/socket.go | 2 +- pkg/sentry/fs/gofer/util.go | 2 +- pkg/sentry/fs/host/control.go | 2 +- pkg/sentry/fs/host/descriptor.go | 2 +- pkg/sentry/fs/host/descriptor_state.go | 2 +- pkg/sentry/fs/host/descriptor_test.go | 2 +- pkg/sentry/fs/host/device.go | 2 +- pkg/sentry/fs/host/file.go | 2 +- pkg/sentry/fs/host/fs.go | 2 +- pkg/sentry/fs/host/fs_test.go | 2 +- pkg/sentry/fs/host/inode.go | 2 +- pkg/sentry/fs/host/inode_state.go | 2 +- pkg/sentry/fs/host/inode_test.go | 2 +- pkg/sentry/fs/host/ioctl_unsafe.go | 2 +- pkg/sentry/fs/host/socket.go | 2 +- pkg/sentry/fs/host/socket_iovec.go | 2 +- pkg/sentry/fs/host/socket_state.go | 2 +- pkg/sentry/fs/host/socket_test.go | 2 +- pkg/sentry/fs/host/socket_unsafe.go | 2 +- pkg/sentry/fs/host/tty.go | 2 +- pkg/sentry/fs/host/util.go | 2 +- pkg/sentry/fs/host/util_unsafe.go | 2 +- pkg/sentry/fs/host/wait_test.go | 2 +- pkg/sentry/fs/inode.go | 2 +- pkg/sentry/fs/inode_inotify.go | 2 +- pkg/sentry/fs/inode_operations.go | 2 +- pkg/sentry/fs/inode_overlay.go | 2 +- pkg/sentry/fs/inode_overlay_test.go | 2 +- pkg/sentry/fs/inotify.go | 2 +- pkg/sentry/fs/inotify_event.go | 2 +- pkg/sentry/fs/inotify_watch.go | 2 +- pkg/sentry/fs/lock/lock.go | 2 +- pkg/sentry/fs/lock/lock_range_test.go | 2 +- pkg/sentry/fs/lock/lock_set_functions.go | 2 +- pkg/sentry/fs/lock/lock_test.go | 2 +- pkg/sentry/fs/mock.go | 2 +- pkg/sentry/fs/mount.go | 2 +- pkg/sentry/fs/mount_overlay.go | 2 +- pkg/sentry/fs/mount_test.go | 2 +- pkg/sentry/fs/mounts.go | 2 +- pkg/sentry/fs/mounts_test.go | 2 +- pkg/sentry/fs/offset.go | 2 +- pkg/sentry/fs/overlay.go | 2 +- pkg/sentry/fs/path.go | 2 +- pkg/sentry/fs/path_test.go | 2 +- pkg/sentry/fs/proc/cpuinfo.go | 2 +- pkg/sentry/fs/proc/device/device.go | 2 +- pkg/sentry/fs/proc/exec_args.go | 2 +- pkg/sentry/fs/proc/fds.go | 2 +- pkg/sentry/fs/proc/filesystems.go | 2 +- pkg/sentry/fs/proc/fs.go | 2 +- pkg/sentry/fs/proc/inode.go | 2 +- pkg/sentry/fs/proc/loadavg.go | 2 +- pkg/sentry/fs/proc/meminfo.go | 2 +- pkg/sentry/fs/proc/mounts.go | 2 +- pkg/sentry/fs/proc/net.go | 2 +- pkg/sentry/fs/proc/net_test.go | 2 +- pkg/sentry/fs/proc/proc.go | 2 +- pkg/sentry/fs/proc/rpcinet_proc.go | 2 +- pkg/sentry/fs/proc/seqfile/seqfile.go | 2 +- pkg/sentry/fs/proc/seqfile/seqfile_test.go | 2 +- pkg/sentry/fs/proc/stat.go | 2 +- pkg/sentry/fs/proc/sys.go | 2 +- pkg/sentry/fs/proc/sys_net.go | 2 +- pkg/sentry/fs/proc/sys_net_state.go | 2 +- pkg/sentry/fs/proc/sys_net_test.go | 2 +- pkg/sentry/fs/proc/task.go | 2 +- pkg/sentry/fs/proc/uid_gid_map.go | 2 +- pkg/sentry/fs/proc/uptime.go | 2 +- pkg/sentry/fs/proc/version.go | 2 +- pkg/sentry/fs/ramfs/dir.go | 2 +- pkg/sentry/fs/ramfs/socket.go | 2 +- pkg/sentry/fs/ramfs/symlink.go | 2 +- pkg/sentry/fs/ramfs/tree.go | 2 +- pkg/sentry/fs/ramfs/tree_test.go | 2 +- pkg/sentry/fs/restore.go | 2 +- pkg/sentry/fs/save.go | 2 +- pkg/sentry/fs/seek.go | 2 +- pkg/sentry/fs/sync.go | 2 +- pkg/sentry/fs/sys/device.go | 2 +- pkg/sentry/fs/sys/devices.go | 2 +- pkg/sentry/fs/sys/fs.go | 2 +- pkg/sentry/fs/sys/sys.go | 2 +- pkg/sentry/fs/timerfd/timerfd.go | 2 +- pkg/sentry/fs/tmpfs/device.go | 2 +- pkg/sentry/fs/tmpfs/file_regular.go | 2 +- pkg/sentry/fs/tmpfs/file_test.go | 2 +- pkg/sentry/fs/tmpfs/fs.go | 2 +- pkg/sentry/fs/tmpfs/inode_file.go | 2 +- pkg/sentry/fs/tmpfs/tmpfs.go | 2 +- pkg/sentry/fs/tty/dir.go | 2 +- pkg/sentry/fs/tty/fs.go | 2 +- pkg/sentry/fs/tty/line_discipline.go | 2 +- pkg/sentry/fs/tty/master.go | 2 +- pkg/sentry/fs/tty/queue.go | 2 +- pkg/sentry/fs/tty/slave.go | 2 +- pkg/sentry/fs/tty/terminal.go | 2 +- pkg/sentry/fs/tty/tty_test.go | 2 +- pkg/sentry/hostcpu/getcpu_amd64.s | 2 +- pkg/sentry/hostcpu/hostcpu.go | 2 +- pkg/sentry/hostcpu/hostcpu_test.go | 2 +- pkg/sentry/inet/context.go | 2 +- pkg/sentry/inet/inet.go | 2 +- pkg/sentry/inet/test_stack.go | 2 +- pkg/sentry/kernel/abstract_socket_namespace.go | 2 +- pkg/sentry/kernel/auth/auth.go | 2 +- pkg/sentry/kernel/auth/capability_set.go | 2 +- pkg/sentry/kernel/auth/context.go | 2 +- pkg/sentry/kernel/auth/credentials.go | 2 +- pkg/sentry/kernel/auth/id.go | 2 +- pkg/sentry/kernel/auth/id_map.go | 2 +- pkg/sentry/kernel/auth/id_map_functions.go | 2 +- pkg/sentry/kernel/auth/user_namespace.go | 2 +- pkg/sentry/kernel/context.go | 2 +- pkg/sentry/kernel/contexttest/contexttest.go | 2 +- pkg/sentry/kernel/epoll/epoll.go | 2 +- pkg/sentry/kernel/epoll/epoll_state.go | 2 +- pkg/sentry/kernel/epoll/epoll_test.go | 2 +- pkg/sentry/kernel/eventfd/eventfd.go | 2 +- pkg/sentry/kernel/eventfd/eventfd_test.go | 2 +- pkg/sentry/kernel/fasync/fasync.go | 2 +- pkg/sentry/kernel/fd_map.go | 2 +- pkg/sentry/kernel/fd_map_test.go | 2 +- pkg/sentry/kernel/fs_context.go | 2 +- pkg/sentry/kernel/futex/futex.go | 2 +- pkg/sentry/kernel/futex/futex_test.go | 2 +- pkg/sentry/kernel/ipc_namespace.go | 2 +- pkg/sentry/kernel/kdefs/kdefs.go | 2 +- pkg/sentry/kernel/kernel.go | 2 +- pkg/sentry/kernel/kernel_state.go | 2 +- pkg/sentry/kernel/memevent/memory_events.go | 2 +- pkg/sentry/kernel/memevent/memory_events.proto | 2 +- pkg/sentry/kernel/pending_signals.go | 2 +- pkg/sentry/kernel/pending_signals_state.go | 2 +- pkg/sentry/kernel/pipe/buffers.go | 2 +- pkg/sentry/kernel/pipe/device.go | 2 +- pkg/sentry/kernel/pipe/node.go | 2 +- pkg/sentry/kernel/pipe/node_test.go | 2 +- pkg/sentry/kernel/pipe/pipe.go | 2 +- pkg/sentry/kernel/pipe/pipe_test.go | 2 +- pkg/sentry/kernel/pipe/reader.go | 2 +- pkg/sentry/kernel/pipe/reader_writer.go | 2 +- pkg/sentry/kernel/pipe/writer.go | 2 +- pkg/sentry/kernel/posixtimer.go | 2 +- pkg/sentry/kernel/ptrace.go | 2 +- pkg/sentry/kernel/ptrace_amd64.go | 2 +- pkg/sentry/kernel/ptrace_arm64.go | 2 +- pkg/sentry/kernel/rseq.go | 2 +- pkg/sentry/kernel/sched/cpuset.go | 2 +- pkg/sentry/kernel/sched/cpuset_test.go | 2 +- pkg/sentry/kernel/sched/sched.go | 2 +- pkg/sentry/kernel/seccomp.go | 2 +- pkg/sentry/kernel/semaphore/semaphore.go | 2 +- pkg/sentry/kernel/semaphore/semaphore_test.go | 2 +- pkg/sentry/kernel/sessions.go | 2 +- pkg/sentry/kernel/shm/device.go | 2 +- pkg/sentry/kernel/shm/shm.go | 2 +- pkg/sentry/kernel/signal.go | 2 +- pkg/sentry/kernel/signal_handlers.go | 2 +- pkg/sentry/kernel/syscalls.go | 2 +- pkg/sentry/kernel/syscalls_state.go | 2 +- pkg/sentry/kernel/syslog.go | 2 +- pkg/sentry/kernel/table_test.go | 2 +- pkg/sentry/kernel/task.go | 2 +- pkg/sentry/kernel/task_acct.go | 2 +- pkg/sentry/kernel/task_block.go | 2 +- pkg/sentry/kernel/task_clone.go | 2 +- pkg/sentry/kernel/task_context.go | 2 +- pkg/sentry/kernel/task_exec.go | 2 +- pkg/sentry/kernel/task_exit.go | 2 +- pkg/sentry/kernel/task_futex.go | 2 +- pkg/sentry/kernel/task_identity.go | 2 +- pkg/sentry/kernel/task_log.go | 2 +- pkg/sentry/kernel/task_net.go | 2 +- pkg/sentry/kernel/task_run.go | 2 +- pkg/sentry/kernel/task_sched.go | 2 +- pkg/sentry/kernel/task_signals.go | 2 +- pkg/sentry/kernel/task_start.go | 2 +- pkg/sentry/kernel/task_stop.go | 2 +- pkg/sentry/kernel/task_syscall.go | 2 +- pkg/sentry/kernel/task_test.go | 2 +- pkg/sentry/kernel/task_usermem.go | 2 +- pkg/sentry/kernel/thread_group.go | 2 +- pkg/sentry/kernel/threads.go | 2 +- pkg/sentry/kernel/time/context.go | 2 +- pkg/sentry/kernel/time/time.go | 2 +- pkg/sentry/kernel/timekeeper.go | 2 +- pkg/sentry/kernel/timekeeper_state.go | 2 +- pkg/sentry/kernel/timekeeper_test.go | 2 +- pkg/sentry/kernel/uncaught_signal.proto | 2 +- pkg/sentry/kernel/uts_namespace.go | 2 +- pkg/sentry/kernel/vdso.go | 2 +- pkg/sentry/kernel/version.go | 2 +- pkg/sentry/limits/context.go | 2 +- pkg/sentry/limits/limits.go | 2 +- pkg/sentry/limits/limits_test.go | 2 +- pkg/sentry/limits/linux.go | 2 +- pkg/sentry/loader/elf.go | 2 +- pkg/sentry/loader/interpreter.go | 2 +- pkg/sentry/loader/loader.go | 2 +- pkg/sentry/loader/vdso.go | 2 +- pkg/sentry/loader/vdso_state.go | 2 +- pkg/sentry/memmap/mapping_set.go | 2 +- pkg/sentry/memmap/mapping_set_test.go | 2 +- pkg/sentry/memmap/memmap.go | 2 +- pkg/sentry/memutil/memutil.go | 2 +- pkg/sentry/memutil/memutil_unsafe.go | 2 +- pkg/sentry/mm/address_space.go | 2 +- pkg/sentry/mm/aio_context.go | 2 +- pkg/sentry/mm/aio_context_state.go | 2 +- pkg/sentry/mm/debug.go | 2 +- pkg/sentry/mm/io.go | 2 +- pkg/sentry/mm/lifecycle.go | 2 +- pkg/sentry/mm/metadata.go | 2 +- pkg/sentry/mm/mm.go | 2 +- pkg/sentry/mm/mm_test.go | 2 +- pkg/sentry/mm/pma.go | 2 +- pkg/sentry/mm/procfs.go | 2 +- pkg/sentry/mm/save_restore.go | 2 +- pkg/sentry/mm/shm.go | 2 +- pkg/sentry/mm/special_mappable.go | 2 +- pkg/sentry/mm/syscalls.go | 2 +- pkg/sentry/mm/vma.go | 2 +- pkg/sentry/pgalloc/context.go | 2 +- pkg/sentry/pgalloc/pgalloc.go | 2 +- pkg/sentry/pgalloc/pgalloc_test.go | 2 +- pkg/sentry/pgalloc/pgalloc_unsafe.go | 2 +- pkg/sentry/pgalloc/save_restore.go | 2 +- pkg/sentry/platform/context.go | 2 +- pkg/sentry/platform/interrupt/interrupt.go | 2 +- pkg/sentry/platform/interrupt/interrupt_test.go | 2 +- pkg/sentry/platform/kvm/address_space.go | 2 +- pkg/sentry/platform/kvm/allocator.go | 2 +- pkg/sentry/platform/kvm/bluepill.go | 2 +- pkg/sentry/platform/kvm/bluepill_amd64.go | 2 +- pkg/sentry/platform/kvm/bluepill_amd64.s | 2 +- pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go | 2 +- pkg/sentry/platform/kvm/bluepill_fault.go | 2 +- pkg/sentry/platform/kvm/bluepill_unsafe.go | 2 +- pkg/sentry/platform/kvm/context.go | 2 +- pkg/sentry/platform/kvm/kvm.go | 2 +- pkg/sentry/platform/kvm/kvm_amd64.go | 2 +- pkg/sentry/platform/kvm/kvm_amd64_unsafe.go | 2 +- pkg/sentry/platform/kvm/kvm_const.go | 2 +- pkg/sentry/platform/kvm/kvm_test.go | 2 +- pkg/sentry/platform/kvm/machine.go | 2 +- pkg/sentry/platform/kvm/machine_amd64.go | 2 +- pkg/sentry/platform/kvm/machine_amd64_unsafe.go | 2 +- pkg/sentry/platform/kvm/machine_unsafe.go | 2 +- pkg/sentry/platform/kvm/physical_map.go | 2 +- pkg/sentry/platform/kvm/testutil/testutil.go | 2 +- pkg/sentry/platform/kvm/testutil/testutil_amd64.go | 2 +- pkg/sentry/platform/kvm/testutil/testutil_amd64.s | 2 +- pkg/sentry/platform/kvm/virtual_map.go | 2 +- pkg/sentry/platform/kvm/virtual_map_test.go | 2 +- pkg/sentry/platform/mmap_min_addr.go | 2 +- pkg/sentry/platform/platform.go | 2 +- pkg/sentry/platform/procid/procid.go | 2 +- pkg/sentry/platform/procid/procid_amd64.s | 2 +- pkg/sentry/platform/procid/procid_arm64.s | 2 +- pkg/sentry/platform/procid/procid_net_test.go | 2 +- pkg/sentry/platform/procid/procid_test.go | 2 +- pkg/sentry/platform/ptrace/ptrace.go | 2 +- pkg/sentry/platform/ptrace/ptrace_unsafe.go | 2 +- pkg/sentry/platform/ptrace/stub_amd64.s | 2 +- pkg/sentry/platform/ptrace/stub_unsafe.go | 2 +- pkg/sentry/platform/ptrace/subprocess.go | 2 +- pkg/sentry/platform/ptrace/subprocess_amd64.go | 2 +- pkg/sentry/platform/ptrace/subprocess_linux.go | 2 +- pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go | 2 +- pkg/sentry/platform/ptrace/subprocess_unsafe.go | 2 +- pkg/sentry/platform/ring0/defs.go | 2 +- pkg/sentry/platform/ring0/defs_amd64.go | 2 +- pkg/sentry/platform/ring0/entry_amd64.go | 2 +- pkg/sentry/platform/ring0/entry_amd64.s | 2 +- pkg/sentry/platform/ring0/gen_offsets/main.go | 2 +- pkg/sentry/platform/ring0/kernel.go | 2 +- pkg/sentry/platform/ring0/kernel_amd64.go | 2 +- pkg/sentry/platform/ring0/kernel_unsafe.go | 2 +- pkg/sentry/platform/ring0/lib_amd64.go | 2 +- pkg/sentry/platform/ring0/lib_amd64.s | 2 +- pkg/sentry/platform/ring0/offsets_amd64.go | 2 +- pkg/sentry/platform/ring0/pagetables/allocator.go | 2 +- pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables_test.go | 2 +- pkg/sentry/platform/ring0/pagetables/pagetables_x86.go | 2 +- pkg/sentry/platform/ring0/pagetables/pcids_x86.go | 2 +- pkg/sentry/platform/ring0/pagetables/walker_amd64.go | 2 +- pkg/sentry/platform/ring0/ring0.go | 2 +- pkg/sentry/platform/ring0/x86.go | 2 +- pkg/sentry/platform/safecopy/atomic_amd64.s | 2 +- pkg/sentry/platform/safecopy/safecopy.go | 2 +- pkg/sentry/platform/safecopy/safecopy_test.go | 2 +- pkg/sentry/platform/safecopy/safecopy_unsafe.go | 2 +- pkg/sentry/platform/safecopy/sighandler_amd64.s | 2 +- pkg/sentry/platform/safecopy/sighandler_arm64.s | 2 +- pkg/sentry/safemem/block_unsafe.go | 2 +- pkg/sentry/safemem/io.go | 2 +- pkg/sentry/safemem/io_test.go | 2 +- pkg/sentry/safemem/safemem.go | 2 +- pkg/sentry/safemem/seq_test.go | 2 +- pkg/sentry/safemem/seq_unsafe.go | 2 +- pkg/sentry/sighandling/sighandling.go | 2 +- pkg/sentry/sighandling/sighandling_unsafe.go | 2 +- pkg/sentry/socket/control/control.go | 2 +- pkg/sentry/socket/epsocket/device.go | 2 +- pkg/sentry/socket/epsocket/epsocket.go | 2 +- pkg/sentry/socket/epsocket/provider.go | 2 +- pkg/sentry/socket/epsocket/save_restore.go | 2 +- pkg/sentry/socket/epsocket/stack.go | 2 +- pkg/sentry/socket/hostinet/device.go | 2 +- pkg/sentry/socket/hostinet/hostinet.go | 2 +- pkg/sentry/socket/hostinet/save_restore.go | 2 +- pkg/sentry/socket/hostinet/socket.go | 2 +- pkg/sentry/socket/hostinet/socket_unsafe.go | 2 +- pkg/sentry/socket/hostinet/stack.go | 2 +- pkg/sentry/socket/netlink/message.go | 2 +- pkg/sentry/socket/netlink/port/port.go | 2 +- pkg/sentry/socket/netlink/port/port_test.go | 2 +- pkg/sentry/socket/netlink/provider.go | 2 +- pkg/sentry/socket/netlink/route/protocol.go | 2 +- pkg/sentry/socket/netlink/socket.go | 2 +- pkg/sentry/socket/rpcinet/conn/conn.go | 2 +- pkg/sentry/socket/rpcinet/device.go | 2 +- pkg/sentry/socket/rpcinet/notifier/notifier.go | 2 +- pkg/sentry/socket/rpcinet/rpcinet.go | 2 +- pkg/sentry/socket/rpcinet/socket.go | 2 +- pkg/sentry/socket/rpcinet/stack.go | 2 +- pkg/sentry/socket/rpcinet/stack_unsafe.go | 2 +- pkg/sentry/socket/socket.go | 2 +- pkg/sentry/socket/unix/device.go | 2 +- pkg/sentry/socket/unix/io.go | 2 +- pkg/sentry/socket/unix/transport/connectioned.go | 2 +- pkg/sentry/socket/unix/transport/connectioned_state.go | 2 +- pkg/sentry/socket/unix/transport/connectionless.go | 2 +- pkg/sentry/socket/unix/transport/queue.go | 2 +- pkg/sentry/socket/unix/transport/unix.go | 2 +- pkg/sentry/socket/unix/unix.go | 2 +- pkg/sentry/state/state.go | 2 +- pkg/sentry/state/state_metadata.go | 2 +- pkg/sentry/state/state_unsafe.go | 2 +- pkg/sentry/strace/capability.go | 2 +- pkg/sentry/strace/clone.go | 2 +- pkg/sentry/strace/futex.go | 2 +- pkg/sentry/strace/linux64.go | 2 +- pkg/sentry/strace/open.go | 2 +- pkg/sentry/strace/poll.go | 2 +- pkg/sentry/strace/ptrace.go | 2 +- pkg/sentry/strace/signal.go | 2 +- pkg/sentry/strace/socket.go | 2 +- pkg/sentry/strace/strace.go | 2 +- pkg/sentry/strace/strace.proto | 2 +- pkg/sentry/strace/syscalls.go | 2 +- pkg/sentry/syscalls/epoll.go | 2 +- pkg/sentry/syscalls/linux/error.go | 2 +- pkg/sentry/syscalls/linux/flags.go | 2 +- pkg/sentry/syscalls/linux/linux64.go | 2 +- pkg/sentry/syscalls/linux/sigset.go | 2 +- pkg/sentry/syscalls/linux/sys_aio.go | 2 +- pkg/sentry/syscalls/linux/sys_capability.go | 2 +- pkg/sentry/syscalls/linux/sys_epoll.go | 2 +- pkg/sentry/syscalls/linux/sys_eventfd.go | 2 +- pkg/sentry/syscalls/linux/sys_file.go | 2 +- pkg/sentry/syscalls/linux/sys_futex.go | 2 +- pkg/sentry/syscalls/linux/sys_getdents.go | 2 +- pkg/sentry/syscalls/linux/sys_identity.go | 2 +- pkg/sentry/syscalls/linux/sys_inotify.go | 2 +- pkg/sentry/syscalls/linux/sys_lseek.go | 2 +- pkg/sentry/syscalls/linux/sys_mmap.go | 2 +- pkg/sentry/syscalls/linux/sys_mount.go | 2 +- pkg/sentry/syscalls/linux/sys_pipe.go | 2 +- pkg/sentry/syscalls/linux/sys_poll.go | 2 +- pkg/sentry/syscalls/linux/sys_prctl.go | 2 +- pkg/sentry/syscalls/linux/sys_random.go | 2 +- pkg/sentry/syscalls/linux/sys_read.go | 2 +- pkg/sentry/syscalls/linux/sys_rlimit.go | 2 +- pkg/sentry/syscalls/linux/sys_rusage.go | 2 +- pkg/sentry/syscalls/linux/sys_sched.go | 2 +- pkg/sentry/syscalls/linux/sys_seccomp.go | 2 +- pkg/sentry/syscalls/linux/sys_sem.go | 2 +- pkg/sentry/syscalls/linux/sys_shm.go | 2 +- pkg/sentry/syscalls/linux/sys_signal.go | 2 +- pkg/sentry/syscalls/linux/sys_socket.go | 2 +- pkg/sentry/syscalls/linux/sys_stat.go | 2 +- pkg/sentry/syscalls/linux/sys_sync.go | 2 +- pkg/sentry/syscalls/linux/sys_sysinfo.go | 2 +- pkg/sentry/syscalls/linux/sys_syslog.go | 2 +- pkg/sentry/syscalls/linux/sys_thread.go | 2 +- pkg/sentry/syscalls/linux/sys_time.go | 2 +- pkg/sentry/syscalls/linux/sys_timer.go | 2 +- pkg/sentry/syscalls/linux/sys_timerfd.go | 2 +- pkg/sentry/syscalls/linux/sys_tls.go | 2 +- pkg/sentry/syscalls/linux/sys_utsname.go | 2 +- pkg/sentry/syscalls/linux/sys_write.go | 2 +- pkg/sentry/syscalls/linux/timespec.go | 2 +- pkg/sentry/syscalls/syscalls.go | 2 +- pkg/sentry/time/calibrated_clock.go | 2 +- pkg/sentry/time/calibrated_clock_test.go | 2 +- pkg/sentry/time/clock_id.go | 2 +- pkg/sentry/time/clocks.go | 2 +- pkg/sentry/time/muldiv_amd64.s | 2 +- pkg/sentry/time/muldiv_arm64.s | 2 +- pkg/sentry/time/parameters.go | 2 +- pkg/sentry/time/parameters_test.go | 2 +- pkg/sentry/time/sampler.go | 2 +- pkg/sentry/time/sampler_test.go | 2 +- pkg/sentry/time/sampler_unsafe.go | 2 +- pkg/sentry/time/tsc_amd64.s | 2 +- pkg/sentry/time/tsc_arm64.s | 2 +- pkg/sentry/unimpl/events.go | 2 +- pkg/sentry/unimpl/unimplemented_syscall.proto | 2 +- pkg/sentry/uniqueid/context.go | 2 +- pkg/sentry/usage/cpu.go | 2 +- pkg/sentry/usage/io.go | 2 +- pkg/sentry/usage/memory.go | 2 +- pkg/sentry/usage/memory_unsafe.go | 2 +- pkg/sentry/usage/usage.go | 2 +- pkg/sentry/usermem/access_type.go | 2 +- pkg/sentry/usermem/addr.go | 2 +- pkg/sentry/usermem/addr_range_seq_test.go | 2 +- pkg/sentry/usermem/addr_range_seq_unsafe.go | 2 +- pkg/sentry/usermem/bytes_io.go | 2 +- pkg/sentry/usermem/bytes_io_unsafe.go | 2 +- pkg/sentry/usermem/usermem.go | 2 +- pkg/sentry/usermem/usermem_arm64.go | 2 +- pkg/sentry/usermem/usermem_test.go | 2 +- pkg/sentry/usermem/usermem_unsafe.go | 2 +- pkg/sentry/usermem/usermem_x86.go | 2 +- pkg/sentry/watchdog/watchdog.go | 2 +- pkg/sleep/commit_amd64.s | 2 +- pkg/sleep/commit_asm.go | 2 +- pkg/sleep/commit_noasm.go | 2 +- pkg/sleep/empty.s | 2 +- pkg/sleep/sleep_test.go | 2 +- pkg/sleep/sleep_unsafe.go | 2 +- pkg/state/decode.go | 2 +- pkg/state/encode.go | 2 +- pkg/state/encode_unsafe.go | 2 +- pkg/state/map.go | 2 +- pkg/state/object.proto | 2 +- pkg/state/printer.go | 2 +- pkg/state/state.go | 2 +- pkg/state/state_test.go | 2 +- pkg/state/statefile/statefile.go | 2 +- pkg/state/statefile/statefile_test.go | 2 +- pkg/state/stats.go | 2 +- pkg/syserr/host_linux.go | 2 +- pkg/syserr/netstack.go | 2 +- pkg/syserr/syserr.go | 2 +- pkg/syserror/syserror.go | 2 +- pkg/syserror/syserror_test.go | 2 +- pkg/tcpip/adapters/gonet/gonet.go | 2 +- pkg/tcpip/adapters/gonet/gonet_test.go | 2 +- pkg/tcpip/buffer/prependable.go | 2 +- pkg/tcpip/buffer/view.go | 2 +- pkg/tcpip/buffer/view_test.go | 2 +- pkg/tcpip/checker/checker.go | 2 +- pkg/tcpip/hash/jenkins/jenkins.go | 2 +- pkg/tcpip/hash/jenkins/jenkins_test.go | 2 +- pkg/tcpip/header/arp.go | 2 +- pkg/tcpip/header/checksum.go | 2 +- pkg/tcpip/header/eth.go | 2 +- pkg/tcpip/header/gue.go | 2 +- pkg/tcpip/header/icmpv4.go | 2 +- pkg/tcpip/header/icmpv6.go | 2 +- pkg/tcpip/header/interfaces.go | 2 +- pkg/tcpip/header/ipv4.go | 2 +- pkg/tcpip/header/ipv6.go | 2 +- pkg/tcpip/header/ipv6_fragment.go | 2 +- pkg/tcpip/header/ipversion_test.go | 2 +- pkg/tcpip/header/tcp.go | 2 +- pkg/tcpip/header/tcp_test.go | 2 +- pkg/tcpip/header/udp.go | 2 +- pkg/tcpip/link/channel/channel.go | 2 +- pkg/tcpip/link/fdbased/endpoint.go | 2 +- pkg/tcpip/link/fdbased/endpoint_test.go | 2 +- pkg/tcpip/link/fdbased/endpoint_unsafe.go | 2 +- pkg/tcpip/link/fdbased/mmap.go | 2 +- pkg/tcpip/link/fdbased/mmap_amd64_unsafe.go | 2 +- pkg/tcpip/link/loopback/loopback.go | 2 +- pkg/tcpip/link/muxed/injectable.go | 2 +- pkg/tcpip/link/muxed/injectable_test.go | 2 +- pkg/tcpip/link/rawfile/blockingpoll_amd64.s | 2 +- pkg/tcpip/link/rawfile/blockingpoll_amd64_unsafe.go | 2 +- pkg/tcpip/link/rawfile/blockingpoll_unsafe.go | 2 +- pkg/tcpip/link/rawfile/errors.go | 2 +- pkg/tcpip/link/rawfile/rawfile_unsafe.go | 2 +- pkg/tcpip/link/sharedmem/pipe/pipe.go | 2 +- pkg/tcpip/link/sharedmem/pipe/pipe_test.go | 2 +- pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go | 2 +- pkg/tcpip/link/sharedmem/pipe/rx.go | 2 +- pkg/tcpip/link/sharedmem/pipe/tx.go | 2 +- pkg/tcpip/link/sharedmem/queue/queue_test.go | 2 +- pkg/tcpip/link/sharedmem/queue/rx.go | 2 +- pkg/tcpip/link/sharedmem/queue/tx.go | 2 +- pkg/tcpip/link/sharedmem/rx.go | 2 +- pkg/tcpip/link/sharedmem/sharedmem.go | 2 +- pkg/tcpip/link/sharedmem/sharedmem_test.go | 2 +- pkg/tcpip/link/sharedmem/sharedmem_unsafe.go | 2 +- pkg/tcpip/link/sharedmem/tx.go | 2 +- pkg/tcpip/link/sniffer/pcap.go | 2 +- pkg/tcpip/link/sniffer/sniffer.go | 2 +- pkg/tcpip/link/tun/tun_unsafe.go | 2 +- pkg/tcpip/link/waitable/waitable.go | 2 +- pkg/tcpip/link/waitable/waitable_test.go | 2 +- pkg/tcpip/network/arp/arp.go | 2 +- pkg/tcpip/network/arp/arp_test.go | 2 +- pkg/tcpip/network/fragmentation/frag_heap.go | 2 +- pkg/tcpip/network/fragmentation/frag_heap_test.go | 2 +- pkg/tcpip/network/fragmentation/fragmentation.go | 2 +- pkg/tcpip/network/fragmentation/fragmentation_test.go | 2 +- pkg/tcpip/network/fragmentation/reassembler.go | 2 +- pkg/tcpip/network/fragmentation/reassembler_test.go | 2 +- pkg/tcpip/network/hash/hash.go | 2 +- pkg/tcpip/network/ip_test.go | 2 +- pkg/tcpip/network/ipv4/icmp.go | 2 +- pkg/tcpip/network/ipv4/ipv4.go | 2 +- pkg/tcpip/network/ipv4/ipv4_test.go | 2 +- pkg/tcpip/network/ipv6/icmp.go | 2 +- pkg/tcpip/network/ipv6/icmp_test.go | 2 +- pkg/tcpip/network/ipv6/ipv6.go | 2 +- pkg/tcpip/ports/ports.go | 2 +- pkg/tcpip/ports/ports_test.go | 2 +- pkg/tcpip/sample/tun_tcp_connect/main.go | 2 +- pkg/tcpip/sample/tun_tcp_echo/main.go | 2 +- pkg/tcpip/seqnum/seqnum.go | 2 +- pkg/tcpip/stack/linkaddrcache.go | 2 +- pkg/tcpip/stack/linkaddrcache_test.go | 2 +- pkg/tcpip/stack/nic.go | 2 +- pkg/tcpip/stack/registration.go | 2 +- pkg/tcpip/stack/route.go | 2 +- pkg/tcpip/stack/stack.go | 2 +- pkg/tcpip/stack/stack_global_state.go | 2 +- pkg/tcpip/stack/stack_test.go | 2 +- pkg/tcpip/stack/transport_demuxer.go | 2 +- pkg/tcpip/stack/transport_test.go | 2 +- pkg/tcpip/tcpip.go | 2 +- pkg/tcpip/tcpip_test.go | 2 +- pkg/tcpip/time.s | 2 +- pkg/tcpip/time_unsafe.go | 2 +- pkg/tcpip/transport/icmp/endpoint.go | 2 +- pkg/tcpip/transport/icmp/endpoint_state.go | 2 +- pkg/tcpip/transport/icmp/protocol.go | 2 +- pkg/tcpip/transport/raw/raw.go | 2 +- pkg/tcpip/transport/raw/state.go | 2 +- pkg/tcpip/transport/tcp/accept.go | 2 +- pkg/tcpip/transport/tcp/connect.go | 2 +- pkg/tcpip/transport/tcp/cubic.go | 2 +- pkg/tcpip/transport/tcp/dual_stack_test.go | 2 +- pkg/tcpip/transport/tcp/endpoint.go | 2 +- pkg/tcpip/transport/tcp/endpoint_state.go | 2 +- pkg/tcpip/transport/tcp/forwarder.go | 2 +- pkg/tcpip/transport/tcp/protocol.go | 2 +- pkg/tcpip/transport/tcp/rcv.go | 2 +- pkg/tcpip/transport/tcp/reno.go | 2 +- pkg/tcpip/transport/tcp/sack.go | 2 +- pkg/tcpip/transport/tcp/sack_scoreboard.go | 2 +- pkg/tcpip/transport/tcp/sack_scoreboard_test.go | 2 +- pkg/tcpip/transport/tcp/segment.go | 2 +- pkg/tcpip/transport/tcp/segment_heap.go | 2 +- pkg/tcpip/transport/tcp/segment_queue.go | 2 +- pkg/tcpip/transport/tcp/segment_state.go | 2 +- pkg/tcpip/transport/tcp/snd.go | 2 +- pkg/tcpip/transport/tcp/snd_state.go | 2 +- pkg/tcpip/transport/tcp/tcp_sack_test.go | 2 +- pkg/tcpip/transport/tcp/tcp_test.go | 2 +- pkg/tcpip/transport/tcp/tcp_timestamp_test.go | 2 +- pkg/tcpip/transport/tcp/testing/context/context.go | 2 +- pkg/tcpip/transport/tcp/timer.go | 2 +- pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go | 2 +- pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go | 2 +- pkg/tcpip/transport/udp/endpoint.go | 2 +- pkg/tcpip/transport/udp/endpoint_state.go | 2 +- pkg/tcpip/transport/udp/forwarder.go | 2 +- pkg/tcpip/transport/udp/protocol.go | 2 +- pkg/tcpip/transport/udp/udp_test.go | 2 +- pkg/tmutex/tmutex.go | 2 +- pkg/tmutex/tmutex_test.go | 2 +- pkg/unet/unet.go | 2 +- pkg/unet/unet_test.go | 2 +- pkg/unet/unet_unsafe.go | 2 +- pkg/urpc/urpc.go | 2 +- pkg/urpc/urpc_test.go | 2 +- pkg/waiter/waiter.go | 2 +- pkg/waiter/waiter_test.go | 2 +- runsc/boot/compat.go | 2 +- runsc/boot/compat_amd64.go | 2 +- runsc/boot/compat_test.go | 2 +- runsc/boot/config.go | 2 +- runsc/boot/controller.go | 2 +- runsc/boot/debug.go | 2 +- runsc/boot/events.go | 2 +- runsc/boot/fds.go | 2 +- runsc/boot/filter/config.go | 2 +- runsc/boot/filter/extra_filters.go | 2 +- runsc/boot/filter/extra_filters_msan.go | 2 +- runsc/boot/filter/extra_filters_race.go | 2 +- runsc/boot/filter/filter.go | 2 +- runsc/boot/fs.go | 2 +- runsc/boot/limits.go | 2 +- runsc/boot/loader.go | 2 +- runsc/boot/loader_test.go | 2 +- runsc/boot/network.go | 2 +- runsc/boot/strace.go | 2 +- runsc/cgroup/cgroup.go | 2 +- runsc/cgroup/cgroup_test.go | 2 +- runsc/cmd/boot.go | 2 +- runsc/cmd/capability.go | 2 +- runsc/cmd/capability_test.go | 2 +- runsc/cmd/checkpoint.go | 2 +- runsc/cmd/chroot.go | 2 +- runsc/cmd/cmd.go | 2 +- runsc/cmd/create.go | 2 +- runsc/cmd/debug.go | 2 +- runsc/cmd/delete.go | 2 +- runsc/cmd/delete_test.go | 2 +- runsc/cmd/do.go | 2 +- runsc/cmd/events.go | 2 +- runsc/cmd/exec.go | 2 +- runsc/cmd/exec_test.go | 2 +- runsc/cmd/gofer.go | 2 +- runsc/cmd/gofer_test.go | 2 +- runsc/cmd/kill.go | 2 +- runsc/cmd/list.go | 2 +- runsc/cmd/path.go | 2 +- runsc/cmd/pause.go | 2 +- runsc/cmd/ps.go | 2 +- runsc/cmd/restore.go | 2 +- runsc/cmd/resume.go | 2 +- runsc/cmd/run.go | 2 +- runsc/cmd/spec.go | 2 +- runsc/cmd/start.go | 2 +- runsc/cmd/state.go | 2 +- runsc/cmd/wait.go | 2 +- runsc/console/console.go | 2 +- runsc/container/console_test.go | 2 +- runsc/container/container.go | 2 +- runsc/container/container_test.go | 2 +- runsc/container/hook.go | 2 +- runsc/container/multi_container_test.go | 2 +- runsc/container/shared_volume_test.go | 2 +- runsc/container/status.go | 2 +- runsc/container/test_app.go | 2 +- runsc/fsgofer/filter/config.go | 2 +- runsc/fsgofer/filter/extra_filters.go | 2 +- runsc/fsgofer/filter/extra_filters_msan.go | 2 +- runsc/fsgofer/filter/extra_filters_race.go | 2 +- runsc/fsgofer/filter/filter.go | 2 +- runsc/fsgofer/fsgofer.go | 2 +- runsc/fsgofer/fsgofer_test.go | 2 +- runsc/fsgofer/fsgofer_unsafe.go | 2 +- runsc/main.go | 2 +- runsc/sandbox/network.go | 2 +- runsc/sandbox/network_unsafe.go | 2 +- runsc/sandbox/sandbox.go | 2 +- runsc/specutils/fs.go | 2 +- runsc/specutils/namespace.go | 2 +- runsc/specutils/specutils.go | 2 +- runsc/specutils/specutils_test.go | 2 +- runsc/test/image/image.go | 2 +- runsc/test/image/image_test.go | 2 +- runsc/test/image/mysql.sql | 2 +- runsc/test/image/ruby.rb | 2 +- runsc/test/image/ruby.sh | 2 +- runsc/test/install.sh | 2 +- runsc/test/integration/exec_test.go | 2 +- runsc/test/integration/integration.go | 2 +- runsc/test/integration/integration_test.go | 2 +- runsc/test/root/cgroup_test.go | 2 +- runsc/test/root/chroot_test.go | 2 +- runsc/test/root/crictl_test.go | 2 +- runsc/test/root/root.go | 2 +- runsc/test/root/testdata/busybox.go | 2 +- runsc/test/root/testdata/containerd_config.go | 2 +- runsc/test/root/testdata/httpd.go | 2 +- runsc/test/root/testdata/httpd_mount_paths.go | 2 +- runsc/test/root/testdata/sandbox.go | 2 +- runsc/test/testutil/crictl.go | 2 +- runsc/test/testutil/docker.go | 2 +- runsc/test/testutil/testutil.go | 2 +- runsc/test/testutil/testutil_race.go | 2 +- runsc/tools/dockercfg/dockercfg.go | 2 +- runsc/version.go | 2 +- test/syscalls/gtest/gtest.go | 2 +- test/syscalls/linux/32bit.cc | 2 +- test/syscalls/linux/accept_bind.cc | 2 +- test/syscalls/linux/accept_bind_stream.cc | 2 +- test/syscalls/linux/access.cc | 2 +- test/syscalls/linux/affinity.cc | 2 +- test/syscalls/linux/aio.cc | 2 +- test/syscalls/linux/alarm.cc | 2 +- test/syscalls/linux/arch_prctl.cc | 2 +- test/syscalls/linux/bad.cc | 2 +- test/syscalls/linux/base_poll_test.cc | 2 +- test/syscalls/linux/base_poll_test.h | 2 +- test/syscalls/linux/bind.cc | 2 +- test/syscalls/linux/brk.cc | 2 +- test/syscalls/linux/chdir.cc | 2 +- test/syscalls/linux/chmod.cc | 2 +- test/syscalls/linux/chown.cc | 2 +- test/syscalls/linux/chroot.cc | 2 +- test/syscalls/linux/clock_getres.cc | 2 +- test/syscalls/linux/clock_gettime.cc | 2 +- test/syscalls/linux/clock_nanosleep.cc | 2 +- test/syscalls/linux/concurrency.cc | 2 +- test/syscalls/linux/creat.cc | 2 +- test/syscalls/linux/dev.cc | 2 +- test/syscalls/linux/dup.cc | 2 +- test/syscalls/linux/epoll.cc | 2 +- test/syscalls/linux/eventfd.cc | 2 +- test/syscalls/linux/exceptions.cc | 2 +- test/syscalls/linux/exec.cc | 2 +- test/syscalls/linux/exec.h | 2 +- test/syscalls/linux/exec_assert_closed_workload.cc | 2 +- test/syscalls/linux/exec_basic_workload.cc | 2 +- test/syscalls/linux/exec_binary.cc | 2 +- test/syscalls/linux/exec_proc_exe_workload.cc | 2 +- test/syscalls/linux/exec_state_workload.cc | 2 +- test/syscalls/linux/exit.cc | 2 +- test/syscalls/linux/exit_script.sh | 2 +- test/syscalls/linux/fadvise64.cc | 2 +- test/syscalls/linux/fallocate.cc | 2 +- test/syscalls/linux/fault.cc | 2 +- test/syscalls/linux/fchdir.cc | 2 +- test/syscalls/linux/fcntl.cc | 2 +- test/syscalls/linux/file_base.h | 2 +- test/syscalls/linux/flock.cc | 2 +- test/syscalls/linux/fork.cc | 2 +- test/syscalls/linux/fpsig_fork.cc | 2 +- test/syscalls/linux/fpsig_nested.cc | 2 +- test/syscalls/linux/fsync.cc | 2 +- test/syscalls/linux/futex.cc | 2 +- test/syscalls/linux/getcpu.cc | 2 +- test/syscalls/linux/getdents.cc | 2 +- test/syscalls/linux/getrandom.cc | 2 +- test/syscalls/linux/getrusage.cc | 2 +- test/syscalls/linux/inotify.cc | 2 +- test/syscalls/linux/ioctl.cc | 2 +- test/syscalls/linux/ip_socket_test_util.cc | 2 +- test/syscalls/linux/ip_socket_test_util.h | 2 +- test/syscalls/linux/itimer.cc | 2 +- test/syscalls/linux/kill.cc | 2 +- test/syscalls/linux/link.cc | 2 +- test/syscalls/linux/lseek.cc | 2 +- test/syscalls/linux/madvise.cc | 2 +- test/syscalls/linux/memfd.cc | 2 +- test/syscalls/linux/memory_accounting.cc | 2 +- test/syscalls/linux/mempolicy.cc | 2 +- test/syscalls/linux/mincore.cc | 2 +- test/syscalls/linux/mkdir.cc | 2 +- test/syscalls/linux/mknod.cc | 2 +- test/syscalls/linux/mlock.cc | 2 +- test/syscalls/linux/mmap.cc | 2 +- test/syscalls/linux/mount.cc | 2 +- test/syscalls/linux/mremap.cc | 2 +- test/syscalls/linux/msync.cc | 2 +- test/syscalls/linux/munmap.cc | 2 +- test/syscalls/linux/open.cc | 2 +- test/syscalls/linux/open_create.cc | 2 +- test/syscalls/linux/partial_bad_buffer.cc | 2 +- test/syscalls/linux/pause.cc | 2 +- test/syscalls/linux/pipe.cc | 2 +- test/syscalls/linux/poll.cc | 2 +- test/syscalls/linux/ppoll.cc | 2 +- test/syscalls/linux/prctl.cc | 2 +- test/syscalls/linux/prctl_setuid.cc | 2 +- test/syscalls/linux/pread64.cc | 2 +- test/syscalls/linux/preadv.cc | 2 +- test/syscalls/linux/preadv2.cc | 2 +- test/syscalls/linux/priority.cc | 2 +- test/syscalls/linux/priority_execve.cc | 2 +- test/syscalls/linux/proc.cc | 2 +- test/syscalls/linux/proc_net.cc | 2 +- test/syscalls/linux/proc_net_unix.cc | 2 +- test/syscalls/linux/proc_pid_smaps.cc | 2 +- test/syscalls/linux/proc_pid_uid_gid_map.cc | 2 +- test/syscalls/linux/pselect.cc | 2 +- test/syscalls/linux/ptrace.cc | 2 +- test/syscalls/linux/pty.cc | 2 +- test/syscalls/linux/pwrite64.cc | 2 +- test/syscalls/linux/pwritev2.cc | 2 +- test/syscalls/linux/raw_socket_ipv4.cc | 2 +- test/syscalls/linux/read.cc | 2 +- test/syscalls/linux/readv.cc | 2 +- test/syscalls/linux/readv_common.cc | 2 +- test/syscalls/linux/readv_common.h | 2 +- test/syscalls/linux/readv_socket.cc | 2 +- test/syscalls/linux/rename.cc | 2 +- test/syscalls/linux/rlimits.cc | 2 +- test/syscalls/linux/rtsignal.cc | 2 +- test/syscalls/linux/sched.cc | 2 +- test/syscalls/linux/sched_yield.cc | 2 +- test/syscalls/linux/seccomp.cc | 2 +- test/syscalls/linux/select.cc | 2 +- test/syscalls/linux/semaphore.cc | 2 +- test/syscalls/linux/sendfile.cc | 2 +- test/syscalls/linux/sendfile_socket.cc | 2 +- test/syscalls/linux/shm.cc | 2 +- test/syscalls/linux/sigaction.cc | 2 +- test/syscalls/linux/sigaltstack.cc | 2 +- test/syscalls/linux/sigaltstack_check.cc | 2 +- test/syscalls/linux/sigiret.cc | 2 +- test/syscalls/linux/sigprocmask.cc | 2 +- test/syscalls/linux/sigstop.cc | 2 +- test/syscalls/linux/sigtimedwait.cc | 2 +- test/syscalls/linux/socket_abstract.cc | 2 +- test/syscalls/linux/socket_blocking.cc | 2 +- test/syscalls/linux/socket_blocking.h | 2 +- test/syscalls/linux/socket_filesystem.cc | 2 +- test/syscalls/linux/socket_generic.cc | 2 +- test/syscalls/linux/socket_generic.h | 2 +- test/syscalls/linux/socket_inet_loopback.cc | 2 +- test/syscalls/linux/socket_ip_loopback_blocking.cc | 2 +- test/syscalls/linux/socket_ip_tcp_generic.cc | 2 +- test/syscalls/linux/socket_ip_tcp_generic.h | 2 +- test/syscalls/linux/socket_ip_tcp_generic_loopback.cc | 2 +- test/syscalls/linux/socket_ip_tcp_loopback.cc | 2 +- test/syscalls/linux/socket_ip_tcp_loopback_blocking.cc | 2 +- test/syscalls/linux/socket_ip_tcp_loopback_nonblock.cc | 2 +- test/syscalls/linux/socket_ip_tcp_udp_generic.cc | 2 +- test/syscalls/linux/socket_ip_udp_generic.cc | 2 +- test/syscalls/linux/socket_ip_udp_generic.h | 2 +- test/syscalls/linux/socket_ip_udp_loopback.cc | 2 +- test/syscalls/linux/socket_ip_udp_loopback_blocking.cc | 2 +- test/syscalls/linux/socket_ip_udp_loopback_nonblock.cc | 2 +- .../syscalls/linux/socket_ipv4_tcp_unbound_external_networking.cc | 2 +- test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.h | 2 +- .../linux/socket_ipv4_tcp_unbound_external_networking_test.cc | 2 +- test/syscalls/linux/socket_ipv4_udp_unbound.cc | 2 +- test/syscalls/linux/socket_ipv4_udp_unbound.h | 2 +- .../syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc | 2 +- test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.h | 2 +- .../linux/socket_ipv4_udp_unbound_external_networking_test.cc | 2 +- test/syscalls/linux/socket_ipv4_udp_unbound_loopback.cc | 2 +- test/syscalls/linux/socket_netdevice.cc | 2 +- test/syscalls/linux/socket_netlink_route.cc | 2 +- test/syscalls/linux/socket_netlink_util.cc | 2 +- test/syscalls/linux/socket_netlink_util.h | 2 +- test/syscalls/linux/socket_non_blocking.cc | 2 +- test/syscalls/linux/socket_non_blocking.h | 2 +- test/syscalls/linux/socket_non_stream.cc | 2 +- test/syscalls/linux/socket_non_stream.h | 2 +- test/syscalls/linux/socket_non_stream_blocking.cc | 2 +- test/syscalls/linux/socket_non_stream_blocking.h | 2 +- test/syscalls/linux/socket_stream.cc | 2 +- test/syscalls/linux/socket_stream.h | 2 +- test/syscalls/linux/socket_stream_blocking.cc | 2 +- test/syscalls/linux/socket_stream_blocking.h | 2 +- test/syscalls/linux/socket_stream_nonblock.cc | 2 +- test/syscalls/linux/socket_stream_nonblock.h | 2 +- test/syscalls/linux/socket_test_util.cc | 2 +- test/syscalls/linux/socket_test_util.h | 2 +- test/syscalls/linux/socket_unix.cc | 2 +- test/syscalls/linux/socket_unix.h | 2 +- test/syscalls/linux/socket_unix_abstract.cc | 2 +- test/syscalls/linux/socket_unix_abstract_nonblock.cc | 2 +- test/syscalls/linux/socket_unix_blocking_local.cc | 2 +- test/syscalls/linux/socket_unix_dgram.cc | 2 +- test/syscalls/linux/socket_unix_dgram.h | 2 +- test/syscalls/linux/socket_unix_dgram_local.cc | 2 +- test/syscalls/linux/socket_unix_dgram_non_blocking.cc | 2 +- test/syscalls/linux/socket_unix_domain.cc | 2 +- test/syscalls/linux/socket_unix_filesystem.cc | 2 +- test/syscalls/linux/socket_unix_filesystem_nonblock.cc | 2 +- test/syscalls/linux/socket_unix_non_stream.cc | 2 +- test/syscalls/linux/socket_unix_non_stream.h | 2 +- test/syscalls/linux/socket_unix_non_stream_blocking_local.cc | 2 +- test/syscalls/linux/socket_unix_pair.cc | 2 +- test/syscalls/linux/socket_unix_pair_nonblock.cc | 2 +- test/syscalls/linux/socket_unix_seqpacket.cc | 2 +- test/syscalls/linux/socket_unix_seqpacket.h | 2 +- test/syscalls/linux/socket_unix_seqpacket_local.cc | 2 +- test/syscalls/linux/socket_unix_stream.cc | 2 +- test/syscalls/linux/socket_unix_stream_blocking_local.cc | 2 +- test/syscalls/linux/socket_unix_stream_local.cc | 2 +- test/syscalls/linux/socket_unix_stream_nonblock_local.cc | 2 +- test/syscalls/linux/socket_unix_unbound_abstract.cc | 2 +- test/syscalls/linux/socket_unix_unbound_dgram.cc | 2 +- test/syscalls/linux/socket_unix_unbound_filesystem.cc | 2 +- test/syscalls/linux/socket_unix_unbound_seqpacket.cc | 2 +- test/syscalls/linux/socket_unix_unbound_stream.cc | 2 +- test/syscalls/linux/stat.cc | 2 +- test/syscalls/linux/stat_times.cc | 2 +- test/syscalls/linux/statfs.cc | 2 +- test/syscalls/linux/sticky.cc | 2 +- test/syscalls/linux/symlink.cc | 2 +- test/syscalls/linux/sync.cc | 2 +- test/syscalls/linux/sync_file_range.cc | 2 +- test/syscalls/linux/sysinfo.cc | 2 +- test/syscalls/linux/syslog.cc | 2 +- test/syscalls/linux/sysret.cc | 2 +- test/syscalls/linux/tcp_socket.cc | 2 +- test/syscalls/linux/temp_umask.h | 2 +- test/syscalls/linux/tgkill.cc | 2 +- test/syscalls/linux/time.cc | 2 +- test/syscalls/linux/timerfd.cc | 2 +- test/syscalls/linux/timers.cc | 2 +- test/syscalls/linux/tkill.cc | 2 +- test/syscalls/linux/truncate.cc | 2 +- test/syscalls/linux/udp_bind.cc | 2 +- test/syscalls/linux/udp_socket.cc | 2 +- test/syscalls/linux/uidgid.cc | 2 +- test/syscalls/linux/uname.cc | 2 +- test/syscalls/linux/unix_domain_socket_test_util.cc | 2 +- test/syscalls/linux/unix_domain_socket_test_util.h | 2 +- test/syscalls/linux/unlink.cc | 2 +- test/syscalls/linux/unshare.cc | 2 +- test/syscalls/linux/utimes.cc | 2 +- test/syscalls/linux/vdso.cc | 2 +- test/syscalls/linux/vdso_clock_gettime.cc | 2 +- test/syscalls/linux/vfork.cc | 2 +- test/syscalls/linux/vsyscall.cc | 2 +- test/syscalls/linux/wait.cc | 2 +- test/syscalls/linux/write.cc | 2 +- test/syscalls/syscall_test_runner.go | 2 +- test/syscalls/syscall_test_runner.sh | 2 +- test/util/capability_util.cc | 2 +- test/util/capability_util.h | 2 +- test/util/cleanup.h | 2 +- test/util/epoll_util.cc | 2 +- test/util/epoll_util.h | 2 +- test/util/eventfd_util.h | 2 +- test/util/file_descriptor.h | 2 +- test/util/fs_util.cc | 2 +- test/util/fs_util.h | 2 +- test/util/fs_util_test.cc | 2 +- test/util/logging.cc | 2 +- test/util/logging.h | 2 +- test/util/memory_util.h | 2 +- test/util/mount_util.h | 2 +- test/util/multiprocess_util.cc | 2 +- test/util/multiprocess_util.h | 2 +- test/util/posix_error.cc | 2 +- test/util/posix_error.h | 2 +- test/util/posix_error_test.cc | 2 +- test/util/proc_util.cc | 2 +- test/util/proc_util.h | 2 +- test/util/proc_util_test.cc | 2 +- test/util/rlimit_util.cc | 2 +- test/util/rlimit_util.h | 2 +- test/util/save_util.cc | 2 +- test/util/save_util.h | 2 +- test/util/signal_util.cc | 2 +- test/util/signal_util.h | 2 +- test/util/temp_path.cc | 2 +- test/util/temp_path.h | 2 +- test/util/test_main.cc | 2 +- test/util/test_util.cc | 2 +- test/util/test_util.h | 2 +- test/util/test_util_test.cc | 2 +- test/util/thread_util.h | 2 +- test/util/timer_util.cc | 2 +- test/util/timer_util.h | 2 +- third_party/gvsync/atomicptr_unsafe.go | 2 +- third_party/gvsync/atomicptrtest/atomicptr_test.go | 2 +- third_party/gvsync/downgradable_rwmutex_test.go | 2 +- third_party/gvsync/downgradable_rwmutex_unsafe.go | 2 +- third_party/gvsync/gvsync.go | 2 +- third_party/gvsync/memmove_unsafe.go | 2 +- third_party/gvsync/norace_unsafe.go | 2 +- third_party/gvsync/race_unsafe.go | 2 +- third_party/gvsync/seqatomic_unsafe.go | 2 +- third_party/gvsync/seqatomictest/seqatomic_test.go | 2 +- third_party/gvsync/seqcount.go | 2 +- third_party/gvsync/seqcount_test.go | 2 +- tools/go_generics/generics.go | 2 +- tools/go_generics/generics_tests/all_stmts/input.go | 2 +- tools/go_generics/generics_tests/all_stmts/output/output.go | 2 +- tools/go_generics/generics_tests/all_types/input.go | 2 +- tools/go_generics/generics_tests/all_types/lib/lib.go | 2 +- tools/go_generics/generics_tests/all_types/output/output.go | 2 +- tools/go_generics/generics_tests/consts/input.go | 2 +- tools/go_generics/generics_tests/consts/output/output.go | 2 +- tools/go_generics/generics_tests/imports/input.go | 2 +- tools/go_generics/generics_tests/imports/output/output.go | 2 +- tools/go_generics/generics_tests/remove_typedef/input.go | 2 +- tools/go_generics/generics_tests/remove_typedef/output/output.go | 2 +- tools/go_generics/generics_tests/simple/input.go | 2 +- tools/go_generics/generics_tests/simple/output/output.go | 2 +- tools/go_generics/globals/globals_visitor.go | 2 +- tools/go_generics/globals/scope.go | 2 +- tools/go_generics/go_generics_unittest.sh | 2 +- tools/go_generics/go_merge/main.go | 2 +- tools/go_generics/imports.go | 2 +- tools/go_generics/remove.go | 2 +- tools/go_generics/rules_tests/template.go | 2 +- tools/go_generics/rules_tests/template_test.go | 2 +- tools/go_stateify/main.go | 2 +- tools/tag_release.sh | 2 +- tools/workspace_status.sh | 2 +- vdso/barrier.h | 2 +- vdso/check_vdso.py | 2 +- vdso/compiler.h | 2 +- vdso/cycle_clock.h | 2 +- vdso/seqlock.h | 2 +- vdso/syscalls.h | 2 +- vdso/vdso.cc | 2 +- vdso/vdso_time.cc | 2 +- vdso/vdso_time.h | 2 +- 1235 files changed, 1242 insertions(+), 1234 deletions(-) create mode 100644 AUTHORS (limited to 'pkg/sentry/kernel') diff --git a/AUTHORS b/AUTHORS new file mode 100644 index 000000000..01ba46567 --- /dev/null +++ b/AUTHORS @@ -0,0 +1,8 @@ +# This is the list of gVisor authors for copyright purposes. +# +# This does not necessarily list everyone who has contributed code, since in +# some cases, their employer may be the copyright holder. To see the full list +# of contributors, see the revision history in source control. +# +# Please send a patch if you would like to be included in this list. +Google LLC diff --git a/kokoro/run_build.sh b/kokoro/run_build.sh index 89e24b037..63fffda48 100755 --- a/kokoro/run_build.sh +++ b/kokoro/run_build.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/kokoro/run_tests.sh b/kokoro/run_tests.sh index 8a3ce7402..08f678e39 100755 --- a/kokoro/run_tests.sh +++ b/kokoro/run_tests.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/pkg/abi/abi.go b/pkg/abi/abi.go index 7770f0405..d56c481c9 100644 --- a/pkg/abi/abi.go +++ b/pkg/abi/abi.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/abi_linux.go b/pkg/abi/abi_linux.go index 9d9f361a4..3059479bd 100644 --- a/pkg/abi/abi_linux.go +++ b/pkg/abi/abi_linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/flag.go b/pkg/abi/flag.go index b48757da8..dcdd66d4e 100644 --- a/pkg/abi/flag.go +++ b/pkg/abi/flag.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/aio.go b/pkg/abi/linux/aio.go index 1b7ca714a..3c6e0079d 100644 --- a/pkg/abi/linux/aio.go +++ b/pkg/abi/linux/aio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ashmem.go b/pkg/abi/linux/ashmem.go index ced1e44d4..2a722abe0 100644 --- a/pkg/abi/linux/ashmem.go +++ b/pkg/abi/linux/ashmem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/audit.go b/pkg/abi/linux/audit.go index b39ba4515..6cca69af9 100644 --- a/pkg/abi/linux/audit.go +++ b/pkg/abi/linux/audit.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/binder.go b/pkg/abi/linux/binder.go index 522dc6f53..63b08324a 100644 --- a/pkg/abi/linux/binder.go +++ b/pkg/abi/linux/binder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/bpf.go b/pkg/abi/linux/bpf.go index d9cd09948..aa3d3ce70 100644 --- a/pkg/abi/linux/bpf.go +++ b/pkg/abi/linux/bpf.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/capability.go b/pkg/abi/linux/capability.go index 7d96f013e..c120cac64 100644 --- a/pkg/abi/linux/capability.go +++ b/pkg/abi/linux/capability.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/dev.go b/pkg/abi/linux/dev.go index 5b1199aac..421e11256 100644 --- a/pkg/abi/linux/dev.go +++ b/pkg/abi/linux/dev.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/elf.go b/pkg/abi/linux/elf.go index 928067c04..fb1c679d2 100644 --- a/pkg/abi/linux/elf.go +++ b/pkg/abi/linux/elf.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/errors.go b/pkg/abi/linux/errors.go index e5f6f3f07..93f85a864 100644 --- a/pkg/abi/linux/errors.go +++ b/pkg/abi/linux/errors.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/eventfd.go b/pkg/abi/linux/eventfd.go index 5614f5cf1..9c479fc8f 100644 --- a/pkg/abi/linux/eventfd.go +++ b/pkg/abi/linux/eventfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/exec.go b/pkg/abi/linux/exec.go index a07c29243..579d46c41 100644 --- a/pkg/abi/linux/exec.go +++ b/pkg/abi/linux/exec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/fcntl.go b/pkg/abi/linux/fcntl.go index c8558933a..cc8f2702d 100644 --- a/pkg/abi/linux/fcntl.go +++ b/pkg/abi/linux/fcntl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/file.go b/pkg/abi/linux/file.go index 46b10ca97..753fec3ed 100644 --- a/pkg/abi/linux/file.go +++ b/pkg/abi/linux/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/fs.go b/pkg/abi/linux/fs.go index a9f2ba132..c82ab9b5b 100644 --- a/pkg/abi/linux/fs.go +++ b/pkg/abi/linux/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/futex.go b/pkg/abi/linux/futex.go index afdf4123b..08bfde3b5 100644 --- a/pkg/abi/linux/futex.go +++ b/pkg/abi/linux/futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/inotify.go b/pkg/abi/linux/inotify.go index 79c5d3593..2d08194ba 100644 --- a/pkg/abi/linux/inotify.go +++ b/pkg/abi/linux/inotify.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ioctl.go b/pkg/abi/linux/ioctl.go index 191b26e4d..04bb767dc 100644 --- a/pkg/abi/linux/ioctl.go +++ b/pkg/abi/linux/ioctl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ip.go b/pkg/abi/linux/ip.go index 77ac1062c..31e56ffa6 100644 --- a/pkg/abi/linux/ip.go +++ b/pkg/abi/linux/ip.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ipc.go b/pkg/abi/linux/ipc.go index 10681768b..2ef8d6cbb 100644 --- a/pkg/abi/linux/ipc.go +++ b/pkg/abi/linux/ipc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/limits.go b/pkg/abi/linux/limits.go index e0aa5b31d..c74dfcd53 100644 --- a/pkg/abi/linux/limits.go +++ b/pkg/abi/linux/limits.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/linux.go b/pkg/abi/linux/linux.go index d365f693d..8a8f831cd 100644 --- a/pkg/abi/linux/linux.go +++ b/pkg/abi/linux/linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/mm.go b/pkg/abi/linux/mm.go index eda8d9788..0b02f938a 100644 --- a/pkg/abi/linux/mm.go +++ b/pkg/abi/linux/mm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/netdevice.go b/pkg/abi/linux/netdevice.go index e3b6b1e40..aef1acf75 100644 --- a/pkg/abi/linux/netdevice.go +++ b/pkg/abi/linux/netdevice.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/netlink.go b/pkg/abi/linux/netlink.go index 25c5e17fd..5e718c363 100644 --- a/pkg/abi/linux/netlink.go +++ b/pkg/abi/linux/netlink.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/netlink_route.go b/pkg/abi/linux/netlink_route.go index 4200b6506..630dc339a 100644 --- a/pkg/abi/linux/netlink_route.go +++ b/pkg/abi/linux/netlink_route.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/poll.go b/pkg/abi/linux/poll.go index 9f0b15d1c..c04d26e4c 100644 --- a/pkg/abi/linux/poll.go +++ b/pkg/abi/linux/poll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/prctl.go b/pkg/abi/linux/prctl.go index db3206f36..dae2de290 100644 --- a/pkg/abi/linux/prctl.go +++ b/pkg/abi/linux/prctl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/ptrace.go b/pkg/abi/linux/ptrace.go index 7db4f5464..23e605ab2 100644 --- a/pkg/abi/linux/ptrace.go +++ b/pkg/abi/linux/ptrace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/rusage.go b/pkg/abi/linux/rusage.go index 7fea4b589..d8302dc85 100644 --- a/pkg/abi/linux/rusage.go +++ b/pkg/abi/linux/rusage.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/sched.go b/pkg/abi/linux/sched.go index ef96a3801..193d9a242 100644 --- a/pkg/abi/linux/sched.go +++ b/pkg/abi/linux/sched.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/seccomp.go b/pkg/abi/linux/seccomp.go index 8673a27bf..4eeb5cd7a 100644 --- a/pkg/abi/linux/seccomp.go +++ b/pkg/abi/linux/seccomp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/sem.go b/pkg/abi/linux/sem.go index b80c93daf..de422c519 100644 --- a/pkg/abi/linux/sem.go +++ b/pkg/abi/linux/sem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/shm.go b/pkg/abi/linux/shm.go index 82a80e609..e45aadb10 100644 --- a/pkg/abi/linux/shm.go +++ b/pkg/abi/linux/shm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/signal.go b/pkg/abi/linux/signal.go index 395f9f31e..9cbd77dda 100644 --- a/pkg/abi/linux/signal.go +++ b/pkg/abi/linux/signal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/socket.go b/pkg/abi/linux/socket.go index 6fa4e7c3e..417840731 100644 --- a/pkg/abi/linux/socket.go +++ b/pkg/abi/linux/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/tcp.go b/pkg/abi/linux/tcp.go index 67908deb9..174d470e2 100644 --- a/pkg/abi/linux/tcp.go +++ b/pkg/abi/linux/tcp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/time.go b/pkg/abi/linux/time.go index bbd21e726..fa9ee27e1 100644 --- a/pkg/abi/linux/time.go +++ b/pkg/abi/linux/time.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/timer.go b/pkg/abi/linux/timer.go index a6f420bdb..e32d09e10 100644 --- a/pkg/abi/linux/timer.go +++ b/pkg/abi/linux/timer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/tty.go b/pkg/abi/linux/tty.go index bff882d89..8ac02aee8 100644 --- a/pkg/abi/linux/tty.go +++ b/pkg/abi/linux/tty.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/uio.go b/pkg/abi/linux/uio.go index 7e00d9959..1fd1e9802 100644 --- a/pkg/abi/linux/uio.go +++ b/pkg/abi/linux/uio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/abi/linux/utsname.go b/pkg/abi/linux/utsname.go index f80ed7d4a..60f220a67 100644 --- a/pkg/abi/linux/utsname.go +++ b/pkg/abi/linux/utsname.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/amutex/amutex.go b/pkg/amutex/amutex.go index 26b674435..85e819304 100644 --- a/pkg/amutex/amutex.go +++ b/pkg/amutex/amutex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/amutex/amutex_test.go b/pkg/amutex/amutex_test.go index 104e0dab1..6a0af006e 100644 --- a/pkg/amutex/amutex_test.go +++ b/pkg/amutex/amutex_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/atomicbitops/atomic_bitops.go b/pkg/atomicbitops/atomic_bitops.go index 9a57f9599..63aa2b7f1 100644 --- a/pkg/atomicbitops/atomic_bitops.go +++ b/pkg/atomicbitops/atomic_bitops.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/atomicbitops/atomic_bitops_amd64.s b/pkg/atomicbitops/atomic_bitops_amd64.s index b37e3aad3..db0972001 100644 --- a/pkg/atomicbitops/atomic_bitops_amd64.s +++ b/pkg/atomicbitops/atomic_bitops_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/atomicbitops/atomic_bitops_common.go b/pkg/atomicbitops/atomic_bitops_common.go index b03242baa..b2a943dcb 100644 --- a/pkg/atomicbitops/atomic_bitops_common.go +++ b/pkg/atomicbitops/atomic_bitops_common.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/atomicbitops/atomic_bitops_test.go b/pkg/atomicbitops/atomic_bitops_test.go index ee6207cb3..965e9be79 100644 --- a/pkg/atomicbitops/atomic_bitops_test.go +++ b/pkg/atomicbitops/atomic_bitops_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/binary/binary.go b/pkg/binary/binary.go index 02f7e9fb8..631785f7b 100644 --- a/pkg/binary/binary.go +++ b/pkg/binary/binary.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/binary/binary_test.go b/pkg/binary/binary_test.go index 200961c70..4d609a438 100644 --- a/pkg/binary/binary_test.go +++ b/pkg/binary/binary_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/bits.go b/pkg/bits/bits.go index eb3c80f49..a26433ad6 100644 --- a/pkg/bits/bits.go +++ b/pkg/bits/bits.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/bits_template.go b/pkg/bits/bits_template.go index 8c578cca2..93a435b80 100644 --- a/pkg/bits/bits_template.go +++ b/pkg/bits/bits_template.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/uint64_arch_amd64.go b/pkg/bits/uint64_arch_amd64.go index 1fef89394..faccaa61a 100644 --- a/pkg/bits/uint64_arch_amd64.go +++ b/pkg/bits/uint64_arch_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/uint64_arch_amd64_asm.s b/pkg/bits/uint64_arch_amd64_asm.s index 8c7322f0f..8ff364181 100644 --- a/pkg/bits/uint64_arch_amd64_asm.s +++ b/pkg/bits/uint64_arch_amd64_asm.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/uint64_arch_generic.go b/pkg/bits/uint64_arch_generic.go index cfb47400b..7dd2d1480 100644 --- a/pkg/bits/uint64_arch_generic.go +++ b/pkg/bits/uint64_arch_generic.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bits/uint64_test.go b/pkg/bits/uint64_test.go index d6dbaf602..1b018d808 100644 --- a/pkg/bits/uint64_test.go +++ b/pkg/bits/uint64_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/bpf.go b/pkg/bpf/bpf.go index 98d44d911..eb546f48f 100644 --- a/pkg/bpf/bpf.go +++ b/pkg/bpf/bpf.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/decoder.go b/pkg/bpf/decoder.go index ae6b8839a..45c192215 100644 --- a/pkg/bpf/decoder.go +++ b/pkg/bpf/decoder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/decoder_test.go b/pkg/bpf/decoder_test.go index f093e1e41..8c4bdad21 100644 --- a/pkg/bpf/decoder_test.go +++ b/pkg/bpf/decoder_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/input_bytes.go b/pkg/bpf/input_bytes.go index 745c0749b..86b216cfc 100644 --- a/pkg/bpf/input_bytes.go +++ b/pkg/bpf/input_bytes.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/interpreter.go b/pkg/bpf/interpreter.go index 86c7add4d..86de523a2 100644 --- a/pkg/bpf/interpreter.go +++ b/pkg/bpf/interpreter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/interpreter_test.go b/pkg/bpf/interpreter_test.go index c46a43991..67b00ffe3 100644 --- a/pkg/bpf/interpreter_test.go +++ b/pkg/bpf/interpreter_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/program_builder.go b/pkg/bpf/program_builder.go index b4ce228e1..fc9d27203 100644 --- a/pkg/bpf/program_builder.go +++ b/pkg/bpf/program_builder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/bpf/program_builder_test.go b/pkg/bpf/program_builder_test.go index 0e0b79d88..5b2ad67de 100644 --- a/pkg/bpf/program_builder_test.go +++ b/pkg/bpf/program_builder_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/compressio/compressio.go b/pkg/compressio/compressio.go index 4daaa82b6..8c14ccbfa 100644 --- a/pkg/compressio/compressio.go +++ b/pkg/compressio/compressio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/compressio/compressio_test.go b/pkg/compressio/compressio_test.go index 1bbabee79..86dc47e44 100644 --- a/pkg/compressio/compressio_test.go +++ b/pkg/compressio/compressio_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/control/client/client.go b/pkg/control/client/client.go index 0d0c9f148..3fec27846 100644 --- a/pkg/control/client/client.go +++ b/pkg/control/client/client.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/control/server/server.go b/pkg/control/server/server.go index c46b5d70b..1a15da1a8 100644 --- a/pkg/control/server/server.go +++ b/pkg/control/server/server.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/cpuid/cpu_amd64.s b/pkg/cpuid/cpu_amd64.s index 905c1d12e..ac80d3c8a 100644 --- a/pkg/cpuid/cpu_amd64.s +++ b/pkg/cpuid/cpu_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/cpuid/cpuid.go b/pkg/cpuid/cpuid.go index 61441150e..3eb2bcd2b 100644 --- a/pkg/cpuid/cpuid.go +++ b/pkg/cpuid/cpuid.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/cpuid/cpuid_parse_test.go b/pkg/cpuid/cpuid_parse_test.go index e8f87a10e..dd9969db4 100644 --- a/pkg/cpuid/cpuid_parse_test.go +++ b/pkg/cpuid/cpuid_parse_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/cpuid/cpuid_test.go b/pkg/cpuid/cpuid_test.go index 64ade1cbe..6ae14d2da 100644 --- a/pkg/cpuid/cpuid_test.go +++ b/pkg/cpuid/cpuid_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/client.go b/pkg/dhcp/client.go index 2ba79be32..b7cde3819 100644 --- a/pkg/dhcp/client.go +++ b/pkg/dhcp/client.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/dhcp.go b/pkg/dhcp/dhcp.go index 6945bcd35..f96ffd891 100644 --- a/pkg/dhcp/dhcp.go +++ b/pkg/dhcp/dhcp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/dhcp_string.go b/pkg/dhcp/dhcp_string.go index 8533895bd..29ce98593 100644 --- a/pkg/dhcp/dhcp_string.go +++ b/pkg/dhcp/dhcp_string.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/dhcp_test.go b/pkg/dhcp/dhcp_test.go index e1d8ef603..751626bb0 100644 --- a/pkg/dhcp/dhcp_test.go +++ b/pkg/dhcp/dhcp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/dhcp/server.go b/pkg/dhcp/server.go index 9549ff705..6a1972860 100644 --- a/pkg/dhcp/server.go +++ b/pkg/dhcp/server.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/eventchannel/event.go b/pkg/eventchannel/event.go index 41a7b5ed3..4c8ae573b 100644 --- a/pkg/eventchannel/event.go +++ b/pkg/eventchannel/event.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/eventchannel/event.proto b/pkg/eventchannel/event.proto index c1679c7e7..34468f072 100644 --- a/pkg/eventchannel/event.proto +++ b/pkg/eventchannel/event.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/fd/fd.go b/pkg/fd/fd.go index d40758c22..2785243a2 100644 --- a/pkg/fd/fd.go +++ b/pkg/fd/fd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/fd/fd_test.go b/pkg/fd/fd_test.go index 42bb3ef6c..5fb0ad47d 100644 --- a/pkg/fd/fd_test.go +++ b/pkg/fd/fd_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/fdnotifier/fdnotifier.go b/pkg/fdnotifier/fdnotifier.go index aa4906ca0..f0b028b0b 100644 --- a/pkg/fdnotifier/fdnotifier.go +++ b/pkg/fdnotifier/fdnotifier.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/fdnotifier/poll_unsafe.go b/pkg/fdnotifier/poll_unsafe.go index 05be9aeb5..bc5e0ac44 100644 --- a/pkg/fdnotifier/poll_unsafe.go +++ b/pkg/fdnotifier/poll_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/gate/gate.go b/pkg/gate/gate.go index 48122bf5a..bda6aae09 100644 --- a/pkg/gate/gate.go +++ b/pkg/gate/gate.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/gate/gate_test.go b/pkg/gate/gate_test.go index 95620fa8e..7467e7d07 100644 --- a/pkg/gate/gate_test.go +++ b/pkg/gate/gate_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/ilist/list.go b/pkg/ilist/list.go index 51c9b6df3..019caadca 100644 --- a/pkg/ilist/list.go +++ b/pkg/ilist/list.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/ilist/list_test.go b/pkg/ilist/list_test.go index f37946dc2..3f9abfb56 100644 --- a/pkg/ilist/list_test.go +++ b/pkg/ilist/list_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/linewriter/linewriter.go b/pkg/linewriter/linewriter.go index 5fbd4e779..cd6e4e2ce 100644 --- a/pkg/linewriter/linewriter.go +++ b/pkg/linewriter/linewriter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/linewriter/linewriter_test.go b/pkg/linewriter/linewriter_test.go index 9140ee6af..96dc7e6e0 100644 --- a/pkg/linewriter/linewriter_test.go +++ b/pkg/linewriter/linewriter_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/glog.go b/pkg/log/glog.go index 24d5390d7..5732785b4 100644 --- a/pkg/log/glog.go +++ b/pkg/log/glog.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/glog_unsafe.go b/pkg/log/glog_unsafe.go index bb06aa7d3..ea17ae349 100644 --- a/pkg/log/glog_unsafe.go +++ b/pkg/log/glog_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/json.go b/pkg/log/json.go index 96bd13d87..a278c8fc8 100644 --- a/pkg/log/json.go +++ b/pkg/log/json.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/json_k8s.go b/pkg/log/json_k8s.go index 9c2f8d2b7..c2c019915 100644 --- a/pkg/log/json_k8s.go +++ b/pkg/log/json_k8s.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/json_test.go b/pkg/log/json_test.go index b8c7a795e..f25224fe1 100644 --- a/pkg/log/json_test.go +++ b/pkg/log/json_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/log.go b/pkg/log/log.go index b8d456aae..7d563241e 100644 --- a/pkg/log/log.go +++ b/pkg/log/log.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/log/log_test.go b/pkg/log/log_test.go index a59d457dd..0634e7c1f 100644 --- a/pkg/log/log_test.go +++ b/pkg/log/log_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/metric/metric.go b/pkg/metric/metric.go index e5eb95f89..803709cc4 100644 --- a/pkg/metric/metric.go +++ b/pkg/metric/metric.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/metric/metric.proto b/pkg/metric/metric.proto index 917fda1ac..a2c2bd1ba 100644 --- a/pkg/metric/metric.proto +++ b/pkg/metric/metric.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/metric/metric_test.go b/pkg/metric/metric_test.go index 40034a589..b8b124c83 100644 --- a/pkg/metric/metric_test.go +++ b/pkg/metric/metric_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/buffer.go b/pkg/p9/buffer.go index b7bb14ef9..4c8c6555d 100644 --- a/pkg/p9/buffer.go +++ b/pkg/p9/buffer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/buffer_test.go b/pkg/p9/buffer_test.go index 18d55e5c0..a9c75f86b 100644 --- a/pkg/p9/buffer_test.go +++ b/pkg/p9/buffer_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/client.go b/pkg/p9/client.go index 67887874a..2f9c716d0 100644 --- a/pkg/p9/client.go +++ b/pkg/p9/client.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/client_file.go b/pkg/p9/client_file.go index 992d1daf7..63c65129a 100644 --- a/pkg/p9/client_file.go +++ b/pkg/p9/client_file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/client_test.go b/pkg/p9/client_test.go index f7145452d..fc49729d8 100644 --- a/pkg/p9/client_test.go +++ b/pkg/p9/client_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/file.go b/pkg/p9/file.go index 55ceb52e1..a52a0f3e7 100644 --- a/pkg/p9/file.go +++ b/pkg/p9/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/handlers.go b/pkg/p9/handlers.go index c1d1ac1e8..6da2ce4e3 100644 --- a/pkg/p9/handlers.go +++ b/pkg/p9/handlers.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/local_server/local_server.go b/pkg/p9/local_server/local_server.go index 69b90c6cd..f4077a9d4 100644 --- a/pkg/p9/local_server/local_server.go +++ b/pkg/p9/local_server/local_server.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/messages.go b/pkg/p9/messages.go index 97decd3cc..833defbd6 100644 --- a/pkg/p9/messages.go +++ b/pkg/p9/messages.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/messages_test.go b/pkg/p9/messages_test.go index 68395a396..10a0587cf 100644 --- a/pkg/p9/messages_test.go +++ b/pkg/p9/messages_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/p9.go b/pkg/p9/p9.go index 4ea9f2f9a..78c7d3f86 100644 --- a/pkg/p9/p9.go +++ b/pkg/p9/p9.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/p9_test.go b/pkg/p9/p9_test.go index 02498346c..8dda6cc64 100644 --- a/pkg/p9/p9_test.go +++ b/pkg/p9/p9_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/p9test/client_test.go b/pkg/p9/p9test/client_test.go index 242d81b95..e00dd03ab 100644 --- a/pkg/p9/p9test/client_test.go +++ b/pkg/p9/p9test/client_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/p9test/p9test.go b/pkg/p9/p9test/p9test.go index f9bacbf84..1c8eff200 100644 --- a/pkg/p9/p9test/p9test.go +++ b/pkg/p9/p9test/p9test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/path_tree.go b/pkg/p9/path_tree.go index 60b20578e..f37ad4ab2 100644 --- a/pkg/p9/path_tree.go +++ b/pkg/p9/path_tree.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/pool.go b/pkg/p9/pool.go index 34ed898e8..52de889e1 100644 --- a/pkg/p9/pool.go +++ b/pkg/p9/pool.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/pool_test.go b/pkg/p9/pool_test.go index 71052d8c4..e4746b8da 100644 --- a/pkg/p9/pool_test.go +++ b/pkg/p9/pool_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/server.go b/pkg/p9/server.go index 3ef151595..b2a86d8fa 100644 --- a/pkg/p9/server.go +++ b/pkg/p9/server.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/transport.go b/pkg/p9/transport.go index bafb377de..ef59077ff 100644 --- a/pkg/p9/transport.go +++ b/pkg/p9/transport.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/transport_test.go b/pkg/p9/transport_test.go index b7b7825bd..c833d1c9c 100644 --- a/pkg/p9/transport_test.go +++ b/pkg/p9/transport_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/version.go b/pkg/p9/version.go index ceb6fabbf..a36a499a1 100644 --- a/pkg/p9/version.go +++ b/pkg/p9/version.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/p9/version_test.go b/pkg/p9/version_test.go index c053614c9..291e8580e 100644 --- a/pkg/p9/version_test.go +++ b/pkg/p9/version_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/rand/rand.go b/pkg/rand/rand.go index 593a14380..a2714784d 100644 --- a/pkg/rand/rand.go +++ b/pkg/rand/rand.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/rand/rand_linux.go b/pkg/rand/rand_linux.go index 7ebe8f3b0..2b92db3e6 100644 --- a/pkg/rand/rand_linux.go +++ b/pkg/rand/rand_linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/refs/refcounter.go b/pkg/refs/refcounter.go index 8f08c74c7..20f515391 100644 --- a/pkg/refs/refcounter.go +++ b/pkg/refs/refcounter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/refs/refcounter_state.go b/pkg/refs/refcounter_state.go index 136f06fbf..7c99fd2b5 100644 --- a/pkg/refs/refcounter_state.go +++ b/pkg/refs/refcounter_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/refs/refcounter_test.go b/pkg/refs/refcounter_test.go index abaa87453..ffd3d3f07 100644 --- a/pkg/refs/refcounter_test.go +++ b/pkg/refs/refcounter_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp.go b/pkg/seccomp/seccomp.go index e113f3574..50c9409e4 100644 --- a/pkg/seccomp/seccomp.go +++ b/pkg/seccomp/seccomp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp_rules.go b/pkg/seccomp/seccomp_rules.go index a9278c64b..29eec8db1 100644 --- a/pkg/seccomp/seccomp_rules.go +++ b/pkg/seccomp/seccomp_rules.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp_test.go b/pkg/seccomp/seccomp_test.go index 11ed90eb4..47ecac6f7 100644 --- a/pkg/seccomp/seccomp_test.go +++ b/pkg/seccomp/seccomp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp_test_victim.go b/pkg/seccomp/seccomp_test_victim.go index dd5ed0041..afc2f755f 100644 --- a/pkg/seccomp/seccomp_test_victim.go +++ b/pkg/seccomp/seccomp_test_victim.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/seccomp/seccomp_unsafe.go b/pkg/seccomp/seccomp_unsafe.go index a31c6471d..ccd40d9db 100644 --- a/pkg/seccomp/seccomp_unsafe.go +++ b/pkg/seccomp/seccomp_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/secio/full_reader.go b/pkg/secio/full_reader.go index 90b1772a7..aed2564bd 100644 --- a/pkg/secio/full_reader.go +++ b/pkg/secio/full_reader.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/secio/secio.go b/pkg/secio/secio.go index e5f74a497..b43226035 100644 --- a/pkg/secio/secio.go +++ b/pkg/secio/secio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/secio/secio_test.go b/pkg/secio/secio_test.go index 8304c4f74..d1d905187 100644 --- a/pkg/secio/secio_test.go +++ b/pkg/secio/secio_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/range.go b/pkg/segment/range.go index 057bcd7ff..4d4aeffef 100644 --- a/pkg/segment/range.go +++ b/pkg/segment/range.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/set.go b/pkg/segment/set.go index 74a916ea3..982eb3fdd 100644 --- a/pkg/segment/set.go +++ b/pkg/segment/set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/set_state.go b/pkg/segment/set_state.go index b86e1b75f..76de92591 100644 --- a/pkg/segment/set_state.go +++ b/pkg/segment/set_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/test/segment_test.go b/pkg/segment/test/segment_test.go index 0825105db..f19a005f3 100644 --- a/pkg/segment/test/segment_test.go +++ b/pkg/segment/test/segment_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/segment/test/set_functions.go b/pkg/segment/test/set_functions.go index 41f649011..bcddb39bb 100644 --- a/pkg/segment/test/set_functions.go +++ b/pkg/segment/test/set_functions.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/aligned.go b/pkg/sentry/arch/aligned.go index c88c034f6..df01a903d 100644 --- a/pkg/sentry/arch/aligned.go +++ b/pkg/sentry/arch/aligned.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch.go b/pkg/sentry/arch/arch.go index 16d8eb2b2..53f0c9018 100644 --- a/pkg/sentry/arch/arch.go +++ b/pkg/sentry/arch/arch.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch_amd64.go b/pkg/sentry/arch/arch_amd64.go index 7ec2f2c84..135c2ee1f 100644 --- a/pkg/sentry/arch/arch_amd64.go +++ b/pkg/sentry/arch/arch_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch_amd64.s b/pkg/sentry/arch/arch_amd64.s index fa9857df7..bd61402cf 100644 --- a/pkg/sentry/arch/arch_amd64.s +++ b/pkg/sentry/arch/arch_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch_state_x86.go b/pkg/sentry/arch/arch_state_x86.go index 01949049d..bb52d8db0 100644 --- a/pkg/sentry/arch/arch_state_x86.go +++ b/pkg/sentry/arch/arch_state_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/arch_x86.go b/pkg/sentry/arch/arch_x86.go index 4305fe2cb..4d167ce98 100644 --- a/pkg/sentry/arch/arch_x86.go +++ b/pkg/sentry/arch/arch_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/auxv.go b/pkg/sentry/arch/auxv.go index 5df65a691..80c923103 100644 --- a/pkg/sentry/arch/auxv.go +++ b/pkg/sentry/arch/auxv.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/registers.proto b/pkg/sentry/arch/registers.proto index f4c2f7043..9dc83e241 100644 --- a/pkg/sentry/arch/registers.proto +++ b/pkg/sentry/arch/registers.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/signal_act.go b/pkg/sentry/arch/signal_act.go index ad098c746..f9ca2e74e 100644 --- a/pkg/sentry/arch/signal_act.go +++ b/pkg/sentry/arch/signal_act.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/signal_amd64.go b/pkg/sentry/arch/signal_amd64.go index 7f76eba27..aa030fd70 100644 --- a/pkg/sentry/arch/signal_amd64.go +++ b/pkg/sentry/arch/signal_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/signal_info.go b/pkg/sentry/arch/signal_info.go index fa0ecbec5..f93ee8b46 100644 --- a/pkg/sentry/arch/signal_info.go +++ b/pkg/sentry/arch/signal_info.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/signal_stack.go b/pkg/sentry/arch/signal_stack.go index c02ae3b7c..a442f9fdc 100644 --- a/pkg/sentry/arch/signal_stack.go +++ b/pkg/sentry/arch/signal_stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/stack.go b/pkg/sentry/arch/stack.go index 2e33ccdf5..7e6324e82 100644 --- a/pkg/sentry/arch/stack.go +++ b/pkg/sentry/arch/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/arch/syscalls_amd64.go b/pkg/sentry/arch/syscalls_amd64.go index 47c31d4b9..8b4f23007 100644 --- a/pkg/sentry/arch/syscalls_amd64.go +++ b/pkg/sentry/arch/syscalls_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/context/context.go b/pkg/sentry/context/context.go index eefc3e1b4..d70f3a5c3 100644 --- a/pkg/sentry/context/context.go +++ b/pkg/sentry/context/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/context/contexttest/contexttest.go b/pkg/sentry/context/contexttest/contexttest.go index a29087775..a42038711 100644 --- a/pkg/sentry/context/contexttest/contexttest.go +++ b/pkg/sentry/context/contexttest/contexttest.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/control.go b/pkg/sentry/control/control.go index 32d30b6ea..6060b9b4f 100644 --- a/pkg/sentry/control/control.go +++ b/pkg/sentry/control/control.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/pprof.go b/pkg/sentry/control/pprof.go index 1af092af3..94ed149f2 100644 --- a/pkg/sentry/control/pprof.go +++ b/pkg/sentry/control/pprof.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/proc.go b/pkg/sentry/control/proc.go index aca2267a7..f7f02a3e1 100644 --- a/pkg/sentry/control/proc.go +++ b/pkg/sentry/control/proc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/proc_test.go b/pkg/sentry/control/proc_test.go index 5d52cd829..b7895d03c 100644 --- a/pkg/sentry/control/proc_test.go +++ b/pkg/sentry/control/proc_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/control/state.go b/pkg/sentry/control/state.go index b6bbf69fa..11efcaba1 100644 --- a/pkg/sentry/control/state.go +++ b/pkg/sentry/control/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/device/device.go b/pkg/sentry/device/device.go index ae4fa1d93..458d03b30 100644 --- a/pkg/sentry/device/device.go +++ b/pkg/sentry/device/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/device/device_test.go b/pkg/sentry/device/device_test.go index 5d8805c2f..e3f51ce4f 100644 --- a/pkg/sentry/device/device_test.go +++ b/pkg/sentry/device/device_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/anon/anon.go b/pkg/sentry/fs/anon/anon.go index a5e8c4f0d..a6ea8b9e7 100644 --- a/pkg/sentry/fs/anon/anon.go +++ b/pkg/sentry/fs/anon/anon.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/anon/device.go b/pkg/sentry/fs/anon/device.go index 2d1249299..5927bd11e 100644 --- a/pkg/sentry/fs/anon/device.go +++ b/pkg/sentry/fs/anon/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index 1f61c5711..b53746519 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ashmem/device.go b/pkg/sentry/fs/ashmem/device.go index 5369d1b0d..5e005bc2e 100644 --- a/pkg/sentry/fs/ashmem/device.go +++ b/pkg/sentry/fs/ashmem/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ashmem/pin_board.go b/pkg/sentry/fs/ashmem/pin_board.go index 7c997f533..bdf23b371 100644 --- a/pkg/sentry/fs/ashmem/pin_board.go +++ b/pkg/sentry/fs/ashmem/pin_board.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ashmem/pin_board_test.go b/pkg/sentry/fs/ashmem/pin_board_test.go index 736e628dc..24f5d86d6 100644 --- a/pkg/sentry/fs/ashmem/pin_board_test.go +++ b/pkg/sentry/fs/ashmem/pin_board_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/attr.go b/pkg/sentry/fs/attr.go index 3523b068a..591e35e6a 100644 --- a/pkg/sentry/fs/attr.go +++ b/pkg/sentry/fs/attr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index d9f1559de..acbbd5466 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/context.go b/pkg/sentry/fs/context.go index 4869428a8..c80ea0175 100644 --- a/pkg/sentry/fs/context.go +++ b/pkg/sentry/fs/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/copy_up.go b/pkg/sentry/fs/copy_up.go index ba69e718d..ee2d3d115 100644 --- a/pkg/sentry/fs/copy_up.go +++ b/pkg/sentry/fs/copy_up.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/copy_up_test.go b/pkg/sentry/fs/copy_up_test.go index 98a0b7638..54810afca 100644 --- a/pkg/sentry/fs/copy_up_test.go +++ b/pkg/sentry/fs/copy_up_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dentry.go b/pkg/sentry/fs/dentry.go index 29fb155a4..fe656cc24 100644 --- a/pkg/sentry/fs/dentry.go +++ b/pkg/sentry/fs/dentry.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/dev.go b/pkg/sentry/fs/dev/dev.go index fbc750a71..34ac01173 100644 --- a/pkg/sentry/fs/dev/dev.go +++ b/pkg/sentry/fs/dev/dev.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/device.go b/pkg/sentry/fs/dev/device.go index 3cecdf6e2..9f4e41fc9 100644 --- a/pkg/sentry/fs/dev/device.go +++ b/pkg/sentry/fs/dev/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/fs.go b/pkg/sentry/fs/dev/fs.go index cf4e7d00f..6096a40f8 100644 --- a/pkg/sentry/fs/dev/fs.go +++ b/pkg/sentry/fs/dev/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index 82da9aae9..6b11afa44 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 5d306d352..069212b6d 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index ffd5cf6c3..de0f3e5e5 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent.go b/pkg/sentry/fs/dirent.go index 54fc11fe1..c0bc261a2 100644 --- a/pkg/sentry/fs/dirent.go +++ b/pkg/sentry/fs/dirent.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_cache.go b/pkg/sentry/fs/dirent_cache.go index d26a06971..71f2d11de 100644 --- a/pkg/sentry/fs/dirent_cache.go +++ b/pkg/sentry/fs/dirent_cache.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_cache_limiter.go b/pkg/sentry/fs/dirent_cache_limiter.go index 024c7b2d5..ebb80bd50 100644 --- a/pkg/sentry/fs/dirent_cache_limiter.go +++ b/pkg/sentry/fs/dirent_cache_limiter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_cache_test.go b/pkg/sentry/fs/dirent_cache_test.go index 93e8d415f..395c879f5 100644 --- a/pkg/sentry/fs/dirent_cache_test.go +++ b/pkg/sentry/fs/dirent_cache_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_refs_test.go b/pkg/sentry/fs/dirent_refs_test.go index 325404e27..db88d850e 100644 --- a/pkg/sentry/fs/dirent_refs_test.go +++ b/pkg/sentry/fs/dirent_refs_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/dirent_state.go b/pkg/sentry/fs/dirent_state.go index 5cf151dab..18652b809 100644 --- a/pkg/sentry/fs/dirent_state.go +++ b/pkg/sentry/fs/dirent_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index 98483ab68..95e66ea8d 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe_opener.go b/pkg/sentry/fs/fdpipe/pipe_opener.go index 92ab6ff0e..0cabe2e18 100644 --- a/pkg/sentry/fs/fdpipe/pipe_opener.go +++ b/pkg/sentry/fs/fdpipe/pipe_opener.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe_opener_test.go b/pkg/sentry/fs/fdpipe/pipe_opener_test.go index 69516e048..8c8b1b40c 100644 --- a/pkg/sentry/fs/fdpipe/pipe_opener_test.go +++ b/pkg/sentry/fs/fdpipe/pipe_opener_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe_state.go b/pkg/sentry/fs/fdpipe/pipe_state.go index 4395666ad..8b347aa11 100644 --- a/pkg/sentry/fs/fdpipe/pipe_state.go +++ b/pkg/sentry/fs/fdpipe/pipe_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fdpipe/pipe_test.go b/pkg/sentry/fs/fdpipe/pipe_test.go index 7e3ee5257..b59a6aa0e 100644 --- a/pkg/sentry/fs/fdpipe/pipe_test.go +++ b/pkg/sentry/fs/fdpipe/pipe_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 5d5026661..62b35dabc 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_operations.go b/pkg/sentry/fs/file_operations.go index e0fa5135f..ab0acb6eb 100644 --- a/pkg/sentry/fs/file_operations.go +++ b/pkg/sentry/fs/file_operations.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index 6e680f0a4..948ce9c6f 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_overlay_test.go b/pkg/sentry/fs/file_overlay_test.go index a4ac58763..6a2b8007c 100644 --- a/pkg/sentry/fs/file_overlay_test.go +++ b/pkg/sentry/fs/file_overlay_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_state.go b/pkg/sentry/fs/file_state.go index 1c3bae3e8..523182d59 100644 --- a/pkg/sentry/fs/file_state.go +++ b/pkg/sentry/fs/file_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/file_test.go b/pkg/sentry/fs/file_test.go index f3ed9a70b..d867a0257 100644 --- a/pkg/sentry/fs/file_test.go +++ b/pkg/sentry/fs/file_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/filesystems.go b/pkg/sentry/fs/filesystems.go index a6b27c402..acd84dfcc 100644 --- a/pkg/sentry/fs/filesystems.go +++ b/pkg/sentry/fs/filesystems.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/filetest/filetest.go b/pkg/sentry/fs/filetest/filetest.go index 388a1ce36..f6b827800 100644 --- a/pkg/sentry/fs/filetest/filetest.go +++ b/pkg/sentry/fs/filetest/filetest.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/flags.go b/pkg/sentry/fs/flags.go index bf2a20b33..5c8cb773f 100644 --- a/pkg/sentry/fs/flags.go +++ b/pkg/sentry/fs/flags.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fs.go b/pkg/sentry/fs/fs.go index 119689776..632055cce 100644 --- a/pkg/sentry/fs/fs.go +++ b/pkg/sentry/fs/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/dirty_set.go b/pkg/sentry/fs/fsutil/dirty_set.go index 5add16ac4..9cd196d7d 100644 --- a/pkg/sentry/fs/fsutil/dirty_set.go +++ b/pkg/sentry/fs/fsutil/dirty_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/dirty_set_test.go b/pkg/sentry/fs/fsutil/dirty_set_test.go index f5c9d9215..d9c68baa3 100644 --- a/pkg/sentry/fs/fsutil/dirty_set_test.go +++ b/pkg/sentry/fs/fsutil/dirty_set_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/file.go b/pkg/sentry/fs/fsutil/file.go index 42afdd11c..e355d8594 100644 --- a/pkg/sentry/fs/fsutil/file.go +++ b/pkg/sentry/fs/fsutil/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/file_range_set.go b/pkg/sentry/fs/fsutil/file_range_set.go index 32ebf64ff..b5ac6c71c 100644 --- a/pkg/sentry/fs/fsutil/file_range_set.go +++ b/pkg/sentry/fs/fsutil/file_range_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/frame_ref_set.go b/pkg/sentry/fs/fsutil/frame_ref_set.go index b6e783614..6565c28c8 100644 --- a/pkg/sentry/fs/fsutil/frame_ref_set.go +++ b/pkg/sentry/fs/fsutil/frame_ref_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/fsutil.go b/pkg/sentry/fs/fsutil/fsutil.go index 319c4841b..c9587b1d9 100644 --- a/pkg/sentry/fs/fsutil/fsutil.go +++ b/pkg/sentry/fs/fsutil/fsutil.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/host_file_mapper.go b/pkg/sentry/fs/fsutil/host_file_mapper.go index 9599665f0..2bdfc0db6 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/host_file_mapper_state.go b/pkg/sentry/fs/fsutil/host_file_mapper_state.go index bbd15b30b..576d2a3df 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper_state.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go b/pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go index 86df76822..7167be263 100644 --- a/pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go +++ b/pkg/sentry/fs/fsutil/host_file_mapper_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/host_mappable.go b/pkg/sentry/fs/fsutil/host_mappable.go index 4a182baa1..28686f3b3 100644 --- a/pkg/sentry/fs/fsutil/host_mappable.go +++ b/pkg/sentry/fs/fsutil/host_mappable.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index 468171a9b..b6366d906 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index ba33b9912..919d2534c 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/fsutil/inode_cached_test.go b/pkg/sentry/fs/fsutil/inode_cached_test.go index 2a8a1639c..661ec41f6 100644 --- a/pkg/sentry/fs/fsutil/inode_cached_test.go +++ b/pkg/sentry/fs/fsutil/inode_cached_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/attr.go b/pkg/sentry/fs/gofer/attr.go index 98700d014..c572f3396 100644 --- a/pkg/sentry/fs/gofer/attr.go +++ b/pkg/sentry/fs/gofer/attr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/cache_policy.go b/pkg/sentry/fs/gofer/cache_policy.go index 51c573aef..35cd0c1d6 100644 --- a/pkg/sentry/fs/gofer/cache_policy.go +++ b/pkg/sentry/fs/gofer/cache_policy.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/context_file.go b/pkg/sentry/fs/gofer/context_file.go index 455953237..d512afefc 100644 --- a/pkg/sentry/fs/gofer/context_file.go +++ b/pkg/sentry/fs/gofer/context_file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/device.go b/pkg/sentry/fs/gofer/device.go index 52c5acf48..1de6c247c 100644 --- a/pkg/sentry/fs/gofer/device.go +++ b/pkg/sentry/fs/gofer/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index 35caa42cd..bc2be546e 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/file_state.go b/pkg/sentry/fs/gofer/file_state.go index d0c64003c..31264e065 100644 --- a/pkg/sentry/fs/gofer/file_state.go +++ b/pkg/sentry/fs/gofer/file_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/fs.go b/pkg/sentry/fs/gofer/fs.go index adff0abac..6ab89fcc2 100644 --- a/pkg/sentry/fs/gofer/fs.go +++ b/pkg/sentry/fs/gofer/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/gofer_test.go b/pkg/sentry/fs/gofer/gofer_test.go index 36201f017..29d34da7e 100644 --- a/pkg/sentry/fs/gofer/gofer_test.go +++ b/pkg/sentry/fs/gofer/gofer_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/handles.go b/pkg/sentry/fs/gofer/handles.go index 0b33e80c3..c7098cd36 100644 --- a/pkg/sentry/fs/gofer/handles.go +++ b/pkg/sentry/fs/gofer/handles.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go index 1181a24cc..f6f20844d 100644 --- a/pkg/sentry/fs/gofer/inode.go +++ b/pkg/sentry/fs/gofer/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/inode_state.go b/pkg/sentry/fs/gofer/inode_state.go index 44d76ba9f..ac22ee4b1 100644 --- a/pkg/sentry/fs/gofer/inode_state.go +++ b/pkg/sentry/fs/gofer/inode_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/path.go b/pkg/sentry/fs/gofer/path.go index 8ae33d286..4cbf9e9d9 100644 --- a/pkg/sentry/fs/gofer/path.go +++ b/pkg/sentry/fs/gofer/path.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/session.go b/pkg/sentry/fs/gofer/session.go index 4ed688ce5..4cb65e7c6 100644 --- a/pkg/sentry/fs/gofer/session.go +++ b/pkg/sentry/fs/gofer/session.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/session_state.go b/pkg/sentry/fs/gofer/session_state.go index b1f299be5..68fbf3417 100644 --- a/pkg/sentry/fs/gofer/session_state.go +++ b/pkg/sentry/fs/gofer/session_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/socket.go b/pkg/sentry/fs/gofer/socket.go index ce6d3d5c3..cbd5b9a84 100644 --- a/pkg/sentry/fs/gofer/socket.go +++ b/pkg/sentry/fs/gofer/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/gofer/util.go b/pkg/sentry/fs/gofer/util.go index 1a759370d..d0e1096ce 100644 --- a/pkg/sentry/fs/gofer/util.go +++ b/pkg/sentry/fs/gofer/util.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/control.go b/pkg/sentry/fs/host/control.go index 0753640a2..480f0c8f4 100644 --- a/pkg/sentry/fs/host/control.go +++ b/pkg/sentry/fs/host/control.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/descriptor.go b/pkg/sentry/fs/host/descriptor.go index 554e1693a..ffcd57a94 100644 --- a/pkg/sentry/fs/host/descriptor.go +++ b/pkg/sentry/fs/host/descriptor.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/descriptor_state.go b/pkg/sentry/fs/host/descriptor_state.go index 530c0109f..8167390a9 100644 --- a/pkg/sentry/fs/host/descriptor_state.go +++ b/pkg/sentry/fs/host/descriptor_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/descriptor_test.go b/pkg/sentry/fs/host/descriptor_test.go index 5dec84ab2..ff08e43af 100644 --- a/pkg/sentry/fs/host/descriptor_test.go +++ b/pkg/sentry/fs/host/descriptor_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/device.go b/pkg/sentry/fs/host/device.go index b5adedf44..055024c44 100644 --- a/pkg/sentry/fs/host/device.go +++ b/pkg/sentry/fs/host/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index 2a8f285ff..82e2ae3b9 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/fs.go b/pkg/sentry/fs/host/fs.go index de349a41a..b1b8dc0b6 100644 --- a/pkg/sentry/fs/host/fs.go +++ b/pkg/sentry/fs/host/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/fs_test.go b/pkg/sentry/fs/host/fs_test.go index c83b29a16..16c89ddf1 100644 --- a/pkg/sentry/fs/host/fs_test.go +++ b/pkg/sentry/fs/host/fs_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index 69c648f67..20e077f77 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/inode_state.go b/pkg/sentry/fs/host/inode_state.go index b7c1a9581..26cc755bc 100644 --- a/pkg/sentry/fs/host/inode_state.go +++ b/pkg/sentry/fs/host/inode_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/inode_test.go b/pkg/sentry/fs/host/inode_test.go index 9f1561bd5..ad1878b5a 100644 --- a/pkg/sentry/fs/host/inode_test.go +++ b/pkg/sentry/fs/host/inode_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/ioctl_unsafe.go b/pkg/sentry/fs/host/ioctl_unsafe.go index 175dca613..b5a85c4d9 100644 --- a/pkg/sentry/fs/host/ioctl_unsafe.go +++ b/pkg/sentry/fs/host/ioctl_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket.go b/pkg/sentry/fs/host/socket.go index be2c3581f..3034e9441 100644 --- a/pkg/sentry/fs/host/socket.go +++ b/pkg/sentry/fs/host/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket_iovec.go b/pkg/sentry/fs/host/socket_iovec.go index d4ce4a8c1..5efbb3ae8 100644 --- a/pkg/sentry/fs/host/socket_iovec.go +++ b/pkg/sentry/fs/host/socket_iovec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket_state.go b/pkg/sentry/fs/host/socket_state.go index 2932c1f16..5676c451a 100644 --- a/pkg/sentry/fs/host/socket_state.go +++ b/pkg/sentry/fs/host/socket_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket_test.go b/pkg/sentry/fs/host/socket_test.go index 83e8e1b3c..cc760a7e1 100644 --- a/pkg/sentry/fs/host/socket_test.go +++ b/pkg/sentry/fs/host/socket_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/socket_unsafe.go b/pkg/sentry/fs/host/socket_unsafe.go index f35e2492d..8873705c0 100644 --- a/pkg/sentry/fs/host/socket_unsafe.go +++ b/pkg/sentry/fs/host/socket_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/tty.go b/pkg/sentry/fs/host/tty.go index c5cb75df7..e45b339f5 100644 --- a/pkg/sentry/fs/host/tty.go +++ b/pkg/sentry/fs/host/tty.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/util.go b/pkg/sentry/fs/host/util.go index 40c450660..94ff7708e 100644 --- a/pkg/sentry/fs/host/util.go +++ b/pkg/sentry/fs/host/util.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/util_unsafe.go b/pkg/sentry/fs/host/util_unsafe.go index a8721d197..b95a57c3f 100644 --- a/pkg/sentry/fs/host/util_unsafe.go +++ b/pkg/sentry/fs/host/util_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/host/wait_test.go b/pkg/sentry/fs/host/wait_test.go index 9ca8c399f..afcb74724 100644 --- a/pkg/sentry/fs/host/wait_test.go +++ b/pkg/sentry/fs/host/wait_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index fe411a766..d764ef93d 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode_inotify.go b/pkg/sentry/fs/inode_inotify.go index d2b653bc7..0f2a66a79 100644 --- a/pkg/sentry/fs/inode_inotify.go +++ b/pkg/sentry/fs/inode_inotify.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode_operations.go b/pkg/sentry/fs/inode_operations.go index ff8b75f31..ac287e1e4 100644 --- a/pkg/sentry/fs/inode_operations.go +++ b/pkg/sentry/fs/inode_operations.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode_overlay.go b/pkg/sentry/fs/inode_overlay.go index bda3e1861..3d015328e 100644 --- a/pkg/sentry/fs/inode_overlay.go +++ b/pkg/sentry/fs/inode_overlay.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inode_overlay_test.go b/pkg/sentry/fs/inode_overlay_test.go index fa8accf6c..66b3da2d0 100644 --- a/pkg/sentry/fs/inode_overlay_test.go +++ b/pkg/sentry/fs/inode_overlay_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 59fa662f3..2652582c3 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inotify_event.go b/pkg/sentry/fs/inotify_event.go index f09928b68..d52f956e4 100644 --- a/pkg/sentry/fs/inotify_event.go +++ b/pkg/sentry/fs/inotify_event.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/inotify_watch.go b/pkg/sentry/fs/inotify_watch.go index d33e7e498..a0b488467 100644 --- a/pkg/sentry/fs/inotify_watch.go +++ b/pkg/sentry/fs/inotify_watch.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/lock/lock.go b/pkg/sentry/fs/lock/lock.go index 5ff800d2d..f2aee4512 100644 --- a/pkg/sentry/fs/lock/lock.go +++ b/pkg/sentry/fs/lock/lock.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/lock/lock_range_test.go b/pkg/sentry/fs/lock/lock_range_test.go index b0ab882b9..6221199d1 100644 --- a/pkg/sentry/fs/lock/lock_range_test.go +++ b/pkg/sentry/fs/lock/lock_range_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/lock/lock_set_functions.go b/pkg/sentry/fs/lock/lock_set_functions.go index 395592a4b..8a3ace0c1 100644 --- a/pkg/sentry/fs/lock/lock_set_functions.go +++ b/pkg/sentry/fs/lock/lock_set_functions.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/lock/lock_test.go b/pkg/sentry/fs/lock/lock_test.go index 67fa4b1dd..ba002aeb7 100644 --- a/pkg/sentry/fs/lock/lock_test.go +++ b/pkg/sentry/fs/lock/lock_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mock.go b/pkg/sentry/fs/mock.go index 118e30f63..cf359a1f1 100644 --- a/pkg/sentry/fs/mock.go +++ b/pkg/sentry/fs/mock.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mount.go b/pkg/sentry/fs/mount.go index 4d1693204..a169ea4c9 100644 --- a/pkg/sentry/fs/mount.go +++ b/pkg/sentry/fs/mount.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mount_overlay.go b/pkg/sentry/fs/mount_overlay.go index fb60a1aec..535f812c8 100644 --- a/pkg/sentry/fs/mount_overlay.go +++ b/pkg/sentry/fs/mount_overlay.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mount_test.go b/pkg/sentry/fs/mount_test.go index d7605b2c9..9f7fbeff2 100644 --- a/pkg/sentry/fs/mount_test.go +++ b/pkg/sentry/fs/mount_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mounts.go b/pkg/sentry/fs/mounts.go index f6f7be0aa..01eb4607e 100644 --- a/pkg/sentry/fs/mounts.go +++ b/pkg/sentry/fs/mounts.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/mounts_test.go b/pkg/sentry/fs/mounts_test.go index 54000614f..56d726dd1 100644 --- a/pkg/sentry/fs/mounts_test.go +++ b/pkg/sentry/fs/mounts_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/offset.go b/pkg/sentry/fs/offset.go index 38aee765a..3f68da149 100644 --- a/pkg/sentry/fs/offset.go +++ b/pkg/sentry/fs/offset.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/overlay.go b/pkg/sentry/fs/overlay.go index f3e2d5cbe..db89a5f70 100644 --- a/pkg/sentry/fs/overlay.go +++ b/pkg/sentry/fs/overlay.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/path.go b/pkg/sentry/fs/path.go index 52139b648..e4dc02dbb 100644 --- a/pkg/sentry/fs/path.go +++ b/pkg/sentry/fs/path.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/path_test.go b/pkg/sentry/fs/path_test.go index 4ba1498f6..e6f57ebba 100644 --- a/pkg/sentry/fs/path_test.go +++ b/pkg/sentry/fs/path_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/cpuinfo.go b/pkg/sentry/fs/proc/cpuinfo.go index f756c45bf..15031234e 100644 --- a/pkg/sentry/fs/proc/cpuinfo.go +++ b/pkg/sentry/fs/proc/cpuinfo.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/device/device.go b/pkg/sentry/fs/proc/device/device.go index 04b687bcf..0de466c73 100644 --- a/pkg/sentry/fs/proc/device/device.go +++ b/pkg/sentry/fs/proc/device/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/exec_args.go b/pkg/sentry/fs/proc/exec_args.go index fc21dfbbd..d49dad685 100644 --- a/pkg/sentry/fs/proc/exec_args.go +++ b/pkg/sentry/fs/proc/exec_args.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/fds.go b/pkg/sentry/fs/proc/fds.go index f2329e623..744b31c74 100644 --- a/pkg/sentry/fs/proc/fds.go +++ b/pkg/sentry/fs/proc/fds.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/filesystems.go b/pkg/sentry/fs/proc/filesystems.go index c050a00be..7bb081d0e 100644 --- a/pkg/sentry/fs/proc/filesystems.go +++ b/pkg/sentry/fs/proc/filesystems.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/fs.go b/pkg/sentry/fs/proc/fs.go index 666a2d054..7c5f8484a 100644 --- a/pkg/sentry/fs/proc/fs.go +++ b/pkg/sentry/fs/proc/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/inode.go b/pkg/sentry/fs/proc/inode.go index 8dde2ea46..b03807043 100644 --- a/pkg/sentry/fs/proc/inode.go +++ b/pkg/sentry/fs/proc/inode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/loadavg.go b/pkg/sentry/fs/proc/loadavg.go index 3ee0e570a..2dfe7089a 100644 --- a/pkg/sentry/fs/proc/loadavg.go +++ b/pkg/sentry/fs/proc/loadavg.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/meminfo.go b/pkg/sentry/fs/proc/meminfo.go index 75cbf3e77..d2b9b92c7 100644 --- a/pkg/sentry/fs/proc/meminfo.go +++ b/pkg/sentry/fs/proc/meminfo.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/mounts.go b/pkg/sentry/fs/proc/mounts.go index fe62b167b..37ed30724 100644 --- a/pkg/sentry/fs/proc/mounts.go +++ b/pkg/sentry/fs/proc/mounts.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/net.go b/pkg/sentry/fs/proc/net.go index d24b2d370..4a107c739 100644 --- a/pkg/sentry/fs/proc/net.go +++ b/pkg/sentry/fs/proc/net.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/net_test.go b/pkg/sentry/fs/proc/net_test.go index 94677cc1d..9aed5fdca 100644 --- a/pkg/sentry/fs/proc/net_test.go +++ b/pkg/sentry/fs/proc/net_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/proc.go b/pkg/sentry/fs/proc/proc.go index 64e1e1998..196fa5128 100644 --- a/pkg/sentry/fs/proc/proc.go +++ b/pkg/sentry/fs/proc/proc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/rpcinet_proc.go b/pkg/sentry/fs/proc/rpcinet_proc.go index 81f64a28b..db53686f6 100644 --- a/pkg/sentry/fs/proc/rpcinet_proc.go +++ b/pkg/sentry/fs/proc/rpcinet_proc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index 0a0eb45e2..10ea1f55d 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/seqfile/seqfile_test.go b/pkg/sentry/fs/proc/seqfile/seqfile_test.go index 35403ab7f..c4de565eb 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile_test.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/stat.go b/pkg/sentry/fs/proc/stat.go index 18bd8e9b6..397f9ec6b 100644 --- a/pkg/sentry/fs/proc/stat.go +++ b/pkg/sentry/fs/proc/stat.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index a7bc9198e..b889ed625 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index 0ce77f04f..e49794a48 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/sys_net_state.go b/pkg/sentry/fs/proc/sys_net_state.go index 5f481a1cf..6eba709c6 100644 --- a/pkg/sentry/fs/proc/sys_net_state.go +++ b/pkg/sentry/fs/proc/sys_net_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/sys_net_test.go b/pkg/sentry/fs/proc/sys_net_test.go index ea0d94fce..78135ba13 100644 --- a/pkg/sentry/fs/proc/sys_net_test.go +++ b/pkg/sentry/fs/proc/sys_net_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 9f65a8337..0f400e80f 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index d433632cf..d649da0f1 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/uptime.go b/pkg/sentry/fs/proc/uptime.go index d7ae26fcf..1ddf9fafa 100644 --- a/pkg/sentry/fs/proc/uptime.go +++ b/pkg/sentry/fs/proc/uptime.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/proc/version.go b/pkg/sentry/fs/proc/version.go index 58e0c793c..a5479990c 100644 --- a/pkg/sentry/fs/proc/version.go +++ b/pkg/sentry/fs/proc/version.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index c0400b67d..a6b6a5c33 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index 5bcb6c364..9406a07ca 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index 35dabdad2..f7835fe05 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/tree.go b/pkg/sentry/fs/ramfs/tree.go index c1ac8a78b..8c6b31f70 100644 --- a/pkg/sentry/fs/ramfs/tree.go +++ b/pkg/sentry/fs/ramfs/tree.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/ramfs/tree_test.go b/pkg/sentry/fs/ramfs/tree_test.go index 8bee9cfc1..27abeb6ba 100644 --- a/pkg/sentry/fs/ramfs/tree_test.go +++ b/pkg/sentry/fs/ramfs/tree_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/restore.go b/pkg/sentry/fs/restore.go index a6645b41e..f10168125 100644 --- a/pkg/sentry/fs/restore.go +++ b/pkg/sentry/fs/restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/save.go b/pkg/sentry/fs/save.go index 90988d385..2eaf6ab69 100644 --- a/pkg/sentry/fs/save.go +++ b/pkg/sentry/fs/save.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/seek.go b/pkg/sentry/fs/seek.go index 72f3fb632..0f43918ad 100644 --- a/pkg/sentry/fs/seek.go +++ b/pkg/sentry/fs/seek.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sync.go b/pkg/sentry/fs/sync.go index 6dcc2fe8d..1fff8059c 100644 --- a/pkg/sentry/fs/sync.go +++ b/pkg/sentry/fs/sync.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sys/device.go b/pkg/sentry/fs/sys/device.go index 38ecd0c18..128d3a9d9 100644 --- a/pkg/sentry/fs/sys/device.go +++ b/pkg/sentry/fs/sys/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sys/devices.go b/pkg/sentry/fs/sys/devices.go index 8b728a4e4..db91de435 100644 --- a/pkg/sentry/fs/sys/devices.go +++ b/pkg/sentry/fs/sys/devices.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sys/fs.go b/pkg/sentry/fs/sys/fs.go index 44ae43754..f0c2322e0 100644 --- a/pkg/sentry/fs/sys/fs.go +++ b/pkg/sentry/fs/sys/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/sys/sys.go b/pkg/sentry/fs/sys/sys.go index c5b56fe69..d20ef91fa 100644 --- a/pkg/sentry/fs/sys/sys.go +++ b/pkg/sentry/fs/sys/sys.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/timerfd/timerfd.go b/pkg/sentry/fs/timerfd/timerfd.go index ef9a08854..749961f51 100644 --- a/pkg/sentry/fs/timerfd/timerfd.go +++ b/pkg/sentry/fs/timerfd/timerfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/device.go b/pkg/sentry/fs/tmpfs/device.go index aade93c26..179c3a46f 100644 --- a/pkg/sentry/fs/tmpfs/device.go +++ b/pkg/sentry/fs/tmpfs/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/file_regular.go b/pkg/sentry/fs/tmpfs/file_regular.go index d0c9b8bea..1ef256511 100644 --- a/pkg/sentry/fs/tmpfs/file_regular.go +++ b/pkg/sentry/fs/tmpfs/file_regular.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/file_test.go b/pkg/sentry/fs/tmpfs/file_test.go index 743061190..b44c06556 100644 --- a/pkg/sentry/fs/tmpfs/file_test.go +++ b/pkg/sentry/fs/tmpfs/file_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/fs.go b/pkg/sentry/fs/tmpfs/fs.go index 8e44421b6..b7c29a4d1 100644 --- a/pkg/sentry/fs/tmpfs/fs.go +++ b/pkg/sentry/fs/tmpfs/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index 4450e1363..f89d86c83 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 5bb4922cb..832914453 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index f8713471a..0fc777e67 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/fs.go b/pkg/sentry/fs/tty/fs.go index a53448c47..701b2f7d9 100644 --- a/pkg/sentry/fs/tty/fs.go +++ b/pkg/sentry/fs/tty/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/line_discipline.go b/pkg/sentry/fs/tty/line_discipline.go index c4a364edb..20d29d130 100644 --- a/pkg/sentry/fs/tty/line_discipline.go +++ b/pkg/sentry/fs/tty/line_discipline.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index e2686a074..45e167e5f 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/queue.go b/pkg/sentry/fs/tty/queue.go index 5e88d84d9..11fb92be3 100644 --- a/pkg/sentry/fs/tty/queue.go +++ b/pkg/sentry/fs/tty/queue.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index ed080ca0f..0ae57a02c 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/terminal.go b/pkg/sentry/fs/tty/terminal.go index 79f9d76d7..2b4160ba5 100644 --- a/pkg/sentry/fs/tty/terminal.go +++ b/pkg/sentry/fs/tty/terminal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/fs/tty/tty_test.go b/pkg/sentry/fs/tty/tty_test.go index ad535838f..d2e75a511 100644 --- a/pkg/sentry/fs/tty/tty_test.go +++ b/pkg/sentry/fs/tty/tty_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/hostcpu/getcpu_amd64.s b/pkg/sentry/hostcpu/getcpu_amd64.s index 409db1450..aa00316da 100644 --- a/pkg/sentry/hostcpu/getcpu_amd64.s +++ b/pkg/sentry/hostcpu/getcpu_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/hostcpu/hostcpu.go b/pkg/sentry/hostcpu/hostcpu.go index 3adc847bb..d78f78402 100644 --- a/pkg/sentry/hostcpu/hostcpu.go +++ b/pkg/sentry/hostcpu/hostcpu.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/hostcpu/hostcpu_test.go b/pkg/sentry/hostcpu/hostcpu_test.go index 38de0e1f6..7d6885c9e 100644 --- a/pkg/sentry/hostcpu/hostcpu_test.go +++ b/pkg/sentry/hostcpu/hostcpu_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/inet/context.go b/pkg/sentry/inet/context.go index d05e96f15..8550c4793 100644 --- a/pkg/sentry/inet/context.go +++ b/pkg/sentry/inet/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/inet/inet.go b/pkg/sentry/inet/inet.go index 8206377cc..7c104fd47 100644 --- a/pkg/sentry/inet/inet.go +++ b/pkg/sentry/inet/inet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/inet/test_stack.go b/pkg/sentry/inet/test_stack.go index 05c1a1792..624371eb6 100644 --- a/pkg/sentry/inet/test_stack.go +++ b/pkg/sentry/inet/test_stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/abstract_socket_namespace.go b/pkg/sentry/kernel/abstract_socket_namespace.go index 1ea2cee36..5ce52e66c 100644 --- a/pkg/sentry/kernel/abstract_socket_namespace.go +++ b/pkg/sentry/kernel/abstract_socket_namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/auth.go b/pkg/sentry/kernel/auth/auth.go index 19f15fd36..847d121aa 100644 --- a/pkg/sentry/kernel/auth/auth.go +++ b/pkg/sentry/kernel/auth/auth.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/capability_set.go b/pkg/sentry/kernel/auth/capability_set.go index 88d6243aa..7a0c967cd 100644 --- a/pkg/sentry/kernel/auth/capability_set.go +++ b/pkg/sentry/kernel/auth/capability_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/context.go b/pkg/sentry/kernel/auth/context.go index f7e945599..16d110610 100644 --- a/pkg/sentry/kernel/auth/context.go +++ b/pkg/sentry/kernel/auth/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/credentials.go b/pkg/sentry/kernel/auth/credentials.go index 2055da196..1511a0324 100644 --- a/pkg/sentry/kernel/auth/credentials.go +++ b/pkg/sentry/kernel/auth/credentials.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/id.go b/pkg/sentry/kernel/auth/id.go index e5bed44d7..0a58ba17c 100644 --- a/pkg/sentry/kernel/auth/id.go +++ b/pkg/sentry/kernel/auth/id.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/id_map.go b/pkg/sentry/kernel/auth/id_map.go index 43f439825..e5d6028d6 100644 --- a/pkg/sentry/kernel/auth/id_map.go +++ b/pkg/sentry/kernel/auth/id_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/id_map_functions.go b/pkg/sentry/kernel/auth/id_map_functions.go index 8f1a189ec..432dbfb6d 100644 --- a/pkg/sentry/kernel/auth/id_map_functions.go +++ b/pkg/sentry/kernel/auth/id_map_functions.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/auth/user_namespace.go b/pkg/sentry/kernel/auth/user_namespace.go index 159940a69..a40dd668f 100644 --- a/pkg/sentry/kernel/auth/user_namespace.go +++ b/pkg/sentry/kernel/auth/user_namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/context.go b/pkg/sentry/kernel/context.go index b629521eb..a1a084eab 100644 --- a/pkg/sentry/kernel/context.go +++ b/pkg/sentry/kernel/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/contexttest/contexttest.go b/pkg/sentry/kernel/contexttest/contexttest.go index eb56a6a07..ae67e2a25 100644 --- a/pkg/sentry/kernel/contexttest/contexttest.go +++ b/pkg/sentry/kernel/contexttest/contexttest.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index befefb11c..2399ae6f2 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/epoll/epoll_state.go b/pkg/sentry/kernel/epoll/epoll_state.go index f6e3e4825..4c3c38f9e 100644 --- a/pkg/sentry/kernel/epoll/epoll_state.go +++ b/pkg/sentry/kernel/epoll/epoll_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/epoll/epoll_test.go b/pkg/sentry/kernel/epoll/epoll_test.go index d89c1b745..49b781b69 100644 --- a/pkg/sentry/kernel/epoll/epoll_test.go +++ b/pkg/sentry/kernel/epoll/epoll_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index b448ad813..5d3139eef 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/eventfd/eventfd_test.go b/pkg/sentry/kernel/eventfd/eventfd_test.go index 14e8996d9..1159638e5 100644 --- a/pkg/sentry/kernel/eventfd/eventfd_test.go +++ b/pkg/sentry/kernel/eventfd/eventfd_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/fasync/fasync.go b/pkg/sentry/kernel/fasync/fasync.go index 298d988ea..84cd08501 100644 --- a/pkg/sentry/kernel/fasync/fasync.go +++ b/pkg/sentry/kernel/fasync/fasync.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/fd_map.go b/pkg/sentry/kernel/fd_map.go index 715f4714d..c5636d233 100644 --- a/pkg/sentry/kernel/fd_map.go +++ b/pkg/sentry/kernel/fd_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/fd_map_test.go b/pkg/sentry/kernel/fd_map_test.go index 9e76f0a2d..22db4c7cf 100644 --- a/pkg/sentry/kernel/fd_map_test.go +++ b/pkg/sentry/kernel/fd_map_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/fs_context.go b/pkg/sentry/kernel/fs_context.go index 3cf0db280..d8115f59a 100644 --- a/pkg/sentry/kernel/fs_context.go +++ b/pkg/sentry/kernel/fs_context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/futex/futex.go b/pkg/sentry/kernel/futex/futex.go index cd7d51621..bb38eb81e 100644 --- a/pkg/sentry/kernel/futex/futex.go +++ b/pkg/sentry/kernel/futex/futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/futex/futex_test.go b/pkg/sentry/kernel/futex/futex_test.go index 9d44ee8e5..2de5239bf 100644 --- a/pkg/sentry/kernel/futex/futex_test.go +++ b/pkg/sentry/kernel/futex/futex_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/ipc_namespace.go b/pkg/sentry/kernel/ipc_namespace.go index 9ceb9bd92..ebe12812c 100644 --- a/pkg/sentry/kernel/ipc_namespace.go +++ b/pkg/sentry/kernel/ipc_namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/kdefs/kdefs.go b/pkg/sentry/kernel/kdefs/kdefs.go index 8eafe810b..304da2032 100644 --- a/pkg/sentry/kernel/kdefs/kdefs.go +++ b/pkg/sentry/kernel/kdefs/kdefs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index a1b2d7161..0468dd678 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/kernel_state.go b/pkg/sentry/kernel/kernel_state.go index aae6f9ad2..48c3ff5a9 100644 --- a/pkg/sentry/kernel/kernel_state.go +++ b/pkg/sentry/kernel/kernel_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/memevent/memory_events.go b/pkg/sentry/kernel/memevent/memory_events.go index d09d6debf..0e2cee807 100644 --- a/pkg/sentry/kernel/memevent/memory_events.go +++ b/pkg/sentry/kernel/memevent/memory_events.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/memevent/memory_events.proto b/pkg/sentry/kernel/memevent/memory_events.proto index 43b8deb76..bf8029ff5 100644 --- a/pkg/sentry/kernel/memevent/memory_events.proto +++ b/pkg/sentry/kernel/memevent/memory_events.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pending_signals.go b/pkg/sentry/kernel/pending_signals.go index deff6def9..c93f6598a 100644 --- a/pkg/sentry/kernel/pending_signals.go +++ b/pkg/sentry/kernel/pending_signals.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pending_signals_state.go b/pkg/sentry/kernel/pending_signals_state.go index 72be6702f..2c902c7e3 100644 --- a/pkg/sentry/kernel/pending_signals_state.go +++ b/pkg/sentry/kernel/pending_signals_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/buffers.go b/pkg/sentry/kernel/pipe/buffers.go index 54e059f8b..ba53fd482 100644 --- a/pkg/sentry/kernel/pipe/buffers.go +++ b/pkg/sentry/kernel/pipe/buffers.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/device.go b/pkg/sentry/kernel/pipe/device.go index eec5c5de8..eb59e15a1 100644 --- a/pkg/sentry/kernel/pipe/device.go +++ b/pkg/sentry/kernel/pipe/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index 1336b6293..99188dddf 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/node_test.go b/pkg/sentry/kernel/pipe/node_test.go index ad103b195..7ddecdad8 100644 --- a/pkg/sentry/kernel/pipe/node_test.go +++ b/pkg/sentry/kernel/pipe/node_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index 357d1162e..bd7649d2f 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/pipe_test.go b/pkg/sentry/kernel/pipe/pipe_test.go index 3b9895927..de340c40c 100644 --- a/pkg/sentry/kernel/pipe/pipe_test.go +++ b/pkg/sentry/kernel/pipe/pipe_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/reader.go b/pkg/sentry/kernel/pipe/reader.go index f27379969..48fab45d1 100644 --- a/pkg/sentry/kernel/pipe/reader.go +++ b/pkg/sentry/kernel/pipe/reader.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index 1090432d7..ddcc5e09a 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/pipe/writer.go b/pkg/sentry/kernel/pipe/writer.go index 6fea9769c..0f29fbc43 100644 --- a/pkg/sentry/kernel/pipe/writer.go +++ b/pkg/sentry/kernel/pipe/writer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/posixtimer.go b/pkg/sentry/kernel/posixtimer.go index 40b5acca3..a016b4087 100644 --- a/pkg/sentry/kernel/posixtimer.go +++ b/pkg/sentry/kernel/posixtimer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/ptrace.go b/pkg/sentry/kernel/ptrace.go index 15f2e2964..4423e7efd 100644 --- a/pkg/sentry/kernel/ptrace.go +++ b/pkg/sentry/kernel/ptrace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/ptrace_amd64.go b/pkg/sentry/kernel/ptrace_amd64.go index 1f88efca3..048eeaa3f 100644 --- a/pkg/sentry/kernel/ptrace_amd64.go +++ b/pkg/sentry/kernel/ptrace_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google Inc. +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/ptrace_arm64.go b/pkg/sentry/kernel/ptrace_arm64.go index 4636405e6..4899c813f 100644 --- a/pkg/sentry/kernel/ptrace_arm64.go +++ b/pkg/sentry/kernel/ptrace_arm64.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google Inc. +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/rseq.go b/pkg/sentry/kernel/rseq.go index 6d3314e81..c4fb2c56c 100644 --- a/pkg/sentry/kernel/rseq.go +++ b/pkg/sentry/kernel/rseq.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/sched/cpuset.go b/pkg/sentry/kernel/sched/cpuset.go index 41ac1067d..c6c436690 100644 --- a/pkg/sentry/kernel/sched/cpuset.go +++ b/pkg/sentry/kernel/sched/cpuset.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/sched/cpuset_test.go b/pkg/sentry/kernel/sched/cpuset_test.go index a036ed513..3af9f1197 100644 --- a/pkg/sentry/kernel/sched/cpuset_test.go +++ b/pkg/sentry/kernel/sched/cpuset_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/sched/sched.go b/pkg/sentry/kernel/sched/sched.go index e59909baf..de18c9d02 100644 --- a/pkg/sentry/kernel/sched/sched.go +++ b/pkg/sentry/kernel/sched/sched.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/seccomp.go b/pkg/sentry/kernel/seccomp.go index 4bed4d373..cc75eb08a 100644 --- a/pkg/sentry/kernel/seccomp.go +++ b/pkg/sentry/kernel/seccomp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/semaphore/semaphore.go b/pkg/sentry/kernel/semaphore/semaphore.go index 2b7c1a9bc..9d0620e02 100644 --- a/pkg/sentry/kernel/semaphore/semaphore.go +++ b/pkg/sentry/kernel/semaphore/semaphore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/semaphore/semaphore_test.go b/pkg/sentry/kernel/semaphore/semaphore_test.go index 2e51e6ee5..abfcd0fb4 100644 --- a/pkg/sentry/kernel/semaphore/semaphore_test.go +++ b/pkg/sentry/kernel/semaphore/semaphore_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/sessions.go b/pkg/sentry/kernel/sessions.go index 070c2f930..610e199da 100644 --- a/pkg/sentry/kernel/sessions.go +++ b/pkg/sentry/kernel/sessions.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/shm/device.go b/pkg/sentry/kernel/shm/device.go index bbc653ed8..3cb759072 100644 --- a/pkg/sentry/kernel/shm/device.go +++ b/pkg/sentry/kernel/shm/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/shm/shm.go b/pkg/sentry/kernel/shm/shm.go index d4812a065..00393b5f0 100644 --- a/pkg/sentry/kernel/shm/shm.go +++ b/pkg/sentry/kernel/shm/shm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/signal.go b/pkg/sentry/kernel/signal.go index 22a56c6fc..b528ec0dc 100644 --- a/pkg/sentry/kernel/signal.go +++ b/pkg/sentry/kernel/signal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/signal_handlers.go b/pkg/sentry/kernel/signal_handlers.go index 60cbe85b8..ce8bcb5e5 100644 --- a/pkg/sentry/kernel/signal_handlers.go +++ b/pkg/sentry/kernel/signal_handlers.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/syscalls.go b/pkg/sentry/kernel/syscalls.go index 293b21249..0572053db 100644 --- a/pkg/sentry/kernel/syscalls.go +++ b/pkg/sentry/kernel/syscalls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/syscalls_state.go b/pkg/sentry/kernel/syscalls_state.go index 981455d46..00358326b 100644 --- a/pkg/sentry/kernel/syscalls_state.go +++ b/pkg/sentry/kernel/syscalls_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/syslog.go b/pkg/sentry/kernel/syslog.go index 2aecf3eea..175d1b247 100644 --- a/pkg/sentry/kernel/syslog.go +++ b/pkg/sentry/kernel/syslog.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/table_test.go b/pkg/sentry/kernel/table_test.go index 3b29d3c6a..8f7cdb9f3 100644 --- a/pkg/sentry/kernel/table_test.go +++ b/pkg/sentry/kernel/table_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go index ed2175c37..f9378c2de 100644 --- a/pkg/sentry/kernel/task.go +++ b/pkg/sentry/kernel/task.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_acct.go b/pkg/sentry/kernel/task_acct.go index 24230af89..1ca2a82eb 100644 --- a/pkg/sentry/kernel/task_acct.go +++ b/pkg/sentry/kernel/task_acct.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_block.go b/pkg/sentry/kernel/task_block.go index e5027e551..30a7f6b1e 100644 --- a/pkg/sentry/kernel/task_block.go +++ b/pkg/sentry/kernel/task_block.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_clone.go b/pkg/sentry/kernel/task_clone.go index daf974920..bba8ddd39 100644 --- a/pkg/sentry/kernel/task_clone.go +++ b/pkg/sentry/kernel/task_clone.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_context.go b/pkg/sentry/kernel/task_context.go index ac38dd157..bbd294141 100644 --- a/pkg/sentry/kernel/task_context.go +++ b/pkg/sentry/kernel/task_context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_exec.go b/pkg/sentry/kernel/task_exec.go index b49f902a5..5d1425d5c 100644 --- a/pkg/sentry/kernel/task_exec.go +++ b/pkg/sentry/kernel/task_exec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index a07956208..6e9701b01 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_futex.go b/pkg/sentry/kernel/task_futex.go index 351cf47d7..f98097c2c 100644 --- a/pkg/sentry/kernel/task_futex.go +++ b/pkg/sentry/kernel/task_futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_identity.go b/pkg/sentry/kernel/task_identity.go index 6c9608f8d..17f08729a 100644 --- a/pkg/sentry/kernel/task_identity.go +++ b/pkg/sentry/kernel/task_identity.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_log.go b/pkg/sentry/kernel/task_log.go index f4c881c2d..e0e57e8bd 100644 --- a/pkg/sentry/kernel/task_log.go +++ b/pkg/sentry/kernel/task_log.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_net.go b/pkg/sentry/kernel/task_net.go index fc7cefc1f..04c684c1a 100644 --- a/pkg/sentry/kernel/task_net.go +++ b/pkg/sentry/kernel/task_net.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 7115aa967..4549b437e 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_sched.go b/pkg/sentry/kernel/task_sched.go index 3d654bf93..5455f6ea9 100644 --- a/pkg/sentry/kernel/task_sched.go +++ b/pkg/sentry/kernel/task_sched.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_signals.go b/pkg/sentry/kernel/task_signals.go index 7f2e0df72..654cf7525 100644 --- a/pkg/sentry/kernel/task_signals.go +++ b/pkg/sentry/kernel/task_signals.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_start.go b/pkg/sentry/kernel/task_start.go index b7534c0a2..b42531e57 100644 --- a/pkg/sentry/kernel/task_start.go +++ b/pkg/sentry/kernel/task_start.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_stop.go b/pkg/sentry/kernel/task_stop.go index 1302cadc1..e735a5dd0 100644 --- a/pkg/sentry/kernel/task_stop.go +++ b/pkg/sentry/kernel/task_stop.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_syscall.go b/pkg/sentry/kernel/task_syscall.go index 52f5fde8d..a9283d0df 100644 --- a/pkg/sentry/kernel/task_syscall.go +++ b/pkg/sentry/kernel/task_syscall.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_test.go b/pkg/sentry/kernel/task_test.go index 3f37f505d..b895361d0 100644 --- a/pkg/sentry/kernel/task_test.go +++ b/pkg/sentry/kernel/task_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/task_usermem.go b/pkg/sentry/kernel/task_usermem.go index cb68799d3..461bd7316 100644 --- a/pkg/sentry/kernel/task_usermem.go +++ b/pkg/sentry/kernel/task_usermem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/thread_group.go b/pkg/sentry/kernel/thread_group.go index 58f3a7ec9..8bd53928e 100644 --- a/pkg/sentry/kernel/thread_group.go +++ b/pkg/sentry/kernel/thread_group.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/threads.go b/pkg/sentry/kernel/threads.go index 4fd6cf4e2..656bbd46c 100644 --- a/pkg/sentry/kernel/threads.go +++ b/pkg/sentry/kernel/threads.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/time/context.go b/pkg/sentry/kernel/time/context.go index 3675ea20d..c0660d362 100644 --- a/pkg/sentry/kernel/time/context.go +++ b/pkg/sentry/kernel/time/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/time/time.go b/pkg/sentry/kernel/time/time.go index ca0f4ba2e..3846cf1ea 100644 --- a/pkg/sentry/kernel/time/time.go +++ b/pkg/sentry/kernel/time/time.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/timekeeper.go b/pkg/sentry/kernel/timekeeper.go index d7bd85e78..505a4fa4f 100644 --- a/pkg/sentry/kernel/timekeeper.go +++ b/pkg/sentry/kernel/timekeeper.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/timekeeper_state.go b/pkg/sentry/kernel/timekeeper_state.go index f3a3ed543..6ce358a05 100644 --- a/pkg/sentry/kernel/timekeeper_state.go +++ b/pkg/sentry/kernel/timekeeper_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/timekeeper_test.go b/pkg/sentry/kernel/timekeeper_test.go index 6084bcb18..a92ad689e 100644 --- a/pkg/sentry/kernel/timekeeper_test.go +++ b/pkg/sentry/kernel/timekeeper_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/uncaught_signal.proto b/pkg/sentry/kernel/uncaught_signal.proto index c7f6a1978..0bdb062cb 100644 --- a/pkg/sentry/kernel/uncaught_signal.proto +++ b/pkg/sentry/kernel/uncaught_signal.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/uts_namespace.go b/pkg/sentry/kernel/uts_namespace.go index ed5f0c031..96fe3cbb9 100644 --- a/pkg/sentry/kernel/uts_namespace.go +++ b/pkg/sentry/kernel/uts_namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/vdso.go b/pkg/sentry/kernel/vdso.go index 3a35f1d00..d40ad74f4 100644 --- a/pkg/sentry/kernel/vdso.go +++ b/pkg/sentry/kernel/vdso.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/kernel/version.go b/pkg/sentry/kernel/version.go index 8d2f14209..5640dd71d 100644 --- a/pkg/sentry/kernel/version.go +++ b/pkg/sentry/kernel/version.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/limits/context.go b/pkg/sentry/limits/context.go index bf413eb7d..9200edb52 100644 --- a/pkg/sentry/limits/context.go +++ b/pkg/sentry/limits/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/limits/limits.go b/pkg/sentry/limits/limits.go index b0571739f..b6c22656b 100644 --- a/pkg/sentry/limits/limits.go +++ b/pkg/sentry/limits/limits.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/limits/limits_test.go b/pkg/sentry/limits/limits_test.go index 945428163..658a20f56 100644 --- a/pkg/sentry/limits/limits_test.go +++ b/pkg/sentry/limits/limits_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/limits/linux.go b/pkg/sentry/limits/linux.go index e09d0d2fb..a2b401e3d 100644 --- a/pkg/sentry/limits/linux.go +++ b/pkg/sentry/limits/linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/elf.go b/pkg/sentry/loader/elf.go index 385ad0102..97e32c8ba 100644 --- a/pkg/sentry/loader/elf.go +++ b/pkg/sentry/loader/elf.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/interpreter.go b/pkg/sentry/loader/interpreter.go index 35b83654d..b88062ae5 100644 --- a/pkg/sentry/loader/interpreter.go +++ b/pkg/sentry/loader/interpreter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/loader.go b/pkg/sentry/loader/loader.go index 79051befa..dc1a52398 100644 --- a/pkg/sentry/loader/loader.go +++ b/pkg/sentry/loader/loader.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index 8c196df84..207d8ed3d 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/loader/vdso_state.go b/pkg/sentry/loader/vdso_state.go index b327f0e1e..db378e90a 100644 --- a/pkg/sentry/loader/vdso_state.go +++ b/pkg/sentry/loader/vdso_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memmap/mapping_set.go b/pkg/sentry/memmap/mapping_set.go index bd07e9aac..3cf2b338f 100644 --- a/pkg/sentry/memmap/mapping_set.go +++ b/pkg/sentry/memmap/mapping_set.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memmap/mapping_set_test.go b/pkg/sentry/memmap/mapping_set_test.go index 45d1d4688..c702555ce 100644 --- a/pkg/sentry/memmap/mapping_set_test.go +++ b/pkg/sentry/memmap/mapping_set_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memmap/memmap.go b/pkg/sentry/memmap/memmap.go index 3f6f7ebd0..0106c857d 100644 --- a/pkg/sentry/memmap/memmap.go +++ b/pkg/sentry/memmap/memmap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memutil/memutil.go b/pkg/sentry/memutil/memutil.go index 286d50ca4..a4154c42a 100644 --- a/pkg/sentry/memutil/memutil.go +++ b/pkg/sentry/memutil/memutil.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/memutil/memutil_unsafe.go b/pkg/sentry/memutil/memutil_unsafe.go index bc2c72f55..92eab8a26 100644 --- a/pkg/sentry/memutil/memutil_unsafe.go +++ b/pkg/sentry/memutil/memutil_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/address_space.go b/pkg/sentry/mm/address_space.go index 4dddcf7b5..06f587fde 100644 --- a/pkg/sentry/mm/address_space.go +++ b/pkg/sentry/mm/address_space.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/aio_context.go b/pkg/sentry/mm/aio_context.go index 7075792e0..5c61acf36 100644 --- a/pkg/sentry/mm/aio_context.go +++ b/pkg/sentry/mm/aio_context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/aio_context_state.go b/pkg/sentry/mm/aio_context_state.go index 192a6f744..c37fc9f7b 100644 --- a/pkg/sentry/mm/aio_context_state.go +++ b/pkg/sentry/mm/aio_context_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/debug.go b/pkg/sentry/mm/debug.go index d075ee1ca..fe58cfc4c 100644 --- a/pkg/sentry/mm/debug.go +++ b/pkg/sentry/mm/debug.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/io.go b/pkg/sentry/mm/io.go index 81787a6fd..e4c057d28 100644 --- a/pkg/sentry/mm/io.go +++ b/pkg/sentry/mm/io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/lifecycle.go b/pkg/sentry/mm/lifecycle.go index 2fe03172c..e6aa6f9ef 100644 --- a/pkg/sentry/mm/lifecycle.go +++ b/pkg/sentry/mm/lifecycle.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/metadata.go b/pkg/sentry/mm/metadata.go index 5ef1ba0b1..9768e51f1 100644 --- a/pkg/sentry/mm/metadata.go +++ b/pkg/sentry/mm/metadata.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/mm.go b/pkg/sentry/mm/mm.go index a3417a46e..d25aa5136 100644 --- a/pkg/sentry/mm/mm.go +++ b/pkg/sentry/mm/mm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/mm_test.go b/pkg/sentry/mm/mm_test.go index ae4fba478..f4917419f 100644 --- a/pkg/sentry/mm/mm_test.go +++ b/pkg/sentry/mm/mm_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/pma.go b/pkg/sentry/mm/pma.go index 0cca743ef..ece561ff0 100644 --- a/pkg/sentry/mm/pma.go +++ b/pkg/sentry/mm/pma.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/procfs.go b/pkg/sentry/mm/procfs.go index 7cdbf6e25..c8302a553 100644 --- a/pkg/sentry/mm/procfs.go +++ b/pkg/sentry/mm/procfs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/save_restore.go b/pkg/sentry/mm/save_restore.go index 46e0e0754..0385957bd 100644 --- a/pkg/sentry/mm/save_restore.go +++ b/pkg/sentry/mm/save_restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/shm.go b/pkg/sentry/mm/shm.go index 3bc48c7e7..12913007b 100644 --- a/pkg/sentry/mm/shm.go +++ b/pkg/sentry/mm/shm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/special_mappable.go b/pkg/sentry/mm/special_mappable.go index 3b5161998..687959005 100644 --- a/pkg/sentry/mm/special_mappable.go +++ b/pkg/sentry/mm/special_mappable.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/syscalls.go b/pkg/sentry/mm/syscalls.go index 7b675b9b5..a25318abb 100644 --- a/pkg/sentry/mm/syscalls.go +++ b/pkg/sentry/mm/syscalls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/mm/vma.go b/pkg/sentry/mm/vma.go index 931995254..ad901344b 100644 --- a/pkg/sentry/mm/vma.go +++ b/pkg/sentry/mm/vma.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/pgalloc/context.go b/pkg/sentry/pgalloc/context.go index adc97e78f..cb9809b1f 100644 --- a/pkg/sentry/pgalloc/context.go +++ b/pkg/sentry/pgalloc/context.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google Inc. +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/pgalloc/pgalloc.go b/pkg/sentry/pgalloc/pgalloc.go index 0754e608f..411dafa07 100644 --- a/pkg/sentry/pgalloc/pgalloc.go +++ b/pkg/sentry/pgalloc/pgalloc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/pgalloc/pgalloc_test.go b/pkg/sentry/pgalloc/pgalloc_test.go index 726623c1a..14a39bb9e 100644 --- a/pkg/sentry/pgalloc/pgalloc_test.go +++ b/pkg/sentry/pgalloc/pgalloc_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/pgalloc/pgalloc_unsafe.go b/pkg/sentry/pgalloc/pgalloc_unsafe.go index 33b0a68a8..a4b5d581c 100644 --- a/pkg/sentry/pgalloc/pgalloc_unsafe.go +++ b/pkg/sentry/pgalloc/pgalloc_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/pgalloc/save_restore.go b/pkg/sentry/pgalloc/save_restore.go index 21024e656..cf169af55 100644 --- a/pkg/sentry/pgalloc/save_restore.go +++ b/pkg/sentry/pgalloc/save_restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/context.go b/pkg/sentry/platform/context.go index cca21a23e..793f57fd7 100644 --- a/pkg/sentry/platform/context.go +++ b/pkg/sentry/platform/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/interrupt/interrupt.go b/pkg/sentry/platform/interrupt/interrupt.go index 9c83f41eb..a4651f500 100644 --- a/pkg/sentry/platform/interrupt/interrupt.go +++ b/pkg/sentry/platform/interrupt/interrupt.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/interrupt/interrupt_test.go b/pkg/sentry/platform/interrupt/interrupt_test.go index fb3284395..0ecdf6e7a 100644 --- a/pkg/sentry/platform/interrupt/interrupt_test.go +++ b/pkg/sentry/platform/interrupt/interrupt_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/address_space.go b/pkg/sentry/platform/kvm/address_space.go index f2f7ab1e8..689122175 100644 --- a/pkg/sentry/platform/kvm/address_space.go +++ b/pkg/sentry/platform/kvm/address_space.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/allocator.go b/pkg/sentry/platform/kvm/allocator.go index b25cad155..42bcc9733 100644 --- a/pkg/sentry/platform/kvm/allocator.go +++ b/pkg/sentry/platform/kvm/allocator.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill.go b/pkg/sentry/platform/kvm/bluepill.go index f24f1c662..a926e6f8b 100644 --- a/pkg/sentry/platform/kvm/bluepill.go +++ b/pkg/sentry/platform/kvm/bluepill.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_amd64.go b/pkg/sentry/platform/kvm/bluepill_amd64.go index 6520682d7..c258408f9 100644 --- a/pkg/sentry/platform/kvm/bluepill_amd64.go +++ b/pkg/sentry/platform/kvm/bluepill_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_amd64.s b/pkg/sentry/platform/kvm/bluepill_amd64.s index 65b01f358..2bc34a435 100644 --- a/pkg/sentry/platform/kvm/bluepill_amd64.s +++ b/pkg/sentry/platform/kvm/bluepill_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go b/pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go index 21de2488e..92fde7ee0 100644 --- a/pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go +++ b/pkg/sentry/platform/kvm/bluepill_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_fault.go b/pkg/sentry/platform/kvm/bluepill_fault.go index e79a30ef2..3c452f5ba 100644 --- a/pkg/sentry/platform/kvm/bluepill_fault.go +++ b/pkg/sentry/platform/kvm/bluepill_fault.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/bluepill_unsafe.go b/pkg/sentry/platform/kvm/bluepill_unsafe.go index 2605f8c93..4184939e5 100644 --- a/pkg/sentry/platform/kvm/bluepill_unsafe.go +++ b/pkg/sentry/platform/kvm/bluepill_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/context.go b/pkg/sentry/platform/kvm/context.go index c75a4b415..0eb0020f7 100644 --- a/pkg/sentry/platform/kvm/context.go +++ b/pkg/sentry/platform/kvm/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm.go b/pkg/sentry/platform/kvm/kvm.go index c5a4435b1..ed0521c3f 100644 --- a/pkg/sentry/platform/kvm/kvm.go +++ b/pkg/sentry/platform/kvm/kvm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm_amd64.go b/pkg/sentry/platform/kvm/kvm_amd64.go index 70d0ac63b..61493ccaf 100644 --- a/pkg/sentry/platform/kvm/kvm_amd64.go +++ b/pkg/sentry/platform/kvm/kvm_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go b/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go index d0f6bb225..46c4b9113 100644 --- a/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go +++ b/pkg/sentry/platform/kvm/kvm_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm_const.go b/pkg/sentry/platform/kvm/kvm_const.go index cac8d9937..d05f05c29 100644 --- a/pkg/sentry/platform/kvm/kvm_const.go +++ b/pkg/sentry/platform/kvm/kvm_const.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/kvm_test.go b/pkg/sentry/platform/kvm/kvm_test.go index 361200622..e83db71e9 100644 --- a/pkg/sentry/platform/kvm/kvm_test.go +++ b/pkg/sentry/platform/kvm/kvm_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/machine.go b/pkg/sentry/platform/kvm/machine.go index b8b3c9a4a..f5953b96e 100644 --- a/pkg/sentry/platform/kvm/machine.go +++ b/pkg/sentry/platform/kvm/machine.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/machine_amd64.go b/pkg/sentry/platform/kvm/machine_amd64.go index ccfe837b5..b6821122a 100644 --- a/pkg/sentry/platform/kvm/machine_amd64.go +++ b/pkg/sentry/platform/kvm/machine_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/machine_amd64_unsafe.go b/pkg/sentry/platform/kvm/machine_amd64_unsafe.go index 69ba67ced..06a2e3b0c 100644 --- a/pkg/sentry/platform/kvm/machine_amd64_unsafe.go +++ b/pkg/sentry/platform/kvm/machine_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/machine_unsafe.go b/pkg/sentry/platform/kvm/machine_unsafe.go index 22ae60b63..452d88d7f 100644 --- a/pkg/sentry/platform/kvm/machine_unsafe.go +++ b/pkg/sentry/platform/kvm/machine_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/physical_map.go b/pkg/sentry/platform/kvm/physical_map.go index 9d7dca5b3..450eb8201 100644 --- a/pkg/sentry/platform/kvm/physical_map.go +++ b/pkg/sentry/platform/kvm/physical_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/testutil/testutil.go b/pkg/sentry/platform/kvm/testutil/testutil.go index 0d496561d..6cf2359a3 100644 --- a/pkg/sentry/platform/kvm/testutil/testutil.go +++ b/pkg/sentry/platform/kvm/testutil/testutil.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/testutil/testutil_amd64.go b/pkg/sentry/platform/kvm/testutil/testutil_amd64.go index fcba33813..203d71528 100644 --- a/pkg/sentry/platform/kvm/testutil/testutil_amd64.go +++ b/pkg/sentry/platform/kvm/testutil/testutil_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/testutil/testutil_amd64.s b/pkg/sentry/platform/kvm/testutil/testutil_amd64.s index f1da41a44..491ec0c2a 100644 --- a/pkg/sentry/platform/kvm/testutil/testutil_amd64.s +++ b/pkg/sentry/platform/kvm/testutil/testutil_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/virtual_map.go b/pkg/sentry/platform/kvm/virtual_map.go index 0343e9267..28a1b4414 100644 --- a/pkg/sentry/platform/kvm/virtual_map.go +++ b/pkg/sentry/platform/kvm/virtual_map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/kvm/virtual_map_test.go b/pkg/sentry/platform/kvm/virtual_map_test.go index 935e0eb93..d03ec654a 100644 --- a/pkg/sentry/platform/kvm/virtual_map_test.go +++ b/pkg/sentry/platform/kvm/virtual_map_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/mmap_min_addr.go b/pkg/sentry/platform/mmap_min_addr.go index 1bcc1f8e9..90976735b 100644 --- a/pkg/sentry/platform/mmap_min_addr.go +++ b/pkg/sentry/platform/mmap_min_addr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/platform.go b/pkg/sentry/platform/platform.go index 0e48417b9..ae37276ad 100644 --- a/pkg/sentry/platform/platform.go +++ b/pkg/sentry/platform/platform.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid.go b/pkg/sentry/platform/procid/procid.go index 3f49ab093..78b92422c 100644 --- a/pkg/sentry/platform/procid/procid.go +++ b/pkg/sentry/platform/procid/procid.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid_amd64.s b/pkg/sentry/platform/procid/procid_amd64.s index ef3439c03..272c9fc14 100644 --- a/pkg/sentry/platform/procid/procid_amd64.s +++ b/pkg/sentry/platform/procid/procid_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid_arm64.s b/pkg/sentry/platform/procid/procid_arm64.s index 02e907b6b..7a1684a18 100644 --- a/pkg/sentry/platform/procid/procid_arm64.s +++ b/pkg/sentry/platform/procid/procid_arm64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid_net_test.go b/pkg/sentry/platform/procid/procid_net_test.go index e8dcc479d..b628e2285 100644 --- a/pkg/sentry/platform/procid/procid_net_test.go +++ b/pkg/sentry/platform/procid/procid_net_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/procid/procid_test.go b/pkg/sentry/platform/procid/procid_test.go index 7a57c7cdc..88dd0b3ae 100644 --- a/pkg/sentry/platform/procid/procid_test.go +++ b/pkg/sentry/platform/procid/procid_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/ptrace.go b/pkg/sentry/platform/ptrace/ptrace.go index 3c0713e95..6a890dd81 100644 --- a/pkg/sentry/platform/ptrace/ptrace.go +++ b/pkg/sentry/platform/ptrace/ptrace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/ptrace_unsafe.go b/pkg/sentry/platform/ptrace/ptrace_unsafe.go index 223b23199..585f6c1fb 100644 --- a/pkg/sentry/platform/ptrace/ptrace_unsafe.go +++ b/pkg/sentry/platform/ptrace/ptrace_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/stub_amd64.s b/pkg/sentry/platform/ptrace/stub_amd64.s index 63f98e40d..64c718d21 100644 --- a/pkg/sentry/platform/ptrace/stub_amd64.s +++ b/pkg/sentry/platform/ptrace/stub_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/stub_unsafe.go b/pkg/sentry/platform/ptrace/stub_unsafe.go index 48c16c4a1..54d5021a9 100644 --- a/pkg/sentry/platform/ptrace/stub_unsafe.go +++ b/pkg/sentry/platform/ptrace/stub_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess.go b/pkg/sentry/platform/ptrace/subprocess.go index 2a5d699ec..83b43057f 100644 --- a/pkg/sentry/platform/ptrace/subprocess.go +++ b/pkg/sentry/platform/ptrace/subprocess.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess_amd64.go b/pkg/sentry/platform/ptrace/subprocess_amd64.go index d23a1133e..77a0e908f 100644 --- a/pkg/sentry/platform/ptrace/subprocess_amd64.go +++ b/pkg/sentry/platform/ptrace/subprocess_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess_linux.go b/pkg/sentry/platform/ptrace/subprocess_linux.go index e2aab8135..2c07b4ac3 100644 --- a/pkg/sentry/platform/ptrace/subprocess_linux.go +++ b/pkg/sentry/platform/ptrace/subprocess_linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go b/pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go index 0c9263060..1bf7eab28 100644 --- a/pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go +++ b/pkg/sentry/platform/ptrace/subprocess_linux_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ptrace/subprocess_unsafe.go b/pkg/sentry/platform/ptrace/subprocess_unsafe.go index ca6c4ac97..17736b05b 100644 --- a/pkg/sentry/platform/ptrace/subprocess_unsafe.go +++ b/pkg/sentry/platform/ptrace/subprocess_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/defs.go b/pkg/sentry/platform/ring0/defs.go index 98d0a6de0..5bbd4612d 100644 --- a/pkg/sentry/platform/ring0/defs.go +++ b/pkg/sentry/platform/ring0/defs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/defs_amd64.go b/pkg/sentry/platform/ring0/defs_amd64.go index 67242b92b..413c3dbc4 100644 --- a/pkg/sentry/platform/ring0/defs_amd64.go +++ b/pkg/sentry/platform/ring0/defs_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/entry_amd64.go b/pkg/sentry/platform/ring0/entry_amd64.go index 4a9affe64..a5ce67885 100644 --- a/pkg/sentry/platform/ring0/entry_amd64.go +++ b/pkg/sentry/platform/ring0/entry_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/entry_amd64.s b/pkg/sentry/platform/ring0/entry_amd64.s index afb040a6f..8cb8c4996 100644 --- a/pkg/sentry/platform/ring0/entry_amd64.s +++ b/pkg/sentry/platform/ring0/entry_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/gen_offsets/main.go b/pkg/sentry/platform/ring0/gen_offsets/main.go index 11c49855f..a4927da2f 100644 --- a/pkg/sentry/platform/ring0/gen_offsets/main.go +++ b/pkg/sentry/platform/ring0/gen_offsets/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/kernel.go b/pkg/sentry/platform/ring0/kernel.go index 19ac6eb7c..900c0bba7 100644 --- a/pkg/sentry/platform/ring0/kernel.go +++ b/pkg/sentry/platform/ring0/kernel.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/kernel_amd64.go b/pkg/sentry/platform/ring0/kernel_amd64.go index 5ed4342dd..3577b5127 100644 --- a/pkg/sentry/platform/ring0/kernel_amd64.go +++ b/pkg/sentry/platform/ring0/kernel_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/kernel_unsafe.go b/pkg/sentry/platform/ring0/kernel_unsafe.go index faf4240e5..16955ad91 100644 --- a/pkg/sentry/platform/ring0/kernel_unsafe.go +++ b/pkg/sentry/platform/ring0/kernel_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/lib_amd64.go b/pkg/sentry/platform/ring0/lib_amd64.go index 2b95a0141..9c5f26962 100644 --- a/pkg/sentry/platform/ring0/lib_amd64.go +++ b/pkg/sentry/platform/ring0/lib_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/lib_amd64.s b/pkg/sentry/platform/ring0/lib_amd64.s index 98a130525..75d742750 100644 --- a/pkg/sentry/platform/ring0/lib_amd64.s +++ b/pkg/sentry/platform/ring0/lib_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/offsets_amd64.go b/pkg/sentry/platform/ring0/offsets_amd64.go index 806e07ec0..85cc3fdad 100644 --- a/pkg/sentry/platform/ring0/offsets_amd64.go +++ b/pkg/sentry/platform/ring0/offsets_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/allocator.go b/pkg/sentry/platform/ring0/pagetables/allocator.go index ee6e90a11..23fd5c352 100644 --- a/pkg/sentry/platform/ring0/pagetables/allocator.go +++ b/pkg/sentry/platform/ring0/pagetables/allocator.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go b/pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go index f48647b3a..1b996b4e2 100644 --- a/pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go +++ b/pkg/sentry/platform/ring0/pagetables/allocator_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables.go b/pkg/sentry/platform/ring0/pagetables/pagetables.go index c7207ec18..e5dcaada7 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go index 746f614e5..7aa6c524e 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go index 2f82c4353..a1ec4b109 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_amd64_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_test.go b/pkg/sentry/platform/ring0/pagetables/pagetables_test.go index 3e5dc7dc7..36e424495 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_test.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pagetables_x86.go b/pkg/sentry/platform/ring0/pagetables/pagetables_x86.go index 6bd8c3584..ff427fbe9 100644 --- a/pkg/sentry/platform/ring0/pagetables/pagetables_x86.go +++ b/pkg/sentry/platform/ring0/pagetables/pagetables_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/pcids_x86.go b/pkg/sentry/platform/ring0/pagetables/pcids_x86.go index 0d9a51aa5..0f029f25d 100644 --- a/pkg/sentry/platform/ring0/pagetables/pcids_x86.go +++ b/pkg/sentry/platform/ring0/pagetables/pcids_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/pagetables/walker_amd64.go b/pkg/sentry/platform/ring0/pagetables/walker_amd64.go index c4c71d23e..8f9dacd93 100644 --- a/pkg/sentry/platform/ring0/pagetables/walker_amd64.go +++ b/pkg/sentry/platform/ring0/pagetables/walker_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/ring0.go b/pkg/sentry/platform/ring0/ring0.go index 10c51e88d..cdeb1b43a 100644 --- a/pkg/sentry/platform/ring0/ring0.go +++ b/pkg/sentry/platform/ring0/ring0.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/ring0/x86.go b/pkg/sentry/platform/ring0/x86.go index 4c6daec22..7e5ceafdb 100644 --- a/pkg/sentry/platform/ring0/x86.go +++ b/pkg/sentry/platform/ring0/x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/atomic_amd64.s b/pkg/sentry/platform/safecopy/atomic_amd64.s index f90b4bfd1..a0cd78f33 100644 --- a/pkg/sentry/platform/safecopy/atomic_amd64.s +++ b/pkg/sentry/platform/safecopy/atomic_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/safecopy.go b/pkg/sentry/platform/safecopy/safecopy.go index 69c66a3b7..5126871eb 100644 --- a/pkg/sentry/platform/safecopy/safecopy.go +++ b/pkg/sentry/platform/safecopy/safecopy.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/safecopy_test.go b/pkg/sentry/platform/safecopy/safecopy_test.go index 1a682d28a..5818f7f9b 100644 --- a/pkg/sentry/platform/safecopy/safecopy_test.go +++ b/pkg/sentry/platform/safecopy/safecopy_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/safecopy_unsafe.go b/pkg/sentry/platform/safecopy/safecopy_unsafe.go index f84527484..eef028e68 100644 --- a/pkg/sentry/platform/safecopy/safecopy_unsafe.go +++ b/pkg/sentry/platform/safecopy/safecopy_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/sighandler_amd64.s b/pkg/sentry/platform/safecopy/sighandler_amd64.s index db7701a29..475ae48e9 100644 --- a/pkg/sentry/platform/safecopy/sighandler_amd64.s +++ b/pkg/sentry/platform/safecopy/sighandler_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/platform/safecopy/sighandler_arm64.s b/pkg/sentry/platform/safecopy/sighandler_arm64.s index cdfca8207..53e4ac2c1 100644 --- a/pkg/sentry/platform/safecopy/sighandler_arm64.s +++ b/pkg/sentry/platform/safecopy/sighandler_arm64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/block_unsafe.go b/pkg/sentry/safemem/block_unsafe.go index c3a9780d2..1f72deb61 100644 --- a/pkg/sentry/safemem/block_unsafe.go +++ b/pkg/sentry/safemem/block_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/io.go b/pkg/sentry/safemem/io.go index 6cb52439f..5c3d73eb7 100644 --- a/pkg/sentry/safemem/io.go +++ b/pkg/sentry/safemem/io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/io_test.go b/pkg/sentry/safemem/io_test.go index 2eda8c3bb..629741bee 100644 --- a/pkg/sentry/safemem/io_test.go +++ b/pkg/sentry/safemem/io_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/safemem.go b/pkg/sentry/safemem/safemem.go index 090932d3e..3e70d33a2 100644 --- a/pkg/sentry/safemem/safemem.go +++ b/pkg/sentry/safemem/safemem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/seq_test.go b/pkg/sentry/safemem/seq_test.go index fddcaf714..eba4bb535 100644 --- a/pkg/sentry/safemem/seq_test.go +++ b/pkg/sentry/safemem/seq_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/safemem/seq_unsafe.go b/pkg/sentry/safemem/seq_unsafe.go index 83a6b7183..354a95dde 100644 --- a/pkg/sentry/safemem/seq_unsafe.go +++ b/pkg/sentry/safemem/seq_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/sighandling/sighandling.go b/pkg/sentry/sighandling/sighandling.go index 571245ce5..659b43363 100644 --- a/pkg/sentry/sighandling/sighandling.go +++ b/pkg/sentry/sighandling/sighandling.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/sighandling/sighandling_unsafe.go b/pkg/sentry/sighandling/sighandling_unsafe.go index db6e71487..aca77888a 100644 --- a/pkg/sentry/sighandling/sighandling_unsafe.go +++ b/pkg/sentry/sighandling/sighandling_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/control/control.go b/pkg/sentry/socket/control/control.go index d44f5e88a..abda364c9 100644 --- a/pkg/sentry/socket/control/control.go +++ b/pkg/sentry/socket/control/control.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/device.go b/pkg/sentry/socket/epsocket/device.go index 3cc138eb0..ab4083efe 100644 --- a/pkg/sentry/socket/epsocket/device.go +++ b/pkg/sentry/socket/epsocket/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index 768fa0dfa..520d82f68 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/provider.go b/pkg/sentry/socket/epsocket/provider.go index 0d9c2df24..5a89a63fb 100644 --- a/pkg/sentry/socket/epsocket/provider.go +++ b/pkg/sentry/socket/epsocket/provider.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/save_restore.go b/pkg/sentry/socket/epsocket/save_restore.go index f19afb6c0..feaafb7cc 100644 --- a/pkg/sentry/socket/epsocket/save_restore.go +++ b/pkg/sentry/socket/epsocket/save_restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/epsocket/stack.go b/pkg/sentry/socket/epsocket/stack.go index 37c48f4bc..edefa225b 100644 --- a/pkg/sentry/socket/epsocket/stack.go +++ b/pkg/sentry/socket/epsocket/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/device.go b/pkg/sentry/socket/hostinet/device.go index c5133f3bb..4267e3691 100644 --- a/pkg/sentry/socket/hostinet/device.go +++ b/pkg/sentry/socket/hostinet/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/hostinet.go b/pkg/sentry/socket/hostinet/hostinet.go index 7858892ab..0d6f51d2b 100644 --- a/pkg/sentry/socket/hostinet/hostinet.go +++ b/pkg/sentry/socket/hostinet/hostinet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/save_restore.go b/pkg/sentry/socket/hostinet/save_restore.go index 3827f082a..1dec33897 100644 --- a/pkg/sentry/socket/hostinet/save_restore.go +++ b/pkg/sentry/socket/hostinet/save_restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index 49349074f..71884d3db 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/socket_unsafe.go b/pkg/sentry/socket/hostinet/socket_unsafe.go index 59c8910ca..eed0c7837 100644 --- a/pkg/sentry/socket/hostinet/socket_unsafe.go +++ b/pkg/sentry/socket/hostinet/socket_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/hostinet/stack.go b/pkg/sentry/socket/hostinet/stack.go index 4ce73c1f1..9c45991ba 100644 --- a/pkg/sentry/socket/hostinet/stack.go +++ b/pkg/sentry/socket/hostinet/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/message.go b/pkg/sentry/socket/netlink/message.go index a95172cba..5bd3b49ce 100644 --- a/pkg/sentry/socket/netlink/message.go +++ b/pkg/sentry/socket/netlink/message.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/port/port.go b/pkg/sentry/socket/netlink/port/port.go index 20b9a6e37..e9d3275b1 100644 --- a/pkg/sentry/socket/netlink/port/port.go +++ b/pkg/sentry/socket/netlink/port/port.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/port/port_test.go b/pkg/sentry/socket/netlink/port/port_test.go index 49b3b48ab..516f6cd6c 100644 --- a/pkg/sentry/socket/netlink/port/port_test.go +++ b/pkg/sentry/socket/netlink/port/port_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/provider.go b/pkg/sentry/socket/netlink/provider.go index 06786bd50..76cf12fd4 100644 --- a/pkg/sentry/socket/netlink/provider.go +++ b/pkg/sentry/socket/netlink/provider.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/route/protocol.go b/pkg/sentry/socket/netlink/route/protocol.go index e414b829b..9f0a81403 100644 --- a/pkg/sentry/socket/netlink/route/protocol.go +++ b/pkg/sentry/socket/netlink/route/protocol.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index a34f9d3ca..dc688eb00 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/conn/conn.go b/pkg/sentry/socket/rpcinet/conn/conn.go index 64106c4b5..f537c7f63 100644 --- a/pkg/sentry/socket/rpcinet/conn/conn.go +++ b/pkg/sentry/socket/rpcinet/conn/conn.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/device.go b/pkg/sentry/socket/rpcinet/device.go index d2b9f9222..44c0a39b7 100644 --- a/pkg/sentry/socket/rpcinet/device.go +++ b/pkg/sentry/socket/rpcinet/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/notifier/notifier.go b/pkg/sentry/socket/rpcinet/notifier/notifier.go index f06d12231..601e05994 100644 --- a/pkg/sentry/socket/rpcinet/notifier/notifier.go +++ b/pkg/sentry/socket/rpcinet/notifier/notifier.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/rpcinet.go b/pkg/sentry/socket/rpcinet/rpcinet.go index 6c98e6acb..5d4fd4dac 100644 --- a/pkg/sentry/socket/rpcinet/rpcinet.go +++ b/pkg/sentry/socket/rpcinet/rpcinet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index cf8f69efb..c028ed4dd 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/stack.go b/pkg/sentry/socket/rpcinet/stack.go index cb8344ec6..a1be711df 100644 --- a/pkg/sentry/socket/rpcinet/stack.go +++ b/pkg/sentry/socket/rpcinet/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/rpcinet/stack_unsafe.go b/pkg/sentry/socket/rpcinet/stack_unsafe.go index d04fb2069..e53f578ba 100644 --- a/pkg/sentry/socket/rpcinet/stack_unsafe.go +++ b/pkg/sentry/socket/rpcinet/stack_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/socket.go b/pkg/sentry/socket/socket.go index 62ba13782..7e840b452 100644 --- a/pkg/sentry/socket/socket.go +++ b/pkg/sentry/socket/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/device.go b/pkg/sentry/socket/unix/device.go index 41820dbb3..734d39ee6 100644 --- a/pkg/sentry/socket/unix/device.go +++ b/pkg/sentry/socket/unix/device.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/io.go b/pkg/sentry/socket/unix/io.go index 7d80e4393..382911d51 100644 --- a/pkg/sentry/socket/unix/io.go +++ b/pkg/sentry/socket/unix/io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/connectioned.go b/pkg/sentry/socket/unix/transport/connectioned.go index 62641bb34..18e492862 100644 --- a/pkg/sentry/socket/unix/transport/connectioned.go +++ b/pkg/sentry/socket/unix/transport/connectioned.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/connectioned_state.go b/pkg/sentry/socket/unix/transport/connectioned_state.go index 608a6a97a..7e02a5db8 100644 --- a/pkg/sentry/socket/unix/transport/connectioned_state.go +++ b/pkg/sentry/socket/unix/transport/connectioned_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/connectionless.go b/pkg/sentry/socket/unix/transport/connectionless.go index 728863f3f..43ff875e4 100644 --- a/pkg/sentry/socket/unix/transport/connectionless.go +++ b/pkg/sentry/socket/unix/transport/connectionless.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/queue.go b/pkg/sentry/socket/unix/transport/queue.go index 45a58c600..b650caae7 100644 --- a/pkg/sentry/socket/unix/transport/queue.go +++ b/pkg/sentry/socket/unix/transport/queue.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/transport/unix.go b/pkg/sentry/socket/unix/transport/unix.go index 12b1576bd..d5f7f7aa8 100644 --- a/pkg/sentry/socket/unix/transport/unix.go +++ b/pkg/sentry/socket/unix/transport/unix.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 01efd24d3..e9607aa01 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/state/state.go b/pkg/sentry/state/state.go index 224f8b709..27fde505b 100644 --- a/pkg/sentry/state/state.go +++ b/pkg/sentry/state/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/state/state_metadata.go b/pkg/sentry/state/state_metadata.go index 7f047b808..b8e128c40 100644 --- a/pkg/sentry/state/state_metadata.go +++ b/pkg/sentry/state/state_metadata.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/state/state_unsafe.go b/pkg/sentry/state/state_unsafe.go index f02e12b2a..7745b6ac6 100644 --- a/pkg/sentry/state/state_unsafe.go +++ b/pkg/sentry/state/state_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/capability.go b/pkg/sentry/strace/capability.go index 9001181e7..f85d6636e 100644 --- a/pkg/sentry/strace/capability.go +++ b/pkg/sentry/strace/capability.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/clone.go b/pkg/sentry/strace/clone.go index e18ce84dc..ff6a432c6 100644 --- a/pkg/sentry/strace/clone.go +++ b/pkg/sentry/strace/clone.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/futex.go b/pkg/sentry/strace/futex.go index f4aa7fcad..24301bda6 100644 --- a/pkg/sentry/strace/futex.go +++ b/pkg/sentry/strace/futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/linux64.go b/pkg/sentry/strace/linux64.go index 6043b8cb1..3650fd6e1 100644 --- a/pkg/sentry/strace/linux64.go +++ b/pkg/sentry/strace/linux64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/open.go b/pkg/sentry/strace/open.go index 3bf348d7a..140727b02 100644 --- a/pkg/sentry/strace/open.go +++ b/pkg/sentry/strace/open.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/poll.go b/pkg/sentry/strace/poll.go index b6b05423c..15605187d 100644 --- a/pkg/sentry/strace/poll.go +++ b/pkg/sentry/strace/poll.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/ptrace.go b/pkg/sentry/strace/ptrace.go index 8c4b79227..485aacb8a 100644 --- a/pkg/sentry/strace/ptrace.go +++ b/pkg/sentry/strace/ptrace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/signal.go b/pkg/sentry/strace/signal.go index 524be0e15..f82460e1c 100644 --- a/pkg/sentry/strace/signal.go +++ b/pkg/sentry/strace/signal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/socket.go b/pkg/sentry/strace/socket.go index 4c1a9d469..dbe53b9a2 100644 --- a/pkg/sentry/strace/socket.go +++ b/pkg/sentry/strace/socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/strace.go b/pkg/sentry/strace/strace.go index 434a200d9..f4c1be4ce 100644 --- a/pkg/sentry/strace/strace.go +++ b/pkg/sentry/strace/strace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/strace.proto b/pkg/sentry/strace/strace.proto index f1fc539d6..4b2f73a5f 100644 --- a/pkg/sentry/strace/strace.proto +++ b/pkg/sentry/strace/strace.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/strace/syscalls.go b/pkg/sentry/strace/syscalls.go index 8c897fcbe..eae2d6c12 100644 --- a/pkg/sentry/strace/syscalls.go +++ b/pkg/sentry/strace/syscalls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/epoll.go b/pkg/sentry/syscalls/epoll.go index b90d191b7..ec1eab331 100644 --- a/pkg/sentry/syscalls/epoll.go +++ b/pkg/sentry/syscalls/epoll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/error.go b/pkg/sentry/syscalls/linux/error.go index 304a12dde..1ba3695fb 100644 --- a/pkg/sentry/syscalls/linux/error.go +++ b/pkg/sentry/syscalls/linux/error.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/flags.go b/pkg/sentry/syscalls/linux/flags.go index d2aec963a..d83e12971 100644 --- a/pkg/sentry/syscalls/linux/flags.go +++ b/pkg/sentry/syscalls/linux/flags.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index b9b4ccbd1..9a460ebdf 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sigset.go b/pkg/sentry/syscalls/linux/sigset.go index a033b7c70..5438b664b 100644 --- a/pkg/sentry/syscalls/linux/sigset.go +++ b/pkg/sentry/syscalls/linux/sigset.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_aio.go b/pkg/sentry/syscalls/linux/sys_aio.go index 61c2647bf..1b27b2415 100644 --- a/pkg/sentry/syscalls/linux/sys_aio.go +++ b/pkg/sentry/syscalls/linux/sys_aio.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_capability.go b/pkg/sentry/syscalls/linux/sys_capability.go index cf972dc28..622cb8d0d 100644 --- a/pkg/sentry/syscalls/linux/sys_capability.go +++ b/pkg/sentry/syscalls/linux/sys_capability.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_epoll.go b/pkg/sentry/syscalls/linux/sys_epoll.go index 200c46355..1467feb4e 100644 --- a/pkg/sentry/syscalls/linux/sys_epoll.go +++ b/pkg/sentry/syscalls/linux/sys_epoll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_eventfd.go b/pkg/sentry/syscalls/linux/sys_eventfd.go index 903172890..ca4ead488 100644 --- a/pkg/sentry/syscalls/linux/sys_eventfd.go +++ b/pkg/sentry/syscalls/linux/sys_eventfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 967464c85..893322647 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_futex.go b/pkg/sentry/syscalls/linux/sys_futex.go index f0c89cba4..7cef4b50c 100644 --- a/pkg/sentry/syscalls/linux/sys_futex.go +++ b/pkg/sentry/syscalls/linux/sys_futex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_getdents.go b/pkg/sentry/syscalls/linux/sys_getdents.go index 4b441b31b..1b597d5bc 100644 --- a/pkg/sentry/syscalls/linux/sys_getdents.go +++ b/pkg/sentry/syscalls/linux/sys_getdents.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_identity.go b/pkg/sentry/syscalls/linux/sys_identity.go index 8d594aa83..27e765a2d 100644 --- a/pkg/sentry/syscalls/linux/sys_identity.go +++ b/pkg/sentry/syscalls/linux/sys_identity.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_inotify.go b/pkg/sentry/syscalls/linux/sys_inotify.go index 26a505782..20269a769 100644 --- a/pkg/sentry/syscalls/linux/sys_inotify.go +++ b/pkg/sentry/syscalls/linux/sys_inotify.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_lseek.go b/pkg/sentry/syscalls/linux/sys_lseek.go index ad3bfd761..8aadc6d8c 100644 --- a/pkg/sentry/syscalls/linux/sys_lseek.go +++ b/pkg/sentry/syscalls/linux/sys_lseek.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_mmap.go b/pkg/sentry/syscalls/linux/sys_mmap.go index 805b251b1..64a6e639c 100644 --- a/pkg/sentry/syscalls/linux/sys_mmap.go +++ b/pkg/sentry/syscalls/linux/sys_mmap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_mount.go b/pkg/sentry/syscalls/linux/sys_mount.go index e110a553f..cf613bad0 100644 --- a/pkg/sentry/syscalls/linux/sys_mount.go +++ b/pkg/sentry/syscalls/linux/sys_mount.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_pipe.go b/pkg/sentry/syscalls/linux/sys_pipe.go index 3652c429e..036845c13 100644 --- a/pkg/sentry/syscalls/linux/sys_pipe.go +++ b/pkg/sentry/syscalls/linux/sys_pipe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_poll.go b/pkg/sentry/syscalls/linux/sys_poll.go index 17b6768e5..e32099dd4 100644 --- a/pkg/sentry/syscalls/linux/sys_poll.go +++ b/pkg/sentry/syscalls/linux/sys_poll.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_prctl.go b/pkg/sentry/syscalls/linux/sys_prctl.go index 7a29bd9b7..117ae1a0e 100644 --- a/pkg/sentry/syscalls/linux/sys_prctl.go +++ b/pkg/sentry/syscalls/linux/sys_prctl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_random.go b/pkg/sentry/syscalls/linux/sys_random.go index 452dff058..fc3959a7e 100644 --- a/pkg/sentry/syscalls/linux/sys_random.go +++ b/pkg/sentry/syscalls/linux/sys_random.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_read.go b/pkg/sentry/syscalls/linux/sys_read.go index 50c7d7a74..48b0fd49d 100644 --- a/pkg/sentry/syscalls/linux/sys_read.go +++ b/pkg/sentry/syscalls/linux/sys_read.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_rlimit.go b/pkg/sentry/syscalls/linux/sys_rlimit.go index 443334693..8b0379779 100644 --- a/pkg/sentry/syscalls/linux/sys_rlimit.go +++ b/pkg/sentry/syscalls/linux/sys_rlimit.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_rusage.go b/pkg/sentry/syscalls/linux/sys_rusage.go index ab07c77f9..003d718da 100644 --- a/pkg/sentry/syscalls/linux/sys_rusage.go +++ b/pkg/sentry/syscalls/linux/sys_rusage.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_sched.go b/pkg/sentry/syscalls/linux/sys_sched.go index e679a6694..8aea03abe 100644 --- a/pkg/sentry/syscalls/linux/sys_sched.go +++ b/pkg/sentry/syscalls/linux/sys_sched.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_seccomp.go b/pkg/sentry/syscalls/linux/sys_seccomp.go index f08fdf5cb..b4262162a 100644 --- a/pkg/sentry/syscalls/linux/sys_seccomp.go +++ b/pkg/sentry/syscalls/linux/sys_seccomp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_sem.go b/pkg/sentry/syscalls/linux/sys_sem.go index 86f850ef1..5bd61ab87 100644 --- a/pkg/sentry/syscalls/linux/sys_sem.go +++ b/pkg/sentry/syscalls/linux/sys_sem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_shm.go b/pkg/sentry/syscalls/linux/sys_shm.go index a0d3a73c5..d0eceac7c 100644 --- a/pkg/sentry/syscalls/linux/sys_shm.go +++ b/pkg/sentry/syscalls/linux/sys_shm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_signal.go b/pkg/sentry/syscalls/linux/sys_signal.go index a539354c5..7fbeb4fcd 100644 --- a/pkg/sentry/syscalls/linux/sys_signal.go +++ b/pkg/sentry/syscalls/linux/sys_signal.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_socket.go b/pkg/sentry/syscalls/linux/sys_socket.go index c8748958a..69862f110 100644 --- a/pkg/sentry/syscalls/linux/sys_socket.go +++ b/pkg/sentry/syscalls/linux/sys_socket.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_stat.go b/pkg/sentry/syscalls/linux/sys_stat.go index 49c225011..10fc201ef 100644 --- a/pkg/sentry/syscalls/linux/sys_stat.go +++ b/pkg/sentry/syscalls/linux/sys_stat.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_sync.go b/pkg/sentry/syscalls/linux/sys_sync.go index 68488330f..4352482fb 100644 --- a/pkg/sentry/syscalls/linux/sys_sync.go +++ b/pkg/sentry/syscalls/linux/sys_sync.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_sysinfo.go b/pkg/sentry/syscalls/linux/sys_sysinfo.go index 6f7acf98f..ecf88edc1 100644 --- a/pkg/sentry/syscalls/linux/sys_sysinfo.go +++ b/pkg/sentry/syscalls/linux/sys_sysinfo.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_syslog.go b/pkg/sentry/syscalls/linux/sys_syslog.go index 7193b7aed..9efc58d34 100644 --- a/pkg/sentry/syscalls/linux/sys_syslog.go +++ b/pkg/sentry/syscalls/linux/sys_syslog.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index ddcb5b789..23c2f7035 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_time.go b/pkg/sentry/syscalls/linux/sys_time.go index 063fbb106..b4f2609c0 100644 --- a/pkg/sentry/syscalls/linux/sys_time.go +++ b/pkg/sentry/syscalls/linux/sys_time.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_timer.go b/pkg/sentry/syscalls/linux/sys_timer.go index 6baf4599b..04ea7a4e9 100644 --- a/pkg/sentry/syscalls/linux/sys_timer.go +++ b/pkg/sentry/syscalls/linux/sys_timer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_timerfd.go b/pkg/sentry/syscalls/linux/sys_timerfd.go index f70d13682..ec0155cbb 100644 --- a/pkg/sentry/syscalls/linux/sys_timerfd.go +++ b/pkg/sentry/syscalls/linux/sys_timerfd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_tls.go b/pkg/sentry/syscalls/linux/sys_tls.go index 8ea78093b..1e8312e00 100644 --- a/pkg/sentry/syscalls/linux/sys_tls.go +++ b/pkg/sentry/syscalls/linux/sys_tls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_utsname.go b/pkg/sentry/syscalls/linux/sys_utsname.go index f7545b965..fa81fe10e 100644 --- a/pkg/sentry/syscalls/linux/sys_utsname.go +++ b/pkg/sentry/syscalls/linux/sys_utsname.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/sys_write.go b/pkg/sentry/syscalls/linux/sys_write.go index e405608c4..1da72d606 100644 --- a/pkg/sentry/syscalls/linux/sys_write.go +++ b/pkg/sentry/syscalls/linux/sys_write.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/linux/timespec.go b/pkg/sentry/syscalls/linux/timespec.go index 752ec326d..fa6fcdc0b 100644 --- a/pkg/sentry/syscalls/linux/timespec.go +++ b/pkg/sentry/syscalls/linux/timespec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/syscalls/syscalls.go b/pkg/sentry/syscalls/syscalls.go index 425ce900c..5d10b3824 100644 --- a/pkg/sentry/syscalls/syscalls.go +++ b/pkg/sentry/syscalls/syscalls.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/calibrated_clock.go b/pkg/sentry/time/calibrated_clock.go index a98bcd7de..c27e391c9 100644 --- a/pkg/sentry/time/calibrated_clock.go +++ b/pkg/sentry/time/calibrated_clock.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/calibrated_clock_test.go b/pkg/sentry/time/calibrated_clock_test.go index a9237630e..d6622bfe2 100644 --- a/pkg/sentry/time/calibrated_clock_test.go +++ b/pkg/sentry/time/calibrated_clock_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/clock_id.go b/pkg/sentry/time/clock_id.go index 1317a5dad..724f59dd9 100644 --- a/pkg/sentry/time/clock_id.go +++ b/pkg/sentry/time/clock_id.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/clocks.go b/pkg/sentry/time/clocks.go index e26386520..837e86094 100644 --- a/pkg/sentry/time/clocks.go +++ b/pkg/sentry/time/clocks.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/muldiv_amd64.s b/pkg/sentry/time/muldiv_amd64.s index bfcb8c724..028c6684e 100644 --- a/pkg/sentry/time/muldiv_amd64.s +++ b/pkg/sentry/time/muldiv_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/muldiv_arm64.s b/pkg/sentry/time/muldiv_arm64.s index 5fa82a136..5ad57a8a3 100644 --- a/pkg/sentry/time/muldiv_arm64.s +++ b/pkg/sentry/time/muldiv_arm64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/parameters.go b/pkg/sentry/time/parameters.go index 8568b1193..63cf7c4a3 100644 --- a/pkg/sentry/time/parameters.go +++ b/pkg/sentry/time/parameters.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/parameters_test.go b/pkg/sentry/time/parameters_test.go index 4a0c4e880..e1b9084ac 100644 --- a/pkg/sentry/time/parameters_test.go +++ b/pkg/sentry/time/parameters_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/sampler.go b/pkg/sentry/time/sampler.go index 445690d49..2140a99b7 100644 --- a/pkg/sentry/time/sampler.go +++ b/pkg/sentry/time/sampler.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/sampler_test.go b/pkg/sentry/time/sampler_test.go index ec0e442b6..3e70a1134 100644 --- a/pkg/sentry/time/sampler_test.go +++ b/pkg/sentry/time/sampler_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/sampler_unsafe.go b/pkg/sentry/time/sampler_unsafe.go index 0f8eb4fc8..e76180217 100644 --- a/pkg/sentry/time/sampler_unsafe.go +++ b/pkg/sentry/time/sampler_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/tsc_amd64.s b/pkg/sentry/time/tsc_amd64.s index e53d477f7..6a8eed664 100644 --- a/pkg/sentry/time/tsc_amd64.s +++ b/pkg/sentry/time/tsc_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/time/tsc_arm64.s b/pkg/sentry/time/tsc_arm64.s index c1c9760ef..da9fa4112 100644 --- a/pkg/sentry/time/tsc_arm64.s +++ b/pkg/sentry/time/tsc_arm64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/unimpl/events.go b/pkg/sentry/unimpl/events.go index f78f8c981..d92766e2d 100644 --- a/pkg/sentry/unimpl/events.go +++ b/pkg/sentry/unimpl/events.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/unimpl/unimplemented_syscall.proto b/pkg/sentry/unimpl/unimplemented_syscall.proto index 41579b016..0d7a94be7 100644 --- a/pkg/sentry/unimpl/unimplemented_syscall.proto +++ b/pkg/sentry/unimpl/unimplemented_syscall.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/uniqueid/context.go b/pkg/sentry/uniqueid/context.go index 399d98c29..e55b89689 100644 --- a/pkg/sentry/uniqueid/context.go +++ b/pkg/sentry/uniqueid/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/cpu.go b/pkg/sentry/usage/cpu.go index cbd7cfe19..bfc282d69 100644 --- a/pkg/sentry/usage/cpu.go +++ b/pkg/sentry/usage/cpu.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/io.go b/pkg/sentry/usage/io.go index 8e27a0a88..dfcd3a49d 100644 --- a/pkg/sentry/usage/io.go +++ b/pkg/sentry/usage/io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/memory.go b/pkg/sentry/usage/memory.go index 5be9ed9c6..c316f1597 100644 --- a/pkg/sentry/usage/memory.go +++ b/pkg/sentry/usage/memory.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/memory_unsafe.go b/pkg/sentry/usage/memory_unsafe.go index a3ae668a5..9e0014ca0 100644 --- a/pkg/sentry/usage/memory_unsafe.go +++ b/pkg/sentry/usage/memory_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usage/usage.go b/pkg/sentry/usage/usage.go index ab327f8e2..e3d33a965 100644 --- a/pkg/sentry/usage/usage.go +++ b/pkg/sentry/usage/usage.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/access_type.go b/pkg/sentry/usermem/access_type.go index 9e6a27bcf..9c1742a59 100644 --- a/pkg/sentry/usermem/access_type.go +++ b/pkg/sentry/usermem/access_type.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/addr.go b/pkg/sentry/usermem/addr.go index 2a75aa60c..e79210804 100644 --- a/pkg/sentry/usermem/addr.go +++ b/pkg/sentry/usermem/addr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/addr_range_seq_test.go b/pkg/sentry/usermem/addr_range_seq_test.go index bd6a1ec8a..82f735026 100644 --- a/pkg/sentry/usermem/addr_range_seq_test.go +++ b/pkg/sentry/usermem/addr_range_seq_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/addr_range_seq_unsafe.go b/pkg/sentry/usermem/addr_range_seq_unsafe.go index f5fd446fa..c09337c15 100644 --- a/pkg/sentry/usermem/addr_range_seq_unsafe.go +++ b/pkg/sentry/usermem/addr_range_seq_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/bytes_io.go b/pkg/sentry/usermem/bytes_io.go index 274f568d0..f98d82168 100644 --- a/pkg/sentry/usermem/bytes_io.go +++ b/pkg/sentry/usermem/bytes_io.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/bytes_io_unsafe.go b/pkg/sentry/usermem/bytes_io_unsafe.go index 7add8bc82..bb49d2ff3 100644 --- a/pkg/sentry/usermem/bytes_io_unsafe.go +++ b/pkg/sentry/usermem/bytes_io_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/usermem.go b/pkg/sentry/usermem/usermem.go index 4c7d5014a..31e4d6ada 100644 --- a/pkg/sentry/usermem/usermem.go +++ b/pkg/sentry/usermem/usermem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/usermem_arm64.go b/pkg/sentry/usermem/usermem_arm64.go index 7fd4ce963..fdfc30a66 100644 --- a/pkg/sentry/usermem/usermem_arm64.go +++ b/pkg/sentry/usermem/usermem_arm64.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/usermem_test.go b/pkg/sentry/usermem/usermem_test.go index 1991a9641..4a07118b7 100644 --- a/pkg/sentry/usermem/usermem_test.go +++ b/pkg/sentry/usermem/usermem_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/usermem_unsafe.go b/pkg/sentry/usermem/usermem_unsafe.go index 3895e7871..876783e78 100644 --- a/pkg/sentry/usermem/usermem_unsafe.go +++ b/pkg/sentry/usermem/usermem_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/usermem/usermem_x86.go b/pkg/sentry/usermem/usermem_x86.go index 9ec90f9ff..8059b72d2 100644 --- a/pkg/sentry/usermem/usermem_x86.go +++ b/pkg/sentry/usermem/usermem_x86.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sentry/watchdog/watchdog.go b/pkg/sentry/watchdog/watchdog.go index b4f1e3a4f..2fc4472dd 100644 --- a/pkg/sentry/watchdog/watchdog.go +++ b/pkg/sentry/watchdog/watchdog.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/commit_amd64.s b/pkg/sleep/commit_amd64.s index d08df7f37..bc4ac2c3c 100644 --- a/pkg/sleep/commit_amd64.s +++ b/pkg/sleep/commit_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/commit_asm.go b/pkg/sleep/commit_asm.go index 90eef4cbc..35e2cc337 100644 --- a/pkg/sleep/commit_asm.go +++ b/pkg/sleep/commit_asm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/commit_noasm.go b/pkg/sleep/commit_noasm.go index 967d22e24..686b1da3d 100644 --- a/pkg/sleep/commit_noasm.go +++ b/pkg/sleep/commit_noasm.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/empty.s b/pkg/sleep/empty.s index 85d52cd9c..fb37360ac 100644 --- a/pkg/sleep/empty.s +++ b/pkg/sleep/empty.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/sleep_test.go b/pkg/sleep/sleep_test.go index 8feb9ffc2..130806c86 100644 --- a/pkg/sleep/sleep_test.go +++ b/pkg/sleep/sleep_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/sleep/sleep_unsafe.go b/pkg/sleep/sleep_unsafe.go index 45fb6f0ea..62e0abc34 100644 --- a/pkg/sleep/sleep_unsafe.go +++ b/pkg/sleep/sleep_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/decode.go b/pkg/state/decode.go index 54b5ad8b8..73a59f871 100644 --- a/pkg/state/decode.go +++ b/pkg/state/decode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/encode.go b/pkg/state/encode.go index fe8512bbf..b0714170b 100644 --- a/pkg/state/encode.go +++ b/pkg/state/encode.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/encode_unsafe.go b/pkg/state/encode_unsafe.go index be94742a8..457e6dbb7 100644 --- a/pkg/state/encode_unsafe.go +++ b/pkg/state/encode_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/map.go b/pkg/state/map.go index 0035d7250..1fb9b47b8 100644 --- a/pkg/state/map.go +++ b/pkg/state/map.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/object.proto b/pkg/state/object.proto index d3b46ea97..952289069 100644 --- a/pkg/state/object.proto +++ b/pkg/state/object.proto @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/printer.go b/pkg/state/printer.go index aee4b69fb..5174c3ba3 100644 --- a/pkg/state/printer.go +++ b/pkg/state/printer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/state.go b/pkg/state/state.go index 4486f83a7..cf7df803a 100644 --- a/pkg/state/state.go +++ b/pkg/state/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/state_test.go b/pkg/state/state_test.go index 22bcad9e1..7c24bbcda 100644 --- a/pkg/state/state_test.go +++ b/pkg/state/state_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/statefile/statefile.go b/pkg/state/statefile/statefile.go index c21e3bb0e..ad4e3b43e 100644 --- a/pkg/state/statefile/statefile.go +++ b/pkg/state/statefile/statefile.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/statefile/statefile_test.go b/pkg/state/statefile/statefile_test.go index b4f400e01..60b769895 100644 --- a/pkg/state/statefile/statefile_test.go +++ b/pkg/state/statefile/statefile_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/state/stats.go b/pkg/state/stats.go index 17ca258fc..eb51cda47 100644 --- a/pkg/state/stats.go +++ b/pkg/state/stats.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserr/host_linux.go b/pkg/syserr/host_linux.go index 74bbe9f5b..fc6ef60a1 100644 --- a/pkg/syserr/host_linux.go +++ b/pkg/syserr/host_linux.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserr/netstack.go b/pkg/syserr/netstack.go index 1a23919ef..bd489b424 100644 --- a/pkg/syserr/netstack.go +++ b/pkg/syserr/netstack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserr/syserr.go b/pkg/syserr/syserr.go index 232634dd4..4ddbd3322 100644 --- a/pkg/syserr/syserr.go +++ b/pkg/syserr/syserr.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserror/syserror.go b/pkg/syserror/syserror.go index 5558cccff..345653544 100644 --- a/pkg/syserror/syserror.go +++ b/pkg/syserror/syserror.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/syserror/syserror_test.go b/pkg/syserror/syserror_test.go index 0f0da5781..f2a10ee7b 100644 --- a/pkg/syserror/syserror_test.go +++ b/pkg/syserror/syserror_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/adapters/gonet/gonet.go b/pkg/tcpip/adapters/gonet/gonet.go index 628e28f57..df8bf435d 100644 --- a/pkg/tcpip/adapters/gonet/gonet.go +++ b/pkg/tcpip/adapters/gonet/gonet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/adapters/gonet/gonet_test.go b/pkg/tcpip/adapters/gonet/gonet_test.go index e84f73feb..2c81c5697 100644 --- a/pkg/tcpip/adapters/gonet/gonet_test.go +++ b/pkg/tcpip/adapters/gonet/gonet_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/buffer/prependable.go b/pkg/tcpip/buffer/prependable.go index d3a9a0f88..43cbbc74c 100644 --- a/pkg/tcpip/buffer/prependable.go +++ b/pkg/tcpip/buffer/prependable.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/buffer/view.go b/pkg/tcpip/buffer/view.go index 43cbb9461..1a9d40778 100644 --- a/pkg/tcpip/buffer/view.go +++ b/pkg/tcpip/buffer/view.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/buffer/view_test.go b/pkg/tcpip/buffer/view_test.go index 74a0a96fc..ebc3a17b7 100644 --- a/pkg/tcpip/buffer/view_test.go +++ b/pkg/tcpip/buffer/view_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/checker/checker.go b/pkg/tcpip/checker/checker.go index 5dfb3ca1d..6e7edf3ab 100644 --- a/pkg/tcpip/checker/checker.go +++ b/pkg/tcpip/checker/checker.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/hash/jenkins/jenkins.go b/pkg/tcpip/hash/jenkins/jenkins.go index e66d5f12b..52c22230e 100644 --- a/pkg/tcpip/hash/jenkins/jenkins.go +++ b/pkg/tcpip/hash/jenkins/jenkins.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/hash/jenkins/jenkins_test.go b/pkg/tcpip/hash/jenkins/jenkins_test.go index 9d86174aa..4c78b5808 100644 --- a/pkg/tcpip/hash/jenkins/jenkins_test.go +++ b/pkg/tcpip/hash/jenkins/jenkins_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/arp.go b/pkg/tcpip/header/arp.go index 22b259ccb..55fe7292c 100644 --- a/pkg/tcpip/header/arp.go +++ b/pkg/tcpip/header/arp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/checksum.go b/pkg/tcpip/header/checksum.go index 2e8c65fac..2eaa7938a 100644 --- a/pkg/tcpip/header/checksum.go +++ b/pkg/tcpip/header/checksum.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/eth.go b/pkg/tcpip/header/eth.go index 77365bc41..76143f454 100644 --- a/pkg/tcpip/header/eth.go +++ b/pkg/tcpip/header/eth.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/gue.go b/pkg/tcpip/header/gue.go index 2ad13955a..10d358c0e 100644 --- a/pkg/tcpip/header/gue.go +++ b/pkg/tcpip/header/gue.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/icmpv4.go b/pkg/tcpip/header/icmpv4.go index 3ac89cdae..782e1053c 100644 --- a/pkg/tcpip/header/icmpv4.go +++ b/pkg/tcpip/header/icmpv4.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/icmpv6.go b/pkg/tcpip/header/icmpv6.go index e317975e8..d0b10d849 100644 --- a/pkg/tcpip/header/icmpv6.go +++ b/pkg/tcpip/header/icmpv6.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/interfaces.go b/pkg/tcpip/header/interfaces.go index ac327d8a5..fb250ea30 100644 --- a/pkg/tcpip/header/interfaces.go +++ b/pkg/tcpip/header/interfaces.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/ipv4.go b/pkg/tcpip/header/ipv4.go index c3b8fb00e..96e461491 100644 --- a/pkg/tcpip/header/ipv4.go +++ b/pkg/tcpip/header/ipv4.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/ipv6.go b/pkg/tcpip/header/ipv6.go index 3d24736c7..66820a466 100644 --- a/pkg/tcpip/header/ipv6.go +++ b/pkg/tcpip/header/ipv6.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/ipv6_fragment.go b/pkg/tcpip/header/ipv6_fragment.go index e36d5177b..6d896355a 100644 --- a/pkg/tcpip/header/ipv6_fragment.go +++ b/pkg/tcpip/header/ipv6_fragment.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/ipversion_test.go b/pkg/tcpip/header/ipversion_test.go index 8301ba5cf..0c830180e 100644 --- a/pkg/tcpip/header/ipversion_test.go +++ b/pkg/tcpip/header/ipversion_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/tcp.go b/pkg/tcpip/header/tcp.go index e656ebb15..0cd89b992 100644 --- a/pkg/tcpip/header/tcp.go +++ b/pkg/tcpip/header/tcp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/tcp_test.go b/pkg/tcpip/header/tcp_test.go index 7cd98df3b..9a2b99489 100644 --- a/pkg/tcpip/header/tcp_test.go +++ b/pkg/tcpip/header/tcp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/header/udp.go b/pkg/tcpip/header/udp.go index e8c860436..2205fec18 100644 --- a/pkg/tcpip/header/udp.go +++ b/pkg/tcpip/header/udp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go index f7501a1bc..ee9dd8700 100644 --- a/pkg/tcpip/link/channel/channel.go +++ b/pkg/tcpip/link/channel/channel.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go index 8f4d67074..4da376774 100644 --- a/pkg/tcpip/link/fdbased/endpoint.go +++ b/pkg/tcpip/link/fdbased/endpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/fdbased/endpoint_test.go b/pkg/tcpip/link/fdbased/endpoint_test.go index c8b037d57..31138e4ac 100644 --- a/pkg/tcpip/link/fdbased/endpoint_test.go +++ b/pkg/tcpip/link/fdbased/endpoint_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/fdbased/endpoint_unsafe.go b/pkg/tcpip/link/fdbased/endpoint_unsafe.go index 36e7fe5a9..97a477b61 100644 --- a/pkg/tcpip/link/fdbased/endpoint_unsafe.go +++ b/pkg/tcpip/link/fdbased/endpoint_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/fdbased/mmap.go b/pkg/tcpip/link/fdbased/mmap.go index f1e71c233..430c85a42 100644 --- a/pkg/tcpip/link/fdbased/mmap.go +++ b/pkg/tcpip/link/fdbased/mmap.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/fdbased/mmap_amd64_unsafe.go b/pkg/tcpip/link/fdbased/mmap_amd64_unsafe.go index e5ac7996d..135da2498 100644 --- a/pkg/tcpip/link/fdbased/mmap_amd64_unsafe.go +++ b/pkg/tcpip/link/fdbased/mmap_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/loopback/loopback.go b/pkg/tcpip/link/loopback/loopback.go index 2dc4bcfda..2c1148123 100644 --- a/pkg/tcpip/link/loopback/loopback.go +++ b/pkg/tcpip/link/loopback/loopback.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/muxed/injectable.go b/pkg/tcpip/link/muxed/injectable.go index b3e71c7fc..be07b7c29 100644 --- a/pkg/tcpip/link/muxed/injectable.go +++ b/pkg/tcpip/link/muxed/injectable.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/muxed/injectable_test.go b/pkg/tcpip/link/muxed/injectable_test.go index 031449a05..5d40dfacc 100644 --- a/pkg/tcpip/link/muxed/injectable_test.go +++ b/pkg/tcpip/link/muxed/injectable_test.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/blockingpoll_amd64.s b/pkg/tcpip/link/rawfile/blockingpoll_amd64.s index 9dade5421..b54131573 100644 --- a/pkg/tcpip/link/rawfile/blockingpoll_amd64.s +++ b/pkg/tcpip/link/rawfile/blockingpoll_amd64.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/blockingpoll_amd64_unsafe.go b/pkg/tcpip/link/rawfile/blockingpoll_amd64_unsafe.go index 3ba96a123..0b51982c6 100644 --- a/pkg/tcpip/link/rawfile/blockingpoll_amd64_unsafe.go +++ b/pkg/tcpip/link/rawfile/blockingpoll_amd64_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/blockingpoll_unsafe.go b/pkg/tcpip/link/rawfile/blockingpoll_unsafe.go index 94ddad8ea..4eab77c74 100644 --- a/pkg/tcpip/link/rawfile/blockingpoll_unsafe.go +++ b/pkg/tcpip/link/rawfile/blockingpoll_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/errors.go b/pkg/tcpip/link/rawfile/errors.go index 7359849b1..8bde41637 100644 --- a/pkg/tcpip/link/rawfile/errors.go +++ b/pkg/tcpip/link/rawfile/errors.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/rawfile/rawfile_unsafe.go b/pkg/tcpip/link/rawfile/rawfile_unsafe.go index fe2779125..86db7a487 100644 --- a/pkg/tcpip/link/rawfile/rawfile_unsafe.go +++ b/pkg/tcpip/link/rawfile/rawfile_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/pipe.go b/pkg/tcpip/link/sharedmem/pipe/pipe.go index e014324cc..74c9f0311 100644 --- a/pkg/tcpip/link/sharedmem/pipe/pipe.go +++ b/pkg/tcpip/link/sharedmem/pipe/pipe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/pipe_test.go b/pkg/tcpip/link/sharedmem/pipe/pipe_test.go index 30742ccb1..59ef69a8b 100644 --- a/pkg/tcpip/link/sharedmem/pipe/pipe_test.go +++ b/pkg/tcpip/link/sharedmem/pipe/pipe_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go b/pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go index f491d74a2..62d17029e 100644 --- a/pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go +++ b/pkg/tcpip/link/sharedmem/pipe/pipe_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/rx.go b/pkg/tcpip/link/sharedmem/pipe/rx.go index 8d641c76f..f22e533ac 100644 --- a/pkg/tcpip/link/sharedmem/pipe/rx.go +++ b/pkg/tcpip/link/sharedmem/pipe/rx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/pipe/tx.go b/pkg/tcpip/link/sharedmem/pipe/tx.go index e75175d98..9841eb231 100644 --- a/pkg/tcpip/link/sharedmem/pipe/tx.go +++ b/pkg/tcpip/link/sharedmem/pipe/tx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/queue/queue_test.go b/pkg/tcpip/link/sharedmem/queue/queue_test.go index 391165bc3..d3f8f4b8b 100644 --- a/pkg/tcpip/link/sharedmem/queue/queue_test.go +++ b/pkg/tcpip/link/sharedmem/queue/queue_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/queue/rx.go b/pkg/tcpip/link/sharedmem/queue/rx.go index d3a5da08a..d9aecf2d9 100644 --- a/pkg/tcpip/link/sharedmem/queue/rx.go +++ b/pkg/tcpip/link/sharedmem/queue/rx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/queue/tx.go b/pkg/tcpip/link/sharedmem/queue/tx.go index 845108db1..a24dccd11 100644 --- a/pkg/tcpip/link/sharedmem/queue/tx.go +++ b/pkg/tcpip/link/sharedmem/queue/tx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/rx.go b/pkg/tcpip/link/sharedmem/rx.go index 3eeab769e..215cb607f 100644 --- a/pkg/tcpip/link/sharedmem/rx.go +++ b/pkg/tcpip/link/sharedmem/rx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/sharedmem.go b/pkg/tcpip/link/sharedmem/sharedmem.go index 6e6aa5a13..e34b780f8 100644 --- a/pkg/tcpip/link/sharedmem/sharedmem.go +++ b/pkg/tcpip/link/sharedmem/sharedmem.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/sharedmem_test.go b/pkg/tcpip/link/sharedmem/sharedmem_test.go index 1f44e224c..65b9d7085 100644 --- a/pkg/tcpip/link/sharedmem/sharedmem_test.go +++ b/pkg/tcpip/link/sharedmem/sharedmem_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/sharedmem_unsafe.go b/pkg/tcpip/link/sharedmem/sharedmem_unsafe.go index b91adbaf7..f7e816a41 100644 --- a/pkg/tcpip/link/sharedmem/sharedmem_unsafe.go +++ b/pkg/tcpip/link/sharedmem/sharedmem_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sharedmem/tx.go b/pkg/tcpip/link/sharedmem/tx.go index 37da34831..ac3577aa6 100644 --- a/pkg/tcpip/link/sharedmem/tx.go +++ b/pkg/tcpip/link/sharedmem/tx.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sniffer/pcap.go b/pkg/tcpip/link/sniffer/pcap.go index 3d0d8d852..c16c19647 100644 --- a/pkg/tcpip/link/sniffer/pcap.go +++ b/pkg/tcpip/link/sniffer/pcap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/sniffer/sniffer.go b/pkg/tcpip/link/sniffer/sniffer.go index 462a6e3a3..e87ae07d7 100644 --- a/pkg/tcpip/link/sniffer/sniffer.go +++ b/pkg/tcpip/link/sniffer/sniffer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/tun/tun_unsafe.go b/pkg/tcpip/link/tun/tun_unsafe.go index e4c589dda..09ca9b527 100644 --- a/pkg/tcpip/link/tun/tun_unsafe.go +++ b/pkg/tcpip/link/tun/tun_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/waitable/waitable.go b/pkg/tcpip/link/waitable/waitable.go index bd9f9845b..21690a226 100644 --- a/pkg/tcpip/link/waitable/waitable.go +++ b/pkg/tcpip/link/waitable/waitable.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/link/waitable/waitable_test.go b/pkg/tcpip/link/waitable/waitable_test.go index a2df6be95..62054fb7f 100644 --- a/pkg/tcpip/link/waitable/waitable_test.go +++ b/pkg/tcpip/link/waitable/waitable_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/arp/arp.go b/pkg/tcpip/network/arp/arp.go index 975919e80..a3f2bce3e 100644 --- a/pkg/tcpip/network/arp/arp.go +++ b/pkg/tcpip/network/arp/arp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/arp/arp_test.go b/pkg/tcpip/network/arp/arp_test.go index 14b9cb8b6..1b971b1a3 100644 --- a/pkg/tcpip/network/arp/arp_test.go +++ b/pkg/tcpip/network/arp/arp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/frag_heap.go b/pkg/tcpip/network/fragmentation/frag_heap.go index 55615c8e6..9ad3e5a8a 100644 --- a/pkg/tcpip/network/fragmentation/frag_heap.go +++ b/pkg/tcpip/network/fragmentation/frag_heap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/frag_heap_test.go b/pkg/tcpip/network/fragmentation/frag_heap_test.go index 1b1b72e88..3a2486ba8 100644 --- a/pkg/tcpip/network/fragmentation/frag_heap_test.go +++ b/pkg/tcpip/network/fragmentation/frag_heap_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/fragmentation.go b/pkg/tcpip/network/fragmentation/fragmentation.go index a5dda0398..e90edb375 100644 --- a/pkg/tcpip/network/fragmentation/fragmentation.go +++ b/pkg/tcpip/network/fragmentation/fragmentation.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/fragmentation_test.go b/pkg/tcpip/network/fragmentation/fragmentation_test.go index 5bf3463a9..99ded68a3 100644 --- a/pkg/tcpip/network/fragmentation/fragmentation_test.go +++ b/pkg/tcpip/network/fragmentation/fragmentation_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/reassembler.go b/pkg/tcpip/network/fragmentation/reassembler.go index c9ad2bef6..04f9ab964 100644 --- a/pkg/tcpip/network/fragmentation/reassembler.go +++ b/pkg/tcpip/network/fragmentation/reassembler.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/fragmentation/reassembler_test.go b/pkg/tcpip/network/fragmentation/reassembler_test.go index a2bc9707a..7eee0710d 100644 --- a/pkg/tcpip/network/fragmentation/reassembler_test.go +++ b/pkg/tcpip/network/fragmentation/reassembler_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/hash/hash.go b/pkg/tcpip/network/hash/hash.go index 07960ddf0..0c91905dc 100644 --- a/pkg/tcpip/network/hash/hash.go +++ b/pkg/tcpip/network/hash/hash.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ip_test.go b/pkg/tcpip/network/ip_test.go index 522009fac..4b822e2c6 100644 --- a/pkg/tcpip/network/ip_test.go +++ b/pkg/tcpip/network/ip_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv4/icmp.go b/pkg/tcpip/network/ipv4/icmp.go index 1c3acda4b..9cb81245a 100644 --- a/pkg/tcpip/network/ipv4/icmp.go +++ b/pkg/tcpip/network/ipv4/icmp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go index cbdca98a5..c6af0db79 100644 --- a/pkg/tcpip/network/ipv4/ipv4.go +++ b/pkg/tcpip/network/ipv4/ipv4.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv4/ipv4_test.go b/pkg/tcpip/network/ipv4/ipv4_test.go index 42e85564e..146143ab3 100644 --- a/pkg/tcpip/network/ipv4/ipv4_test.go +++ b/pkg/tcpip/network/ipv4/ipv4_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go index be28be36d..9c011e107 100644 --- a/pkg/tcpip/network/ipv6/icmp.go +++ b/pkg/tcpip/network/ipv6/icmp.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv6/icmp_test.go b/pkg/tcpip/network/ipv6/icmp_test.go index 8b57a0641..d8737a616 100644 --- a/pkg/tcpip/network/ipv6/icmp_test.go +++ b/pkg/tcpip/network/ipv6/icmp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go index 9a743ea80..4b8cd496b 100644 --- a/pkg/tcpip/network/ipv6/ipv6.go +++ b/pkg/tcpip/network/ipv6/ipv6.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/ports/ports.go b/pkg/tcpip/ports/ports.go index d212a5792..a1712b590 100644 --- a/pkg/tcpip/ports/ports.go +++ b/pkg/tcpip/ports/ports.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/ports/ports_test.go b/pkg/tcpip/ports/ports_test.go index 01e7320b4..8466c661b 100644 --- a/pkg/tcpip/ports/ports_test.go +++ b/pkg/tcpip/ports/ports_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/sample/tun_tcp_connect/main.go b/pkg/tcpip/sample/tun_tcp_connect/main.go index cf8900c4d..1681de56e 100644 --- a/pkg/tcpip/sample/tun_tcp_connect/main.go +++ b/pkg/tcpip/sample/tun_tcp_connect/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/sample/tun_tcp_echo/main.go b/pkg/tcpip/sample/tun_tcp_echo/main.go index da6202f97..642607f83 100644 --- a/pkg/tcpip/sample/tun_tcp_echo/main.go +++ b/pkg/tcpip/sample/tun_tcp_echo/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/seqnum/seqnum.go b/pkg/tcpip/seqnum/seqnum.go index f2b988839..b40a3c212 100644 --- a/pkg/tcpip/seqnum/seqnum.go +++ b/pkg/tcpip/seqnum/seqnum.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/linkaddrcache.go b/pkg/tcpip/stack/linkaddrcache.go index 40e4bdb4a..42b9768ae 100644 --- a/pkg/tcpip/stack/linkaddrcache.go +++ b/pkg/tcpip/stack/linkaddrcache.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/linkaddrcache_test.go b/pkg/tcpip/stack/linkaddrcache_test.go index 77a09ca86..91b2ffea8 100644 --- a/pkg/tcpip/stack/linkaddrcache_test.go +++ b/pkg/tcpip/stack/linkaddrcache_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go index c18571b0f..8008d9870 100644 --- a/pkg/tcpip/stack/nic.go +++ b/pkg/tcpip/stack/nic.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/registration.go b/pkg/tcpip/stack/registration.go index 6e1660051..c70533a35 100644 --- a/pkg/tcpip/stack/registration.go +++ b/pkg/tcpip/stack/registration.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/route.go b/pkg/tcpip/stack/route.go index 8ae562dcd..3d4c282a9 100644 --- a/pkg/tcpip/stack/route.go +++ b/pkg/tcpip/stack/route.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index cb9ffe9c2..f204ca790 100644 --- a/pkg/tcpip/stack/stack.go +++ b/pkg/tcpip/stack/stack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/stack_global_state.go b/pkg/tcpip/stack/stack_global_state.go index 3d7e4b719..dfec4258a 100644 --- a/pkg/tcpip/stack/stack_global_state.go +++ b/pkg/tcpip/stack/stack_global_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/stack_test.go b/pkg/tcpip/stack/stack_test.go index b5375df3c..351f63221 100644 --- a/pkg/tcpip/stack/stack_test.go +++ b/pkg/tcpip/stack/stack_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/transport_demuxer.go b/pkg/tcpip/stack/transport_demuxer.go index a8ac18e72..e8b562ad9 100644 --- a/pkg/tcpip/stack/transport_demuxer.go +++ b/pkg/tcpip/stack/transport_demuxer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/stack/transport_test.go b/pkg/tcpip/stack/transport_test.go index 2df974bf2..8d74f1543 100644 --- a/pkg/tcpip/stack/transport_test.go +++ b/pkg/tcpip/stack/transport_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/tcpip.go b/pkg/tcpip/tcpip.go index b09137f08..9367c8c02 100644 --- a/pkg/tcpip/tcpip.go +++ b/pkg/tcpip/tcpip.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/tcpip_test.go b/pkg/tcpip/tcpip_test.go index 1f7b04398..ebb1c1b56 100644 --- a/pkg/tcpip/tcpip_test.go +++ b/pkg/tcpip/tcpip_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/time.s b/pkg/tcpip/time.s index 85d52cd9c..fb37360ac 100644 --- a/pkg/tcpip/time.s +++ b/pkg/tcpip/time.s @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/time_unsafe.go b/pkg/tcpip/time_unsafe.go index 7ec5741af..1a307483b 100644 --- a/pkg/tcpip/time_unsafe.go +++ b/pkg/tcpip/time_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/icmp/endpoint.go b/pkg/tcpip/transport/icmp/endpoint.go index 8f2e3aa20..00840cfcf 100644 --- a/pkg/tcpip/transport/icmp/endpoint.go +++ b/pkg/tcpip/transport/icmp/endpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/icmp/endpoint_state.go b/pkg/tcpip/transport/icmp/endpoint_state.go index 8a7909246..332b3cd33 100644 --- a/pkg/tcpip/transport/icmp/endpoint_state.go +++ b/pkg/tcpip/transport/icmp/endpoint_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/icmp/protocol.go b/pkg/tcpip/transport/icmp/protocol.go index 09ee2f892..954fde9d8 100644 --- a/pkg/tcpip/transport/icmp/protocol.go +++ b/pkg/tcpip/transport/icmp/protocol.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/raw/raw.go b/pkg/tcpip/transport/raw/raw.go index f0f60ce91..7004c7ff4 100644 --- a/pkg/tcpip/transport/raw/raw.go +++ b/pkg/tcpip/transport/raw/raw.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/raw/state.go b/pkg/tcpip/transport/raw/state.go index e3891a8b8..e8907ebb1 100644 --- a/pkg/tcpip/transport/raw/state.go +++ b/pkg/tcpip/transport/raw/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/accept.go b/pkg/tcpip/transport/tcp/accept.go index a3894ed8f..e506d7133 100644 --- a/pkg/tcpip/transport/tcp/accept.go +++ b/pkg/tcpip/transport/tcp/accept.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/connect.go b/pkg/tcpip/transport/tcp/connect.go index 6c4a4d95e..eaa67aeb7 100644 --- a/pkg/tcpip/transport/tcp/connect.go +++ b/pkg/tcpip/transport/tcp/connect.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/cubic.go b/pkg/tcpip/transport/tcp/cubic.go index 003525d86..e618cd2b9 100644 --- a/pkg/tcpip/transport/tcp/cubic.go +++ b/pkg/tcpip/transport/tcp/cubic.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/dual_stack_test.go b/pkg/tcpip/transport/tcp/dual_stack_test.go index 2886cc707..43bcfa070 100644 --- a/pkg/tcpip/transport/tcp/dual_stack_test.go +++ b/pkg/tcpip/transport/tcp/dual_stack_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/endpoint.go b/pkg/tcpip/transport/tcp/endpoint.go index 09eff5be1..982f491cc 100644 --- a/pkg/tcpip/transport/tcp/endpoint.go +++ b/pkg/tcpip/transport/tcp/endpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/endpoint_state.go b/pkg/tcpip/transport/tcp/endpoint_state.go index 7f9dabb4d..27b0be046 100644 --- a/pkg/tcpip/transport/tcp/endpoint_state.go +++ b/pkg/tcpip/transport/tcp/endpoint_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/forwarder.go b/pkg/tcpip/transport/tcp/forwarder.go index 6a7efaf1d..e088e24cb 100644 --- a/pkg/tcpip/transport/tcp/forwarder.go +++ b/pkg/tcpip/transport/tcp/forwarder.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/protocol.go b/pkg/tcpip/transport/tcp/protocol.go index b5fb160bc..b86473891 100644 --- a/pkg/tcpip/transport/tcp/protocol.go +++ b/pkg/tcpip/transport/tcp/protocol.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/rcv.go b/pkg/tcpip/transport/tcp/rcv.go index fa6bdddba..b08a0e356 100644 --- a/pkg/tcpip/transport/tcp/rcv.go +++ b/pkg/tcpip/transport/tcp/rcv.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/reno.go b/pkg/tcpip/transport/tcp/reno.go index e4f8b7d5a..f83ebc717 100644 --- a/pkg/tcpip/transport/tcp/reno.go +++ b/pkg/tcpip/transport/tcp/reno.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/sack.go b/pkg/tcpip/transport/tcp/sack.go index 24e48fe7b..6a013d99b 100644 --- a/pkg/tcpip/transport/tcp/sack.go +++ b/pkg/tcpip/transport/tcp/sack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/sack_scoreboard.go b/pkg/tcpip/transport/tcp/sack_scoreboard.go index 21878ad82..99560d5b4 100644 --- a/pkg/tcpip/transport/tcp/sack_scoreboard.go +++ b/pkg/tcpip/transport/tcp/sack_scoreboard.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/sack_scoreboard_test.go b/pkg/tcpip/transport/tcp/sack_scoreboard_test.go index 3cf2ff451..8f6890cdf 100644 --- a/pkg/tcpip/transport/tcp/sack_scoreboard_test.go +++ b/pkg/tcpip/transport/tcp/sack_scoreboard_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/segment.go b/pkg/tcpip/transport/tcp/segment.go index c603fe713..187effb6b 100644 --- a/pkg/tcpip/transport/tcp/segment.go +++ b/pkg/tcpip/transport/tcp/segment.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/segment_heap.go b/pkg/tcpip/transport/tcp/segment_heap.go index 98422fadf..9fd061d7d 100644 --- a/pkg/tcpip/transport/tcp/segment_heap.go +++ b/pkg/tcpip/transport/tcp/segment_heap.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/segment_queue.go b/pkg/tcpip/transport/tcp/segment_queue.go index 0c637d7ad..3b020e580 100644 --- a/pkg/tcpip/transport/tcp/segment_queue.go +++ b/pkg/tcpip/transport/tcp/segment_queue.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/segment_state.go b/pkg/tcpip/transport/tcp/segment_state.go index 68b049f06..dd7e14aa6 100644 --- a/pkg/tcpip/transport/tcp/segment_state.go +++ b/pkg/tcpip/transport/tcp/segment_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/snd.go b/pkg/tcpip/transport/tcp/snd.go index 6317748cf..50743670e 100644 --- a/pkg/tcpip/transport/tcp/snd.go +++ b/pkg/tcpip/transport/tcp/snd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/snd_state.go b/pkg/tcpip/transport/tcp/snd_state.go index 86bbd643f..12eff8afc 100644 --- a/pkg/tcpip/transport/tcp/snd_state.go +++ b/pkg/tcpip/transport/tcp/snd_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/tcp_sack_test.go b/pkg/tcpip/transport/tcp/tcp_sack_test.go index 06b0702c5..dbfbd5c4f 100644 --- a/pkg/tcpip/transport/tcp/tcp_sack_test.go +++ b/pkg/tcpip/transport/tcp/tcp_sack_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/tcp_test.go b/pkg/tcpip/transport/tcp/tcp_test.go index c5732ad1c..a8b290dae 100644 --- a/pkg/tcpip/transport/tcp/tcp_test.go +++ b/pkg/tcpip/transport/tcp/tcp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/tcp_timestamp_test.go b/pkg/tcpip/transport/tcp/tcp_timestamp_test.go index 87c640967..039bbcfba 100644 --- a/pkg/tcpip/transport/tcp/tcp_timestamp_test.go +++ b/pkg/tcpip/transport/tcp/tcp_timestamp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/testing/context/context.go b/pkg/tcpip/transport/tcp/testing/context/context.go index 6e2fed880..fa721a7f8 100644 --- a/pkg/tcpip/transport/tcp/testing/context/context.go +++ b/pkg/tcpip/transport/tcp/testing/context/context.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcp/timer.go b/pkg/tcpip/transport/tcp/timer.go index 38240d2d5..fc1c7cbd2 100644 --- a/pkg/tcpip/transport/tcp/timer.go +++ b/pkg/tcpip/transport/tcp/timer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go b/pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go index b94568fb1..f1dcd36d5 100644 --- a/pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go +++ b/pkg/tcpip/transport/tcpconntrack/tcp_conntrack.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go b/pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go index aaeae9b18..435e136de 100644 --- a/pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go +++ b/pkg/tcpip/transport/tcpconntrack/tcp_conntrack_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/endpoint.go b/pkg/tcpip/transport/udp/endpoint.go index 1f9251de3..db65a4e88 100644 --- a/pkg/tcpip/transport/udp/endpoint.go +++ b/pkg/tcpip/transport/udp/endpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/endpoint_state.go b/pkg/tcpip/transport/udp/endpoint_state.go index b2daaf751..163dcbc13 100644 --- a/pkg/tcpip/transport/udp/endpoint_state.go +++ b/pkg/tcpip/transport/udp/endpoint_state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/forwarder.go b/pkg/tcpip/transport/udp/forwarder.go index d80c47e34..25bdd2929 100644 --- a/pkg/tcpip/transport/udp/forwarder.go +++ b/pkg/tcpip/transport/udp/forwarder.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/protocol.go b/pkg/tcpip/transport/udp/protocol.go index 616a9f388..8b47cce17 100644 --- a/pkg/tcpip/transport/udp/protocol.go +++ b/pkg/tcpip/transport/udp/protocol.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tcpip/transport/udp/udp_test.go b/pkg/tcpip/transport/udp/udp_test.go index 2f4e94c58..86a8fa19b 100644 --- a/pkg/tcpip/transport/udp/udp_test.go +++ b/pkg/tcpip/transport/udp/udp_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tmutex/tmutex.go b/pkg/tmutex/tmutex.go index df61d89f5..c4685020d 100644 --- a/pkg/tmutex/tmutex.go +++ b/pkg/tmutex/tmutex.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/tmutex/tmutex_test.go b/pkg/tmutex/tmutex_test.go index a4537cb3b..ce34c7962 100644 --- a/pkg/tmutex/tmutex_test.go +++ b/pkg/tmutex/tmutex_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/unet/unet.go b/pkg/unet/unet.go index 114fb8c5b..2aa1af4ff 100644 --- a/pkg/unet/unet.go +++ b/pkg/unet/unet.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/unet/unet_test.go b/pkg/unet/unet_test.go index db5485539..763b23c7c 100644 --- a/pkg/unet/unet_test.go +++ b/pkg/unet/unet_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/unet/unet_unsafe.go b/pkg/unet/unet_unsafe.go index 1d6ec286c..fa0916439 100644 --- a/pkg/unet/unet_unsafe.go +++ b/pkg/unet/unet_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/urpc/urpc.go b/pkg/urpc/urpc.go index 719f0e92f..0f155ec74 100644 --- a/pkg/urpc/urpc.go +++ b/pkg/urpc/urpc.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/urpc/urpc_test.go b/pkg/urpc/urpc_test.go index f1b9a85ca..5bf2c5ed2 100644 --- a/pkg/urpc/urpc_test.go +++ b/pkg/urpc/urpc_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/waiter/waiter.go b/pkg/waiter/waiter.go index a6c9dff3c..8a65ed164 100644 --- a/pkg/waiter/waiter.go +++ b/pkg/waiter/waiter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/pkg/waiter/waiter_test.go b/pkg/waiter/waiter_test.go index 60853f9c1..c1b94a4f3 100644 --- a/pkg/waiter/waiter_test.go +++ b/pkg/waiter/waiter_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/compat.go b/runsc/boot/compat.go index b3499bcde..c1b33c551 100644 --- a/runsc/boot/compat.go +++ b/runsc/boot/compat.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/compat_amd64.go b/runsc/boot/compat_amd64.go index 0c9472f18..99df5e614 100644 --- a/runsc/boot/compat_amd64.go +++ b/runsc/boot/compat_amd64.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/compat_test.go b/runsc/boot/compat_test.go index f1940dd72..ccec3d20c 100644 --- a/runsc/boot/compat_test.go +++ b/runsc/boot/compat_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/config.go b/runsc/boot/config.go index ba47effc1..b6771de30 100644 --- a/runsc/boot/config.go +++ b/runsc/boot/config.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/controller.go b/runsc/boot/controller.go index 712c50ee9..ab7c58838 100644 --- a/runsc/boot/controller.go +++ b/runsc/boot/controller.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/debug.go b/runsc/boot/debug.go index d224d08b7..79f7387ac 100644 --- a/runsc/boot/debug.go +++ b/runsc/boot/debug.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/events.go b/runsc/boot/events.go index 717adfedd..ffd99f5e9 100644 --- a/runsc/boot/events.go +++ b/runsc/boot/events.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/fds.go b/runsc/boot/fds.go index a3d21d963..4e428b49c 100644 --- a/runsc/boot/fds.go +++ b/runsc/boot/fds.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/config.go b/runsc/boot/filter/config.go index 9c72e3b1a..652da1cef 100644 --- a/runsc/boot/filter/config.go +++ b/runsc/boot/filter/config.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/extra_filters.go b/runsc/boot/filter/extra_filters.go index 67f3101fe..5c5ec4e06 100644 --- a/runsc/boot/filter/extra_filters.go +++ b/runsc/boot/filter/extra_filters.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/extra_filters_msan.go b/runsc/boot/filter/extra_filters_msan.go index fb95283ab..ac5a0f1aa 100644 --- a/runsc/boot/filter/extra_filters_msan.go +++ b/runsc/boot/filter/extra_filters_msan.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/extra_filters_race.go b/runsc/boot/filter/extra_filters_race.go index 02a122c95..ba3c1ce87 100644 --- a/runsc/boot/filter/extra_filters_race.go +++ b/runsc/boot/filter/extra_filters_race.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/filter/filter.go b/runsc/boot/filter/filter.go index fb197f9b1..17479e0dd 100644 --- a/runsc/boot/filter/filter.go +++ b/runsc/boot/filter/filter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/fs.go b/runsc/boot/fs.go index 07061b9b3..aeb1c52cc 100644 --- a/runsc/boot/fs.go +++ b/runsc/boot/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/limits.go b/runsc/boot/limits.go index 32e62cdf7..3364aa5e6 100644 --- a/runsc/boot/limits.go +++ b/runsc/boot/limits.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 75ec19c32..0b5be0a42 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/loader_test.go b/runsc/boot/loader_test.go index 01578cfc5..9a864ad3f 100644 --- a/runsc/boot/loader_test.go +++ b/runsc/boot/loader_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/network.go b/runsc/boot/network.go index 35baa36ad..598ec969e 100644 --- a/runsc/boot/network.go +++ b/runsc/boot/network.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/boot/strace.go b/runsc/boot/strace.go index 028bcc1f4..19c7f8fbd 100644 --- a/runsc/boot/strace.go +++ b/runsc/boot/strace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cgroup/cgroup.go b/runsc/cgroup/cgroup.go index 2b338b6c6..7431b17d6 100644 --- a/runsc/cgroup/cgroup.go +++ b/runsc/cgroup/cgroup.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cgroup/cgroup_test.go b/runsc/cgroup/cgroup_test.go index ecc184f74..548c80e9a 100644 --- a/runsc/cgroup/cgroup_test.go +++ b/runsc/cgroup/cgroup_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/boot.go b/runsc/cmd/boot.go index ff2fa2fb9..ac937f7bc 100644 --- a/runsc/cmd/boot.go +++ b/runsc/cmd/boot.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/capability.go b/runsc/cmd/capability.go index e5da021e5..312e5b471 100644 --- a/runsc/cmd/capability.go +++ b/runsc/cmd/capability.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/capability_test.go b/runsc/cmd/capability_test.go index dd278b32d..ee74d33d8 100644 --- a/runsc/cmd/capability_test.go +++ b/runsc/cmd/capability_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/checkpoint.go b/runsc/cmd/checkpoint.go index f722df055..96d3c3378 100644 --- a/runsc/cmd/checkpoint.go +++ b/runsc/cmd/checkpoint.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/chroot.go b/runsc/cmd/chroot.go index ed1dafef1..1a774db04 100644 --- a/runsc/cmd/chroot.go +++ b/runsc/cmd/chroot.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/cmd.go b/runsc/cmd/cmd.go index 208cf5304..aa7b1a636 100644 --- a/runsc/cmd/cmd.go +++ b/runsc/cmd/cmd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/create.go b/runsc/cmd/create.go index 30c8fa283..629c198fd 100644 --- a/runsc/cmd/create.go +++ b/runsc/cmd/create.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/debug.go b/runsc/cmd/debug.go index 3ee9a9b49..000f694c7 100644 --- a/runsc/cmd/debug.go +++ b/runsc/cmd/debug.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/delete.go b/runsc/cmd/delete.go index 3206b267a..9039723e9 100644 --- a/runsc/cmd/delete.go +++ b/runsc/cmd/delete.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/delete_test.go b/runsc/cmd/delete_test.go index 4a5b4774a..45fc91016 100644 --- a/runsc/cmd/delete_test.go +++ b/runsc/cmd/delete_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/do.go b/runsc/cmd/do.go index 343461130..67d415733 100644 --- a/runsc/cmd/do.go +++ b/runsc/cmd/do.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/events.go b/runsc/cmd/events.go index 208d2f74b..c6bc8fc3a 100644 --- a/runsc/cmd/events.go +++ b/runsc/cmd/events.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/exec.go b/runsc/cmd/exec.go index 718d01067..ad2508405 100644 --- a/runsc/cmd/exec.go +++ b/runsc/cmd/exec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/exec_test.go b/runsc/cmd/exec_test.go index 686c5e150..6f0f258c0 100644 --- a/runsc/cmd/exec_test.go +++ b/runsc/cmd/exec_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/gofer.go b/runsc/cmd/gofer.go index 82487887c..bccb29397 100644 --- a/runsc/cmd/gofer.go +++ b/runsc/cmd/gofer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/gofer_test.go b/runsc/cmd/gofer_test.go index 8e692feb9..cbea7f127 100644 --- a/runsc/cmd/gofer_test.go +++ b/runsc/cmd/gofer_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/kill.go b/runsc/cmd/kill.go index e67f82473..aed5f3291 100644 --- a/runsc/cmd/kill.go +++ b/runsc/cmd/kill.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/list.go b/runsc/cmd/list.go index 1dcea2af0..1f5ca2473 100644 --- a/runsc/cmd/list.go +++ b/runsc/cmd/list.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/path.go b/runsc/cmd/path.go index 1276f0dbd..0e9ef7fa5 100644 --- a/runsc/cmd/path.go +++ b/runsc/cmd/path.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/pause.go b/runsc/cmd/pause.go index 2c93e5f3e..11b36aa10 100644 --- a/runsc/cmd/pause.go +++ b/runsc/cmd/pause.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/ps.go b/runsc/cmd/ps.go index 060d796f2..3a3e6f17a 100644 --- a/runsc/cmd/ps.go +++ b/runsc/cmd/ps.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/restore.go b/runsc/cmd/restore.go index 66b23c38e..27b06713a 100644 --- a/runsc/cmd/restore.go +++ b/runsc/cmd/restore.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/resume.go b/runsc/cmd/resume.go index 5551d1450..9a2ade41e 100644 --- a/runsc/cmd/resume.go +++ b/runsc/cmd/resume.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/run.go b/runsc/cmd/run.go index be1c1b678..4d5f5c139 100644 --- a/runsc/cmd/run.go +++ b/runsc/cmd/run.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/spec.go b/runsc/cmd/spec.go index 063bd39c5..344da13ba 100644 --- a/runsc/cmd/spec.go +++ b/runsc/cmd/spec.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/start.go b/runsc/cmd/start.go index 9e2e0c11d..657726251 100644 --- a/runsc/cmd/start.go +++ b/runsc/cmd/start.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/state.go b/runsc/cmd/state.go index c3ef65ab5..f0d449b19 100644 --- a/runsc/cmd/state.go +++ b/runsc/cmd/state.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/cmd/wait.go b/runsc/cmd/wait.go index 6498dd15c..a55a682f3 100644 --- a/runsc/cmd/wait.go +++ b/runsc/cmd/wait.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/console/console.go b/runsc/console/console.go index 2eb9a8807..64b23639a 100644 --- a/runsc/console/console.go +++ b/runsc/console/console.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/console_test.go b/runsc/container/console_test.go index 0b0dfb4cb..b8af27c15 100644 --- a/runsc/container/console_test.go +++ b/runsc/container/console_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/container.go b/runsc/container/container.go index a30c217f7..884bbc0fb 100644 --- a/runsc/container/container.go +++ b/runsc/container/container.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/container_test.go b/runsc/container/container_test.go index 603c4d929..9458dbb90 100644 --- a/runsc/container/container_test.go +++ b/runsc/container/container_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/hook.go b/runsc/container/hook.go index 6b9e5550a..acae6781e 100644 --- a/runsc/container/hook.go +++ b/runsc/container/hook.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/multi_container_test.go b/runsc/container/multi_container_test.go index 8922e6dbe..e554237cf 100644 --- a/runsc/container/multi_container_test.go +++ b/runsc/container/multi_container_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/shared_volume_test.go b/runsc/container/shared_volume_test.go index 8f81ed630..9d5a592a5 100644 --- a/runsc/container/shared_volume_test.go +++ b/runsc/container/shared_volume_test.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/status.go b/runsc/container/status.go index 234ffb0dd..91d9112f1 100644 --- a/runsc/container/status.go +++ b/runsc/container/status.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/container/test_app.go b/runsc/container/test_app.go index b5071ada6..62923f1ef 100644 --- a/runsc/container/test_app.go +++ b/runsc/container/test_app.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/config.go b/runsc/fsgofer/filter/config.go index 75a087848..a1ad49fb2 100644 --- a/runsc/fsgofer/filter/config.go +++ b/runsc/fsgofer/filter/config.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/extra_filters.go b/runsc/fsgofer/filter/extra_filters.go index 67f3101fe..5c5ec4e06 100644 --- a/runsc/fsgofer/filter/extra_filters.go +++ b/runsc/fsgofer/filter/extra_filters.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/extra_filters_msan.go b/runsc/fsgofer/filter/extra_filters_msan.go index 7e142b790..553060bc3 100644 --- a/runsc/fsgofer/filter/extra_filters_msan.go +++ b/runsc/fsgofer/filter/extra_filters_msan.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/extra_filters_race.go b/runsc/fsgofer/filter/extra_filters_race.go index 3cd29472a..28555f898 100644 --- a/runsc/fsgofer/filter/extra_filters_race.go +++ b/runsc/fsgofer/filter/extra_filters_race.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/filter/filter.go b/runsc/fsgofer/filter/filter.go index c120d57a6..ff8154369 100644 --- a/runsc/fsgofer/filter/filter.go +++ b/runsc/fsgofer/filter/filter.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/fsgofer.go b/runsc/fsgofer/fsgofer.go index c964a2a3b..158f22ddc 100644 --- a/runsc/fsgofer/fsgofer.go +++ b/runsc/fsgofer/fsgofer.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/fsgofer_test.go b/runsc/fsgofer/fsgofer_test.go index e74df7ede..695836927 100644 --- a/runsc/fsgofer/fsgofer_test.go +++ b/runsc/fsgofer/fsgofer_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/fsgofer/fsgofer_unsafe.go b/runsc/fsgofer/fsgofer_unsafe.go index 94413db86..58af5e44d 100644 --- a/runsc/fsgofer/fsgofer_unsafe.go +++ b/runsc/fsgofer/fsgofer_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/main.go b/runsc/main.go index b35726a74..11bc73f75 100644 --- a/runsc/main.go +++ b/runsc/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/sandbox/network.go b/runsc/sandbox/network.go index 6c6b665a0..2a68d7043 100644 --- a/runsc/sandbox/network.go +++ b/runsc/sandbox/network.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/sandbox/network_unsafe.go b/runsc/sandbox/network_unsafe.go index f7447f002..2a2a0fb7e 100644 --- a/runsc/sandbox/network_unsafe.go +++ b/runsc/sandbox/network_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/sandbox/sandbox.go b/runsc/sandbox/sandbox.go index 48a0dafe2..dac35ca0b 100644 --- a/runsc/sandbox/sandbox.go +++ b/runsc/sandbox/sandbox.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/specutils/fs.go b/runsc/specutils/fs.go index 98c3b19c0..1f3afb4e4 100644 --- a/runsc/specutils/fs.go +++ b/runsc/specutils/fs.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/specutils/namespace.go b/runsc/specutils/namespace.go index 35da789f4..7d194335c 100644 --- a/runsc/specutils/namespace.go +++ b/runsc/specutils/namespace.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/specutils/specutils.go b/runsc/specutils/specutils.go index ac85bec71..c72207fb4 100644 --- a/runsc/specutils/specutils.go +++ b/runsc/specutils/specutils.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/specutils/specutils_test.go b/runsc/specutils/specutils_test.go index 02af6e6ad..2c86fffe8 100644 --- a/runsc/specutils/specutils_test.go +++ b/runsc/specutils/specutils_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/image/image.go b/runsc/test/image/image.go index bcb6f876f..297f1ab92 100644 --- a/runsc/test/image/image.go +++ b/runsc/test/image/image.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/image/image_test.go b/runsc/test/image/image_test.go index f7e750d71..0c45602f9 100644 --- a/runsc/test/image/image_test.go +++ b/runsc/test/image/image_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/image/mysql.sql b/runsc/test/image/mysql.sql index c1271e719..51554b98d 100644 --- a/runsc/test/image/mysql.sql +++ b/runsc/test/image/mysql.sql @@ -1,4 +1,4 @@ -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/runsc/test/image/ruby.rb b/runsc/test/image/ruby.rb index 25d1ac129..aced49c6d 100644 --- a/runsc/test/image/ruby.rb +++ b/runsc/test/image/ruby.rb @@ -1,4 +1,4 @@ -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/runsc/test/image/ruby.sh b/runsc/test/image/ruby.sh index d3a9b5656..ebe8d5b0e 100644 --- a/runsc/test/image/ruby.sh +++ b/runsc/test/image/ruby.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/runsc/test/install.sh b/runsc/test/install.sh index 32e1e884e..457df2d26 100755 --- a/runsc/test/install.sh +++ b/runsc/test/install.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/runsc/test/integration/exec_test.go b/runsc/test/integration/exec_test.go index d87957e2d..7af064d79 100644 --- a/runsc/test/integration/exec_test.go +++ b/runsc/test/integration/exec_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/integration/integration.go b/runsc/test/integration/integration.go index e15321c87..4cd5f6c24 100644 --- a/runsc/test/integration/integration.go +++ b/runsc/test/integration/integration.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/integration/integration_test.go b/runsc/test/integration/integration_test.go index 4a2770d48..b2e86aacc 100644 --- a/runsc/test/integration/integration_test.go +++ b/runsc/test/integration/integration_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/cgroup_test.go b/runsc/test/root/cgroup_test.go index 91839048c..edb6dee1d 100644 --- a/runsc/test/root/cgroup_test.go +++ b/runsc/test/root/cgroup_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/chroot_test.go b/runsc/test/root/chroot_test.go index 0deca0532..da2f473b9 100644 --- a/runsc/test/root/chroot_test.go +++ b/runsc/test/root/chroot_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/crictl_test.go b/runsc/test/root/crictl_test.go index 37fe53ba3..3cc176104 100644 --- a/runsc/test/root/crictl_test.go +++ b/runsc/test/root/crictl_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/root.go b/runsc/test/root/root.go index 586ea0fe3..349c752cc 100644 --- a/runsc/test/root/root.go +++ b/runsc/test/root/root.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/testdata/busybox.go b/runsc/test/root/testdata/busybox.go index 544571c63..e4dbd2843 100644 --- a/runsc/test/root/testdata/busybox.go +++ b/runsc/test/root/testdata/busybox.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/testdata/containerd_config.go b/runsc/test/root/testdata/containerd_config.go index 949354987..e12f1ec88 100644 --- a/runsc/test/root/testdata/containerd_config.go +++ b/runsc/test/root/testdata/containerd_config.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/testdata/httpd.go b/runsc/test/root/testdata/httpd.go index f65b1da5d..45d5e33d4 100644 --- a/runsc/test/root/testdata/httpd.go +++ b/runsc/test/root/testdata/httpd.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/testdata/httpd_mount_paths.go b/runsc/test/root/testdata/httpd_mount_paths.go index 5ca14340e..ac3f4446a 100644 --- a/runsc/test/root/testdata/httpd_mount_paths.go +++ b/runsc/test/root/testdata/httpd_mount_paths.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/root/testdata/sandbox.go b/runsc/test/root/testdata/sandbox.go index 194242a27..0db210370 100644 --- a/runsc/test/root/testdata/sandbox.go +++ b/runsc/test/root/testdata/sandbox.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/testutil/crictl.go b/runsc/test/testutil/crictl.go index 84bb4475a..4f9ee0c05 100644 --- a/runsc/test/testutil/crictl.go +++ b/runsc/test/testutil/crictl.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/testutil/docker.go b/runsc/test/testutil/docker.go index b651319ed..29ef505b4 100644 --- a/runsc/test/testutil/docker.go +++ b/runsc/test/testutil/docker.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/testutil/testutil.go b/runsc/test/testutil/testutil.go index 79f0a8b6b..6a4c045a8 100644 --- a/runsc/test/testutil/testutil.go +++ b/runsc/test/testutil/testutil.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/test/testutil/testutil_race.go b/runsc/test/testutil/testutil_race.go index 9267af150..86db6ffa1 100644 --- a/runsc/test/testutil/testutil_race.go +++ b/runsc/test/testutil/testutil_race.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/tools/dockercfg/dockercfg.go b/runsc/tools/dockercfg/dockercfg.go index cc7a67816..6fb134558 100644 --- a/runsc/tools/dockercfg/dockercfg.go +++ b/runsc/tools/dockercfg/dockercfg.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/runsc/version.go b/runsc/version.go index 4894f2de6..ce0573a9b 100644 --- a/runsc/version.go +++ b/runsc/version.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/gtest/gtest.go b/test/syscalls/gtest/gtest.go index dfe5037cd..bdec8eb07 100644 --- a/test/syscalls/gtest/gtest.go +++ b/test/syscalls/gtest/gtest.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/32bit.cc b/test/syscalls/linux/32bit.cc index 78baf548e..a7cbee06b 100644 --- a/test/syscalls/linux/32bit.cc +++ b/test/syscalls/linux/32bit.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/accept_bind.cc b/test/syscalls/linux/accept_bind.cc index c2bb4a7ce..56377feab 100644 --- a/test/syscalls/linux/accept_bind.cc +++ b/test/syscalls/linux/accept_bind.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/accept_bind_stream.cc b/test/syscalls/linux/accept_bind_stream.cc index 1501e526e..b6cdb3f4f 100644 --- a/test/syscalls/linux/accept_bind_stream.cc +++ b/test/syscalls/linux/accept_bind_stream.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/access.cc b/test/syscalls/linux/access.cc index 6ea070a5d..bcc25cef4 100644 --- a/test/syscalls/linux/access.cc +++ b/test/syscalls/linux/access.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/affinity.cc b/test/syscalls/linux/affinity.cc index 81bd9bcb5..f2d8375b6 100644 --- a/test/syscalls/linux/affinity.cc +++ b/test/syscalls/linux/affinity.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/aio.cc b/test/syscalls/linux/aio.cc index b96aab9b9..68dc05417 100644 --- a/test/syscalls/linux/aio.cc +++ b/test/syscalls/linux/aio.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/alarm.cc b/test/syscalls/linux/alarm.cc index e0ddbb415..d89269985 100644 --- a/test/syscalls/linux/alarm.cc +++ b/test/syscalls/linux/alarm.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/arch_prctl.cc b/test/syscalls/linux/arch_prctl.cc index 5687ceb86..81bf5a775 100644 --- a/test/syscalls/linux/arch_prctl.cc +++ b/test/syscalls/linux/arch_prctl.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/bad.cc b/test/syscalls/linux/bad.cc index a2634a8bf..f246a799e 100644 --- a/test/syscalls/linux/bad.cc +++ b/test/syscalls/linux/bad.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/base_poll_test.cc b/test/syscalls/linux/base_poll_test.cc index bba0108ea..ab7a19dd0 100644 --- a/test/syscalls/linux/base_poll_test.cc +++ b/test/syscalls/linux/base_poll_test.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/base_poll_test.h b/test/syscalls/linux/base_poll_test.h index 9b9b81933..088831f9f 100644 --- a/test/syscalls/linux/base_poll_test.h +++ b/test/syscalls/linux/base_poll_test.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/bind.cc b/test/syscalls/linux/bind.cc index f5aa9c500..de8cca53b 100644 --- a/test/syscalls/linux/bind.cc +++ b/test/syscalls/linux/bind.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/brk.cc b/test/syscalls/linux/brk.cc index 33d353959..a03a44465 100644 --- a/test/syscalls/linux/brk.cc +++ b/test/syscalls/linux/brk.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/chdir.cc b/test/syscalls/linux/chdir.cc index a4b54f0ee..3182c228b 100644 --- a/test/syscalls/linux/chdir.cc +++ b/test/syscalls/linux/chdir.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/chmod.cc b/test/syscalls/linux/chmod.cc index 2f42fe326..79e98597f 100644 --- a/test/syscalls/linux/chmod.cc +++ b/test/syscalls/linux/chmod.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/chown.cc b/test/syscalls/linux/chown.cc index ad892cf6a..eb1762ddf 100644 --- a/test/syscalls/linux/chown.cc +++ b/test/syscalls/linux/chown.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/chroot.cc b/test/syscalls/linux/chroot.cc index 6c200f63e..a4354ff62 100644 --- a/test/syscalls/linux/chroot.cc +++ b/test/syscalls/linux/chroot.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/clock_getres.cc b/test/syscalls/linux/clock_getres.cc index 8f8842299..c408b936c 100644 --- a/test/syscalls/linux/clock_getres.cc +++ b/test/syscalls/linux/clock_getres.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/clock_gettime.cc b/test/syscalls/linux/clock_gettime.cc index 4ecb5f5b1..082ae1c39 100644 --- a/test/syscalls/linux/clock_gettime.cc +++ b/test/syscalls/linux/clock_gettime.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/clock_nanosleep.cc b/test/syscalls/linux/clock_nanosleep.cc index 61c67a5ff..52a69d230 100644 --- a/test/syscalls/linux/clock_nanosleep.cc +++ b/test/syscalls/linux/clock_nanosleep.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/concurrency.cc b/test/syscalls/linux/concurrency.cc index 7978845c1..4e0a13f8b 100644 --- a/test/syscalls/linux/concurrency.cc +++ b/test/syscalls/linux/concurrency.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/creat.cc b/test/syscalls/linux/creat.cc index df2cc0d5c..3c270d6da 100644 --- a/test/syscalls/linux/creat.cc +++ b/test/syscalls/linux/creat.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/dev.cc b/test/syscalls/linux/dev.cc index a140d3b30..b86ebe233 100644 --- a/test/syscalls/linux/dev.cc +++ b/test/syscalls/linux/dev.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/dup.cc b/test/syscalls/linux/dup.cc index e8de2f4c4..4f773bc75 100644 --- a/test/syscalls/linux/dup.cc +++ b/test/syscalls/linux/dup.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/epoll.cc b/test/syscalls/linux/epoll.cc index b4a3bfcba..a4f8f3cec 100644 --- a/test/syscalls/linux/epoll.cc +++ b/test/syscalls/linux/epoll.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/eventfd.cc b/test/syscalls/linux/eventfd.cc index 8111da30e..5e5c39d44 100644 --- a/test/syscalls/linux/eventfd.cc +++ b/test/syscalls/linux/eventfd.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exceptions.cc b/test/syscalls/linux/exceptions.cc index 3f0aa8bf1..0da4c817d 100644 --- a/test/syscalls/linux/exceptions.cc +++ b/test/syscalls/linux/exceptions.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exec.cc b/test/syscalls/linux/exec.cc index 30bc4b608..06c322a99 100644 --- a/test/syscalls/linux/exec.cc +++ b/test/syscalls/linux/exec.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exec.h b/test/syscalls/linux/exec.h index b82bfffd1..5c0f7e654 100644 --- a/test/syscalls/linux/exec.h +++ b/test/syscalls/linux/exec.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exec_assert_closed_workload.cc b/test/syscalls/linux/exec_assert_closed_workload.cc index 4448431e1..95643618d 100644 --- a/test/syscalls/linux/exec_assert_closed_workload.cc +++ b/test/syscalls/linux/exec_assert_closed_workload.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exec_basic_workload.cc b/test/syscalls/linux/exec_basic_workload.cc index d4bdf511f..1bbd6437e 100644 --- a/test/syscalls/linux/exec_basic_workload.cc +++ b/test/syscalls/linux/exec_basic_workload.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exec_binary.cc b/test/syscalls/linux/exec_binary.cc index c10d85398..bdd6eb10b 100644 --- a/test/syscalls/linux/exec_binary.cc +++ b/test/syscalls/linux/exec_binary.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exec_proc_exe_workload.cc b/test/syscalls/linux/exec_proc_exe_workload.cc index b9a4ac749..b3fbd5042 100644 --- a/test/syscalls/linux/exec_proc_exe_workload.cc +++ b/test/syscalls/linux/exec_proc_exe_workload.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exec_state_workload.cc b/test/syscalls/linux/exec_state_workload.cc index b66e22565..725c2977f 100644 --- a/test/syscalls/linux/exec_state_workload.cc +++ b/test/syscalls/linux/exec_state_workload.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exit.cc b/test/syscalls/linux/exit.cc index 7246a7b3b..99de2b376 100644 --- a/test/syscalls/linux/exit.cc +++ b/test/syscalls/linux/exit.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/exit_script.sh b/test/syscalls/linux/exit_script.sh index f014fcf99..527518e06 100755 --- a/test/syscalls/linux/exit_script.sh +++ b/test/syscalls/linux/exit_script.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fadvise64.cc b/test/syscalls/linux/fadvise64.cc index 041e8b7b6..2af7aa6d9 100644 --- a/test/syscalls/linux/fadvise64.cc +++ b/test/syscalls/linux/fadvise64.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fallocate.cc b/test/syscalls/linux/fallocate.cc index e51538734..61b8acc7a 100644 --- a/test/syscalls/linux/fallocate.cc +++ b/test/syscalls/linux/fallocate.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fault.cc b/test/syscalls/linux/fault.cc index cfa7d0d1f..f6e19026f 100644 --- a/test/syscalls/linux/fault.cc +++ b/test/syscalls/linux/fault.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fchdir.cc b/test/syscalls/linux/fchdir.cc index 2b13e36c3..08bcae1e8 100644 --- a/test/syscalls/linux/fchdir.cc +++ b/test/syscalls/linux/fchdir.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fcntl.cc b/test/syscalls/linux/fcntl.cc index 32a90a163..2f8e7c9dd 100644 --- a/test/syscalls/linux/fcntl.cc +++ b/test/syscalls/linux/fcntl.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/file_base.h b/test/syscalls/linux/file_base.h index 43f568111..b5b972c07 100644 --- a/test/syscalls/linux/file_base.h +++ b/test/syscalls/linux/file_base.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/flock.cc b/test/syscalls/linux/flock.cc index 1388d3839..d89cfcbd7 100644 --- a/test/syscalls/linux/flock.cc +++ b/test/syscalls/linux/flock.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fork.cc b/test/syscalls/linux/fork.cc index 73ac885b5..dd6e1a422 100644 --- a/test/syscalls/linux/fork.cc +++ b/test/syscalls/linux/fork.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fpsig_fork.cc b/test/syscalls/linux/fpsig_fork.cc index e8f1dfa8a..e7e9f06a1 100644 --- a/test/syscalls/linux/fpsig_fork.cc +++ b/test/syscalls/linux/fpsig_fork.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fpsig_nested.cc b/test/syscalls/linux/fpsig_nested.cc index 2fa40b42d..395463aed 100644 --- a/test/syscalls/linux/fpsig_nested.cc +++ b/test/syscalls/linux/fpsig_nested.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/fsync.cc b/test/syscalls/linux/fsync.cc index b34229248..e7e057f06 100644 --- a/test/syscalls/linux/fsync.cc +++ b/test/syscalls/linux/fsync.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/futex.cc b/test/syscalls/linux/futex.cc index c7a709a0a..bfec95466 100644 --- a/test/syscalls/linux/futex.cc +++ b/test/syscalls/linux/futex.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/getcpu.cc b/test/syscalls/linux/getcpu.cc index 3a52b25fa..f4d94bd6a 100644 --- a/test/syscalls/linux/getcpu.cc +++ b/test/syscalls/linux/getcpu.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/getdents.cc b/test/syscalls/linux/getdents.cc index e8a7bcd43..d146c8db7 100644 --- a/test/syscalls/linux/getdents.cc +++ b/test/syscalls/linux/getdents.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/getrandom.cc b/test/syscalls/linux/getrandom.cc index be5325497..f97f60029 100644 --- a/test/syscalls/linux/getrandom.cc +++ b/test/syscalls/linux/getrandom.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/getrusage.cc b/test/syscalls/linux/getrusage.cc index 1ae603858..9bdb1e4cd 100644 --- a/test/syscalls/linux/getrusage.cc +++ b/test/syscalls/linux/getrusage.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/inotify.cc b/test/syscalls/linux/inotify.cc index b99d339e5..6a3539e22 100644 --- a/test/syscalls/linux/inotify.cc +++ b/test/syscalls/linux/inotify.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/ioctl.cc b/test/syscalls/linux/ioctl.cc index c7741a177..c525d41d2 100644 --- a/test/syscalls/linux/ioctl.cc +++ b/test/syscalls/linux/ioctl.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/ip_socket_test_util.cc b/test/syscalls/linux/ip_socket_test_util.cc index 0a149c2e5..7612919d4 100644 --- a/test/syscalls/linux/ip_socket_test_util.cc +++ b/test/syscalls/linux/ip_socket_test_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/ip_socket_test_util.h b/test/syscalls/linux/ip_socket_test_util.h index cac790e64..6898effb8 100644 --- a/test/syscalls/linux/ip_socket_test_util.h +++ b/test/syscalls/linux/ip_socket_test_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/itimer.cc b/test/syscalls/linux/itimer.cc index ddfbc28fc..57ffd1595 100644 --- a/test/syscalls/linux/itimer.cc +++ b/test/syscalls/linux/itimer.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/kill.cc b/test/syscalls/linux/kill.cc index cd98de41f..18ad923b8 100644 --- a/test/syscalls/linux/kill.cc +++ b/test/syscalls/linux/kill.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/link.cc b/test/syscalls/linux/link.cc index ed74437bc..a91703070 100644 --- a/test/syscalls/linux/link.cc +++ b/test/syscalls/linux/link.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/lseek.cc b/test/syscalls/linux/lseek.cc index 6a4f1423c..a8af8e545 100644 --- a/test/syscalls/linux/lseek.cc +++ b/test/syscalls/linux/lseek.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/madvise.cc b/test/syscalls/linux/madvise.cc index a79c8c75d..f6ad4d18b 100644 --- a/test/syscalls/linux/madvise.cc +++ b/test/syscalls/linux/madvise.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/memfd.cc b/test/syscalls/linux/memfd.cc index c2513682d..7e103124b 100644 --- a/test/syscalls/linux/memfd.cc +++ b/test/syscalls/linux/memfd.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/memory_accounting.cc b/test/syscalls/linux/memory_accounting.cc index b4b680c34..a6e20f9c3 100644 --- a/test/syscalls/linux/memory_accounting.cc +++ b/test/syscalls/linux/memory_accounting.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/mempolicy.cc b/test/syscalls/linux/mempolicy.cc index 9f8033bdf..4ac4cb88f 100644 --- a/test/syscalls/linux/mempolicy.cc +++ b/test/syscalls/linux/mempolicy.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/mincore.cc b/test/syscalls/linux/mincore.cc index c572bf5ec..5c1240c89 100644 --- a/test/syscalls/linux/mincore.cc +++ b/test/syscalls/linux/mincore.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/mkdir.cc b/test/syscalls/linux/mkdir.cc index 50807b68f..cf138d328 100644 --- a/test/syscalls/linux/mkdir.cc +++ b/test/syscalls/linux/mkdir.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/mknod.cc b/test/syscalls/linux/mknod.cc index 361ca299b..b1675b9c7 100644 --- a/test/syscalls/linux/mknod.cc +++ b/test/syscalls/linux/mknod.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/mlock.cc b/test/syscalls/linux/mlock.cc index a492b2404..aee4f7d1a 100644 --- a/test/syscalls/linux/mlock.cc +++ b/test/syscalls/linux/mlock.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/mmap.cc b/test/syscalls/linux/mmap.cc index a4fb9d1e0..5b5b4c2e8 100644 --- a/test/syscalls/linux/mmap.cc +++ b/test/syscalls/linux/mmap.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/mount.cc b/test/syscalls/linux/mount.cc index 201b83e87..3a17672aa 100644 --- a/test/syscalls/linux/mount.cc +++ b/test/syscalls/linux/mount.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/mremap.cc b/test/syscalls/linux/mremap.cc index 01116c1ab..7298d4ca8 100644 --- a/test/syscalls/linux/mremap.cc +++ b/test/syscalls/linux/mremap.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/msync.cc b/test/syscalls/linux/msync.cc index 5afbfce72..ac7146017 100644 --- a/test/syscalls/linux/msync.cc +++ b/test/syscalls/linux/msync.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/munmap.cc b/test/syscalls/linux/munmap.cc index e20039950..067241f4d 100644 --- a/test/syscalls/linux/munmap.cc +++ b/test/syscalls/linux/munmap.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/open.cc b/test/syscalls/linux/open.cc index 22e4666c2..42646bb02 100644 --- a/test/syscalls/linux/open.cc +++ b/test/syscalls/linux/open.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/open_create.cc b/test/syscalls/linux/open_create.cc index b2cbd63d1..e5a85ef9d 100644 --- a/test/syscalls/linux/open_create.cc +++ b/test/syscalls/linux/open_create.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/partial_bad_buffer.cc b/test/syscalls/linux/partial_bad_buffer.cc index 71288ebc4..83b1ad4e4 100644 --- a/test/syscalls/linux/partial_bad_buffer.cc +++ b/test/syscalls/linux/partial_bad_buffer.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/pause.cc b/test/syscalls/linux/pause.cc index 4e1148c24..8c05efd6f 100644 --- a/test/syscalls/linux/pause.cc +++ b/test/syscalls/linux/pause.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/pipe.cc b/test/syscalls/linux/pipe.cc index abd10b11b..8698295b3 100644 --- a/test/syscalls/linux/pipe.cc +++ b/test/syscalls/linux/pipe.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/poll.cc b/test/syscalls/linux/poll.cc index cd2161bb1..9e5aa7fd0 100644 --- a/test/syscalls/linux/poll.cc +++ b/test/syscalls/linux/poll.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/ppoll.cc b/test/syscalls/linux/ppoll.cc index f8c388c00..8245a11e8 100644 --- a/test/syscalls/linux/ppoll.cc +++ b/test/syscalls/linux/ppoll.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/prctl.cc b/test/syscalls/linux/prctl.cc index 854dec714..bce42dc74 100644 --- a/test/syscalls/linux/prctl.cc +++ b/test/syscalls/linux/prctl.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/prctl_setuid.cc b/test/syscalls/linux/prctl_setuid.cc index c1b561464..00dd6523e 100644 --- a/test/syscalls/linux/prctl_setuid.cc +++ b/test/syscalls/linux/prctl_setuid.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/pread64.cc b/test/syscalls/linux/pread64.cc index 4e5bcfcde..5e3eb1735 100644 --- a/test/syscalls/linux/pread64.cc +++ b/test/syscalls/linux/pread64.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/preadv.cc b/test/syscalls/linux/preadv.cc index 4a31123d8..eebd129f2 100644 --- a/test/syscalls/linux/preadv.cc +++ b/test/syscalls/linux/preadv.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/preadv2.cc b/test/syscalls/linux/preadv2.cc index 58a4f9224..aac960130 100644 --- a/test/syscalls/linux/preadv2.cc +++ b/test/syscalls/linux/preadv2.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/priority.cc b/test/syscalls/linux/priority.cc index 3906c7132..1d9bdfa70 100644 --- a/test/syscalls/linux/priority.cc +++ b/test/syscalls/linux/priority.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/priority_execve.cc b/test/syscalls/linux/priority_execve.cc index 5604bd3d0..5cb343bad 100644 --- a/test/syscalls/linux/priority_execve.cc +++ b/test/syscalls/linux/priority_execve.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/proc.cc b/test/syscalls/linux/proc.cc index 7ba274226..654f26242 100644 --- a/test/syscalls/linux/proc.cc +++ b/test/syscalls/linux/proc.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/proc_net.cc b/test/syscalls/linux/proc_net.cc index 6060d0644..03d0665eb 100644 --- a/test/syscalls/linux/proc_net.cc +++ b/test/syscalls/linux/proc_net.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/proc_net_unix.cc b/test/syscalls/linux/proc_net_unix.cc index ea7c93012..6d745f728 100644 --- a/test/syscalls/linux/proc_net_unix.cc +++ b/test/syscalls/linux/proc_net_unix.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/proc_pid_smaps.cc b/test/syscalls/linux/proc_pid_smaps.cc index cf5c462f3..7f2e8f203 100644 --- a/test/syscalls/linux/proc_pid_smaps.cc +++ b/test/syscalls/linux/proc_pid_smaps.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/proc_pid_uid_gid_map.cc b/test/syscalls/linux/proc_pid_uid_gid_map.cc index 96c58c564..df70b7eb9 100644 --- a/test/syscalls/linux/proc_pid_uid_gid_map.cc +++ b/test/syscalls/linux/proc_pid_uid_gid_map.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/pselect.cc b/test/syscalls/linux/pselect.cc index 3294f6c14..4e43c4d7f 100644 --- a/test/syscalls/linux/pselect.cc +++ b/test/syscalls/linux/pselect.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/ptrace.cc b/test/syscalls/linux/ptrace.cc index e0c56f1fc..4c212836c 100644 --- a/test/syscalls/linux/ptrace.cc +++ b/test/syscalls/linux/ptrace.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/pty.cc b/test/syscalls/linux/pty.cc index 5b2dc9ccb..0485d187c 100644 --- a/test/syscalls/linux/pty.cc +++ b/test/syscalls/linux/pty.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/pwrite64.cc b/test/syscalls/linux/pwrite64.cc index 485b1e48d..e1603fc2d 100644 --- a/test/syscalls/linux/pwrite64.cc +++ b/test/syscalls/linux/pwrite64.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/pwritev2.cc b/test/syscalls/linux/pwritev2.cc index a6949f08e..db519f4e0 100644 --- a/test/syscalls/linux/pwritev2.cc +++ b/test/syscalls/linux/pwritev2.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/raw_socket_ipv4.cc b/test/syscalls/linux/raw_socket_ipv4.cc index 8b8d032cb..e20b5cb50 100644 --- a/test/syscalls/linux/raw_socket_ipv4.cc +++ b/test/syscalls/linux/raw_socket_ipv4.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/read.cc b/test/syscalls/linux/read.cc index eb1b5bc10..4430fa3c2 100644 --- a/test/syscalls/linux/read.cc +++ b/test/syscalls/linux/read.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/readv.cc b/test/syscalls/linux/readv.cc index 0b933673a..f327ec3a9 100644 --- a/test/syscalls/linux/readv.cc +++ b/test/syscalls/linux/readv.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/readv_common.cc b/test/syscalls/linux/readv_common.cc index 349b80d7f..35d2dd9e3 100644 --- a/test/syscalls/linux/readv_common.cc +++ b/test/syscalls/linux/readv_common.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/readv_common.h b/test/syscalls/linux/readv_common.h index e261d545a..b16179fca 100644 --- a/test/syscalls/linux/readv_common.h +++ b/test/syscalls/linux/readv_common.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/readv_socket.cc b/test/syscalls/linux/readv_socket.cc index cf22c395e..3c315cc02 100644 --- a/test/syscalls/linux/readv_socket.cc +++ b/test/syscalls/linux/readv_socket.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/rename.cc b/test/syscalls/linux/rename.cc index c0cbc7cd9..c9d76c2e2 100644 --- a/test/syscalls/linux/rename.cc +++ b/test/syscalls/linux/rename.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/rlimits.cc b/test/syscalls/linux/rlimits.cc index 7b255d0f6..860f0f688 100644 --- a/test/syscalls/linux/rlimits.cc +++ b/test/syscalls/linux/rlimits.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/rtsignal.cc b/test/syscalls/linux/rtsignal.cc index ff948f9d5..81d193ffd 100644 --- a/test/syscalls/linux/rtsignal.cc +++ b/test/syscalls/linux/rtsignal.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sched.cc b/test/syscalls/linux/sched.cc index 60cb6c443..735e99411 100644 --- a/test/syscalls/linux/sched.cc +++ b/test/syscalls/linux/sched.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sched_yield.cc b/test/syscalls/linux/sched_yield.cc index fc45aa5c2..5d24f5b58 100644 --- a/test/syscalls/linux/sched_yield.cc +++ b/test/syscalls/linux/sched_yield.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/seccomp.cc b/test/syscalls/linux/seccomp.cc index 27740d7ef..e77586852 100644 --- a/test/syscalls/linux/seccomp.cc +++ b/test/syscalls/linux/seccomp.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/select.cc b/test/syscalls/linux/select.cc index 41e6043cc..88c010aec 100644 --- a/test/syscalls/linux/select.cc +++ b/test/syscalls/linux/select.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/semaphore.cc b/test/syscalls/linux/semaphore.cc index 1c47b6851..421318fcb 100644 --- a/test/syscalls/linux/semaphore.cc +++ b/test/syscalls/linux/semaphore.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sendfile.cc b/test/syscalls/linux/sendfile.cc index 15fd01ff0..2fbb3f4ef 100644 --- a/test/syscalls/linux/sendfile.cc +++ b/test/syscalls/linux/sendfile.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sendfile_socket.cc b/test/syscalls/linux/sendfile_socket.cc index e2ccf17ce..66adda515 100644 --- a/test/syscalls/linux/sendfile_socket.cc +++ b/test/syscalls/linux/sendfile_socket.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/shm.cc b/test/syscalls/linux/shm.cc index 2c0f9b04a..eb7a3966f 100644 --- a/test/syscalls/linux/shm.cc +++ b/test/syscalls/linux/shm.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sigaction.cc b/test/syscalls/linux/sigaction.cc index cdd2dbf31..9a53fd3e0 100644 --- a/test/syscalls/linux/sigaction.cc +++ b/test/syscalls/linux/sigaction.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sigaltstack.cc b/test/syscalls/linux/sigaltstack.cc index 5741720f4..7d4a12c1d 100644 --- a/test/syscalls/linux/sigaltstack.cc +++ b/test/syscalls/linux/sigaltstack.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sigaltstack_check.cc b/test/syscalls/linux/sigaltstack_check.cc index b71f812a8..5ac1b661d 100644 --- a/test/syscalls/linux/sigaltstack_check.cc +++ b/test/syscalls/linux/sigaltstack_check.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sigiret.cc b/test/syscalls/linux/sigiret.cc index 1b7cecccb..a47c781ea 100644 --- a/test/syscalls/linux/sigiret.cc +++ b/test/syscalls/linux/sigiret.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sigprocmask.cc b/test/syscalls/linux/sigprocmask.cc index 1aea1ecb8..654c6a47f 100644 --- a/test/syscalls/linux/sigprocmask.cc +++ b/test/syscalls/linux/sigprocmask.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sigstop.cc b/test/syscalls/linux/sigstop.cc index e21d23d51..9c7210e17 100644 --- a/test/syscalls/linux/sigstop.cc +++ b/test/syscalls/linux/sigstop.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sigtimedwait.cc b/test/syscalls/linux/sigtimedwait.cc index 1df9c013f..1e5bf5942 100644 --- a/test/syscalls/linux/sigtimedwait.cc +++ b/test/syscalls/linux/sigtimedwait.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_abstract.cc b/test/syscalls/linux/socket_abstract.cc index 639cd4e59..2faf678f7 100644 --- a/test/syscalls/linux/socket_abstract.cc +++ b/test/syscalls/linux/socket_abstract.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_blocking.cc b/test/syscalls/linux/socket_blocking.cc index c1bca467f..00c50d1bf 100644 --- a/test/syscalls/linux/socket_blocking.cc +++ b/test/syscalls/linux/socket_blocking.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_blocking.h b/test/syscalls/linux/socket_blocking.h index 5cddee54b..db26e5ef5 100644 --- a/test/syscalls/linux/socket_blocking.h +++ b/test/syscalls/linux/socket_blocking.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_filesystem.cc b/test/syscalls/linux/socket_filesystem.cc index 2653be158..f7cb72df4 100644 --- a/test/syscalls/linux/socket_filesystem.cc +++ b/test/syscalls/linux/socket_filesystem.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_generic.cc b/test/syscalls/linux/socket_generic.cc index d04d5abe0..f99f3fe62 100644 --- a/test/syscalls/linux/socket_generic.cc +++ b/test/syscalls/linux/socket_generic.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_generic.h b/test/syscalls/linux/socket_generic.h index cd826abcf..00ae7bfc3 100644 --- a/test/syscalls/linux/socket_generic.h +++ b/test/syscalls/linux/socket_generic.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_inet_loopback.cc b/test/syscalls/linux/socket_inet_loopback.cc index 14d7827c2..f86a0f30c 100644 --- a/test/syscalls/linux/socket_inet_loopback.cc +++ b/test/syscalls/linux/socket_inet_loopback.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_loopback_blocking.cc b/test/syscalls/linux/socket_ip_loopback_blocking.cc index 9cec7a71d..d7fc20aad 100644 --- a/test/syscalls/linux/socket_ip_loopback_blocking.cc +++ b/test/syscalls/linux/socket_ip_loopback_blocking.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_tcp_generic.cc b/test/syscalls/linux/socket_ip_tcp_generic.cc index 54f00cd9b..5b198f49d 100644 --- a/test/syscalls/linux/socket_ip_tcp_generic.cc +++ b/test/syscalls/linux/socket_ip_tcp_generic.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_tcp_generic.h b/test/syscalls/linux/socket_ip_tcp_generic.h index f38500d14..a3eff3c73 100644 --- a/test/syscalls/linux/socket_ip_tcp_generic.h +++ b/test/syscalls/linux/socket_ip_tcp_generic.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_tcp_generic_loopback.cc b/test/syscalls/linux/socket_ip_tcp_generic_loopback.cc index 1963d5deb..2c6ae17bf 100644 --- a/test/syscalls/linux/socket_ip_tcp_generic_loopback.cc +++ b/test/syscalls/linux/socket_ip_tcp_generic_loopback.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_tcp_loopback.cc b/test/syscalls/linux/socket_ip_tcp_loopback.cc index 7e36c35d2..831de53b8 100644 --- a/test/syscalls/linux/socket_ip_tcp_loopback.cc +++ b/test/syscalls/linux/socket_ip_tcp_loopback.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_tcp_loopback_blocking.cc b/test/syscalls/linux/socket_ip_tcp_loopback_blocking.cc index 9e2a18d3e..d1ea8ef12 100644 --- a/test/syscalls/linux/socket_ip_tcp_loopback_blocking.cc +++ b/test/syscalls/linux/socket_ip_tcp_loopback_blocking.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_tcp_loopback_nonblock.cc b/test/syscalls/linux/socket_ip_tcp_loopback_nonblock.cc index 54053360f..96c1b3b3d 100644 --- a/test/syscalls/linux/socket_ip_tcp_loopback_nonblock.cc +++ b/test/syscalls/linux/socket_ip_tcp_loopback_nonblock.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_tcp_udp_generic.cc b/test/syscalls/linux/socket_ip_tcp_udp_generic.cc index 5bf1de7c6..251817a9f 100644 --- a/test/syscalls/linux/socket_ip_tcp_udp_generic.cc +++ b/test/syscalls/linux/socket_ip_tcp_udp_generic.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_udp_generic.cc b/test/syscalls/linux/socket_ip_udp_generic.cc index ac15154f2..044394ba7 100644 --- a/test/syscalls/linux/socket_ip_udp_generic.cc +++ b/test/syscalls/linux/socket_ip_udp_generic.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_udp_generic.h b/test/syscalls/linux/socket_ip_udp_generic.h index 8b8fc7c6e..106c54e9f 100644 --- a/test/syscalls/linux/socket_ip_udp_generic.h +++ b/test/syscalls/linux/socket_ip_udp_generic.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_udp_loopback.cc b/test/syscalls/linux/socket_ip_udp_loopback.cc index 0e4463649..fc124e9ef 100644 --- a/test/syscalls/linux/socket_ip_udp_loopback.cc +++ b/test/syscalls/linux/socket_ip_udp_loopback.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_udp_loopback_blocking.cc b/test/syscalls/linux/socket_ip_udp_loopback_blocking.cc index 0c3b669bf..1c3d1c0ad 100644 --- a/test/syscalls/linux/socket_ip_udp_loopback_blocking.cc +++ b/test/syscalls/linux/socket_ip_udp_loopback_blocking.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ip_udp_loopback_nonblock.cc b/test/syscalls/linux/socket_ip_udp_loopback_nonblock.cc index 7bf8597fe..7554b08d5 100644 --- a/test/syscalls/linux/socket_ip_udp_loopback_nonblock.cc +++ b/test/syscalls/linux/socket_ip_udp_loopback_nonblock.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.cc b/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.cc index 8e1c13ff4..3a068aacf 100644 --- a/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.cc +++ b/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.h b/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.h index b23de08d1..fb582b224 100644 --- a/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.h +++ b/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking.h @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking_test.cc b/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking_test.cc index 773d84b13..040bb176e 100644 --- a/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking_test.cc +++ b/test/syscalls/linux/socket_ipv4_tcp_unbound_external_networking_test.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_udp_unbound.cc b/test/syscalls/linux/socket_ipv4_udp_unbound.cc index c99958ed5..709172580 100644 --- a/test/syscalls/linux/socket_ipv4_udp_unbound.cc +++ b/test/syscalls/linux/socket_ipv4_udp_unbound.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_udp_unbound.h b/test/syscalls/linux/socket_ipv4_udp_unbound.h index a780c0144..8e07bfbbf 100644 --- a/test/syscalls/linux/socket_ipv4_udp_unbound.h +++ b/test/syscalls/linux/socket_ipv4_udp_unbound.h @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc b/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc index 9dd9e1bd6..53dcd58cd 100644 --- a/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc +++ b/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.h b/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.h index 5cf9fa8eb..45e1d37ea 100644 --- a/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.h +++ b/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking.h @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking_test.cc b/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking_test.cc index 535a5fa10..ffbb8e6eb 100644 --- a/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking_test.cc +++ b/test/syscalls/linux/socket_ipv4_udp_unbound_external_networking_test.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_ipv4_udp_unbound_loopback.cc b/test/syscalls/linux/socket_ipv4_udp_unbound_loopback.cc index d6a8e428c..cb0105471 100644 --- a/test/syscalls/linux/socket_ipv4_udp_unbound_loopback.cc +++ b/test/syscalls/linux/socket_ipv4_udp_unbound_loopback.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_netdevice.cc b/test/syscalls/linux/socket_netdevice.cc index b4e9fe51b..6a5fa8965 100644 --- a/test/syscalls/linux/socket_netdevice.cc +++ b/test/syscalls/linux/socket_netdevice.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_netlink_route.cc b/test/syscalls/linux/socket_netlink_route.cc index ed4ae1c71..c8693225f 100644 --- a/test/syscalls/linux/socket_netlink_route.cc +++ b/test/syscalls/linux/socket_netlink_route.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_netlink_util.cc b/test/syscalls/linux/socket_netlink_util.cc index edf549544..728d25434 100644 --- a/test/syscalls/linux/socket_netlink_util.cc +++ b/test/syscalls/linux/socket_netlink_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_netlink_util.h b/test/syscalls/linux/socket_netlink_util.h index 44b1f148c..bea449107 100644 --- a/test/syscalls/linux/socket_netlink_util.h +++ b/test/syscalls/linux/socket_netlink_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_non_blocking.cc b/test/syscalls/linux/socket_non_blocking.cc index 1bcc6fb7f..73e6dc618 100644 --- a/test/syscalls/linux/socket_non_blocking.cc +++ b/test/syscalls/linux/socket_non_blocking.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_non_blocking.h b/test/syscalls/linux/socket_non_blocking.h index 287e096bb..bd3e02fd2 100644 --- a/test/syscalls/linux/socket_non_blocking.h +++ b/test/syscalls/linux/socket_non_blocking.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_non_stream.cc b/test/syscalls/linux/socket_non_stream.cc index d170008a4..3c599b6e8 100644 --- a/test/syscalls/linux/socket_non_stream.cc +++ b/test/syscalls/linux/socket_non_stream.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_non_stream.h b/test/syscalls/linux/socket_non_stream.h index 02dd2a958..469fbe6a2 100644 --- a/test/syscalls/linux/socket_non_stream.h +++ b/test/syscalls/linux/socket_non_stream.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_non_stream_blocking.cc b/test/syscalls/linux/socket_non_stream_blocking.cc index 9e92628c3..76127d181 100644 --- a/test/syscalls/linux/socket_non_stream_blocking.cc +++ b/test/syscalls/linux/socket_non_stream_blocking.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_non_stream_blocking.h b/test/syscalls/linux/socket_non_stream_blocking.h index bde355452..6e205a039 100644 --- a/test/syscalls/linux/socket_non_stream_blocking.h +++ b/test/syscalls/linux/socket_non_stream_blocking.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_stream.cc b/test/syscalls/linux/socket_stream.cc index c8a8ad0f6..0417dd347 100644 --- a/test/syscalls/linux/socket_stream.cc +++ b/test/syscalls/linux/socket_stream.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_stream.h b/test/syscalls/linux/socket_stream.h index 35e591e17..b837b8f8c 100644 --- a/test/syscalls/linux/socket_stream.h +++ b/test/syscalls/linux/socket_stream.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_stream_blocking.cc b/test/syscalls/linux/socket_stream_blocking.cc index f0f86c01c..8367460d2 100644 --- a/test/syscalls/linux/socket_stream_blocking.cc +++ b/test/syscalls/linux/socket_stream_blocking.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_stream_blocking.h b/test/syscalls/linux/socket_stream_blocking.h index 06113ad03..9fd19ff90 100644 --- a/test/syscalls/linux/socket_stream_blocking.h +++ b/test/syscalls/linux/socket_stream_blocking.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_stream_nonblock.cc b/test/syscalls/linux/socket_stream_nonblock.cc index a3202ffe4..b00748b97 100644 --- a/test/syscalls/linux/socket_stream_nonblock.cc +++ b/test/syscalls/linux/socket_stream_nonblock.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_stream_nonblock.h b/test/syscalls/linux/socket_stream_nonblock.h index 491f53848..c3b7fad91 100644 --- a/test/syscalls/linux/socket_stream_nonblock.h +++ b/test/syscalls/linux/socket_stream_nonblock.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_test_util.cc b/test/syscalls/linux/socket_test_util.cc index 0be23e541..da69de37c 100644 --- a/test/syscalls/linux/socket_test_util.cc +++ b/test/syscalls/linux/socket_test_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_test_util.h b/test/syscalls/linux/socket_test_util.h index dfabdf179..058313986 100644 --- a/test/syscalls/linux/socket_test_util.h +++ b/test/syscalls/linux/socket_test_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix.cc b/test/syscalls/linux/socket_unix.cc index fafb23ad1..bb3397fa2 100644 --- a/test/syscalls/linux/socket_unix.cc +++ b/test/syscalls/linux/socket_unix.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix.h b/test/syscalls/linux/socket_unix.h index d2a16afb2..3625cc404 100644 --- a/test/syscalls/linux/socket_unix.h +++ b/test/syscalls/linux/socket_unix.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_abstract.cc b/test/syscalls/linux/socket_unix_abstract.cc index c4a3c889c..8241bf997 100644 --- a/test/syscalls/linux/socket_unix_abstract.cc +++ b/test/syscalls/linux/socket_unix_abstract.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_abstract_nonblock.cc b/test/syscalls/linux/socket_unix_abstract_nonblock.cc index a69ee027e..9de0f6dfe 100644 --- a/test/syscalls/linux/socket_unix_abstract_nonblock.cc +++ b/test/syscalls/linux/socket_unix_abstract_nonblock.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_blocking_local.cc b/test/syscalls/linux/socket_unix_blocking_local.cc index 57af118c5..320915b0f 100644 --- a/test/syscalls/linux/socket_unix_blocking_local.cc +++ b/test/syscalls/linux/socket_unix_blocking_local.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_dgram.cc b/test/syscalls/linux/socket_unix_dgram.cc index 5dd5e6d77..3e0f611d2 100644 --- a/test/syscalls/linux/socket_unix_dgram.cc +++ b/test/syscalls/linux/socket_unix_dgram.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_dgram.h b/test/syscalls/linux/socket_unix_dgram.h index 722a3d8e6..0764ef85b 100644 --- a/test/syscalls/linux/socket_unix_dgram.h +++ b/test/syscalls/linux/socket_unix_dgram.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_dgram_local.cc b/test/syscalls/linux/socket_unix_dgram_local.cc index da8f59704..4ba2c80ae 100644 --- a/test/syscalls/linux/socket_unix_dgram_local.cc +++ b/test/syscalls/linux/socket_unix_dgram_local.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_dgram_non_blocking.cc b/test/syscalls/linux/socket_unix_dgram_non_blocking.cc index 3becb513d..9fe86cee8 100644 --- a/test/syscalls/linux/socket_unix_dgram_non_blocking.cc +++ b/test/syscalls/linux/socket_unix_dgram_non_blocking.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_domain.cc b/test/syscalls/linux/socket_unix_domain.cc index f081c601f..fa3efc7f8 100644 --- a/test/syscalls/linux/socket_unix_domain.cc +++ b/test/syscalls/linux/socket_unix_domain.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_filesystem.cc b/test/syscalls/linux/socket_unix_filesystem.cc index 6a67da75f..5dbe67773 100644 --- a/test/syscalls/linux/socket_unix_filesystem.cc +++ b/test/syscalls/linux/socket_unix_filesystem.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_filesystem_nonblock.cc b/test/syscalls/linux/socket_unix_filesystem_nonblock.cc index c13a1e564..137db53c4 100644 --- a/test/syscalls/linux/socket_unix_filesystem_nonblock.cc +++ b/test/syscalls/linux/socket_unix_filesystem_nonblock.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_non_stream.cc b/test/syscalls/linux/socket_unix_non_stream.cc index a565978f9..dafe82494 100644 --- a/test/syscalls/linux/socket_unix_non_stream.cc +++ b/test/syscalls/linux/socket_unix_non_stream.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_non_stream.h b/test/syscalls/linux/socket_unix_non_stream.h index e4214d949..7478ab172 100644 --- a/test/syscalls/linux/socket_unix_non_stream.h +++ b/test/syscalls/linux/socket_unix_non_stream.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_non_stream_blocking_local.cc b/test/syscalls/linux/socket_unix_non_stream_blocking_local.cc index 6c435669b..98cf1fe8a 100644 --- a/test/syscalls/linux/socket_unix_non_stream_blocking_local.cc +++ b/test/syscalls/linux/socket_unix_non_stream_blocking_local.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_pair.cc b/test/syscalls/linux/socket_unix_pair.cc index c575fdcb2..bacfc11e4 100644 --- a/test/syscalls/linux/socket_unix_pair.cc +++ b/test/syscalls/linux/socket_unix_pair.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_pair_nonblock.cc b/test/syscalls/linux/socket_unix_pair_nonblock.cc index 1ae7f9b5e..583506f08 100644 --- a/test/syscalls/linux/socket_unix_pair_nonblock.cc +++ b/test/syscalls/linux/socket_unix_pair_nonblock.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_seqpacket.cc b/test/syscalls/linux/socket_unix_seqpacket.cc index ad0af77e9..6f6367dd5 100644 --- a/test/syscalls/linux/socket_unix_seqpacket.cc +++ b/test/syscalls/linux/socket_unix_seqpacket.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_seqpacket.h b/test/syscalls/linux/socket_unix_seqpacket.h index da8eb2b2b..30d9b9edf 100644 --- a/test/syscalls/linux/socket_unix_seqpacket.h +++ b/test/syscalls/linux/socket_unix_seqpacket.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_seqpacket_local.cc b/test/syscalls/linux/socket_unix_seqpacket_local.cc index e6484d9b4..b903a9e8f 100644 --- a/test/syscalls/linux/socket_unix_seqpacket_local.cc +++ b/test/syscalls/linux/socket_unix_seqpacket_local.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_stream.cc b/test/syscalls/linux/socket_unix_stream.cc index 95f454251..659c93945 100644 --- a/test/syscalls/linux/socket_unix_stream.cc +++ b/test/syscalls/linux/socket_unix_stream.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_stream_blocking_local.cc b/test/syscalls/linux/socket_unix_stream_blocking_local.cc index ec0fc6955..ce0f1e50d 100644 --- a/test/syscalls/linux/socket_unix_stream_blocking_local.cc +++ b/test/syscalls/linux/socket_unix_stream_blocking_local.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_stream_local.cc b/test/syscalls/linux/socket_unix_stream_local.cc index bf4c5f2eb..6b840189c 100644 --- a/test/syscalls/linux/socket_unix_stream_local.cc +++ b/test/syscalls/linux/socket_unix_stream_local.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_stream_nonblock_local.cc b/test/syscalls/linux/socket_unix_stream_nonblock_local.cc index df80b105a..ebec4e0ec 100644 --- a/test/syscalls/linux/socket_unix_stream_nonblock_local.cc +++ b/test/syscalls/linux/socket_unix_stream_nonblock_local.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_unbound_abstract.cc b/test/syscalls/linux/socket_unix_unbound_abstract.cc index b6fe7a9ce..4b5832de8 100644 --- a/test/syscalls/linux/socket_unix_unbound_abstract.cc +++ b/test/syscalls/linux/socket_unix_unbound_abstract.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_unbound_dgram.cc b/test/syscalls/linux/socket_unix_unbound_dgram.cc index 1ec11a08d..2ddc5c11f 100644 --- a/test/syscalls/linux/socket_unix_unbound_dgram.cc +++ b/test/syscalls/linux/socket_unix_unbound_dgram.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_unbound_filesystem.cc b/test/syscalls/linux/socket_unix_unbound_filesystem.cc index d09142aa6..8cb03c450 100644 --- a/test/syscalls/linux/socket_unix_unbound_filesystem.cc +++ b/test/syscalls/linux/socket_unix_unbound_filesystem.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_unbound_seqpacket.cc b/test/syscalls/linux/socket_unix_unbound_seqpacket.cc index 21209b244..0575f2e1d 100644 --- a/test/syscalls/linux/socket_unix_unbound_seqpacket.cc +++ b/test/syscalls/linux/socket_unix_unbound_seqpacket.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/socket_unix_unbound_stream.cc b/test/syscalls/linux/socket_unix_unbound_stream.cc index b95f9569e..091d546b3 100644 --- a/test/syscalls/linux/socket_unix_unbound_stream.cc +++ b/test/syscalls/linux/socket_unix_unbound_stream.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/stat.cc b/test/syscalls/linux/stat.cc index 746318d09..80ba67496 100644 --- a/test/syscalls/linux/stat.cc +++ b/test/syscalls/linux/stat.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/stat_times.cc b/test/syscalls/linux/stat_times.cc index 8346e9a8e..9b53739a0 100644 --- a/test/syscalls/linux/stat_times.cc +++ b/test/syscalls/linux/stat_times.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/statfs.cc b/test/syscalls/linux/statfs.cc index e1e7fc707..aca51d30f 100644 --- a/test/syscalls/linux/statfs.cc +++ b/test/syscalls/linux/statfs.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sticky.cc b/test/syscalls/linux/sticky.cc index 58cf0d014..59fb5dfe6 100644 --- a/test/syscalls/linux/sticky.cc +++ b/test/syscalls/linux/sticky.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/symlink.cc b/test/syscalls/linux/symlink.cc index 318917f4b..494072a9b 100644 --- a/test/syscalls/linux/symlink.cc +++ b/test/syscalls/linux/symlink.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sync.cc b/test/syscalls/linux/sync.cc index 5b777b6eb..fe479390d 100644 --- a/test/syscalls/linux/sync.cc +++ b/test/syscalls/linux/sync.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sync_file_range.cc b/test/syscalls/linux/sync_file_range.cc index d11f58481..36cc42043 100644 --- a/test/syscalls/linux/sync_file_range.cc +++ b/test/syscalls/linux/sync_file_range.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sysinfo.cc b/test/syscalls/linux/sysinfo.cc index a0dd82640..1a71256da 100644 --- a/test/syscalls/linux/sysinfo.cc +++ b/test/syscalls/linux/sysinfo.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/syslog.cc b/test/syscalls/linux/syslog.cc index 5bd0d1cc3..9a7407d96 100644 --- a/test/syscalls/linux/syslog.cc +++ b/test/syscalls/linux/syslog.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/sysret.cc b/test/syscalls/linux/sysret.cc index 8e10220eb..819fa655a 100644 --- a/test/syscalls/linux/sysret.cc +++ b/test/syscalls/linux/sysret.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/tcp_socket.cc b/test/syscalls/linux/tcp_socket.cc index 33620a874..e3f9f9f9d 100644 --- a/test/syscalls/linux/tcp_socket.cc +++ b/test/syscalls/linux/tcp_socket.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/temp_umask.h b/test/syscalls/linux/temp_umask.h index f202dfa59..81a25440c 100644 --- a/test/syscalls/linux/temp_umask.h +++ b/test/syscalls/linux/temp_umask.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/tgkill.cc b/test/syscalls/linux/tgkill.cc index 2d258ef11..80acae5de 100644 --- a/test/syscalls/linux/tgkill.cc +++ b/test/syscalls/linux/tgkill.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/time.cc b/test/syscalls/linux/time.cc index 5a3dfd026..c7eead17e 100644 --- a/test/syscalls/linux/time.cc +++ b/test/syscalls/linux/time.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/timerfd.cc b/test/syscalls/linux/timerfd.cc index b85321795..9df53612f 100644 --- a/test/syscalls/linux/timerfd.cc +++ b/test/syscalls/linux/timerfd.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/timers.cc b/test/syscalls/linux/timers.cc index 14506eb12..fd42e81e1 100644 --- a/test/syscalls/linux/timers.cc +++ b/test/syscalls/linux/timers.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/tkill.cc b/test/syscalls/linux/tkill.cc index 3e8ce5327..bae377c69 100644 --- a/test/syscalls/linux/tkill.cc +++ b/test/syscalls/linux/tkill.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/truncate.cc b/test/syscalls/linux/truncate.cc index 2616a9147..e5cc5d97c 100644 --- a/test/syscalls/linux/truncate.cc +++ b/test/syscalls/linux/truncate.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/udp_bind.cc b/test/syscalls/linux/udp_bind.cc index 547eb2a6c..6d92bdbeb 100644 --- a/test/syscalls/linux/udp_bind.cc +++ b/test/syscalls/linux/udp_bind.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/udp_socket.cc b/test/syscalls/linux/udp_socket.cc index f39281d5c..31db8a2ad 100644 --- a/test/syscalls/linux/udp_socket.cc +++ b/test/syscalls/linux/udp_socket.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/uidgid.cc b/test/syscalls/linux/uidgid.cc index d78a09b1e..bf1ca8679 100644 --- a/test/syscalls/linux/uidgid.cc +++ b/test/syscalls/linux/uidgid.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/uname.cc b/test/syscalls/linux/uname.cc index d22a34bd7..0a5d91017 100644 --- a/test/syscalls/linux/uname.cc +++ b/test/syscalls/linux/uname.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/unix_domain_socket_test_util.cc b/test/syscalls/linux/unix_domain_socket_test_util.cc index 2d7a530b9..6f49e3660 100644 --- a/test/syscalls/linux/unix_domain_socket_test_util.cc +++ b/test/syscalls/linux/unix_domain_socket_test_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/unix_domain_socket_test_util.h b/test/syscalls/linux/unix_domain_socket_test_util.h index 1b09aeae7..aae990245 100644 --- a/test/syscalls/linux/unix_domain_socket_test_util.h +++ b/test/syscalls/linux/unix_domain_socket_test_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/unlink.cc b/test/syscalls/linux/unlink.cc index b10aae025..b6f65e027 100644 --- a/test/syscalls/linux/unlink.cc +++ b/test/syscalls/linux/unlink.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/unshare.cc b/test/syscalls/linux/unshare.cc index 9dd6ec4b6..e32619efe 100644 --- a/test/syscalls/linux/unshare.cc +++ b/test/syscalls/linux/unshare.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/utimes.cc b/test/syscalls/linux/utimes.cc index bf776cd93..80716859a 100644 --- a/test/syscalls/linux/utimes.cc +++ b/test/syscalls/linux/utimes.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/vdso.cc b/test/syscalls/linux/vdso.cc index 0f6e1c7c6..19c80add8 100644 --- a/test/syscalls/linux/vdso.cc +++ b/test/syscalls/linux/vdso.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/vdso_clock_gettime.cc b/test/syscalls/linux/vdso_clock_gettime.cc index 0e936594b..759a50569 100644 --- a/test/syscalls/linux/vdso_clock_gettime.cc +++ b/test/syscalls/linux/vdso_clock_gettime.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/vfork.cc b/test/syscalls/linux/vfork.cc index 9999a909e..631a53654 100644 --- a/test/syscalls/linux/vfork.cc +++ b/test/syscalls/linux/vfork.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/vsyscall.cc b/test/syscalls/linux/vsyscall.cc index cb6840cc6..2c2303358 100644 --- a/test/syscalls/linux/vsyscall.cc +++ b/test/syscalls/linux/vsyscall.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/wait.cc b/test/syscalls/linux/wait.cc index fcd606bec..50d0725a7 100644 --- a/test/syscalls/linux/wait.cc +++ b/test/syscalls/linux/wait.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/linux/write.cc b/test/syscalls/linux/write.cc index 7f80b2fa8..9b219cfd6 100644 --- a/test/syscalls/linux/write.cc +++ b/test/syscalls/linux/write.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/syscall_test_runner.go b/test/syscalls/syscall_test_runner.go index c4af28103..28f312b8b 100644 --- a/test/syscalls/syscall_test_runner.go +++ b/test/syscalls/syscall_test_runner.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/syscalls/syscall_test_runner.sh b/test/syscalls/syscall_test_runner.sh index 87d62786b..864bb2de4 100755 --- a/test/syscalls/syscall_test_runner.sh +++ b/test/syscalls/syscall_test_runner.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/test/util/capability_util.cc b/test/util/capability_util.cc index d1dd95e76..5d733887b 100644 --- a/test/util/capability_util.cc +++ b/test/util/capability_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/capability_util.h b/test/util/capability_util.h index 8708f5e69..e968a2583 100644 --- a/test/util/capability_util.h +++ b/test/util/capability_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/cleanup.h b/test/util/cleanup.h index fb4724f97..c76482ef4 100644 --- a/test/util/cleanup.h +++ b/test/util/cleanup.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/epoll_util.cc b/test/util/epoll_util.cc index 0b95aa8cd..2e5051468 100644 --- a/test/util/epoll_util.cc +++ b/test/util/epoll_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/epoll_util.h b/test/util/epoll_util.h index 521e7a3d3..f233b37d5 100644 --- a/test/util/epoll_util.h +++ b/test/util/epoll_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/eventfd_util.h b/test/util/eventfd_util.h index 1fdb07d3b..cb9ce829c 100644 --- a/test/util/eventfd_util.h +++ b/test/util/eventfd_util.h @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/file_descriptor.h b/test/util/file_descriptor.h index be8812d01..fc5caa55b 100644 --- a/test/util/file_descriptor.h +++ b/test/util/file_descriptor.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/fs_util.cc b/test/util/fs_util.cc index 6bd424417..bc90bd78e 100644 --- a/test/util/fs_util.cc +++ b/test/util/fs_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/fs_util.h b/test/util/fs_util.h index 9412b2f71..eb7cdaa24 100644 --- a/test/util/fs_util.h +++ b/test/util/fs_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/fs_util_test.cc b/test/util/fs_util_test.cc index ce70d58aa..4e12076a1 100644 --- a/test/util/fs_util_test.cc +++ b/test/util/fs_util_test.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/logging.cc b/test/util/logging.cc index 86ea71df3..cc71d77b0 100644 --- a/test/util/logging.cc +++ b/test/util/logging.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/logging.h b/test/util/logging.h index 6e957b172..589166fab 100644 --- a/test/util/logging.h +++ b/test/util/logging.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/memory_util.h b/test/util/memory_util.h index 8f6e99ba6..8c77778ea 100644 --- a/test/util/memory_util.h +++ b/test/util/memory_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/mount_util.h b/test/util/mount_util.h index 468170646..7782e6bf2 100644 --- a/test/util/mount_util.h +++ b/test/util/mount_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/multiprocess_util.cc b/test/util/multiprocess_util.cc index 12637db8c..95f5f3b4f 100644 --- a/test/util/multiprocess_util.cc +++ b/test/util/multiprocess_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/multiprocess_util.h b/test/util/multiprocess_util.h index ba5f2601f..0aecd3439 100644 --- a/test/util/multiprocess_util.h +++ b/test/util/multiprocess_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/posix_error.cc b/test/util/posix_error.cc index ead9ede16..cebf7e0ac 100644 --- a/test/util/posix_error.cc +++ b/test/util/posix_error.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/posix_error.h b/test/util/posix_error.h index 2a66e2e94..b604f4f8f 100644 --- a/test/util/posix_error.h +++ b/test/util/posix_error.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/posix_error_test.cc b/test/util/posix_error_test.cc index c5427b8e5..d67270842 100644 --- a/test/util/posix_error_test.cc +++ b/test/util/posix_error_test.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/proc_util.cc b/test/util/proc_util.cc index 2d9eb1986..9d4db37c3 100644 --- a/test/util/proc_util.cc +++ b/test/util/proc_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/proc_util.h b/test/util/proc_util.h index e1ee2db9c..af209a51e 100644 --- a/test/util/proc_util.h +++ b/test/util/proc_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/proc_util_test.cc b/test/util/proc_util_test.cc index 75335415a..71dd2355e 100644 --- a/test/util/proc_util_test.cc +++ b/test/util/proc_util_test.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/rlimit_util.cc b/test/util/rlimit_util.cc index a9912c372..684253f78 100644 --- a/test/util/rlimit_util.cc +++ b/test/util/rlimit_util.cc @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/rlimit_util.h b/test/util/rlimit_util.h index fa5cc70dc..873252a32 100644 --- a/test/util/rlimit_util.h +++ b/test/util/rlimit_util.h @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/save_util.cc b/test/util/save_util.cc index 5540e2146..05f52b80d 100644 --- a/test/util/save_util.cc +++ b/test/util/save_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/save_util.h b/test/util/save_util.h index 919e4af3d..90460701e 100644 --- a/test/util/save_util.h +++ b/test/util/save_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/signal_util.cc b/test/util/signal_util.cc index 3e2df32a6..26738864f 100644 --- a/test/util/signal_util.cc +++ b/test/util/signal_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/signal_util.h b/test/util/signal_util.h index 80f1808f6..7fd2af015 100644 --- a/test/util/signal_util.h +++ b/test/util/signal_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/temp_path.cc b/test/util/temp_path.cc index 48ce82d20..c5d8fc635 100644 --- a/test/util/temp_path.cc +++ b/test/util/temp_path.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/temp_path.h b/test/util/temp_path.h index 33eb6a72c..89302e0fd 100644 --- a/test/util/temp_path.h +++ b/test/util/temp_path.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/test_main.cc b/test/util/test_main.cc index 4c6b5e860..5c7ee0064 100644 --- a/test/util/test_main.cc +++ b/test/util/test_main.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/test_util.cc b/test/util/test_util.cc index 9b7cfa4dc..c52fd9a4a 100644 --- a/test/util/test_util.cc +++ b/test/util/test_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/test_util.h b/test/util/test_util.h index 905412b24..8f5eb5089 100644 --- a/test/util/test_util.h +++ b/test/util/test_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/test_util_test.cc b/test/util/test_util_test.cc index 5889651d1..b7300d9e5 100644 --- a/test/util/test_util_test.cc +++ b/test/util/test_util_test.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/thread_util.h b/test/util/thread_util.h index df09ac8cf..860e77531 100644 --- a/test/util/thread_util.h +++ b/test/util/thread_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/timer_util.cc b/test/util/timer_util.cc index 681fafb69..43a26b0d3 100644 --- a/test/util/timer_util.cc +++ b/test/util/timer_util.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/test/util/timer_util.h b/test/util/timer_util.h index 9bdc51a57..2cebfa5d1 100644 --- a/test/util/timer_util.h +++ b/test/util/timer_util.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/third_party/gvsync/atomicptr_unsafe.go b/third_party/gvsync/atomicptr_unsafe.go index da9f16240..53a943282 100644 --- a/third_party/gvsync/atomicptr_unsafe.go +++ b/third_party/gvsync/atomicptr_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/atomicptrtest/atomicptr_test.go b/third_party/gvsync/atomicptrtest/atomicptr_test.go index 15d0936d4..8fdc5112e 100644 --- a/third_party/gvsync/atomicptrtest/atomicptr_test.go +++ b/third_party/gvsync/atomicptrtest/atomicptr_test.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/downgradable_rwmutex_test.go b/third_party/gvsync/downgradable_rwmutex_test.go index 6517dd5dc..40c384b8b 100644 --- a/third_party/gvsync/downgradable_rwmutex_test.go +++ b/third_party/gvsync/downgradable_rwmutex_test.go @@ -1,5 +1,5 @@ // Copyright 2009 The Go Authors. All rights reserved. -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/downgradable_rwmutex_unsafe.go b/third_party/gvsync/downgradable_rwmutex_unsafe.go index 131f0a2ba..4d43eb765 100644 --- a/third_party/gvsync/downgradable_rwmutex_unsafe.go +++ b/third_party/gvsync/downgradable_rwmutex_unsafe.go @@ -1,5 +1,5 @@ // Copyright 2009 The Go Authors. All rights reserved. -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/gvsync.go b/third_party/gvsync/gvsync.go index 46a2565fd..3bbef13c3 100644 --- a/third_party/gvsync/gvsync.go +++ b/third_party/gvsync/gvsync.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/memmove_unsafe.go b/third_party/gvsync/memmove_unsafe.go index d483fc739..4c8aa9ab6 100644 --- a/third_party/gvsync/memmove_unsafe.go +++ b/third_party/gvsync/memmove_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/norace_unsafe.go b/third_party/gvsync/norace_unsafe.go index f9c88d13f..e3852db8c 100644 --- a/third_party/gvsync/norace_unsafe.go +++ b/third_party/gvsync/norace_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/race_unsafe.go b/third_party/gvsync/race_unsafe.go index 2cdcdf7f7..13c02a830 100644 --- a/third_party/gvsync/race_unsafe.go +++ b/third_party/gvsync/race_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/seqatomic_unsafe.go b/third_party/gvsync/seqatomic_unsafe.go index ef61503e2..c52d378f1 100644 --- a/third_party/gvsync/seqatomic_unsafe.go +++ b/third_party/gvsync/seqatomic_unsafe.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/seqatomictest/seqatomic_test.go b/third_party/gvsync/seqatomictest/seqatomic_test.go index d0c373bae..2da73cf96 100644 --- a/third_party/gvsync/seqatomictest/seqatomic_test.go +++ b/third_party/gvsync/seqatomictest/seqatomic_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/third_party/gvsync/seqcount.go b/third_party/gvsync/seqcount.go index c7ae91cfa..2c9c2c3d6 100644 --- a/third_party/gvsync/seqcount.go +++ b/third_party/gvsync/seqcount.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/third_party/gvsync/seqcount_test.go b/third_party/gvsync/seqcount_test.go index ee6579ed8..085e574b3 100644 --- a/third_party/gvsync/seqcount_test.go +++ b/third_party/gvsync/seqcount_test.go @@ -1,4 +1,4 @@ -// Copyright 2019 Google LLC +// Copyright 2019 The gVisor Authors. // // Use of this source code is governed by a BSD-style // license that can be found in the LICENSE file. diff --git a/tools/go_generics/generics.go b/tools/go_generics/generics.go index eaf5c4970..ca414d8cb 100644 --- a/tools/go_generics/generics.go +++ b/tools/go_generics/generics.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_stmts/input.go b/tools/go_generics/generics_tests/all_stmts/input.go index 19184a3fe..4791d1ff1 100644 --- a/tools/go_generics/generics_tests/all_stmts/input.go +++ b/tools/go_generics/generics_tests/all_stmts/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_stmts/output/output.go b/tools/go_generics/generics_tests/all_stmts/output/output.go index 51582346c..a53d84535 100644 --- a/tools/go_generics/generics_tests/all_stmts/output/output.go +++ b/tools/go_generics/generics_tests/all_stmts/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_types/input.go b/tools/go_generics/generics_tests/all_types/input.go index ed6e97c29..3575d02ec 100644 --- a/tools/go_generics/generics_tests/all_types/input.go +++ b/tools/go_generics/generics_tests/all_types/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_types/lib/lib.go b/tools/go_generics/generics_tests/all_types/lib/lib.go index 7e73e678e..988786496 100644 --- a/tools/go_generics/generics_tests/all_types/lib/lib.go +++ b/tools/go_generics/generics_tests/all_types/lib/lib.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/all_types/output/output.go b/tools/go_generics/generics_tests/all_types/output/output.go index ec09a6be4..41fd147a1 100644 --- a/tools/go_generics/generics_tests/all_types/output/output.go +++ b/tools/go_generics/generics_tests/all_types/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/consts/input.go b/tools/go_generics/generics_tests/consts/input.go index 394bcc262..04b95fcc6 100644 --- a/tools/go_generics/generics_tests/consts/input.go +++ b/tools/go_generics/generics_tests/consts/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/consts/output/output.go b/tools/go_generics/generics_tests/consts/output/output.go index 91a07fdc2..18d316cc9 100644 --- a/tools/go_generics/generics_tests/consts/output/output.go +++ b/tools/go_generics/generics_tests/consts/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/imports/input.go b/tools/go_generics/generics_tests/imports/input.go index 22e6641a6..0f032c2a1 100644 --- a/tools/go_generics/generics_tests/imports/input.go +++ b/tools/go_generics/generics_tests/imports/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/imports/output/output.go b/tools/go_generics/generics_tests/imports/output/output.go index 2555c0004..2488ca58c 100644 --- a/tools/go_generics/generics_tests/imports/output/output.go +++ b/tools/go_generics/generics_tests/imports/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/remove_typedef/input.go b/tools/go_generics/generics_tests/remove_typedef/input.go index d9c9b8530..cf632bae7 100644 --- a/tools/go_generics/generics_tests/remove_typedef/input.go +++ b/tools/go_generics/generics_tests/remove_typedef/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/remove_typedef/output/output.go b/tools/go_generics/generics_tests/remove_typedef/output/output.go index f111a9426..d44fd8e1c 100644 --- a/tools/go_generics/generics_tests/remove_typedef/output/output.go +++ b/tools/go_generics/generics_tests/remove_typedef/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/simple/input.go b/tools/go_generics/generics_tests/simple/input.go index 711687cf5..2a917f16c 100644 --- a/tools/go_generics/generics_tests/simple/input.go +++ b/tools/go_generics/generics_tests/simple/input.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/generics_tests/simple/output/output.go b/tools/go_generics/generics_tests/simple/output/output.go index 139c9bf9d..6bfa0b25b 100644 --- a/tools/go_generics/generics_tests/simple/output/output.go +++ b/tools/go_generics/generics_tests/simple/output/output.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/globals/globals_visitor.go b/tools/go_generics/globals/globals_visitor.go index daaa17b1d..7ae48c662 100644 --- a/tools/go_generics/globals/globals_visitor.go +++ b/tools/go_generics/globals/globals_visitor.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/globals/scope.go b/tools/go_generics/globals/scope.go index b75a91689..96c965ea2 100644 --- a/tools/go_generics/globals/scope.go +++ b/tools/go_generics/globals/scope.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/go_generics_unittest.sh b/tools/go_generics/go_generics_unittest.sh index e7553a071..44b22db91 100755 --- a/tools/go_generics/go_generics_unittest.sh +++ b/tools/go_generics/go_generics_unittest.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/tools/go_generics/go_merge/main.go b/tools/go_generics/go_merge/main.go index 2f83facf8..f6a331123 100644 --- a/tools/go_generics/go_merge/main.go +++ b/tools/go_generics/go_merge/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/imports.go b/tools/go_generics/imports.go index 57f7c3dce..3a7230c97 100644 --- a/tools/go_generics/imports.go +++ b/tools/go_generics/imports.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/remove.go b/tools/go_generics/remove.go index 139d03955..568a6bbd3 100644 --- a/tools/go_generics/remove.go +++ b/tools/go_generics/remove.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/rules_tests/template.go b/tools/go_generics/rules_tests/template.go index f3f31ae8e..aace61da1 100644 --- a/tools/go_generics/rules_tests/template.go +++ b/tools/go_generics/rules_tests/template.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_generics/rules_tests/template_test.go b/tools/go_generics/rules_tests/template_test.go index 3a38c8629..b2a3446ef 100644 --- a/tools/go_generics/rules_tests/template_test.go +++ b/tools/go_generics/rules_tests/template_test.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/go_stateify/main.go b/tools/go_stateify/main.go index 9e2c8e106..db7a7107b 100644 --- a/tools/go_stateify/main.go +++ b/tools/go_stateify/main.go @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/tools/tag_release.sh b/tools/tag_release.sh index 6906a952f..02a49cdf1 100755 --- a/tools/tag_release.sh +++ b/tools/tag_release.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2019 Google LLC +# Copyright 2019 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/tools/workspace_status.sh b/tools/workspace_status.sh index a0e646e45..64a905fc9 100755 --- a/tools/workspace_status.sh +++ b/tools/workspace_status.sh @@ -1,6 +1,6 @@ #!/bin/bash -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/vdso/barrier.h b/vdso/barrier.h index 5b6c763f6..edba4afb5 100644 --- a/vdso/barrier.h +++ b/vdso/barrier.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/check_vdso.py b/vdso/check_vdso.py index 6f7d7e7ec..e41b09709 100644 --- a/vdso/check_vdso.py +++ b/vdso/check_vdso.py @@ -1,4 +1,4 @@ -# Copyright 2018 Google LLC +# Copyright 2018 The gVisor Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/vdso/compiler.h b/vdso/compiler.h index d65f148fb..54a510000 100644 --- a/vdso/compiler.h +++ b/vdso/compiler.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/cycle_clock.h b/vdso/cycle_clock.h index 309e07a3f..5d3fbb257 100644 --- a/vdso/cycle_clock.h +++ b/vdso/cycle_clock.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/seqlock.h b/vdso/seqlock.h index ab2f3fda3..7a173174b 100644 --- a/vdso/seqlock.h +++ b/vdso/seqlock.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/syscalls.h b/vdso/syscalls.h index 90fb424ce..f5865bb72 100644 --- a/vdso/syscalls.h +++ b/vdso/syscalls.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/vdso.cc b/vdso/vdso.cc index 550729035..6265ad217 100644 --- a/vdso/vdso.cc +++ b/vdso/vdso.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/vdso_time.cc b/vdso/vdso_time.cc index 9fc262f60..1bb4bb86b 100644 --- a/vdso/vdso_time.cc +++ b/vdso/vdso_time.cc @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. diff --git a/vdso/vdso_time.h b/vdso/vdso_time.h index 464dadff2..70d079efc 100644 --- a/vdso/vdso_time.h +++ b/vdso/vdso_time.h @@ -1,4 +1,4 @@ -// Copyright 2018 Google LLC +// Copyright 2018 The gVisor Authors. // // Licensed under the Apache License, Version 2.0 (the "License"); // you may not use this file except in compliance with the License. -- cgit v1.2.3 From 8bfb83d0acdea553082b897d3fd0ad1c1580eaa9 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Tue, 30 Apr 2019 13:55:41 -0700 Subject: Implement async MemoryFile eviction, and use it in CachingInodeOperations. This feature allows MemoryFile to delay eviction of "optional" allocations, such as unused cached file pages. Note that this incidentally makes CachingInodeOperations writeback asynchronous, in the sense that it doesn't occur until eviction; this is necessary because between when a cached page becomes evictable and when it's evicted, file writes (via CachingInodeOperations.Write) may dirty the page. As currently implemented, this feature won't meaningfully impact steady-state memory usage or caching; the reclaimer goroutine will schedule eviction as soon as it runs out of other work to do. Future CLs increase caching by adding constraints on when eviction is scheduled. PiperOrigin-RevId: 246014822 Change-Id: Ia85feb25a2de92a48359eb84434b6ec6f9bea2cb --- pkg/sentry/context/contexttest/contexttest.go | 2 +- pkg/sentry/fs/fsutil/dirty_set.go | 22 ++ pkg/sentry/fs/fsutil/inode_cached.go | 78 +++- pkg/sentry/fs/fsutil/inode_cached_test.go | 8 +- pkg/sentry/kernel/kernel.go | 5 + pkg/sentry/pgalloc/BUILD | 27 ++ pkg/sentry/pgalloc/pgalloc.go | 504 +++++++++++++++++++------- pkg/sentry/pgalloc/save_restore.go | 14 + runsc/boot/loader.go | 2 +- 9 files changed, 511 insertions(+), 151 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/context/contexttest/contexttest.go b/pkg/sentry/context/contexttest/contexttest.go index a42038711..210a235d2 100644 --- a/pkg/sentry/context/contexttest/contexttest.go +++ b/pkg/sentry/context/contexttest/contexttest.go @@ -44,7 +44,7 @@ func Context(tb testing.TB) context.Context { tb.Fatalf("error creating application memory file: %v", err) } memfile := os.NewFile(uintptr(memfd), memfileName) - mf, err := pgalloc.NewMemoryFile(memfile) + mf, err := pgalloc.NewMemoryFile(memfile, pgalloc.MemoryFileOpts{}) if err != nil { memfile.Close() tb.Fatalf("error creating pgalloc.MemoryFile: %v", err) diff --git a/pkg/sentry/fs/fsutil/dirty_set.go b/pkg/sentry/fs/fsutil/dirty_set.go index 9cd196d7d..f1451d77a 100644 --- a/pkg/sentry/fs/fsutil/dirty_set.go +++ b/pkg/sentry/fs/fsutil/dirty_set.go @@ -107,6 +107,7 @@ func (ds *DirtySet) setDirty(mr memmap.MappableRange, keep bool) { var changedAny bool defer func() { if changedAny { + // Merge segments split by Isolate to reduce cost of iteration. ds.MergeRange(mr) } }() @@ -132,6 +133,26 @@ func (ds *DirtySet) setDirty(mr memmap.MappableRange, keep bool) { } } +// AllowClean allows MarkClean to mark offsets in mr as not dirty, ending the +// effect of a previous call to KeepDirty. (It does not itself mark those +// offsets as not dirty.) +func (ds *DirtySet) AllowClean(mr memmap.MappableRange) { + var changedAny bool + defer func() { + if changedAny { + // Merge segments split by Isolate to reduce cost of iteration. + ds.MergeRange(mr) + } + }() + for seg := ds.LowerBoundSegment(mr.Start); seg.Ok() && seg.Start() < mr.End; seg = seg.NextSegment() { + if seg.Value().Keep { + changedAny = true + seg = ds.Isolate(seg, mr) + seg.ValuePtr().Keep = false + } + } +} + // SyncDirty passes pages in the range mr that are stored in cache and // identified as dirty to writeAt, updating dirty to reflect successful writes. // If writeAt returns a successful partial write, SyncDirty will call it @@ -142,6 +163,7 @@ func SyncDirty(ctx context.Context, mr memmap.MappableRange, cache *FileRangeSet var changedDirty bool defer func() { if changedDirty { + // Merge segments split by Isolate to reduce cost of iteration. dirty.MergeRange(mr) } }() diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index 919d2534c..76644e69d 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -175,11 +175,22 @@ func (c *CachingInodeOperations) Release() { defer c.mapsMu.Unlock() c.dataMu.Lock() defer c.dataMu.Unlock() - // The cache should be empty (something has gone terribly wrong if we're - // releasing an inode that is still memory-mapped). - if !c.mappings.IsEmpty() || !c.cache.IsEmpty() || !c.dirty.IsEmpty() { - panic(fmt.Sprintf("Releasing CachingInodeOperations with mappings:\n%s\ncache contents:\n%s\ndirty segments:\n%s", &c.mappings, &c.cache, &c.dirty)) + + // Something has gone terribly wrong if we're releasing an inode that is + // still memory-mapped. + if !c.mappings.IsEmpty() { + panic(fmt.Sprintf("Releasing CachingInodeOperations with mappings:\n%s", &c.mappings)) + } + + // Drop any cached pages that are still awaiting MemoryFile eviction. (This + // means that MemoryFile no longer needs to evict them.) + mf := c.mfp.MemoryFile() + mf.MarkAllUnevictable(c) + if err := SyncDirtyAll(context.Background(), &c.cache, &c.dirty, uint64(c.attr.Size), mf, c.backingFile.WriteFromBlocksAt); err != nil { + panic(fmt.Sprintf("Failed to writeback cached data: %v", err)) } + c.cache.DropAll(mf) + c.dirty.RemoveAll() } // UnstableAttr implements fs.InodeOperations.UnstableAttr. @@ -679,6 +690,13 @@ func (rw *inodeReadWriter) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error return done, nil } +// useHostPageCache returns true if c uses c.backingFile.FD() for all file I/O +// and memory mappings, and false if c.cache may contain data cached from +// c.backingFile. +func (c *CachingInodeOperations) useHostPageCache() bool { + return !c.forcePageCache && c.backingFile.FD() >= 0 +} + // AddMapping implements memmap.Mappable.AddMapping. func (c *CachingInodeOperations) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar usermem.AddrRange, offset uint64, writable bool) error { // Hot path. Avoid defers. @@ -689,7 +707,15 @@ func (c *CachingInodeOperations) AddMapping(ctx context.Context, ms memmap.Mappi for _, r := range mapped { c.hostFileMapper.IncRefOn(r) } - if !usage.IncrementalMappedAccounting && !c.forcePageCache && c.backingFile.FD() >= 0 { + if !c.useHostPageCache() { + // c.Evict() will refuse to evict memory-mapped pages, so tell the + // MemoryFile to not bother trying. + mf := c.mfp.MemoryFile() + for _, r := range mapped { + mf.MarkUnevictable(c, pgalloc.EvictableRange{r.Start, r.End}) + } + } + if c.useHostPageCache() && !usage.IncrementalMappedAccounting { for _, r := range mapped { usage.MemoryAccounting.Inc(r.Length(), usage.Mapped) } @@ -706,7 +732,7 @@ func (c *CachingInodeOperations) RemoveMapping(ctx context.Context, ms memmap.Ma for _, r := range unmapped { c.hostFileMapper.DecRefOn(r) } - if !c.forcePageCache && c.backingFile.FD() >= 0 { + if c.useHostPageCache() { if !usage.IncrementalMappedAccounting { for _, r := range unmapped { usage.MemoryAccounting.Dec(r.Length(), usage.Mapped) @@ -716,17 +742,16 @@ func (c *CachingInodeOperations) RemoveMapping(ctx context.Context, ms memmap.Ma return } - // Writeback dirty mapped memory now that there are no longer any - // mappings that reference it. This is our naive memory eviction - // strategy. + // Pages that are no longer referenced by any application memory mappings + // are now considered unused; allow MemoryFile to evict them when + // necessary. mf := c.mfp.MemoryFile() c.dataMu.Lock() for _, r := range unmapped { - if err := SyncDirty(ctx, r, &c.cache, &c.dirty, uint64(c.attr.Size), mf, c.backingFile.WriteFromBlocksAt); err != nil { - log.Warningf("Failed to writeback cached data %v: %v", r, err) - } - c.cache.Drop(r, mf) - c.dirty.KeepClean(r) + // Since these pages are no longer mapped, they are no longer + // concurrently dirtyable by a writable memory mapping. + c.dirty.AllowClean(r) + mf.MarkEvictable(c, pgalloc.EvictableRange{r.Start, r.End}) } c.dataMu.Unlock() c.mapsMu.Unlock() @@ -740,7 +765,7 @@ func (c *CachingInodeOperations) CopyMapping(ctx context.Context, ms memmap.Mapp // Translate implements memmap.Mappable.Translate. func (c *CachingInodeOperations) Translate(ctx context.Context, required, optional memmap.MappableRange, at usermem.AccessType) ([]memmap.Translation, error) { // Hot path. Avoid defer. - if !c.forcePageCache && c.backingFile.FD() >= 0 { + if c.useHostPageCache() { return []memmap.Translation{ { Source: optional, @@ -853,6 +878,29 @@ func (c *CachingInodeOperations) InvalidateUnsavable(ctx context.Context) error return nil } +// Evict implements pgalloc.EvictableMemoryUser.Evict. +func (c *CachingInodeOperations) Evict(ctx context.Context, er pgalloc.EvictableRange) { + c.mapsMu.Lock() + defer c.mapsMu.Unlock() + c.dataMu.Lock() + defer c.dataMu.Unlock() + + mr := memmap.MappableRange{er.Start, er.End} + mf := c.mfp.MemoryFile() + // Only allow pages that are no longer memory-mapped to be evicted. + for mgap := c.mappings.LowerBoundGap(mr.Start); mgap.Ok() && mgap.Start() < mr.End; mgap = mgap.NextGap() { + mgapMR := mgap.Range().Intersect(mr) + if mgapMR.Length() == 0 { + continue + } + if err := SyncDirty(ctx, mgapMR, &c.cache, &c.dirty, uint64(c.attr.Size), mf, c.backingFile.WriteFromBlocksAt); err != nil { + log.Warningf("Failed to writeback cached data %v: %v", mgapMR, err) + } + c.cache.Drop(mgapMR, mf) + c.dirty.KeepClean(mgapMR) + } +} + // IncRef implements platform.File.IncRef. This is used when we directly map an // underlying host fd and CachingInodeOperations is used as the platform.File // during translation. diff --git a/pkg/sentry/fs/fsutil/inode_cached_test.go b/pkg/sentry/fs/fsutil/inode_cached_test.go index 661ec41f6..3f10efc12 100644 --- a/pkg/sentry/fs/fsutil/inode_cached_test.go +++ b/pkg/sentry/fs/fsutil/inode_cached_test.go @@ -311,12 +311,10 @@ func TestRead(t *testing.T) { t.Errorf("Read back bytes %v, want %v", rbuf, buf) } - // Delete the memory mapping and expect it to cause the cached page to be - // uncached. + // Delete the memory mapping before iops.Release(). The cached page will + // either be evicted by ctx's pgalloc.MemoryFile, or dropped by + // iops.Release(). iops.RemoveMapping(ctx, ms, ar, usermem.PageSize, true) - if cached := iops.cache.Span(); cached != 0 { - t.Fatalf("Span got %d, want 0", cached) - } } func TestWrite(t *testing.T) { diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 0468dd678..91889b573 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -303,7 +303,12 @@ func (k *Kernel) SaveTo(w io.Writer) error { k.pauseTimeLocked() defer k.resumeTimeLocked() + // Evict all evictable MemoryFile allocations. + k.mf.FlushEvictions() + // Flush write operations on open files so data reaches backing storage. + // This must come after k.mf.FlushEvictions() since eviction may cause file + // writes. if err := k.tasks.flushWritesToFiles(ctx); err != nil { return err } diff --git a/pkg/sentry/pgalloc/BUILD b/pkg/sentry/pgalloc/BUILD index 7efa55c20..8a8a0e4e4 100644 --- a/pkg/sentry/pgalloc/BUILD +++ b/pkg/sentry/pgalloc/BUILD @@ -3,6 +3,31 @@ package(licenses = ["notice"]) load("//tools/go_generics:defs.bzl", "go_template_instance") load("//tools/go_stateify:defs.bzl", "go_library", "go_test") +go_template_instance( + name = "evictable_range", + out = "evictable_range.go", + package = "pgalloc", + prefix = "Evictable", + template = "//pkg/segment:generic_range", + types = { + "T": "uint64", + }, +) + +go_template_instance( + name = "evictable_range_set", + out = "evictable_range_set.go", + package = "pgalloc", + prefix = "evictableRange", + template = "//pkg/segment:generic_set", + types = { + "Key": "uint64", + "Range": "EvictableRange", + "Value": "evictableRangeSetValue", + "Functions": "evictableRangeSetFunctions", + }, +) + go_template_instance( name = "usage_set", out = "usage_set.go", @@ -27,6 +52,8 @@ go_library( name = "pgalloc", srcs = [ "context.go", + "evictable_range.go", + "evictable_range_set.go", "pgalloc.go", "pgalloc_unsafe.go", "save_restore.go", diff --git a/pkg/sentry/pgalloc/pgalloc.go b/pkg/sentry/pgalloc/pgalloc.go index 411dafa07..9c1313f6f 100644 --- a/pkg/sentry/pgalloc/pgalloc.go +++ b/pkg/sentry/pgalloc/pgalloc.go @@ -31,6 +31,7 @@ import ( "time" "gvisor.googlesource.com/gvisor/pkg/log" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usage" @@ -41,6 +42,9 @@ import ( // MemoryFile is a platform.File whose pages may be allocated to arbitrary // users. type MemoryFile struct { + // opts holds options passed to NewMemoryFile. opts is immutable. + opts MemoryFileOpts + // MemoryFile owns a single backing file, which is modeled as follows: // // Each page in the file can be committed or uncommitted. A page is @@ -115,6 +119,24 @@ type MemoryFile struct { // fileSize is protected by mu. fileSize int64 + // Pages from the backing file are mapped into the local address space on + // the granularity of large pieces called chunks. mappings is a []uintptr + // that stores, for each chunk, the start address of a mapping of that + // chunk in the current process' address space, or 0 if no such mapping + // exists. Once a chunk is mapped, it is never remapped or unmapped until + // the MemoryFile is destroyed. + // + // Mutating the mappings slice or its contents requires both holding + // mappingsMu and using atomic memory operations. (The slice is mutated + // whenever the file is expanded. Per the above, the only permitted + // mutation of the slice's contents is the assignment of a mapping to a + // chunk that was previously unmapped.) Reading the slice or its contents + // only requires *either* holding mappingsMu or using atomic memory + // operations. This allows MemoryFile.MapInternal to avoid locking in the + // common case where chunk mappings already exist. + mappingsMu sync.Mutex + mappings atomic.Value + // destroyed is set by Destroy to instruct the reclaimer goroutine to // release resources and exit. destroyed is protected by mu. destroyed bool @@ -133,26 +155,44 @@ type MemoryFile struct { // transitions from false to true. reclaimCond sync.Cond - // Pages from the backing file are mapped into the local address space on - // the granularity of large pieces called chunks. mappings is a []uintptr - // that stores, for each chunk, the start address of a mapping of that - // chunk in the current process' address space, or 0 if no such mapping - // exists. Once a chunk is mapped, it is never remapped or unmapped until - // the MemoryFile is destroyed. + // evictable maps EvictableMemoryUsers to eviction state. // - // Mutating the mappings slice or its contents requires both holding - // mappingsMu and using atomic memory operations. (The slice is mutated - // whenever the file is expanded. Per the above, the only permitted - // mutation of the slice's contents is the assignment of a mapping to a - // chunk that was previously unmapped.) Reading the slice or its contents - // only requires *either* holding mappingsMu or using atomic memory - // operations. This allows MemoryFile.MapInternal to avoid locking in the - // common case where chunk mappings already exist. - mappingsMu sync.Mutex - mappings atomic.Value + // evictable is protected by mu. + evictable map[EvictableMemoryUser]*evictableMemoryUserInfo + + // evictionWG counts the number of goroutines currently performing evictions. + evictionWG sync.WaitGroup +} + +// MemoryFileOpts provides options to NewMemoryFile. +type MemoryFileOpts struct { + // DelayedEviction controls the extent to which the MemoryFile may delay + // eviction of evictable allocations. + DelayedEviction DelayedEvictionType } -// usage tracks usage information. +// DelayedEvictionType is the type of MemoryFileOpts.DelayedEviction. +type DelayedEvictionType int + +const ( + // DelayedEvictionDefault has unspecified behavior. + DelayedEvictionDefault DelayedEvictionType = iota + + // DelayedEvictionDisabled requires that evictable allocations are evicted + // as soon as possible. + DelayedEvictionDisabled + + // DelayedEvictionEnabled requests that the MemoryFile delay eviction of + // evictable allocations until doing so is considered necessary to avoid + // performance degradation due to host memory pressure, or OOM kills. + // + // As of this writing, DelayedEvictionEnabled delays evictions until the + // reclaimer goroutine is out of work (pages to reclaim), then evicts all + // pending evictable allocations immediately. + DelayedEvictionEnabled +) + +// usageInfo tracks usage information. // // +stateify savable type usageInfo struct { @@ -166,6 +206,46 @@ type usageInfo struct { refs uint64 } +// An EvictableMemoryUser represents a user of MemoryFile-allocated memory that +// may be asked to deallocate that memory in the presence of memory pressure. +type EvictableMemoryUser interface { + // Evict requests that the EvictableMemoryUser deallocate memory used by + // er, which was registered as evictable by a previous call to + // MemoryFile.MarkEvictable. + // + // Evict is not required to deallocate memory. In particular, since pgalloc + // must call Evict without holding locks to avoid circular lock ordering, + // it is possible that the passed range has already been marked as + // unevictable by a racing call to MemoryFile.MarkUnevictable. + // Implementations of EvictableMemoryUser must detect such races and handle + // them by making Evict have no effect on unevictable ranges. + // + // After a call to Evict, the MemoryFile will consider the evicted range + // unevictable (i.e. it will not call Evict on the same range again) until + // informed otherwise by a subsequent call to MarkEvictable. + Evict(ctx context.Context, er EvictableRange) +} + +// An EvictableRange represents a range of uint64 offsets in an +// EvictableMemoryUser. +// +// In practice, most EvictableMemoryUsers will probably be implementations of +// memmap.Mappable, and EvictableRange therefore corresponds to +// memmap.MappableRange. However, this package cannot depend on the memmap +// package, since doing so would create a circular dependency. +// +// type EvictableRange + +// evictableMemoryUserInfo is the value type of MemoryFile.evictable. +type evictableMemoryUserInfo struct { + // ranges tracks all evictable ranges for the given user. + ranges evictableRangeSet + + // If evicting is true, there is a goroutine currently evicting all + // evictable ranges for this user. + evicting bool +} + const ( chunkShift = 24 chunkSize = 1 << chunkShift // 16 MB @@ -180,7 +260,15 @@ const ( // NewMemoryFile creates a MemoryFile backed by the given file. If // NewMemoryFile succeeds, ownership of file is transferred to the returned // MemoryFile. -func NewMemoryFile(file *os.File) (*MemoryFile, error) { +func NewMemoryFile(file *os.File, opts MemoryFileOpts) (*MemoryFile, error) { + switch opts.DelayedEviction { + case DelayedEvictionDefault: + opts.DelayedEviction = DelayedEvictionEnabled + case DelayedEvictionDisabled, DelayedEvictionEnabled: + default: + return nil, fmt.Errorf("invalid MemoryFileOpts.DelayedEviction: %v", opts.DelayedEviction) + } + // Truncate the file to 0 bytes first to ensure that it's empty. if err := file.Truncate(0); err != nil { return nil, err @@ -189,14 +277,16 @@ func NewMemoryFile(file *os.File) (*MemoryFile, error) { return nil, err } f := &MemoryFile{ + opts: opts, fileSize: initialSize, file: file, // No pages are reclaimable. DecRef will always be able to // decrease minReclaimablePage from this point. minReclaimablePage: maxPage, + evictable: make(map[EvictableMemoryUser]*evictableMemoryUserInfo), } - f.reclaimCond.L = &f.mu f.mappings.Store(make([]uintptr, initialSize/chunkSize)) + f.reclaimCond.L = &f.mu go f.runReclaim() // S/R-SAFE: f.mu // The Linux kernel contains an optional feature called "Integrity @@ -434,113 +524,6 @@ func (f *MemoryFile) markDecommitted(fr platform.FileRange) { f.usage.MergeRange(fr) } -// runReclaim implements the reclaimer goroutine, which continuously decommits -// reclaimable pages in order to reduce memory usage and make them available -// for allocation. -func (f *MemoryFile) runReclaim() { - for { - fr, ok := f.findReclaimable() - if !ok { - break - } - - if err := f.Decommit(fr); err != nil { - log.Warningf("Reclaim failed to decommit %v: %v", fr, err) - // Zero the pages manually. This won't reduce memory usage, but at - // least ensures that the pages will be zero when reallocated. - f.forEachMappingSlice(fr, func(bs []byte) { - for i := range bs { - bs[i] = 0 - } - }) - // Pretend the pages were decommitted even though they weren't, - // since the memory accounting implementation has no idea how to - // deal with this. - f.markDecommitted(fr) - } - f.markReclaimed(fr) - } - // We only get here if findReclaimable finds f.destroyed set and returns - // false. - f.mu.Lock() - defer f.mu.Unlock() - if !f.destroyed { - panic("findReclaimable broke out of reclaim loop, but destroyed is no longer set") - } - f.file.Close() - // Ensure that any attempts to use f.file.Fd() fail instead of getting a fd - // that has possibly been reassigned. - f.file = nil - mappings := f.mappings.Load().([]uintptr) - for i, m := range mappings { - if m != 0 { - _, _, errno := syscall.Syscall(syscall.SYS_MUNMAP, m, chunkSize, 0) - if errno != 0 { - log.Warningf("Failed to unmap mapping %#x for MemoryFile chunk %d: %v", m, i, errno) - } - } - } - // Similarly, invalidate f.mappings. (atomic.Value.Store(nil) panics.) - f.mappings.Store([]uintptr{}) -} - -func (f *MemoryFile) findReclaimable() (platform.FileRange, bool) { - f.mu.Lock() - defer f.mu.Unlock() - for { - for { - if f.destroyed { - return platform.FileRange{}, false - } - if f.reclaimable { - break - } - f.reclaimCond.Wait() - } - // Allocate returns the first usable range in offset order and is - // currently a linear scan, so reclaiming from the beginning of the - // file minimizes the expected latency of Allocate. - for seg := f.usage.LowerBoundSegment(f.minReclaimablePage); seg.Ok(); seg = seg.NextSegment() { - if seg.ValuePtr().refs == 0 { - f.minReclaimablePage = seg.End() - return seg.Range(), true - } - } - // No pages are reclaimable. - f.reclaimable = false - f.minReclaimablePage = maxPage - } -} - -func (f *MemoryFile) markReclaimed(fr platform.FileRange) { - f.mu.Lock() - defer f.mu.Unlock() - seg := f.usage.FindSegment(fr.Start) - // All of fr should be mapped to a single uncommitted reclaimable segment - // accounted to System. - if !seg.Ok() { - panic(fmt.Sprintf("reclaimed pages %v include unreferenced pages:\n%v", fr, &f.usage)) - } - if !seg.Range().IsSupersetOf(fr) { - panic(fmt.Sprintf("reclaimed pages %v are not entirely contained in segment %v with state %v:\n%v", fr, seg.Range(), seg.Value(), &f.usage)) - } - if got, want := seg.Value(), (usageInfo{ - kind: usage.System, - knownCommitted: false, - refs: 0, - }); got != want { - panic(fmt.Sprintf("reclaimed pages %v in segment %v has incorrect state %v, wanted %v:\n%v", fr, seg.Range(), got, want, &f.usage)) - } - // Deallocate reclaimed pages. Even though all of seg is reclaimable, the - // caller of markReclaimed may not have decommitted it, so we can only mark - // fr as reclaimed. - f.usage.Remove(f.usage.Isolate(seg, fr)) - if fr.Start < f.minUnallocatedPage { - // We've deallocated at least one lower page. - f.minUnallocatedPage = fr.Start - } -} - // IncRef implements platform.File.IncRef. func (f *MemoryFile) IncRef(fr platform.FileRange) { if !fr.WellFormed() || fr.Length() == 0 || fr.Start%usermem.PageSize != 0 || fr.End%usermem.PageSize != 0 { @@ -677,9 +660,82 @@ func (f *MemoryFile) getChunkMapping(chunk int) ([]uintptr, uintptr, error) { return mappings, m, nil } -// FD implements platform.File.FD. -func (f *MemoryFile) FD() int { - return int(f.file.Fd()) +// MarkEvictable allows f to request memory deallocation by calling +// user.Evict(er) in the future. +// +// Redundantly marking an already-evictable range as evictable has no effect. +func (f *MemoryFile) MarkEvictable(user EvictableMemoryUser, er EvictableRange) { + f.mu.Lock() + defer f.mu.Unlock() + info, ok := f.evictable[user] + if !ok { + info = &evictableMemoryUserInfo{} + f.evictable[user] = info + } + gap := info.ranges.LowerBoundGap(er.Start) + for gap.Ok() && gap.Start() < er.End { + gapER := gap.Range().Intersect(er) + if gapER.Length() == 0 { + gap = gap.NextGap() + continue + } + gap = info.ranges.Insert(gap, gapER, evictableRangeSetValue{}).NextGap() + } + if !info.evicting { + switch f.opts.DelayedEviction { + case DelayedEvictionDisabled: + // Kick off eviction immediately. + f.startEvictionGoroutineLocked(user, info) + case DelayedEvictionEnabled: + // Ensure that the reclaimer goroutine is running, so that it can + // start eviction when necessary. + f.reclaimCond.Signal() + } + } +} + +// MarkUnevictable informs f that user no longer considers er to be evictable, +// so the MemoryFile should no longer call user.Evict(er). Note that, per +// EvictableMemoryUser.Evict's documentation, user.Evict(er) may still be +// called even after MarkUnevictable returns due to race conditions, and +// implementations of EvictableMemoryUser must handle this possibility. +// +// Redundantly marking an already-unevictable range as unevictable has no +// effect. +func (f *MemoryFile) MarkUnevictable(user EvictableMemoryUser, er EvictableRange) { + f.mu.Lock() + defer f.mu.Unlock() + info, ok := f.evictable[user] + if !ok { + return + } + seg := info.ranges.LowerBoundSegment(er.Start) + for seg.Ok() && seg.Start() < er.End { + seg = info.ranges.Isolate(seg, er) + seg = info.ranges.Remove(seg).NextSegment() + } + // We can only remove info if there's no eviction goroutine running on its + // behalf. + if !info.evicting && info.ranges.IsEmpty() { + delete(f.evictable, user) + } +} + +// MarkAllUnevictable informs f that user no longer considers any offsets to be +// evictable. It otherwise has the same semantics as MarkUnevictable. +func (f *MemoryFile) MarkAllUnevictable(user EvictableMemoryUser) { + f.mu.Lock() + defer f.mu.Unlock() + info, ok := f.evictable[user] + if !ok { + return + } + info.ranges.RemoveAll() + // We can only remove info if there's no eviction goroutine running on its + // behalf. + if !info.evicting { + delete(f.evictable, user) + } } // UpdateUsage ensures that the memory usage statistics in @@ -889,6 +945,11 @@ func (f *MemoryFile) File() *os.File { return f.file } +// FD implements platform.File.FD. +func (f *MemoryFile) FD() int { + return int(f.file.Fd()) +} + // String implements fmt.Stringer.String. // // Note that because f.String locks f.mu, calling f.String internally @@ -900,6 +961,167 @@ func (f *MemoryFile) String() string { return f.usage.String() } +// runReclaim implements the reclaimer goroutine, which continuously decommits +// reclaimable pages in order to reduce memory usage and make them available +// for allocation. +func (f *MemoryFile) runReclaim() { + for { + fr, ok := f.findReclaimable() + if !ok { + break + } + + if err := f.Decommit(fr); err != nil { + log.Warningf("Reclaim failed to decommit %v: %v", fr, err) + // Zero the pages manually. This won't reduce memory usage, but at + // least ensures that the pages will be zero when reallocated. + f.forEachMappingSlice(fr, func(bs []byte) { + for i := range bs { + bs[i] = 0 + } + }) + // Pretend the pages were decommitted even though they weren't, + // since the memory accounting implementation has no idea how to + // deal with this. + f.markDecommitted(fr) + } + f.markReclaimed(fr) + } + // We only get here if findReclaimable finds f.destroyed set and returns + // false. + f.mu.Lock() + defer f.mu.Unlock() + if !f.destroyed { + panic("findReclaimable broke out of reclaim loop, but destroyed is no longer set") + } + f.file.Close() + // Ensure that any attempts to use f.file.Fd() fail instead of getting a fd + // that has possibly been reassigned. + f.file = nil + f.mappingsMu.Lock() + defer f.mappingsMu.Unlock() + mappings := f.mappings.Load().([]uintptr) + for i, m := range mappings { + if m != 0 { + _, _, errno := syscall.Syscall(syscall.SYS_MUNMAP, m, chunkSize, 0) + if errno != 0 { + log.Warningf("Failed to unmap mapping %#x for MemoryFile chunk %d: %v", m, i, errno) + } + } + } + // Similarly, invalidate f.mappings. (atomic.Value.Store(nil) panics.) + f.mappings.Store([]uintptr{}) +} + +func (f *MemoryFile) findReclaimable() (platform.FileRange, bool) { + f.mu.Lock() + defer f.mu.Unlock() + for { + for { + if f.destroyed { + return platform.FileRange{}, false + } + if f.reclaimable { + break + } + if f.opts.DelayedEviction == DelayedEvictionEnabled { + // No work to do. Evict any pending evictable allocations to + // get more reclaimable pages before going to sleep. + f.startEvictionsLocked() + } + f.reclaimCond.Wait() + } + // Allocate returns the first usable range in offset order and is + // currently a linear scan, so reclaiming from the beginning of the + // file minimizes the expected latency of Allocate. + for seg := f.usage.LowerBoundSegment(f.minReclaimablePage); seg.Ok(); seg = seg.NextSegment() { + if seg.ValuePtr().refs == 0 { + f.minReclaimablePage = seg.End() + return seg.Range(), true + } + } + // No pages are reclaimable. + f.reclaimable = false + f.minReclaimablePage = maxPage + } +} + +func (f *MemoryFile) markReclaimed(fr platform.FileRange) { + f.mu.Lock() + defer f.mu.Unlock() + seg := f.usage.FindSegment(fr.Start) + // All of fr should be mapped to a single uncommitted reclaimable segment + // accounted to System. + if !seg.Ok() { + panic(fmt.Sprintf("reclaimed pages %v include unreferenced pages:\n%v", fr, &f.usage)) + } + if !seg.Range().IsSupersetOf(fr) { + panic(fmt.Sprintf("reclaimed pages %v are not entirely contained in segment %v with state %v:\n%v", fr, seg.Range(), seg.Value(), &f.usage)) + } + if got, want := seg.Value(), (usageInfo{ + kind: usage.System, + knownCommitted: false, + refs: 0, + }); got != want { + panic(fmt.Sprintf("reclaimed pages %v in segment %v has incorrect state %v, wanted %v:\n%v", fr, seg.Range(), got, want, &f.usage)) + } + // Deallocate reclaimed pages. Even though all of seg is reclaimable, the + // caller of markReclaimed may not have decommitted it, so we can only mark + // fr as reclaimed. + f.usage.Remove(f.usage.Isolate(seg, fr)) + if fr.Start < f.minUnallocatedPage { + // We've deallocated at least one lower page. + f.minUnallocatedPage = fr.Start + } +} + +// Preconditions: f.mu must be locked. +func (f *MemoryFile) startEvictionsLocked() { + for user, info := range f.evictable { + // Don't start multiple goroutines to evict the same user's + // allocations. + if !info.evicting { + f.startEvictionGoroutineLocked(user, info) + } + } +} + +// Preconditions: info == f.evictable[user]. !info.evicting. f.mu must be +// locked. +func (f *MemoryFile) startEvictionGoroutineLocked(user EvictableMemoryUser, info *evictableMemoryUserInfo) { + info.evicting = true + f.evictionWG.Add(1) + go func() { // S/R-SAFE: f.evictionWG + defer f.evictionWG.Done() + for { + f.mu.Lock() + info, ok := f.evictable[user] + if !ok { + // This shouldn't happen: only this goroutine is permitted + // to delete this entry. + f.mu.Unlock() + panic(fmt.Sprintf("evictableMemoryUserInfo for EvictableMemoryUser %v deleted while eviction goroutine running", user)) + } + if info.ranges.IsEmpty() { + delete(f.evictable, user) + f.mu.Unlock() + return + } + // Evict from the end of info.ranges, under the assumption that + // if ranges in user start being used again (and are + // consequently marked unevictable), such uses are more likely + // to start from the beginning of user. + seg := info.ranges.LastSegment() + er := seg.Range() + info.ranges.Remove(seg) + // user.Evict() must be called without holding f.mu to avoid + // circular lock ordering. + f.mu.Unlock() + user.Evict(context.Background(), er) + } + }() +} + type usageSetFunctions struct{} func (usageSetFunctions) MinKey() uint64 { @@ -920,3 +1142,27 @@ func (usageSetFunctions) Merge(_ platform.FileRange, val1 usageInfo, _ platform. func (usageSetFunctions) Split(_ platform.FileRange, val usageInfo, _ uint64) (usageInfo, usageInfo) { return val, val } + +// evictableRangeSetValue is the value type of evictableRangeSet. +type evictableRangeSetValue struct{} + +type evictableRangeSetFunctions struct{} + +func (evictableRangeSetFunctions) MinKey() uint64 { + return 0 +} + +func (evictableRangeSetFunctions) MaxKey() uint64 { + return math.MaxUint64 +} + +func (evictableRangeSetFunctions) ClearValue(val *evictableRangeSetValue) { +} + +func (evictableRangeSetFunctions) Merge(_ EvictableRange, _ evictableRangeSetValue, _ EvictableRange, _ evictableRangeSetValue) (evictableRangeSetValue, bool) { + return evictableRangeSetValue{}, true +} + +func (evictableRangeSetFunctions) Split(_ EvictableRange, _ evictableRangeSetValue, _ uint64) (evictableRangeSetValue, evictableRangeSetValue) { + return evictableRangeSetValue{}, evictableRangeSetValue{} +} diff --git a/pkg/sentry/pgalloc/save_restore.go b/pkg/sentry/pgalloc/save_restore.go index cf169af55..9534d1aed 100644 --- a/pkg/sentry/pgalloc/save_restore.go +++ b/pkg/sentry/pgalloc/save_restore.go @@ -28,6 +28,15 @@ import ( "gvisor.googlesource.com/gvisor/pkg/state" ) +// FlushEvictions blocks until f has finished evicting all evictable +// allocations. +func (f *MemoryFile) FlushEvictions() { + f.mu.Lock() + f.startEvictionsLocked() + f.mu.Unlock() + f.evictionWG.Wait() +} + // SaveTo writes f's state to the given stream. func (f *MemoryFile) SaveTo(w io.Writer) error { // Wait for reclaim. @@ -40,6 +49,11 @@ func (f *MemoryFile) SaveTo(w io.Writer) error { f.mu.Lock() } + // Ensure that there are no pending evictions. + if len(f.evictable) != 0 { + panic(fmt.Sprintf("evictions still pending for %d users; call FlushEvictions before SaveTo", len(f.evictable))) + } + // Ensure that all pages that contain data have knownCommitted set, since // we only store knownCommitted pages below. zeroPage := make([]byte, usermem.PageSize) diff --git a/runsc/boot/loader.go b/runsc/boot/loader.go index 0b5be0a42..05122a6a8 100644 --- a/runsc/boot/loader.go +++ b/runsc/boot/loader.go @@ -424,7 +424,7 @@ func createMemoryFile() (*pgalloc.MemoryFile, error) { return nil, fmt.Errorf("error creating memfd: %v", err) } memfile := os.NewFile(uintptr(memfd), memfileName) - mf, err := pgalloc.NewMemoryFile(memfile) + mf, err := pgalloc.NewMemoryFile(memfile, pgalloc.MemoryFileOpts{}) if err != nil { memfile.Close() return nil, fmt.Errorf("error creating pgalloc.MemoryFile: %v", err) -- cgit v1.2.3 From 1bee43be13549b01e18d87df194ac219845de5cf Mon Sep 17 00:00:00 2001 From: Fabricio Voznika Date: Thu, 9 May 2019 15:34:44 -0700 Subject: Implement fallocate(2) Closes #225 PiperOrigin-RevId: 247508791 Change-Id: I04f47cf2770b30043e5a272aba4ba6e11d0476cc --- pkg/abi/linux/file.go | 11 +++ pkg/p9/BUILD | 1 + pkg/p9/client_file.go | 12 +++ pkg/p9/file.go | 4 + pkg/p9/handlers.go | 34 +++++++++ pkg/p9/local_server/local_server.go | 5 ++ pkg/p9/messages.go | 59 ++++++++++++++ pkg/p9/p9.go | 81 ++++++++++++++++++++ pkg/p9/version.go | 7 +- pkg/sentry/fs/ashmem/device.go | 1 + pkg/sentry/fs/binder/binder.go | 1 + pkg/sentry/fs/dev/full.go | 3 +- pkg/sentry/fs/dev/null.go | 9 ++- pkg/sentry/fs/dev/random.go | 9 ++- pkg/sentry/fs/fsutil/BUILD | 1 + pkg/sentry/fs/fsutil/host_mappable.go | 10 ++- pkg/sentry/fs/fsutil/inode.go | 25 ++++++ pkg/sentry/fs/fsutil/inode_cached.go | 28 +++++++ pkg/sentry/fs/fsutil/inode_cached_test.go | 9 +++ pkg/sentry/fs/gofer/context_file.go | 7 ++ pkg/sentry/fs/gofer/inode.go | 24 ++++++ pkg/sentry/fs/host/inode.go | 18 +++++ pkg/sentry/fs/inode.go | 7 ++ pkg/sentry/fs/inode_operations.go | 4 + pkg/sentry/fs/inode_overlay.go | 7 ++ pkg/sentry/fs/inode_overlay_test.go | 1 + pkg/sentry/fs/mock.go | 5 ++ pkg/sentry/fs/proc/inode.go | 1 + pkg/sentry/fs/proc/seqfile/seqfile.go | 3 +- pkg/sentry/fs/proc/uid_gid_map.go | 3 +- pkg/sentry/fs/ramfs/dir.go | 1 + pkg/sentry/fs/ramfs/socket.go | 3 +- pkg/sentry/fs/ramfs/symlink.go | 5 +- pkg/sentry/fs/sys/devices.go | 3 +- pkg/sentry/fs/tmpfs/inode_file.go | 27 +++++++ pkg/sentry/fs/tmpfs/tmpfs.go | 8 +- pkg/sentry/fs/tty/dir.go | 5 +- pkg/sentry/kernel/pipe/node.go | 4 + pkg/sentry/syscalls/linux/sys_file.go | 38 ++++++++- runsc/boot/compat.go | 2 +- runsc/fsgofer/filter/config.go | 10 ++- runsc/fsgofer/fsgofer.go | 12 +++ test/syscalls/linux/BUILD | 2 + test/syscalls/linux/fallocate.cc | 123 +++++++++++++++++++++++++----- 44 files changed, 589 insertions(+), 44 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/file.go b/pkg/abi/linux/file.go index 753fec3ed..81ff9fe9e 100644 --- a/pkg/abi/linux/file.go +++ b/pkg/abi/linux/file.go @@ -254,3 +254,14 @@ const ( F_SEAL_GROW = 0x0004 // Prevent file from growing. F_SEAL_WRITE = 0x0008 // Prevent writes. ) + +// Constants related to fallocate(2). Source: include/uapi/linux/falloc.h +const ( + FALLOC_FL_KEEP_SIZE = 0x01 + FALLOC_FL_PUNCH_HOLE = 0x02 + FALLOC_FL_NO_HIDE_STALE = 0x04 + FALLOC_FL_COLLAPSE_RANGE = 0x08 + FALLOC_FL_ZERO_RANGE = 0x10 + FALLOC_FL_INSERT_RANGE = 0x20 + FALLOC_FL_UNSHARE_RANGE = 0x40 +) diff --git a/pkg/p9/BUILD b/pkg/p9/BUILD index 5d972309d..36b2ec5f6 100644 --- a/pkg/p9/BUILD +++ b/pkg/p9/BUILD @@ -26,6 +26,7 @@ go_library( "//pkg/fd", "//pkg/log", "//pkg/unet", + "@org_golang_x_sys//unix:go_default_library", ], ) diff --git a/pkg/p9/client_file.go b/pkg/p9/client_file.go index 63c65129a..471c3a80b 100644 --- a/pkg/p9/client_file.go +++ b/pkg/p9/client_file.go @@ -171,6 +171,18 @@ func (c *clientFile) SetAttr(valid SetAttrMask, attr SetAttr) error { return c.client.sendRecv(&Tsetattr{FID: c.fid, Valid: valid, SetAttr: attr}, &Rsetattr{}) } +// Allocate implements File.Allocate. +func (c *clientFile) Allocate(mode AllocateMode, offset, length uint64) error { + if atomic.LoadUint32(&c.closed) != 0 { + return syscall.EBADF + } + if !versionSupportsTallocate(c.client.version) { + return syscall.EOPNOTSUPP + } + + return c.client.sendRecv(&Tallocate{FID: c.fid, Mode: mode, Offset: offset, Length: length}, &Rallocate{}) +} + // Remove implements File.Remove. // // N.B. This method is no longer part of the file interface and should be diff --git a/pkg/p9/file.go b/pkg/p9/file.go index a52a0f3e7..89e814d50 100644 --- a/pkg/p9/file.go +++ b/pkg/p9/file.go @@ -89,6 +89,10 @@ type File interface { // On the server, SetAttr has a write concurrency guarantee. SetAttr(valid SetAttrMask, attr SetAttr) error + // Allocate allows the caller to directly manipulate the allocated disk space + // for the file. See fallocate(2) for more details. + Allocate(mode AllocateMode, offset, length uint64) error + // Close is called when all references are dropped on the server side, // and Close should be called by the client to drop all references. // diff --git a/pkg/p9/handlers.go b/pkg/p9/handlers.go index 6da2ce4e3..533ead98a 100644 --- a/pkg/p9/handlers.go +++ b/pkg/p9/handlers.go @@ -877,6 +877,40 @@ func (t *Tsetattr) handle(cs *connState) message { return &Rsetattr{} } +// handle implements handler.handle. +func (t *Tallocate) handle(cs *connState) message { + // Lookup the FID. + ref, ok := cs.LookupFID(t.FID) + if !ok { + return newErr(syscall.EBADF) + } + defer ref.DecRef() + + if err := ref.safelyWrite(func() error { + // Has it been opened already? + openFlags, opened := ref.OpenFlags() + if !opened { + return syscall.EINVAL + } + + // Can it be written? Check permissions. + if openFlags&OpenFlagsModeMask == ReadOnly { + return syscall.EBADF + } + + // We don't allow allocate on files that have been deleted. + if ref.isDeleted() { + return syscall.EINVAL + } + + return ref.file.Allocate(t.Mode, t.Offset, t.Length) + }); err != nil { + return newErr(err) + } + + return &Rallocate{} +} + // handle implements handler.handle. func (t *Txattrwalk) handle(cs *connState) message { // Lookup the FID. diff --git a/pkg/p9/local_server/local_server.go b/pkg/p9/local_server/local_server.go index f4077a9d4..d49d94550 100644 --- a/pkg/p9/local_server/local_server.go +++ b/pkg/p9/local_server/local_server.go @@ -323,6 +323,11 @@ func (l *local) Renamed(parent p9.File, newName string) { l.path = path.Join(parent.(*local).path, newName) } +// Allocate implements p9.File.Allocate. +func (l *local) Allocate(mode p9.AllocateMode, offset, length uint64) error { + return syscall.Fallocate(int(l.file.Fd()), mode.ToLinux(), int64(offset), int64(length)) +} + func main() { log.SetLevel(log.Debug) diff --git a/pkg/p9/messages.go b/pkg/p9/messages.go index 3c7898cc1..703753c31 100644 --- a/pkg/p9/messages.go +++ b/pkg/p9/messages.go @@ -1424,6 +1424,63 @@ func (r *Rsetattr) String() string { return fmt.Sprintf("Rsetattr{}") } +// Tallocate is an allocate request. This is an extension to 9P protocol, not +// present in the 9P2000.L standard. +type Tallocate struct { + FID FID + Mode AllocateMode + Offset uint64 + Length uint64 +} + +// Decode implements encoder.Decode. +func (t *Tallocate) Decode(b *buffer) { + t.FID = b.ReadFID() + t.Mode.Decode(b) + t.Offset = b.Read64() + t.Length = b.Read64() +} + +// Encode implements encoder.Encode. +func (t *Tallocate) Encode(b *buffer) { + b.WriteFID(t.FID) + t.Mode.Encode(b) + b.Write64(t.Offset) + b.Write64(t.Length) +} + +// Type implements message.Type. +func (*Tallocate) Type() MsgType { + return MsgTallocate +} + +// String implements fmt.Stringer. +func (t *Tallocate) String() string { + return fmt.Sprintf("Tallocate{FID: %d, Offset: %d, Length: %d}", t.FID, t.Offset, t.Length) +} + +// Rallocate is an allocate response. +type Rallocate struct { +} + +// Decode implements encoder.Decode. +func (*Rallocate) Decode(b *buffer) { +} + +// Encode implements encoder.Encode. +func (*Rallocate) Encode(b *buffer) { +} + +// Type implements message.Type. +func (*Rallocate) Type() MsgType { + return MsgRallocate +} + +// String implements fmt.Stringer. +func (r *Rallocate) String() string { + return fmt.Sprintf("Rallocate{}") +} + // Txattrwalk walks extended attributes. type Txattrwalk struct { // FID is the FID to check for attributes. @@ -2297,4 +2354,6 @@ func init() { msgRegistry.register(MsgRusymlink, func() message { return &Rusymlink{} }) msgRegistry.register(MsgTlconnect, func() message { return &Tlconnect{} }) msgRegistry.register(MsgRlconnect, func() message { return &Rlconnect{} }) + msgRegistry.register(MsgTallocate, func() message { return &Tallocate{} }) + msgRegistry.register(MsgRallocate, func() message { return &Rallocate{} }) } diff --git a/pkg/p9/p9.go b/pkg/p9/p9.go index 78c7d3f86..4039862e6 100644 --- a/pkg/p9/p9.go +++ b/pkg/p9/p9.go @@ -22,6 +22,8 @@ import ( "strings" "sync/atomic" "syscall" + + "golang.org/x/sys/unix" ) // OpenFlags is the mode passed to Open and Create operations. @@ -374,6 +376,8 @@ const ( MsgRusymlink = 135 MsgTlconnect = 136 MsgRlconnect = 137 + MsgTallocate = 138 + MsgRallocate = 139 ) // QIDType represents the file type for QIDs. @@ -1058,3 +1062,80 @@ func (d *Dirent) Encode(b *buffer) { b.WriteQIDType(d.Type) b.WriteString(d.Name) } + +// AllocateMode are possible modes to p9.File.Allocate(). +type AllocateMode struct { + KeepSize bool + PunchHole bool + NoHideStale bool + CollapseRange bool + ZeroRange bool + InsertRange bool + Unshare bool +} + +// ToLinux converts to a value compatible with fallocate(2)'s mode. +func (a *AllocateMode) ToLinux() uint32 { + rv := uint32(0) + if a.KeepSize { + rv |= unix.FALLOC_FL_KEEP_SIZE + } + if a.PunchHole { + rv |= unix.FALLOC_FL_PUNCH_HOLE + } + if a.NoHideStale { + rv |= unix.FALLOC_FL_NO_HIDE_STALE + } + if a.CollapseRange { + rv |= unix.FALLOC_FL_COLLAPSE_RANGE + } + if a.ZeroRange { + rv |= unix.FALLOC_FL_ZERO_RANGE + } + if a.InsertRange { + rv |= unix.FALLOC_FL_INSERT_RANGE + } + if a.Unshare { + rv |= unix.FALLOC_FL_UNSHARE_RANGE + } + return rv +} + +// Decode implements encoder.Decode. +func (a *AllocateMode) Decode(b *buffer) { + mask := b.Read32() + a.KeepSize = mask&0x01 != 0 + a.PunchHole = mask&0x02 != 0 + a.NoHideStale = mask&0x04 != 0 + a.CollapseRange = mask&0x08 != 0 + a.ZeroRange = mask&0x10 != 0 + a.InsertRange = mask&0x20 != 0 + a.Unshare = mask&0x40 != 0 +} + +// Encode implements encoder.Encode. +func (a *AllocateMode) Encode(b *buffer) { + mask := uint32(0) + if a.KeepSize { + mask |= 0x01 + } + if a.PunchHole { + mask |= 0x02 + } + if a.NoHideStale { + mask |= 0x04 + } + if a.CollapseRange { + mask |= 0x08 + } + if a.ZeroRange { + mask |= 0x10 + } + if a.InsertRange { + mask |= 0x20 + } + if a.Unshare { + mask |= 0x40 + } + b.Write32(mask) +} diff --git a/pkg/p9/version.go b/pkg/p9/version.go index a36a499a1..c2a2885ae 100644 --- a/pkg/p9/version.go +++ b/pkg/p9/version.go @@ -26,7 +26,7 @@ const ( // // Clients are expected to start requesting this version number and // to continuously decrement it until a Tversion request succeeds. - highestSupportedVersion uint32 = 6 + highestSupportedVersion uint32 = 7 // lowestSupportedVersion is the lowest supported version X in a // version string of the format 9P2000.L.Google.X. @@ -143,3 +143,8 @@ func VersionSupportsAnonymous(v uint32) bool { func VersionSupportsMultiUser(v uint32) bool { return v >= 6 } + +// versionSupportsTallocate returns true if version v supports Allocate(). +func versionSupportsTallocate(v uint32) bool { + return v >= 7 +} diff --git a/pkg/sentry/fs/ashmem/device.go b/pkg/sentry/fs/ashmem/device.go index 5e005bc2e..22e1530e9 100644 --- a/pkg/sentry/fs/ashmem/device.go +++ b/pkg/sentry/fs/ashmem/device.go @@ -29,6 +29,7 @@ import ( type Device struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopAllocate `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopTruncate `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index acbbd5466..a992253e6 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -46,6 +46,7 @@ const ( type Device struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopAllocate `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopTruncate `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index 6b11afa44..17d68b5c4 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -30,6 +30,7 @@ import ( type fullDevice struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopAllocate `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopTruncate `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` @@ -59,7 +60,6 @@ func (f *fullDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.Fi // +stateify savable type fullFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` @@ -69,6 +69,7 @@ type fullFileOperations struct { fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` readZeros `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } var _ fs.FileOperations = (*fullFileOperations)(nil) diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index 069212b6d..ee13183c8 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -29,6 +29,7 @@ import ( type nullDevice struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopAllocate `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopTruncate `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` @@ -60,17 +61,17 @@ func (n *nullDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.Fi // +stateify savable type nullFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRead `state:"nosave"` - fsutil.FileNoopWrite `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` + fsutil.FileNoopWrite `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } var _ fs.FileOperations = (*nullFileOperations)(nil) @@ -101,16 +102,16 @@ func (zd *zeroDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.F // +stateify savable type zeroFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNoopWrite `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` readZeros `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } var _ fs.FileOperations = (*zeroFileOperations)(nil) diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index de0f3e5e5..b0a412382 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -29,6 +29,7 @@ import ( type randomDevice struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopAllocate `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopTruncate `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` @@ -57,16 +58,16 @@ func (*randomDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.Fi // +stateify savable type randomFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` - fsutil.FileNotDirReaddir `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` - fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNoopWrite `state:"nosave"` + fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } var _ fs.FileOperations = (*randomFileOperations)(nil) diff --git a/pkg/sentry/fs/fsutil/BUILD b/pkg/sentry/fs/fsutil/BUILD index 01098675d..44f43b965 100644 --- a/pkg/sentry/fs/fsutil/BUILD +++ b/pkg/sentry/fs/fsutil/BUILD @@ -113,5 +113,6 @@ go_test( "//pkg/sentry/memmap", "//pkg/sentry/safemem", "//pkg/sentry/usermem", + "//pkg/syserror", ], ) diff --git a/pkg/sentry/fs/fsutil/host_mappable.go b/pkg/sentry/fs/fsutil/host_mappable.go index 28686f3b3..ad0518b8f 100644 --- a/pkg/sentry/fs/fsutil/host_mappable.go +++ b/pkg/sentry/fs/fsutil/host_mappable.go @@ -149,7 +149,7 @@ func (h *HostMappable) Truncate(ctx context.Context, newSize int64) error { } // Invalidate COW mappings that may exist beyond the new size in case the file - // is being shrunk. Other mappinsg don't need to be invalidated because + // is being shrunk. Other mappings don't need to be invalidated because // translate will just return identical mappings after invalidation anyway, // and SIGBUS will be raised and handled when the mappings are touched. // @@ -167,6 +167,14 @@ func (h *HostMappable) Truncate(ctx context.Context, newSize int64) error { return nil } +// Allocate reserves space in the backing file. +func (h *HostMappable) Allocate(ctx context.Context, offset int64, length int64) error { + h.truncateMu.RLock() + err := h.backingFile.Allocate(ctx, offset, length) + h.truncateMu.RUnlock() + return err +} + // Write writes to the file backing this mappable. func (h *HostMappable) Write(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { h.truncateMu.RLock() diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index b6366d906..151be1d0d 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -34,6 +34,7 @@ type SimpleFileInode struct { InodeNoExtendedAttributes `state:"nosave"` InodeNoopRelease `state:"nosave"` InodeNoopWriteOut `state:"nosave"` + InodeNotAllocatable `state:"nosave"` InodeNotDirectory `state:"nosave"` InodeNotMappable `state:"nosave"` InodeNotOpenable `state:"nosave"` @@ -61,6 +62,7 @@ type NoReadWriteFileInode struct { InodeNoExtendedAttributes `state:"nosave"` InodeNoopRelease `state:"nosave"` InodeNoopWriteOut `state:"nosave"` + InodeNotAllocatable `state:"nosave"` InodeNotDirectory `state:"nosave"` InodeNotMappable `state:"nosave"` InodeNotSocket `state:"nosave"` @@ -465,3 +467,26 @@ func (InodeDenyWriteChecker) Check(ctx context.Context, inode *fs.Inode, p fs.Pe } return fs.ContextCanAccessFile(ctx, inode, p) } + +//InodeNotAllocatable can be used by Inodes that do not support Allocate(). +type InodeNotAllocatable struct{} + +func (InodeNotAllocatable) Allocate(_ context.Context, _ *fs.Inode, _, _ int64) error { + return syserror.EOPNOTSUPP +} + +// InodeNoopAllocate implements fs.InodeOperations.Allocate as a noop. +type InodeNoopAllocate struct{} + +// Allocate implements fs.InodeOperations.Allocate. +func (InodeNoopAllocate) Allocate(_ context.Context, _ *fs.Inode, _, _ int64) error { + return nil +} + +// InodeIsDirAllocate implements fs.InodeOperations.Allocate for directories. +type InodeIsDirAllocate struct{} + +// Allocate implements fs.InodeOperations.Allocate. +func (InodeIsDirAllocate) Allocate(_ context.Context, _ *fs.Inode, _, _ int64) error { + return syserror.EISDIR +} diff --git a/pkg/sentry/fs/fsutil/inode_cached.go b/pkg/sentry/fs/fsutil/inode_cached.go index 76644e69d..03cad37f3 100644 --- a/pkg/sentry/fs/fsutil/inode_cached.go +++ b/pkg/sentry/fs/fsutil/inode_cached.go @@ -135,6 +135,10 @@ type CachedFileObject interface { // the file was opened. SetMaskedAttributes(ctx context.Context, mask fs.AttrMask, attr fs.UnstableAttr) error + // Allocate allows the caller to reserve disk space for the inode. + // It's equivalent to fallocate(2) with 'mode=0'. + Allocate(ctx context.Context, offset int64, length int64) error + // Sync instructs the remote filesystem to sync the file to stable storage. Sync(ctx context.Context) error @@ -336,6 +340,30 @@ func (c *CachingInodeOperations) Truncate(ctx context.Context, inode *fs.Inode, return nil } +// Allocate implements fs.InodeOperations.Allocate. +func (c *CachingInodeOperations) Allocate(ctx context.Context, offset, length int64) error { + newSize := offset + length + + // c.attr.Size is protected by both c.attrMu and c.dataMu. + c.attrMu.Lock() + defer c.attrMu.Unlock() + c.dataMu.Lock() + defer c.dataMu.Unlock() + + if newSize <= c.attr.Size { + return nil + } + + now := ktime.NowFromContext(ctx) + if err := c.backingFile.Allocate(ctx, offset, length); err != nil { + return err + } + + c.attr.Size = newSize + c.touchModificationTimeLocked(now) + return nil +} + // WriteOut implements fs.InodeOperations.WriteOut. func (c *CachingInodeOperations) WriteOut(ctx context.Context, inode *fs.Inode) error { c.attrMu.Lock() diff --git a/pkg/sentry/fs/fsutil/inode_cached_test.go b/pkg/sentry/fs/fsutil/inode_cached_test.go index 3f10efc12..be3d4b6fc 100644 --- a/pkg/sentry/fs/fsutil/inode_cached_test.go +++ b/pkg/sentry/fs/fsutil/inode_cached_test.go @@ -26,6 +26,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" + "gvisor.googlesource.com/gvisor/pkg/syserror" ) type noopBackingFile struct{} @@ -50,6 +51,10 @@ func (noopBackingFile) FD() int { return -1 } +func (noopBackingFile) Allocate(ctx context.Context, offset int64, length int64) error { + return nil +} + func TestSetPermissions(t *testing.T) { ctx := contexttest.Context(t) @@ -237,6 +242,10 @@ func (*sliceBackingFile) FD() int { return -1 } +func (f *sliceBackingFile) Allocate(ctx context.Context, offset int64, length int64) error { + return syserror.EOPNOTSUPP +} + type noopMappingSpace struct{} // Invalidate implements memmap.MappingSpace.Invalidate. diff --git a/pkg/sentry/fs/gofer/context_file.go b/pkg/sentry/fs/gofer/context_file.go index 842a34af8..be53ac4d9 100644 --- a/pkg/sentry/fs/gofer/context_file.go +++ b/pkg/sentry/fs/gofer/context_file.go @@ -59,6 +59,13 @@ func (c *contextFile) setAttr(ctx context.Context, valid p9.SetAttrMask, attr p9 return err } +func (c *contextFile) allocate(ctx context.Context, mode p9.AllocateMode, offset, length uint64) error { + ctx.UninterruptibleSleepStart(false) + err := c.file.Allocate(mode, offset, length) + ctx.UninterruptibleSleepFinish(false) + return err +} + func (c *contextFile) rename(ctx context.Context, directory contextFile, name string) error { ctx.UninterruptibleSleepStart(false) err := c.file.Rename(directory.file, name) diff --git a/pkg/sentry/fs/gofer/inode.go b/pkg/sentry/fs/gofer/inode.go index f6f20844d..dcb3b2880 100644 --- a/pkg/sentry/fs/gofer/inode.go +++ b/pkg/sentry/fs/gofer/inode.go @@ -322,6 +322,15 @@ func (i *inodeFileState) unstableAttr(ctx context.Context) (fs.UnstableAttr, err return unstable(ctx, valid, pattr, i.s.mounter, i.s.client), nil } +func (i *inodeFileState) Allocate(ctx context.Context, offset, length int64) error { + i.handlesMu.RLock() + defer i.handlesMu.RUnlock() + + // No options are supported for now. + mode := p9.AllocateMode{} + return i.writeHandles.File.allocate(ctx, mode, uint64(offset), uint64(length)) +} + // session extracts the gofer's session from the MountSource. func (i *inodeOperations) session() *session { return i.fileState.s @@ -498,6 +507,21 @@ func (i *inodeOperations) Truncate(ctx context.Context, inode *fs.Inode, length return i.fileState.file.setAttr(ctx, p9.SetAttrMask{Size: true}, p9.SetAttr{Size: uint64(length)}) } +// Allocate implements fs.InodeOperations.Allocate. +func (i *inodeOperations) Allocate(ctx context.Context, inode *fs.Inode, offset, length int64) error { + // This can only be called for files anyway. + if i.session().cachePolicy.useCachingInodeOps(inode) { + return i.cachingInodeOps.Allocate(ctx, offset, length) + } + if i.session().cachePolicy == cacheRemoteRevalidating { + return i.fileState.hostMappable.Allocate(ctx, offset, length) + } + + // No options are supported for now. + mode := p9.AllocateMode{} + return i.fileState.file.allocate(ctx, mode, uint64(offset), uint64(length)) +} + // WriteOut implements fs.InodeOperations.WriteOut. func (i *inodeOperations) WriteOut(ctx context.Context, inode *fs.Inode) error { if !i.session().cachePolicy.cacheUAttrs(inode) { diff --git a/pkg/sentry/fs/host/inode.go b/pkg/sentry/fs/host/inode.go index 20e077f77..d36ac9a87 100644 --- a/pkg/sentry/fs/host/inode.go +++ b/pkg/sentry/fs/host/inode.go @@ -163,6 +163,11 @@ func (i *inodeFileState) unstableAttr(ctx context.Context) (fs.UnstableAttr, err return unstableAttr(i.mops, &s), nil } +// SetMaskedAttributes implements fsutil.CachedFileObject.SetMaskedAttributes. +func (i *inodeFileState) Allocate(_ context.Context, offset, length int64) error { + return syscall.Fallocate(i.FD(), 0, offset, length) +} + // inodeOperations implements fs.InodeOperations. var _ fs.InodeOperations = (*inodeOperations)(nil) @@ -397,6 +402,19 @@ func (i *inodeOperations) Truncate(ctx context.Context, inode *fs.Inode, size in return i.cachingInodeOps.Truncate(ctx, inode, size) } +// Allocate implements fs.InodeOperations.Allocate. +func (i *inodeOperations) Allocate(ctx context.Context, inode *fs.Inode, offset, length int64) error { + // Is the file not memory-mappable? + if !canMap(inode) { + // Then just send the call to the FD, the host will synchronize the metadata + // update with any host inode and page cache. + return i.fileState.Allocate(ctx, offset, length) + } + // Otherwise we need to go through cachingInodeOps, even if the host page + // cache is in use, to invalidate private copies of truncated pages. + return i.cachingInodeOps.Allocate(ctx, offset, length) +} + // WriteOut implements fs.InodeOperations.WriteOut. func (i *inodeOperations) WriteOut(ctx context.Context, inode *fs.Inode) error { // Have we been using host kernel metadata caches? diff --git a/pkg/sentry/fs/inode.go b/pkg/sentry/fs/inode.go index d764ef93d..22f316daf 100644 --- a/pkg/sentry/fs/inode.go +++ b/pkg/sentry/fs/inode.go @@ -340,6 +340,13 @@ func (i *Inode) Truncate(ctx context.Context, d *Dirent, size int64) error { return i.InodeOperations.Truncate(ctx, i, size) } +func (i *Inode) Allocate(ctx context.Context, d *Dirent, offset int64, length int64) error { + if i.overlay != nil { + return overlayAllocate(ctx, i.overlay, d, offset, length) + } + return i.InodeOperations.Allocate(ctx, i, offset, length) +} + // Readlink calls i.InodeOperations.Readlnk with i as the Inode. func (i *Inode) Readlink(ctx context.Context) (string, error) { if i.overlay != nil { diff --git a/pkg/sentry/fs/inode_operations.go b/pkg/sentry/fs/inode_operations.go index ac287e1e4..abafe4791 100644 --- a/pkg/sentry/fs/inode_operations.go +++ b/pkg/sentry/fs/inode_operations.go @@ -223,6 +223,10 @@ type InodeOperations interface { // Implementations need not check that length >= 0. Truncate(ctx context.Context, inode *Inode, size int64) error + // Allocate allows the caller to reserve disk space for the inode. + // It's equivalent to fallocate(2) with 'mode=0'. + Allocate(ctx context.Context, inode *Inode, offset int64, length int64) error + // WriteOut writes cached Inode state to a backing filesystem in a // synchronous manner. // diff --git a/pkg/sentry/fs/inode_overlay.go b/pkg/sentry/fs/inode_overlay.go index 3d015328e..ead487097 100644 --- a/pkg/sentry/fs/inode_overlay.go +++ b/pkg/sentry/fs/inode_overlay.go @@ -582,6 +582,13 @@ func overlayTruncate(ctx context.Context, o *overlayEntry, d *Dirent, size int64 return o.upper.InodeOperations.Truncate(ctx, o.upper, size) } +func overlayAllocate(ctx context.Context, o *overlayEntry, d *Dirent, offset, length int64) error { + if err := copyUp(ctx, d); err != nil { + return err + } + return o.upper.InodeOperations.Allocate(ctx, o.upper, offset, length) +} + func overlayReadlink(ctx context.Context, o *overlayEntry) (string, error) { o.copyMu.RLock() defer o.copyMu.RUnlock() diff --git a/pkg/sentry/fs/inode_overlay_test.go b/pkg/sentry/fs/inode_overlay_test.go index 66b3da2d0..52ce1d29e 100644 --- a/pkg/sentry/fs/inode_overlay_test.go +++ b/pkg/sentry/fs/inode_overlay_test.go @@ -422,6 +422,7 @@ type inode struct { fsutil.InodeNoExtendedAttributes `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotAllocatable `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotSocket `state:"nosave"` diff --git a/pkg/sentry/fs/mock.go b/pkg/sentry/fs/mock.go index cf359a1f1..a71144b2c 100644 --- a/pkg/sentry/fs/mock.go +++ b/pkg/sentry/fs/mock.go @@ -150,6 +150,11 @@ func (n *MockInodeOperations) Truncate(ctx context.Context, inode *Inode, size i return nil } +// Allocate implements fs.InodeOperations.Allocate. +func (n *MockInodeOperations) Allocate(ctx context.Context, inode *Inode, offset, length int64) error { + return nil +} + // Remove implements fs.InodeOperations.Remove. func (n *MockInodeOperations) Remove(context.Context, *Inode, string) error { return nil diff --git a/pkg/sentry/fs/proc/inode.go b/pkg/sentry/fs/proc/inode.go index b03807043..379569823 100644 --- a/pkg/sentry/fs/proc/inode.go +++ b/pkg/sentry/fs/proc/inode.go @@ -55,6 +55,7 @@ func (i *taskOwnedInodeOps) UnstableAttr(ctx context.Context, inode *fs.Inode) ( type staticFileInodeOps struct { fsutil.InodeDenyWriteChecker `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` + fsutil.InodeNoopAllocate `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopTruncate `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index 10ea1f55d..6b0ae9e60 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -93,6 +93,7 @@ type SeqFile struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotAllocatable `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotSocket `state:"nosave"` @@ -183,7 +184,6 @@ func (s *SeqFile) updateSourceLocked(ctx context.Context, record int) { // // +stateify savable type seqFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` @@ -192,6 +192,7 @@ type seqFileOperations struct { fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` seqFile *SeqFile } diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index d649da0f1..5df3cee13 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -38,6 +38,7 @@ type idMapInodeOperations struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotAllocatable `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotSocket `state:"nosave"` @@ -81,7 +82,6 @@ func (imio *idMapInodeOperations) GetFile(ctx context.Context, dirent *fs.Dirent // +stateify savable type idMapFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` @@ -90,6 +90,7 @@ type idMapFileOperations struct { fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` iops *idMapInodeOperations } diff --git a/pkg/sentry/fs/ramfs/dir.go b/pkg/sentry/fs/ramfs/dir.go index a6b6a5c33..eb98b59cc 100644 --- a/pkg/sentry/fs/ramfs/dir.go +++ b/pkg/sentry/fs/ramfs/dir.go @@ -50,6 +50,7 @@ type CreateOps struct { // +stateify savable type Dir struct { fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeIsDirAllocate `state:"nosave"` fsutil.InodeIsDirTruncate `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index 9406a07ca..a7cb1bb86 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -30,6 +30,7 @@ type Socket struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotAllocatable `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotSymlink `state:"nosave"` @@ -67,7 +68,6 @@ func (s *Socket) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFl // +stateify savable type socketFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` @@ -78,6 +78,7 @@ type socketFileOperations struct { fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoWrite `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } var _ fs.FileOperations = (*socketFileOperations)(nil) diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index f7835fe05..dd2585b02 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -29,10 +29,11 @@ type Symlink struct { fsutil.InodeGenericChecker `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotAllocatable `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` fsutil.InodeNotMappable `state:"nosave"` - fsutil.InodeNotTruncatable `state:"nosave"` fsutil.InodeNotSocket `state:"nosave"` + fsutil.InodeNotTruncatable `state:"nosave"` fsutil.InodeVirtual `state:"nosave"` fsutil.InodeSimpleAttributes @@ -88,7 +89,6 @@ func (s *Symlink) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileF // +stateify savable type symlinkFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` @@ -99,6 +99,7 @@ type symlinkFileOperations struct { fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoWrite `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } var _ fs.FileOperations = (*symlinkFileOperations)(nil) diff --git a/pkg/sentry/fs/sys/devices.go b/pkg/sentry/fs/sys/devices.go index db91de435..bacc93af8 100644 --- a/pkg/sentry/fs/sys/devices.go +++ b/pkg/sentry/fs/sys/devices.go @@ -30,12 +30,13 @@ type cpunum struct { fsutil.InodeNoExtendedAttributes `state:"nosave"` fsutil.InodeNoopRelease `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` + fsutil.InodeNotAllocatable `state:"nosave"` fsutil.InodeNotDirectory `state:"nosave"` fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotSocket `state:"nosave"` fsutil.InodeNotSymlink `state:"nosave"` - fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeNotTruncatable `state:"nosave"` + fsutil.InodeNotVirtual `state:"nosave"` fsutil.InodeSimpleAttributes fsutil.InodeStaticFileGetter diff --git a/pkg/sentry/fs/tmpfs/inode_file.go b/pkg/sentry/fs/tmpfs/inode_file.go index f89d86c83..c90062a22 100644 --- a/pkg/sentry/fs/tmpfs/inode_file.go +++ b/pkg/sentry/fs/tmpfs/inode_file.go @@ -259,6 +259,33 @@ func (f *fileInodeOperations) Truncate(ctx context.Context, _ *fs.Inode, size in return nil } +// Allocate implements fs.InodeOperations.Allocate. +func (f *fileInodeOperations) Allocate(ctx context.Context, _ *fs.Inode, offset, length int64) error { + newSize := offset + length + + f.attrMu.Lock() + defer f.attrMu.Unlock() + f.dataMu.Lock() + defer f.dataMu.Unlock() + + if newSize <= f.attr.Size { + return nil + } + + // Check if current seals allow growth. + if f.seals&linux.F_SEAL_GROW != 0 { + return syserror.EPERM + } + + f.attr.Size = newSize + + now := ktime.NowFromContext(ctx) + f.attr.ModificationTime = now + f.attr.StatusChangeTime = now + + return nil +} + // AddLink implements fs.InodeOperations.AddLink. func (f *fileInodeOperations) AddLink() { f.attrMu.Lock() diff --git a/pkg/sentry/fs/tmpfs/tmpfs.go b/pkg/sentry/fs/tmpfs/tmpfs.go index 832914453..6ad5c5adb 100644 --- a/pkg/sentry/fs/tmpfs/tmpfs.go +++ b/pkg/sentry/fs/tmpfs/tmpfs.go @@ -242,11 +242,16 @@ func (d *Dir) Rename(ctx context.Context, oldParent *fs.Inode, oldName string, n return rename(ctx, oldParent, oldName, newParent, newName, replacement) } -// StatFS implments fs.InodeOperations.StatFS. +// StatFS implements fs.InodeOperations.StatFS. func (*Dir) StatFS(context.Context) (fs.Info, error) { return fsInfo, nil } +// Allocate implements fs.InodeOperations.Allocate. +func (d *Dir) Allocate(ctx context.Context, node *fs.Inode, offset, length int64) error { + return d.ramfsDir.Allocate(ctx, node, offset, length) +} + // Symlink is a symlink. // // +stateify savable @@ -281,6 +286,7 @@ func (s *Symlink) StatFS(context.Context) (fs.Info, error) { type Socket struct { ramfs.Socket fsutil.InodeNotTruncatable `state:"nosave"` + fsutil.InodeNotAllocatable `state:"nosave"` } // NewSocket returns a new socket with the provided permissions. diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index 0fc777e67..8dc40e1f2 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -53,13 +53,14 @@ import ( // +stateify savable type dirInodeOperations struct { fsutil.InodeGenericChecker `state:"nosave"` + fsutil.InodeIsDirAllocate `state:"nosave"` + fsutil.InodeIsDirTruncate `state:"nosave"` fsutil.InodeNoExtendedAttributes `state:"nosave"` fsutil.InodeNoopWriteOut `state:"nosave"` fsutil.InodeNotMappable `state:"nosave"` fsutil.InodeNotRenameable `state:"nosave"` - fsutil.InodeNotSymlink `state:"nosave"` fsutil.InodeNotSocket `state:"nosave"` - fsutil.InodeNotTruncatable `state:"nosave"` + fsutil.InodeNotSymlink `state:"nosave"` fsutil.InodeVirtual `state:"nosave"` fsutil.InodeSimpleAttributes diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index 99188dddf..7c3739360 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -191,3 +191,7 @@ func (*inodeOperations) newHandleLocked(wakeupChan *chan struct{}) { *wakeupChan = nil } } + +func (*inodeOperations) Allocate(_ context.Context, _ *fs.Inode, _, _ int64) error { + return syserror.EPIPE +} diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 893322647..1764bb4b6 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -1900,9 +1900,9 @@ func Renameat(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sysc } // Fallocate implements linux system call fallocate(2). -// (well, not really, but at least we return the expected error codes) func Fallocate(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { fd := kdefs.FD(args[0].Int()) + mode := args[1].Int64() offset := args[2].Int64() length := args[3].Int64() @@ -1915,8 +1915,42 @@ func Fallocate(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Sys if offset < 0 || length <= 0 { return 0, nil, syserror.EINVAL } + if mode != 0 { + t.Kernel().EmitUnimplementedEvent(t) + return 0, nil, syserror.ENOTSUP + } + if !file.Flags().Write { + return 0, nil, syserror.EBADF + } + if fs.IsPipe(file.Dirent.Inode.StableAttr) { + return 0, nil, syserror.ESPIPE + } + if fs.IsDir(file.Dirent.Inode.StableAttr) { + return 0, nil, syserror.EISDIR + } + if !fs.IsRegular(file.Dirent.Inode.StableAttr) { + return 0, nil, syserror.ENODEV + } + size := offset + length + if size < 0 { + return 0, nil, syserror.EFBIG + } + if uint64(size) >= t.ThreadGroup().Limits().Get(limits.FileSize).Cur { + t.SendSignal(&arch.SignalInfo{ + Signo: int32(syscall.SIGXFSZ), + Code: arch.SignalInfoUser, + }) + return 0, nil, syserror.EFBIG + } + + if err := file.Dirent.Inode.Allocate(t, file.Dirent, offset, length); err != nil { + return 0, nil, err + } + + // File length modified, generate notification. + file.Dirent.InotifyEvent(linux.IN_MODIFY, 0) - return 0, nil, syserror.EOPNOTSUPP + return 0, nil, nil } // Flock implements linux syscall flock(2). diff --git a/runsc/boot/compat.go b/runsc/boot/compat.go index c1b33c551..c369e4d64 100644 --- a/runsc/boot/compat.go +++ b/runsc/boot/compat.go @@ -99,7 +99,7 @@ func (c *compatEmitter) emitUnimplementedSyscall(us *spb.UnimplementedSyscall) { // args: cmd, ... tr = newArgsTracker(0) - case syscall.SYS_IOCTL, syscall.SYS_EPOLL_CTL, syscall.SYS_SHMCTL, syscall.SYS_FUTEX: + case syscall.SYS_IOCTL, syscall.SYS_EPOLL_CTL, syscall.SYS_SHMCTL, syscall.SYS_FUTEX, syscall.SYS_FALLOCATE: // args: fd/addr, cmd, ... tr = newArgsTracker(1) diff --git a/runsc/fsgofer/filter/config.go b/runsc/fsgofer/filter/config.go index a1ad49fb2..4faab2946 100644 --- a/runsc/fsgofer/filter/config.go +++ b/runsc/fsgofer/filter/config.go @@ -62,8 +62,14 @@ var allowedSyscalls = seccomp.SyscallRules{ }, syscall.SYS_EXIT: {}, syscall.SYS_EXIT_GROUP: {}, - syscall.SYS_FCHMOD: {}, - syscall.SYS_FCHOWNAT: {}, + syscall.SYS_FALLOCATE: []seccomp.Rule{ + { + seccomp.AllowAny{}, + seccomp.AllowValue(0), + }, + }, + syscall.SYS_FCHMOD: {}, + syscall.SYS_FCHOWNAT: {}, syscall.SYS_FCNTL: []seccomp.Rule{ { seccomp.AllowAny{}, diff --git a/runsc/fsgofer/fsgofer.go b/runsc/fsgofer/fsgofer.go index 3a0806837..b185015b6 100644 --- a/runsc/fsgofer/fsgofer.go +++ b/runsc/fsgofer/fsgofer.go @@ -731,6 +731,18 @@ func (l *localFile) SetAttr(valid p9.SetAttrMask, attr p9.SetAttr) error { return err } +// Allocate implements p9.File. +func (l *localFile) Allocate(mode p9.AllocateMode, offset, length uint64) error { + if !l.isOpen() { + return syscall.EBADF + } + + if err := syscall.Fallocate(l.file.FD(), mode.ToLinux(), int64(offset), int64(length)); err != nil { + return extractErrno(err) + } + return nil +} + // Rename implements p9.File; this should never be called. func (l *localFile) Rename(p9.File, string) error { panic("rename called directly") diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index d99733fc9..7ff4e4883 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -649,6 +649,8 @@ cc_binary( srcs = ["fallocate.cc"], linkstatic = 1, deps = [ + ":file_base", + "//test/util:cleanup", "//test/util:file_descriptor", "//test/util:temp_path", "//test/util:test_main", diff --git a/test/syscalls/linux/fallocate.cc b/test/syscalls/linux/fallocate.cc index 61b8acc7a..1c3d00287 100644 --- a/test/syscalls/linux/fallocate.cc +++ b/test/syscalls/linux/fallocate.cc @@ -12,45 +12,130 @@ // See the License for the specific language governing permissions and // limitations under the License. +#include #include +#include +#include +#include +#include +#include #include #include "gtest/gtest.h" +#include "test/syscalls/linux/file_base.h" +#include "test/util/cleanup.h" #include "test/util/file_descriptor.h" #include "test/util/temp_path.h" #include "test/util/test_util.h" namespace gvisor { namespace testing { - namespace { -// These tests are very rudimentary because fallocate is not -// implemented. We just want to make sure the expected error codes are -// returned. +int fallocate(int fd, int mode, off_t offset, off_t len) { + return syscall(__NR_fallocate, fd, mode, offset, len); +} + +class AllocateTest : public FileTest { + void SetUp() override { FileTest::SetUp(); } +}; + +TEST_F(AllocateTest, Fallocate) { + // Check that it starts at size zero. + struct stat buf; + ASSERT_THAT(fstat(test_file_fd_.get(), &buf), SyscallSucceeds()); + EXPECT_EQ(buf.st_size, 0); + + // Grow to ten bytes. + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, 0, 10), SyscallSucceeds()); + ASSERT_THAT(fstat(test_file_fd_.get(), &buf), SyscallSucceeds()); + EXPECT_EQ(buf.st_size, 10); -TEST(FallocateTest, NotImplemented) { - auto temp_path = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); - FileDescriptor fd = ASSERT_NO_ERRNO_AND_VALUE(Open(temp_path.path(), O_RDWR)); + // Allocate to a smaller size should be noop. + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, 0, 5), SyscallSucceeds()); + ASSERT_THAT(fstat(test_file_fd_.get(), &buf), SyscallSucceeds()); + EXPECT_EQ(buf.st_size, 10); - // Test that a completely unassigned fallocate mode returns EOPNOTSUPP. - ASSERT_THAT(fallocate(fd.get(), 0x80, 0, 32768), - SyscallFailsWithErrno(EOPNOTSUPP)); + // Grow again. + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, 0, 20), SyscallSucceeds()); + ASSERT_THAT(fstat(test_file_fd_.get(), &buf), SyscallSucceeds()); + EXPECT_EQ(buf.st_size, 20); + + // Grow with offset. + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, 10, 20), SyscallSucceeds()); + ASSERT_THAT(fstat(test_file_fd_.get(), &buf), SyscallSucceeds()); + EXPECT_EQ(buf.st_size, 30); + + // Grow with offset beyond EOF. + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, 39, 1), SyscallSucceeds()); + ASSERT_THAT(fstat(test_file_fd_.get(), &buf), SyscallSucceeds()); + EXPECT_EQ(buf.st_size, 40); } -TEST(FallocateTest, BadOffset) { - auto temp_path = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); - FileDescriptor fd = ASSERT_NO_ERRNO_AND_VALUE(Open(temp_path.path(), O_RDWR)); - ASSERT_THAT(fallocate(fd.get(), 0, -1, 32768), SyscallFailsWithErrno(EINVAL)); +TEST_F(AllocateTest, FallocateInvalid) { + // Invalid FD + EXPECT_THAT(fallocate(-1, 0, 0, 10), SyscallFailsWithErrno(EBADF)); + + // Negative offset and size. + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, -1, 10), + SyscallFailsWithErrno(EINVAL)); + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, 0, -1), + SyscallFailsWithErrno(EINVAL)); + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, -1, -1), + SyscallFailsWithErrno(EINVAL)); } -TEST(FallocateTest, BadLength) { - auto temp_path = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); - FileDescriptor fd = ASSERT_NO_ERRNO_AND_VALUE(Open(temp_path.path(), O_RDWR)); - ASSERT_THAT(fallocate(fd.get(), 0, 0, -1), SyscallFailsWithErrno(EINVAL)); +TEST_F(AllocateTest, FallocateReadonly) { + auto file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + FileDescriptor fd = ASSERT_NO_ERRNO_AND_VALUE(Open(file.path(), O_RDONLY)); + EXPECT_THAT(fallocate(fd.get(), 0, 0, 10), SyscallFailsWithErrno(EBADF)); } -} // namespace +TEST_F(AllocateTest, FallocatePipe) { + int pipes[2]; + EXPECT_THAT(pipe(pipes), SyscallSucceeds()); + auto cleanup = Cleanup([&pipes] { + EXPECT_THAT(close(pipes[0]), SyscallSucceeds()); + EXPECT_THAT(close(pipes[1]), SyscallSucceeds()); + }); + + EXPECT_THAT(fallocate(pipes[1], 0, 0, 10), SyscallFailsWithErrno(ESPIPE)); +} + +TEST_F(AllocateTest, FallocateChar) { + const FileDescriptor fd = + ASSERT_NO_ERRNO_AND_VALUE(Open("/dev/null", O_RDWR)); + EXPECT_THAT(fallocate(fd.get(), 0, 0, 10), SyscallFailsWithErrno(ENODEV)); +} + +TEST_F(AllocateTest, FallocateRlimit) { + // Get the current rlimit and restore after test run. + struct rlimit initial_lim; + ASSERT_THAT(getrlimit(RLIMIT_FSIZE, &initial_lim), SyscallSucceeds()); + auto cleanup = Cleanup([&initial_lim] { + EXPECT_THAT(setrlimit(RLIMIT_FSIZE, &initial_lim), SyscallSucceeds()); + }); + + // Try growing past the file size limit. + sigset_t new_mask; + sigemptyset(&new_mask); + sigaddset(&new_mask, SIGXFSZ); + sigprocmask(SIG_BLOCK, &new_mask, nullptr); + struct rlimit setlim = {}; + setlim.rlim_cur = 1024; + setlim.rlim_max = RLIM_INFINITY; + ASSERT_THAT(setrlimit(RLIMIT_FSIZE, &setlim), SyscallSucceeds()); + + EXPECT_THAT(fallocate(test_file_fd_.get(), 0, 0, 1025), + SyscallFailsWithErrno(EFBIG)); + + struct timespec timelimit = {}; + timelimit.tv_sec = 10; + EXPECT_EQ(sigtimedwait(&new_mask, nullptr, &timelimit), SIGXFSZ); + ASSERT_THAT(sigprocmask(SIG_UNBLOCK, &new_mask, nullptr), SyscallSucceeds()); +} + +} // namespace } // namespace testing } // namespace gvisor -- cgit v1.2.3 From 5ee8218483ce172400c21780d5dbc1ec2ba54d63 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Fri, 10 May 2019 13:36:56 -0700 Subject: Add pgalloc.DelayedEvictionManual. PiperOrigin-RevId: 247667272 Change-Id: I16b04e11bb93f50b7e05e888992303f730e4a877 --- pkg/sentry/kernel/kernel.go | 5 +++-- pkg/sentry/pgalloc/pgalloc.go | 21 ++++++++++++++++++++- pkg/sentry/pgalloc/save_restore.go | 11 +---------- 3 files changed, 24 insertions(+), 13 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/kernel.go b/pkg/sentry/kernel/kernel.go index 91889b573..85d73ace2 100644 --- a/pkg/sentry/kernel/kernel.go +++ b/pkg/sentry/kernel/kernel.go @@ -304,10 +304,11 @@ func (k *Kernel) SaveTo(w io.Writer) error { defer k.resumeTimeLocked() // Evict all evictable MemoryFile allocations. - k.mf.FlushEvictions() + k.mf.StartEvictions() + k.mf.WaitForEvictions() // Flush write operations on open files so data reaches backing storage. - // This must come after k.mf.FlushEvictions() since eviction may cause file + // This must come after MemoryFile eviction since eviction may cause file // writes. if err := k.tasks.flushWritesToFiles(ctx); err != nil { return err diff --git a/pkg/sentry/pgalloc/pgalloc.go b/pkg/sentry/pgalloc/pgalloc.go index 9c1313f6f..2b9924ad7 100644 --- a/pkg/sentry/pgalloc/pgalloc.go +++ b/pkg/sentry/pgalloc/pgalloc.go @@ -190,6 +190,11 @@ const ( // reclaimer goroutine is out of work (pages to reclaim), then evicts all // pending evictable allocations immediately. DelayedEvictionEnabled + + // DelayedEvictionManual requires that evictable allocations are only + // evicted when MemoryFile.StartEvictions() is called. This is extremely + // dangerous outside of tests. + DelayedEvictionManual ) // usageInfo tracks usage information. @@ -264,7 +269,7 @@ func NewMemoryFile(file *os.File, opts MemoryFileOpts) (*MemoryFile, error) { switch opts.DelayedEviction { case DelayedEvictionDefault: opts.DelayedEviction = DelayedEvictionEnabled - case DelayedEvictionDisabled, DelayedEvictionEnabled: + case DelayedEvictionDisabled, DelayedEvictionEnabled, DelayedEvictionManual: default: return nil, fmt.Errorf("invalid MemoryFileOpts.DelayedEviction: %v", opts.DelayedEviction) } @@ -1075,6 +1080,14 @@ func (f *MemoryFile) markReclaimed(fr platform.FileRange) { } } +// StartEvictions requests that f evict all evictable allocations. It does not +// wait for eviction to complete; for this, see MemoryFile.WaitForEvictions. +func (f *MemoryFile) StartEvictions() { + f.mu.Lock() + defer f.mu.Unlock() + f.startEvictionsLocked() +} + // Preconditions: f.mu must be locked. func (f *MemoryFile) startEvictionsLocked() { for user, info := range f.evictable { @@ -1122,6 +1135,12 @@ func (f *MemoryFile) startEvictionGoroutineLocked(user EvictableMemoryUser, info }() } +// WaitForEvictions blocks until f is no longer evicting any evictable +// allocations. +func (f *MemoryFile) WaitForEvictions() { + f.evictionWG.Wait() +} + type usageSetFunctions struct{} func (usageSetFunctions) MinKey() uint64 { diff --git a/pkg/sentry/pgalloc/save_restore.go b/pkg/sentry/pgalloc/save_restore.go index 9534d1aed..d4ba384b1 100644 --- a/pkg/sentry/pgalloc/save_restore.go +++ b/pkg/sentry/pgalloc/save_restore.go @@ -28,15 +28,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/state" ) -// FlushEvictions blocks until f has finished evicting all evictable -// allocations. -func (f *MemoryFile) FlushEvictions() { - f.mu.Lock() - f.startEvictionsLocked() - f.mu.Unlock() - f.evictionWG.Wait() -} - // SaveTo writes f's state to the given stream. func (f *MemoryFile) SaveTo(w io.Writer) error { // Wait for reclaim. @@ -51,7 +42,7 @@ func (f *MemoryFile) SaveTo(w io.Writer) error { // Ensure that there are no pending evictions. if len(f.evictable) != 0 { - panic(fmt.Sprintf("evictions still pending for %d users; call FlushEvictions before SaveTo", len(f.evictable))) + panic(fmt.Sprintf("evictions still pending for %d users; call StartEvictions and WaitForEvictions before SaveTo", len(f.evictable))) } // Ensure that all pages that contain data have knownCommitted set, since -- cgit v1.2.3 From 9cdae51feca5cee9faa198161b92a0aeece52d6c Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Tue, 21 May 2019 15:17:05 -0700 Subject: Add basic plumbing for splice and stub implementation. This does not actually implement an efficient splice or sendfile. Rather, it adds a generic plumbing to the file internals so that this can be added. All file implementations use the stub fileutil.NoSplice implementation, which causes sendfile and splice to fall back to an internal copy. A basic splice system call interface is added, along with a test. PiperOrigin-RevId: 249335960 Change-Id: Ic5568be2af0a505c19e7aec66d5af2480ab0939b --- pkg/abi/linux/BUILD | 1 + pkg/abi/linux/splice.go | 23 ++ pkg/sentry/fs/BUILD | 4 +- pkg/sentry/fs/ashmem/area.go | 3 +- pkg/sentry/fs/binder/binder.go | 3 +- pkg/sentry/fs/dev/full.go | 3 +- pkg/sentry/fs/dev/null.go | 4 +- pkg/sentry/fs/dev/random.go | 1 + pkg/sentry/fs/fdpipe/pipe.go | 1 + pkg/sentry/fs/file.go | 139 +++++++---- pkg/sentry/fs/file_operations.go | 47 ++++ pkg/sentry/fs/file_overlay.go | 79 ++++--- pkg/sentry/fs/file_test.go | 24 -- pkg/sentry/fs/filetest/filetest.go | 1 + pkg/sentry/fs/fsutil/file.go | 16 ++ pkg/sentry/fs/fsutil/inode.go | 3 +- pkg/sentry/fs/gofer/file.go | 5 +- pkg/sentry/fs/host/file.go | 1 + pkg/sentry/fs/inotify.go | 10 + pkg/sentry/fs/proc/exec_args.go | 3 +- pkg/sentry/fs/proc/rpcinet_proc.go | 3 +- pkg/sentry/fs/proc/seqfile/seqfile.go | 1 + pkg/sentry/fs/proc/sys.go | 3 +- pkg/sentry/fs/proc/sys_net.go | 6 +- pkg/sentry/fs/proc/task.go | 10 +- pkg/sentry/fs/proc/uid_gid_map.go | 1 + pkg/sentry/fs/proc/uptime.go | 5 +- pkg/sentry/fs/ramfs/socket.go | 7 +- pkg/sentry/fs/ramfs/symlink.go | 7 +- pkg/sentry/fs/splice.go | 187 +++++++++++++++ pkg/sentry/fs/timerfd/timerfd.go | 5 +- pkg/sentry/fs/tmpfs/file_regular.go | 3 +- pkg/sentry/fs/tty/dir.go | 3 +- pkg/sentry/fs/tty/master.go | 3 +- pkg/sentry/fs/tty/slave.go | 3 +- pkg/sentry/kernel/epoll/epoll.go | 3 +- pkg/sentry/kernel/eventfd/eventfd.go | 5 +- pkg/sentry/kernel/pipe/reader_writer.go | 3 +- pkg/sentry/loader/vdso.go | 3 +- pkg/sentry/socket/epsocket/epsocket.go | 3 +- pkg/sentry/socket/hostinet/socket.go | 3 +- pkg/sentry/socket/netlink/socket.go | 3 +- pkg/sentry/socket/rpcinet/socket.go | 3 +- pkg/sentry/socket/unix/unix.go | 3 +- pkg/sentry/syscalls/linux/BUILD | 1 + pkg/sentry/syscalls/linux/linux64.go | 2 +- pkg/sentry/syscalls/linux/sys_file.go | 98 -------- pkg/sentry/syscalls/linux/sys_splice.go | 293 +++++++++++++++++++++++ test/syscalls/BUILD | 2 + test/syscalls/linux/BUILD | 16 ++ test/syscalls/linux/splice.cc | 404 ++++++++++++++++++++++++++++++++ 51 files changed, 1221 insertions(+), 242 deletions(-) create mode 100644 pkg/abi/linux/splice.go delete mode 100644 pkg/sentry/fs/file_test.go create mode 100644 pkg/sentry/fs/splice.go create mode 100644 pkg/sentry/syscalls/linux/sys_splice.go create mode 100644 test/syscalls/linux/splice.cc (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/BUILD b/pkg/abi/linux/BUILD index fdf193873..96e8d4641 100644 --- a/pkg/abi/linux/BUILD +++ b/pkg/abi/linux/BUILD @@ -45,6 +45,7 @@ go_library( "shm.go", "signal.go", "socket.go", + "splice.go", "tcp.go", "time.go", "timer.go", diff --git a/pkg/abi/linux/splice.go b/pkg/abi/linux/splice.go new file mode 100644 index 000000000..650eb87e8 --- /dev/null +++ b/pkg/abi/linux/splice.go @@ -0,0 +1,23 @@ +// Copyright 2019 The gVisor Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package linux + +// Constants for splice(2), sendfile(2) and tee(2). +const ( + SPLICE_F_MOVE = 1 << iota + SPLICE_F_NONBLOCK + SPLICE_F_MORE + SPLICE_F_GIFT +) diff --git a/pkg/sentry/fs/BUILD b/pkg/sentry/fs/BUILD index 1fd9e30f6..142a00840 100644 --- a/pkg/sentry/fs/BUILD +++ b/pkg/sentry/fs/BUILD @@ -40,6 +40,7 @@ go_library( "restore.go", "save.go", "seek.go", + "splice.go", "sync.go", ], importpath = "gvisor.googlesource.com/gvisor/pkg/sentry/fs", @@ -51,6 +52,7 @@ go_library( "//pkg/metric", "//pkg/p9", "//pkg/refs", + "//pkg/secio", "//pkg/sentry/arch", "//pkg/sentry/context", "//pkg/sentry/device", @@ -66,7 +68,6 @@ go_library( "//pkg/sentry/usermem", "//pkg/state", "//pkg/syserror", - "//pkg/tcpip", "//pkg/waiter", ], ) @@ -122,7 +123,6 @@ go_test( srcs = [ "dirent_cache_test.go", "dirent_refs_test.go", - "file_test.go", "mount_test.go", "path_test.go", ], diff --git a/pkg/sentry/fs/ashmem/area.go b/pkg/sentry/fs/ashmem/area.go index b53746519..b4b0cc08b 100644 --- a/pkg/sentry/fs/ashmem/area.go +++ b/pkg/sentry/fs/ashmem/area.go @@ -42,11 +42,12 @@ const ( // // +stateify savable type Area struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` ad *Device diff --git a/pkg/sentry/fs/binder/binder.go b/pkg/sentry/fs/binder/binder.go index a992253e6..c78f1fc40 100644 --- a/pkg/sentry/fs/binder/binder.go +++ b/pkg/sentry/fs/binder/binder.go @@ -86,10 +86,11 @@ func (bd *Device) GetFile(ctx context.Context, d *fs.Dirent, flags fs.FileFlags) // // +stateify savable type Proc struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` bd *Device task *kernel.Task diff --git a/pkg/sentry/fs/dev/full.go b/pkg/sentry/fs/dev/full.go index 17d68b5c4..8f6c6da2d 100644 --- a/pkg/sentry/fs/dev/full.go +++ b/pkg/sentry/fs/dev/full.go @@ -60,6 +60,7 @@ func (f *fullDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.Fi // +stateify savable type fullFileOperations struct { + waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` @@ -68,8 +69,8 @@ type fullFileOperations struct { fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` readZeros `state:"nosave"` - waiter.AlwaysReady `state:"nosave"` } var _ fs.FileOperations = (*fullFileOperations)(nil) diff --git a/pkg/sentry/fs/dev/null.go b/pkg/sentry/fs/dev/null.go index ee13183c8..3f1accef8 100644 --- a/pkg/sentry/fs/dev/null.go +++ b/pkg/sentry/fs/dev/null.go @@ -64,6 +64,7 @@ type nullFileOperations struct { fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRead `state:"nosave"` @@ -104,14 +105,15 @@ func (zd *zeroDevice) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.F type zeroFileOperations struct { fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNoopWrite `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` - readZeros `state:"nosave"` waiter.AlwaysReady `state:"nosave"` + readZeros `state:"nosave"` } var _ fs.FileOperations = (*zeroFileOperations)(nil) diff --git a/pkg/sentry/fs/dev/random.go b/pkg/sentry/fs/dev/random.go index b0a412382..e5a01a906 100644 --- a/pkg/sentry/fs/dev/random.go +++ b/pkg/sentry/fs/dev/random.go @@ -61,6 +61,7 @@ type randomFileOperations struct { fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/fdpipe/pipe.go b/pkg/sentry/fs/fdpipe/pipe.go index 95e66ea8d..4ef7ea08a 100644 --- a/pkg/sentry/fs/fdpipe/pipe.go +++ b/pkg/sentry/fs/fdpipe/pipe.go @@ -43,6 +43,7 @@ type pipeOperations struct { fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` waiter.Queue `state:"nosave"` diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 62b35dabc..8f1baca23 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -21,7 +21,6 @@ import ( "time" "gvisor.googlesource.com/gvisor/pkg/amutex" - "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/metric" "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/sentry/context" @@ -35,8 +34,13 @@ import ( ) var ( - // RecordWaitTime controls writing metrics for filesystem reads. Enabling this comes at a small - // CPU cost due to performing two monotonic clock reads per read call. + // RecordWaitTime controls writing metrics for filesystem reads. + // Enabling this comes at a small CPU cost due to performing two + // monotonic clock reads per read call. + // + // Note that this is only performed in the direct read path, and may + // not be consistently applied for other forms of reads, such as + // splice. RecordWaitTime = false reads = metric.MustCreateNewUint64Metric("/fs/reads", false /* sync */, "Number of file reads.") @@ -306,14 +310,28 @@ func (f *File) Writev(ctx context.Context, src usermem.IOSequence) (int64, error return 0, syserror.ErrInterrupted } - offset, err := f.checkWriteLocked(ctx, &src, f.offset) - if err != nil { + // Handle append mode. + if f.Flags().Append { + if err := f.offsetForAppend(ctx, &f.offset); err != nil { + f.mu.Unlock() + return 0, err + } + } + + // Enforce file limits. + limit, ok := f.checkLimit(ctx, f.offset) + switch { + case ok && limit == 0: f.mu.Unlock() - return 0, err + return 0, syserror.ErrExceedsFileSizeLimit + case ok: + src = src.TakeFirst64(limit) } - n, err := f.FileOperations.Write(ctx, f, src, offset) + + // We must hold the lock during the write. + n, err := f.FileOperations.Write(ctx, f, src, f.offset) if n >= 0 { - atomic.StoreInt64(&f.offset, offset+n) + atomic.StoreInt64(&f.offset, f.offset+n) } f.mu.Unlock() return n, err @@ -325,51 +343,67 @@ func (f *File) Writev(ctx context.Context, src usermem.IOSequence) (int64, error // // Otherwise same as Writev. func (f *File) Pwritev(ctx context.Context, src usermem.IOSequence, offset int64) (int64, error) { - if !f.mu.Lock(ctx) { - return 0, syserror.ErrInterrupted + // "POSIX requires that opening a file with the O_APPEND flag should + // have no effect on the location at which pwrite() writes data. + // However, on Linux, if a file is opened with O_APPEND, pwrite() + // appends data to the end of the file, regardless of the value of + // offset." + if f.Flags().Append { + if !f.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer f.mu.Unlock() + if err := f.offsetForAppend(ctx, &offset); err != nil { + f.mu.Unlock() + return 0, err + } } - offset, err := f.checkWriteLocked(ctx, &src, offset) - if err != nil { - f.mu.Unlock() - return 0, err + // Enforce file limits. + limit, ok := f.checkLimit(ctx, offset) + switch { + case ok && limit == 0: + return 0, syserror.ErrExceedsFileSizeLimit + case ok: + src = src.TakeFirst64(limit) } - n, err := f.FileOperations.Write(ctx, f, src, offset) - f.mu.Unlock() - return n, err + + return f.FileOperations.Write(ctx, f, src, offset) } -// checkWriteLocked returns the offset to write at or an error if the write -// would not succeed. May update src to fit a write operation into a file -// size limit. -func (f *File) checkWriteLocked(ctx context.Context, src *usermem.IOSequence, offset int64) (int64, error) { - // Handle append only files. Note that this is still racy for network - // filesystems. - if f.Flags().Append { - uattr, err := f.Dirent.Inode.UnstableAttr(ctx) - if err != nil { - // This is an odd error, most likely it is evidence - // that something is terribly wrong with the filesystem. - // Return a generic EIO error. - log.Warningf("Failed to check write of inode %#v: %v", f.Dirent.Inode.StableAttr, err) - return offset, syserror.EIO - } - offset = uattr.Size +// offsetForAppend sets the given offset to the end of the file. +// +// Precondition: the underlying file mutex should be held. +func (f *File) offsetForAppend(ctx context.Context, offset *int64) error { + uattr, err := f.Dirent.Inode.UnstableAttr(ctx) + if err != nil { + // This is an odd error, we treat it as evidence that + // something is terribly wrong with the filesystem. + return syserror.EIO } - // Is this a regular file? + // Update the offset. + *offset = uattr.Size + + return nil +} + +// checkLimit checks the offset that the write will be performed at. The +// returned boolean indicates that the write must be limited. The returned +// integer indicates the new maximum write length. +func (f *File) checkLimit(ctx context.Context, offset int64) (int64, bool) { if IsRegular(f.Dirent.Inode.StableAttr) { // Enforce size limits. fileSizeLimit := limits.FromContext(ctx).Get(limits.FileSize).Cur if fileSizeLimit <= math.MaxInt64 { if offset >= int64(fileSizeLimit) { - return offset, syserror.ErrExceedsFileSizeLimit + return 0, true } - *src = src.TakeFirst64(int64(fileSizeLimit) - offset) + return int64(fileSizeLimit) - offset, true } } - return offset, nil + return 0, false } // Fsync calls f.FileOperations.Fsync with f as the File. @@ -466,8 +500,13 @@ func (f *File) Async(newAsync func() FileAsync) FileAsync { return f.async } -// FileReader implements io.Reader and io.ReaderAt. -type FileReader struct { +// lockedReader implements io.Reader and io.ReaderAt. +// +// Note this reads the underlying file using the file operations directly. It +// is the responsibility of the caller to ensure that locks are appropriately +// held and offsets updated if required. This should be used only by internal +// functions that perform these operations and checks at other times. +type lockedReader struct { // Ctx is the context for the file reader. Ctx context.Context @@ -476,19 +515,21 @@ type FileReader struct { } // Read implements io.Reader.Read. -func (r *FileReader) Read(buf []byte) (int, error) { - n, err := r.File.Readv(r.Ctx, usermem.BytesIOSequence(buf)) +func (r *lockedReader) Read(buf []byte) (int, error) { + n, err := r.File.FileOperations.Read(r.Ctx, r.File, usermem.BytesIOSequence(buf), r.File.offset) return int(n), err } // ReadAt implements io.Reader.ReadAt. -func (r *FileReader) ReadAt(buf []byte, offset int64) (int, error) { - n, err := r.File.Preadv(r.Ctx, usermem.BytesIOSequence(buf), offset) +func (r *lockedReader) ReadAt(buf []byte, offset int64) (int, error) { + n, err := r.File.FileOperations.Read(r.Ctx, r.File, usermem.BytesIOSequence(buf), offset) return int(n), err } -// FileWriter implements io.Writer and io.WriterAt. -type FileWriter struct { +// lockedWriter implements io.Writer and io.WriterAt. +// +// The same constraints as lockedReader apply; see above. +type lockedWriter struct { // Ctx is the context for the file writer. Ctx context.Context @@ -497,13 +538,13 @@ type FileWriter struct { } // Write implements io.Writer.Write. -func (w *FileWriter) Write(buf []byte) (int, error) { - n, err := w.File.Writev(w.Ctx, usermem.BytesIOSequence(buf)) +func (w *lockedWriter) Write(buf []byte) (int, error) { + n, err := w.File.FileOperations.Write(w.Ctx, w.File, usermem.BytesIOSequence(buf), w.File.offset) return int(n), err } // WriteAt implements io.Writer.WriteAt. -func (w *FileWriter) WriteAt(buf []byte, offset int64) (int, error) { - n, err := w.File.Pwritev(w.Ctx, usermem.BytesIOSequence(buf), offset) +func (w *lockedWriter) WriteAt(buf []byte, offset int64) (int, error) { + n, err := w.File.FileOperations.Write(w.Ctx, w.File, usermem.BytesIOSequence(buf), offset) return int(n), err } diff --git a/pkg/sentry/fs/file_operations.go b/pkg/sentry/fs/file_operations.go index ab0acb6eb..0f2dfa273 100644 --- a/pkg/sentry/fs/file_operations.go +++ b/pkg/sentry/fs/file_operations.go @@ -22,6 +22,38 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) +// SpliceOpts define how a splice works. +type SpliceOpts struct { + // Length is the length of the splice operation. + Length int64 + + // SrcOffset indicates whether the existing source file offset should + // be used. If this is true, then the Start value below is used. + // + // When passed to FileOperations object, this should always be true as + // the offset will be provided by a layer above, unless the object in + // question is a pipe or socket. This value can be relied upon for such + // an indicator. + SrcOffset bool + + // SrcStart is the start of the source file. This is used only if + // SrcOffset is false. + SrcStart int64 + + // Dup indicates that the contents should not be consumed from the + // source (e.g. in the case of a socket or a pipe), but duplicated. + Dup bool + + // DstOffset indicates that the destination file offset should be used. + // + // See SrcOffset for additional information. + DstOffset bool + + // DstStart is the start of the destination file. This is used only if + // DstOffset is false. + DstStart int64 +} + // FileOperations are operations on a File that diverge per file system. // // Operations that take a *File may use only the following interfaces: @@ -67,6 +99,15 @@ type FileOperations interface { // Read must not be called if !FileFlags.Read. Read(ctx context.Context, file *File, dst usermem.IOSequence, offset int64) (int64, error) + // WriteTo is a variant of read that takes another file as a + // destination. For a splice (copy or move from one file to another), + // first a WriteTo on the source is attempted, followed by a ReadFrom + // on the destination, following by a buffered copy with standard Read + // and Write operations. + // + // The same preconditions as Read apply. + WriteTo(ctx context.Context, file *File, dst *File, opts SpliceOpts) (int64, error) + // Write writes src to file at offset and returns the number of bytes // written which must be greater than or equal to 0. Like Read, file // systems that do not support writing at an offset (i.e. pipefs, sockfs) @@ -81,6 +122,12 @@ type FileOperations interface { // Write must not be called if !FileFlags.Write. Write(ctx context.Context, file *File, src usermem.IOSequence, offset int64) (int64, error) + // ReadFrom is a variant of write that takes a another file as a + // source. See WriteTo for details regarding how this is called. + // + // The same preconditions as Write apply; FileFlags.Write must be set. + ReadFrom(ctx context.Context, file *File, src *File, opts SpliceOpts) (int64, error) + // Fsync writes buffered modifications of file and/or flushes in-flight // operations to backing storage based on syncType. The range to sync is // [start, end]. The end is inclusive so that the last byte of a maximally diff --git a/pkg/sentry/fs/file_overlay.go b/pkg/sentry/fs/file_overlay.go index 948ce9c6f..273de1e14 100644 --- a/pkg/sentry/fs/file_overlay.go +++ b/pkg/sentry/fs/file_overlay.go @@ -17,7 +17,6 @@ package fs import ( "sync" - "gvisor.googlesource.com/gvisor/pkg/log" "gvisor.googlesource.com/gvisor/pkg/refs" "gvisor.googlesource.com/gvisor/pkg/sentry/arch" "gvisor.googlesource.com/gvisor/pkg/sentry/context" @@ -222,31 +221,50 @@ func (f *overlayFileOperations) IterateDir(ctx context.Context, dirCtx *DirCtx, return offset + n, err } -// Read implements FileOperations.Read. -func (f *overlayFileOperations) Read(ctx context.Context, file *File, dst usermem.IOSequence, offset int64) (int64, error) { - o := file.Dirent.Inode.overlay +// onTop performs the given operation on the top-most available layer. +func (f *overlayFileOperations) onTop(ctx context.Context, file *File, fn func(*File, FileOperations) error) error { + file.Dirent.Inode.overlay.copyMu.RLock() + defer file.Dirent.Inode.overlay.copyMu.RUnlock() - o.copyMu.RLock() - defer o.copyMu.RUnlock() + // Only lower layer is available. + if file.Dirent.Inode.overlay.upper == nil { + return fn(f.lower, f.lower.FileOperations) + } - if o.upper != nil { - // We may need to acquire an open file handle to read from if - // copy up has occurred. Otherwise we risk reading from the - // wrong source. - f.upperMu.Lock() - if f.upper == nil { - var err error - f.upper, err = overlayFile(ctx, o.upper, file.Flags()) - if err != nil { - f.upperMu.Unlock() - log.Warningf("failed to acquire handle with flags %v: %v", file.Flags(), err) - return 0, syserror.EIO - } + f.upperMu.Lock() + if f.upper == nil { + upper, err := overlayFile(ctx, file.Dirent.Inode.overlay.upper, file.Flags()) + if err != nil { + // Something very wrong; return a generic filesystem + // error to avoid propagating internals. + f.upperMu.Unlock() + return syserror.EIO } - f.upperMu.Unlock() - return f.upper.FileOperations.Read(ctx, f.upper, dst, offset) + + // Save upper file. + f.upper = upper } - return f.lower.FileOperations.Read(ctx, f.lower, dst, offset) + f.upperMu.Unlock() + + return fn(f.upper, f.upper.FileOperations) +} + +// Read implements FileOperations.Read. +func (f *overlayFileOperations) Read(ctx context.Context, file *File, dst usermem.IOSequence, offset int64) (n int64, err error) { + err = f.onTop(ctx, file, func(file *File, ops FileOperations) error { + n, err = ops.Read(ctx, file, dst, offset) + return err // Will overwrite itself. + }) + return +} + +// WriteTo implements FileOperations.WriteTo. +func (f *overlayFileOperations) WriteTo(ctx context.Context, file *File, dst *File, opts SpliceOpts) (n int64, err error) { + err = f.onTop(ctx, file, func(file *File, ops FileOperations) error { + n, err = ops.WriteTo(ctx, file, dst, opts) + return err // Will overwrite itself. + }) + return } // Write implements FileOperations.Write. @@ -257,15 +275,20 @@ func (f *overlayFileOperations) Write(ctx context.Context, file *File, src userm return f.upper.FileOperations.Write(ctx, f.upper, src, offset) } +// ReadFrom implements FileOperations.ReadFrom. +func (f *overlayFileOperations) ReadFrom(ctx context.Context, file *File, src *File, opts SpliceOpts) (n int64, err error) { + // See above; f.upper must be non-nil. + return f.upper.FileOperations.ReadFrom(ctx, f.upper, src, opts) +} + // Fsync implements FileOperations.Fsync. -func (f *overlayFileOperations) Fsync(ctx context.Context, file *File, start, end int64, syncType SyncType) error { - var err error +func (f *overlayFileOperations) Fsync(ctx context.Context, file *File, start, end int64, syncType SyncType) (err error) { f.upperMu.Lock() if f.upper != nil { err = f.upper.FileOperations.Fsync(ctx, f.upper, start, end, syncType) } f.upperMu.Unlock() - if f.lower != nil { + if err == nil && f.lower != nil { // N.B. Fsync on the lower filesystem can cause writes of file // attributes (i.e. access time) despite the fact that we must // treat the lower filesystem as read-only. @@ -277,15 +300,14 @@ func (f *overlayFileOperations) Fsync(ctx context.Context, file *File, start, en } // Flush implements FileOperations.Flush. -func (f *overlayFileOperations) Flush(ctx context.Context, file *File) error { +func (f *overlayFileOperations) Flush(ctx context.Context, file *File) (err error) { // Flush whatever handles we have. - var err error f.upperMu.Lock() if f.upper != nil { err = f.upper.FileOperations.Flush(ctx, f.upper) } f.upperMu.Unlock() - if f.lower != nil { + if err == nil && f.lower != nil { err = f.lower.FileOperations.Flush(ctx, f.lower) } return err @@ -329,6 +351,7 @@ func (*overlayFileOperations) ConfigureMMap(ctx context.Context, file *File, opt if !o.isMappableLocked() { return syserror.ENODEV } + // FIXME(jamieliu): This is a copy/paste of fsutil.GenericConfigureMMap, // which we can't use because the overlay implementation is in package fs, // so depending on fs/fsutil would create a circular dependency. Move diff --git a/pkg/sentry/fs/file_test.go b/pkg/sentry/fs/file_test.go deleted file mode 100644 index d867a0257..000000000 --- a/pkg/sentry/fs/file_test.go +++ /dev/null @@ -1,24 +0,0 @@ -// Copyright 2018 The gVisor Authors. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package fs - -import "io" - -var ( - _ = io.Reader(&FileReader{}) - _ = io.ReaderAt(&FileReader{}) - _ = io.Writer(&FileWriter{}) - _ = io.WriterAt(&FileWriter{}) -) diff --git a/pkg/sentry/fs/filetest/filetest.go b/pkg/sentry/fs/filetest/filetest.go index f6b827800..c0b1b088d 100644 --- a/pkg/sentry/fs/filetest/filetest.go +++ b/pkg/sentry/fs/filetest/filetest.go @@ -38,6 +38,7 @@ type TestFileOperations struct { fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` waiter.AlwaysReady `state:"nosave"` } diff --git a/pkg/sentry/fs/fsutil/file.go b/pkg/sentry/fs/fsutil/file.go index e355d8594..9381963d0 100644 --- a/pkg/sentry/fs/fsutil/file.go +++ b/pkg/sentry/fs/fsutil/file.go @@ -223,6 +223,20 @@ func (FileNoIoctl) Ioctl(ctx context.Context, io usermem.IO, args arch.SyscallAr return 0, syserror.ENOTTY } +// FileNoSplice implements fs.FileOperations.ReadFrom and +// fs.FileOperations.WriteTo for files that don't support splice. +type FileNoSplice struct{} + +// WriteTo implements fs.FileOperations.WriteTo. +func (FileNoSplice) WriteTo(context.Context, *fs.File, *fs.File, fs.SpliceOpts) (int64, error) { + return 0, syserror.ENOSYS +} + +// ReadFrom implements fs.FileOperations.ReadFrom. +func (FileNoSplice) ReadFrom(context.Context, *fs.File, *fs.File, fs.SpliceOpts) (int64, error) { + return 0, syserror.ENOSYS +} + // DirFileOperations implements most of fs.FileOperations for directories, // except for Readdir and UnstableAttr which the embedding type must implement. type DirFileOperations struct { @@ -233,6 +247,7 @@ type DirFileOperations struct { FileNoopFlush FileNoopFsync FileNoopRelease + FileNoSplice } // Read implements fs.FileOperations.Read @@ -303,6 +318,7 @@ type NoReadWriteFile struct { FileNoWrite `state:"nosave"` FileNotDirReaddir `state:"nosave"` FileUseInodeUnstableAttr `state:"nosave"` + FileNoSplice `state:"nosave"` } var _ fs.FileOperations = (*NoReadWriteFile)(nil) diff --git a/pkg/sentry/fs/fsutil/inode.go b/pkg/sentry/fs/fsutil/inode.go index a22b6ce9c..925887335 100644 --- a/pkg/sentry/fs/fsutil/inode.go +++ b/pkg/sentry/fs/fsutil/inode.go @@ -250,16 +250,17 @@ func (i *InodeSimpleExtendedAttributes) Listxattr(_ *fs.Inode) (map[string]struc // // +stateify savable type staticFile struct { - waiter.AlwaysReady `state:"nosave"` FileGenericSeek `state:"nosave"` FileNoIoctl `state:"nosave"` FileNoMMap `state:"nosave"` + FileNoSplice `state:"nosave"` FileNoopFsync `state:"nosave"` FileNoopFlush `state:"nosave"` FileNoopRelease `state:"nosave"` FileNoopWrite `state:"nosave"` FileNotDirReaddir `state:"nosave"` FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` FileStaticContentReader } diff --git a/pkg/sentry/fs/gofer/file.go b/pkg/sentry/fs/gofer/file.go index bc2be546e..fb4f50113 100644 --- a/pkg/sentry/fs/gofer/file.go +++ b/pkg/sentry/fs/gofer/file.go @@ -46,8 +46,9 @@ var ( // // +stateify savable type fileOperations struct { - fsutil.FileNoIoctl `state:"nosave"` - waiter.AlwaysReady `state:"nosave"` + fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoSplice `state:"nosplice"` + waiter.AlwaysReady `state:"nosave"` // inodeOperations is the inodeOperations backing the file. It is protected // by a reference held by File.Dirent.Inode which is stable until diff --git a/pkg/sentry/fs/host/file.go b/pkg/sentry/fs/host/file.go index 82e2ae3b9..ad0a3ec85 100644 --- a/pkg/sentry/fs/host/file.go +++ b/pkg/sentry/fs/host/file.go @@ -37,6 +37,7 @@ import ( // +stateify savable type fileOperations struct { fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoSplice `state:"nosplice"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` diff --git a/pkg/sentry/fs/inotify.go b/pkg/sentry/fs/inotify.go index 2652582c3..7dfd31020 100644 --- a/pkg/sentry/fs/inotify.go +++ b/pkg/sentry/fs/inotify.go @@ -171,11 +171,21 @@ func (i *Inotify) Read(ctx context.Context, _ *File, dst usermem.IOSequence, _ i return writeLen, nil } +// WriteTo implements FileOperations.WriteTo. +func (*Inotify) WriteTo(context.Context, *File, *File, SpliceOpts) (int64, error) { + return 0, syserror.ENOSYS +} + // Fsync implements FileOperations.Fsync. func (*Inotify) Fsync(context.Context, *File, int64, int64, SyncType) error { return syserror.EINVAL } +// ReadFrom implements FileOperations.ReadFrom. +func (*Inotify) ReadFrom(context.Context, *File, *File, SpliceOpts) (int64, error) { + return 0, syserror.ENOSYS +} + // Flush implements FileOperations.Flush. func (*Inotify) Flush(context.Context, *File) error { return nil diff --git a/pkg/sentry/fs/proc/exec_args.go b/pkg/sentry/fs/proc/exec_args.go index d49dad685..cb28f6bc3 100644 --- a/pkg/sentry/fs/proc/exec_args.go +++ b/pkg/sentry/fs/proc/exec_args.go @@ -77,16 +77,17 @@ func (i *execArgInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs. // +stateify savable type execArgFile struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopWrite `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` // arg is the type of exec argument this file contains. arg execArgType diff --git a/pkg/sentry/fs/proc/rpcinet_proc.go b/pkg/sentry/fs/proc/rpcinet_proc.go index db53686f6..e36c0bfa6 100644 --- a/pkg/sentry/fs/proc/rpcinet_proc.go +++ b/pkg/sentry/fs/proc/rpcinet_proc.go @@ -60,15 +60,16 @@ func (i *rpcInetInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs. // rpcInetFile implements fs.FileOperations as RPCs. type rpcInetFile struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` inode *rpcInetInode } diff --git a/pkg/sentry/fs/proc/seqfile/seqfile.go b/pkg/sentry/fs/proc/seqfile/seqfile.go index 6b0ae9e60..8364d86ed 100644 --- a/pkg/sentry/fs/proc/seqfile/seqfile.go +++ b/pkg/sentry/fs/proc/seqfile/seqfile.go @@ -187,6 +187,7 @@ type seqFileOperations struct { fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/proc/sys.go b/pkg/sentry/fs/proc/sys.go index b889ed625..59846af4f 100644 --- a/pkg/sentry/fs/proc/sys.go +++ b/pkg/sentry/fs/proc/sys.go @@ -134,7 +134,6 @@ var _ fs.InodeOperations = (*hostname)(nil) // +stateify savable type hostnameFile struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoSeek `state:"nosave"` @@ -143,7 +142,9 @@ type hostnameFile struct { fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoWrite `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` } // Read implements fs.FileOperations.Read. diff --git a/pkg/sentry/fs/proc/sys_net.go b/pkg/sentry/fs/proc/sys_net.go index e49794a48..dbf1a987c 100644 --- a/pkg/sentry/fs/proc/sys_net.go +++ b/pkg/sentry/fs/proc/sys_net.go @@ -85,15 +85,16 @@ func (m *tcpMemInode) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.F // +stateify savable type tcpMemFile struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` tcpMemInode *tcpMemInode } @@ -198,15 +199,16 @@ func (s *tcpSack) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileF // +stateify savable type tcpSackFile struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` tcpSack *tcpSack diff --git a/pkg/sentry/fs/proc/task.go b/pkg/sentry/fs/proc/task.go index 66d76d194..494b195cd 100644 --- a/pkg/sentry/fs/proc/task.go +++ b/pkg/sentry/fs/proc/task.go @@ -672,16 +672,17 @@ func (c *comm) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFlag // +stateify savable type commFile struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` t *kernel.Task } @@ -728,16 +729,17 @@ func (a *auxvec) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFl // +stateify savable type auxvecFile struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` t *kernel.Task } diff --git a/pkg/sentry/fs/proc/uid_gid_map.go b/pkg/sentry/fs/proc/uid_gid_map.go index 5df3cee13..a14b1b45f 100644 --- a/pkg/sentry/fs/proc/uid_gid_map.go +++ b/pkg/sentry/fs/proc/uid_gid_map.go @@ -85,6 +85,7 @@ type idMapFileOperations struct { fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` diff --git a/pkg/sentry/fs/proc/uptime.go b/pkg/sentry/fs/proc/uptime.go index 1ddf9fafa..35c3851e1 100644 --- a/pkg/sentry/fs/proc/uptime.go +++ b/pkg/sentry/fs/proc/uptime.go @@ -54,16 +54,17 @@ func (u *uptime) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFl // +stateify savable type uptimeFile struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` startTime ktime.Time } diff --git a/pkg/sentry/fs/ramfs/socket.go b/pkg/sentry/fs/ramfs/socket.go index a7cb1bb86..7d8bca70e 100644 --- a/pkg/sentry/fs/ramfs/socket.go +++ b/pkg/sentry/fs/ramfs/socket.go @@ -70,13 +70,14 @@ func (s *Socket) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileFl type socketFileOperations struct { fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoRead `state:"nosave"` + fsutil.FileNoSeek `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoRead `state:"nosave"` - fsutil.FileNoSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` waiter.AlwaysReady `state:"nosave"` } diff --git a/pkg/sentry/fs/ramfs/symlink.go b/pkg/sentry/fs/ramfs/symlink.go index dd2585b02..21c246169 100644 --- a/pkg/sentry/fs/ramfs/symlink.go +++ b/pkg/sentry/fs/ramfs/symlink.go @@ -91,13 +91,14 @@ func (s *Symlink) GetFile(ctx context.Context, dirent *fs.Dirent, flags fs.FileF type symlinkFileOperations struct { fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoRead `state:"nosave"` + fsutil.FileNoSeek `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoWrite `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` - fsutil.FileNoRead `state:"nosave"` - fsutil.FileNoSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoWrite `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` waiter.AlwaysReady `state:"nosave"` } diff --git a/pkg/sentry/fs/splice.go b/pkg/sentry/fs/splice.go new file mode 100644 index 000000000..65937f44d --- /dev/null +++ b/pkg/sentry/fs/splice.go @@ -0,0 +1,187 @@ +// Copyright 2019 The gVisor Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package fs + +import ( + "io" + "sync/atomic" + + "gvisor.googlesource.com/gvisor/pkg/secio" + "gvisor.googlesource.com/gvisor/pkg/sentry/context" + "gvisor.googlesource.com/gvisor/pkg/syserror" +) + +// Splice moves data to this file, directly from another. +// +// Offsets are updated only if DstOffset and SrcOffset are set. +func Splice(ctx context.Context, dst *File, src *File, opts SpliceOpts) (int64, error) { + // Verify basic file flag permissions. + if !dst.Flags().Write || !src.Flags().Read { + return 0, syserror.EBADF + } + + // Check whether or not the objects being sliced are stream-oriented + // (i.e. pipes or sockets). If yes, we elide checks and offset locks. + srcPipe := IsPipe(src.Dirent.Inode.StableAttr) || IsSocket(src.Dirent.Inode.StableAttr) + dstPipe := IsPipe(dst.Dirent.Inode.StableAttr) || IsSocket(dst.Dirent.Inode.StableAttr) + + if !dstPipe && !opts.DstOffset && !srcPipe && !opts.SrcOffset { + switch { + case dst.UniqueID < src.UniqueID: + // Acquire dst first. + if !dst.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer dst.mu.Unlock() + if !src.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer src.mu.Unlock() + case dst.UniqueID > src.UniqueID: + // Acquire src first. + if !src.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer src.mu.Unlock() + if !dst.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer dst.mu.Unlock() + case dst.UniqueID == src.UniqueID: + // Acquire only one lock; it's the same file. This is a + // bit of a edge case, but presumably it's possible. + if !dst.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer dst.mu.Unlock() + } + // Use both offsets (locked). + opts.DstStart = dst.offset + opts.SrcStart = src.offset + } else if !dstPipe && !opts.DstOffset { + // Acquire only dst. + if !dst.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer dst.mu.Unlock() + opts.DstStart = dst.offset // Safe: locked. + } else if !srcPipe && !opts.SrcOffset { + // Acquire only src. + if !src.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer src.mu.Unlock() + opts.SrcStart = src.offset // Safe: locked. + } + + // Check append-only mode and the limit. + if !dstPipe { + if dst.Flags().Append { + if opts.DstOffset { + // We need to acquire the lock. + if !dst.mu.Lock(ctx) { + return 0, syserror.ErrInterrupted + } + defer dst.mu.Unlock() + } + // Figure out the appropriate offset to use. + if err := dst.offsetForAppend(ctx, &opts.DstStart); err != nil { + return 0, err + } + } + + // Enforce file limits. + limit, ok := dst.checkLimit(ctx, opts.DstStart) + switch { + case ok && limit == 0: + return 0, syserror.ErrExceedsFileSizeLimit + case ok && limit < opts.Length: + opts.Length = limit // Cap the write. + } + } + + // Attempt to do a WriteTo; this is likely the most efficient. + // + // The underlying implementation may be able to donate buffers. + newOpts := SpliceOpts{ + Length: opts.Length, + SrcStart: opts.SrcStart, + SrcOffset: !srcPipe, + Dup: opts.Dup, + DstStart: opts.DstStart, + DstOffset: !dstPipe, + } + n, err := src.FileOperations.WriteTo(ctx, src, dst, newOpts) + if n == 0 && err != nil { + // Attempt as a ReadFrom. If a WriteTo, a ReadFrom may also + // be more efficient than a copy if buffers are cached or readily + // available. (It's unlikely that they can actually be donate + n, err = dst.FileOperations.ReadFrom(ctx, dst, src, newOpts) + } + if n == 0 && err != nil { + // If we've failed up to here, and at least one of the sources + // is a pipe or socket, then we can't properly support dup. + // Return an error indicating that this operation is not + // supported. + if (srcPipe || dstPipe) && newOpts.Dup { + return 0, syserror.EINVAL + } + + // We failed to splice the files. But that's fine; we just fall + // back to a slow path in this case. This copies without doing + // any mode changes, so should still be more efficient. + var ( + r io.Reader + w io.Writer + ) + fw := &lockedWriter{ + Ctx: ctx, + File: dst, + } + if newOpts.DstOffset { + // Use the provided offset. + w = secio.NewOffsetWriter(fw, newOpts.DstStart) + } else { + // Writes will proceed with no offset. + w = fw + } + fr := &lockedReader{ + Ctx: ctx, + File: src, + } + if newOpts.SrcOffset { + // Limit to the given offset and length. + r = io.NewSectionReader(fr, opts.SrcStart, opts.Length) + } else { + // Limit just to the given length. + r = &io.LimitedReader{fr, opts.Length} + } + + // Copy between the two. + n, err = io.Copy(w, r) + } + + // Update offsets, if required. + if n > 0 { + if !dstPipe && !opts.DstOffset { + atomic.StoreInt64(&dst.offset, dst.offset+n) + } + if !srcPipe && !opts.SrcOffset { + atomic.StoreInt64(&src.offset, src.offset+n) + } + } + + return n, err +} diff --git a/pkg/sentry/fs/timerfd/timerfd.go b/pkg/sentry/fs/timerfd/timerfd.go index 749961f51..bce5f091d 100644 --- a/pkg/sentry/fs/timerfd/timerfd.go +++ b/pkg/sentry/fs/timerfd/timerfd.go @@ -36,9 +36,10 @@ type TimerOperations struct { fsutil.FileZeroSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` events waiter.Queue `state:"zerovalue"` diff --git a/pkg/sentry/fs/tmpfs/file_regular.go b/pkg/sentry/fs/tmpfs/file_regular.go index 1ef256511..d1c163879 100644 --- a/pkg/sentry/fs/tmpfs/file_regular.go +++ b/pkg/sentry/fs/tmpfs/file_regular.go @@ -28,14 +28,15 @@ import ( // // +stateify savable type regularFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoopFsync `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` // iops is the InodeOperations of a regular tmpfs file. It is // guaranteed to be the same as file.Dirent.Inode.InodeOperations, diff --git a/pkg/sentry/fs/tty/dir.go b/pkg/sentry/fs/tty/dir.go index 8dc40e1f2..2603354c4 100644 --- a/pkg/sentry/fs/tty/dir.go +++ b/pkg/sentry/fs/tty/dir.go @@ -286,14 +286,15 @@ func (d *dirInodeOperations) masterClose(t *Terminal) { // // +stateify savable type dirFileOperations struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileGenericSeek `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` // di is the inode operations. di *dirInodeOperations diff --git a/pkg/sentry/fs/tty/master.go b/pkg/sentry/fs/tty/master.go index 45e167e5f..afdf44cd1 100644 --- a/pkg/sentry/fs/tty/master.go +++ b/pkg/sentry/fs/tty/master.go @@ -98,8 +98,9 @@ type masterFileOperations struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` // d is the containing dir. diff --git a/pkg/sentry/fs/tty/slave.go b/pkg/sentry/fs/tty/slave.go index 0ae57a02c..2abf32e57 100644 --- a/pkg/sentry/fs/tty/slave.go +++ b/pkg/sentry/fs/tty/slave.go @@ -87,8 +87,9 @@ type slaveFileOperations struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` // si is the inode operations. diff --git a/pkg/sentry/kernel/epoll/epoll.go b/pkg/sentry/kernel/epoll/epoll.go index 2399ae6f2..bbacba1f4 100644 --- a/pkg/sentry/kernel/epoll/epoll.go +++ b/pkg/sentry/kernel/epoll/epoll.go @@ -102,8 +102,9 @@ type EventPoll struct { fsutil.FileNotDirReaddir `state:"zerovalue"` fsutil.FileNoFsync `state:"zerovalue"` fsutil.FileNoopFlush `state:"zerovalue"` - fsutil.FileNoMMap `state:"zerovalue"` fsutil.FileNoIoctl `state:"zerovalue"` + fsutil.FileNoMMap `state:"zerovalue"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` // Wait queue is used to notify interested parties when the event poll diff --git a/pkg/sentry/kernel/eventfd/eventfd.go b/pkg/sentry/kernel/eventfd/eventfd.go index 5d3139eef..2f900be38 100644 --- a/pkg/sentry/kernel/eventfd/eventfd.go +++ b/pkg/sentry/kernel/eventfd/eventfd.go @@ -42,9 +42,10 @@ type EventOperations struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` - fsutil.FileNoMMap `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` + fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` // Mutex that protects accesses to the fields of this event. diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index ddcc5e09a..59899be49 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -38,8 +38,9 @@ type ReaderWriter struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` *Pipe } diff --git a/pkg/sentry/loader/vdso.go b/pkg/sentry/loader/vdso.go index 207d8ed3d..4e73527cf 100644 --- a/pkg/sentry/loader/vdso.go +++ b/pkg/sentry/loader/vdso.go @@ -52,15 +52,16 @@ func (f *fileContext) Value(key interface{}) interface{} { // byteReader implements fs.FileOperations for reading from a []byte source. type byteReader struct { - waiter.AlwaysReady `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` fsutil.FileNoIoctl `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoopRelease `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FilePipeSeek `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` + waiter.AlwaysReady `state:"nosave"` data []byte } diff --git a/pkg/sentry/socket/epsocket/epsocket.go b/pkg/sentry/socket/epsocket/epsocket.go index 520d82f68..31a449cf2 100644 --- a/pkg/sentry/socket/epsocket/epsocket.go +++ b/pkg/sentry/socket/epsocket/epsocket.go @@ -212,9 +212,10 @@ type commonEndpoint interface { type SocketOperations struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` - fsutil.FileNoFsync `state:"nosave"` fsutil.FileNoopFlush `state:"nosave"` + fsutil.FileNoFsync `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` socket.SendReceiveTimeout *waiter.Queue diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index 71884d3db..41f9693bb 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -49,8 +49,9 @@ type socketOperations struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` socket.SendReceiveTimeout diff --git a/pkg/sentry/socket/netlink/socket.go b/pkg/sentry/socket/netlink/socket.go index dc688eb00..afd06ca33 100644 --- a/pkg/sentry/socket/netlink/socket.go +++ b/pkg/sentry/socket/netlink/socket.go @@ -68,8 +68,9 @@ type Socket struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` socket.SendReceiveTimeout diff --git a/pkg/sentry/socket/rpcinet/socket.go b/pkg/sentry/socket/rpcinet/socket.go index c028ed4dd..55e0b6665 100644 --- a/pkg/sentry/socket/rpcinet/socket.go +++ b/pkg/sentry/socket/rpcinet/socket.go @@ -48,8 +48,9 @@ type socketOperations struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` socket.SendReceiveTimeout diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 26788ec31..931056d51 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -48,8 +48,9 @@ type SocketOperations struct { fsutil.FilePipeSeek `state:"nosave"` fsutil.FileNotDirReaddir `state:"nosave"` fsutil.FileNoFsync `state:"nosave"` - fsutil.FileNoopFlush `state:"nosave"` fsutil.FileNoMMap `state:"nosave"` + fsutil.FileNoSplice `state:"nosave"` + fsutil.FileNoopFlush `state:"nosave"` fsutil.FileUseInodeUnstableAttr `state:"nosave"` refs.AtomicRefCount socket.SendReceiveTimeout diff --git a/pkg/sentry/syscalls/linux/BUILD b/pkg/sentry/syscalls/linux/BUILD index 6e2843b36..f76989ae2 100644 --- a/pkg/sentry/syscalls/linux/BUILD +++ b/pkg/sentry/syscalls/linux/BUILD @@ -34,6 +34,7 @@ go_library( "sys_shm.go", "sys_signal.go", "sys_socket.go", + "sys_splice.go", "sys_stat.go", "sys_sync.go", "sys_sysinfo.go", diff --git a/pkg/sentry/syscalls/linux/linux64.go b/pkg/sentry/syscalls/linux/linux64.go index 9a460ebdf..3e4d312af 100644 --- a/pkg/sentry/syscalls/linux/linux64.go +++ b/pkg/sentry/syscalls/linux/linux64.go @@ -407,7 +407,7 @@ var AMD64 = &kernel.SyscallTable{ 273: syscalls.Error(syscall.ENOSYS), // @Syscall(GetRobustList, note:Obsolete) 274: syscalls.Error(syscall.ENOSYS), - // 275: @Syscall(Splice), TODO(b/29354098) + 275: Splice, // 276: @Syscall(Tee), TODO(b/29354098) 277: SyncFileRange, // 278: @Syscall(Vmsplice), TODO(b/29354098) diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 1764bb4b6..8a80cd430 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -15,7 +15,6 @@ package linux import ( - "io" "syscall" "gvisor.googlesource.com/gvisor/pkg/abi/linux" @@ -2025,103 +2024,6 @@ func Flock(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall return 0, nil, nil } -// Sendfile implements linux system call sendfile(2). -func Sendfile(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { - outFD := kdefs.FD(args[0].Int()) - inFD := kdefs.FD(args[1].Int()) - offsetAddr := args[2].Pointer() - count := int64(args[3].SizeT()) - - // Don't send a negative number of bytes. - if count < 0 { - return 0, nil, syserror.EINVAL - } - - if count > int64(kernel.MAX_RW_COUNT) { - count = int64(kernel.MAX_RW_COUNT) - } - - // Get files. - outFile := t.FDMap().GetFile(outFD) - if outFile == nil { - return 0, nil, syserror.EBADF - } - defer outFile.DecRef() - - inFile := t.FDMap().GetFile(inFD) - if inFile == nil { - return 0, nil, syserror.EBADF - } - defer inFile.DecRef() - - // Verify that the outfile is writable. - outFlags := outFile.Flags() - if !outFlags.Write { - return 0, nil, syserror.EBADF - } - - // Verify that the outfile Append flag is not set. - if outFlags.Append { - return 0, nil, syserror.EINVAL - } - - // Verify that we have a regular infile. - // http://elixir.free-electrons.com/linux/latest/source/fs/splice.c#L933 - if !fs.IsRegular(inFile.Dirent.Inode.StableAttr) { - return 0, nil, syserror.EINVAL - } - - // Verify that the infile is readable. - if !inFile.Flags().Read { - return 0, nil, syserror.EBADF - } - - // Setup for sending data. - var n int64 - var err error - w := &fs.FileWriter{t, outFile} - hasOffset := offsetAddr != 0 - // If we have a provided offset. - if hasOffset { - // Verify that when offset address is not null, infile must be seekable - if !inFile.Flags().Pread { - return 0, nil, syserror.ESPIPE - } - // Copy in the offset. - var offset int64 - if _, err := t.CopyIn(offsetAddr, &offset); err != nil { - return 0, nil, err - } - if offset < 0 { - return 0, nil, syserror.EINVAL - } - // Send data using Preadv. - r := io.NewSectionReader(&fs.FileReader{t, inFile}, offset, count) - n, err = io.Copy(w, r) - // Copy out the new offset. - if _, err := t.CopyOut(offsetAddr, n+offset); err != nil { - return 0, nil, err - } - // If we don't have a provided offset. - } else { - // Send data using readv. - inOff := inFile.Offset() - r := &io.LimitedReader{R: &fs.FileReader{t, inFile}, N: count} - n, err = io.Copy(w, r) - inOff += n - if inFile.Offset() != inOff { - // Adjust file position in case more bytes were read than written. - if _, err := inFile.Seek(t, fs.SeekSet, inOff); err != nil { - return 0, nil, syserror.EIO - } - } - } - - // We can only pass a single file to handleIOError, so pick inFile - // arbitrarily. - return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "sendfile", inFile) -} - const ( memfdPrefix = "/memfd:" memfdAllFlags = uint32(linux.MFD_CLOEXEC | linux.MFD_ALLOW_SEALING) diff --git a/pkg/sentry/syscalls/linux/sys_splice.go b/pkg/sentry/syscalls/linux/sys_splice.go new file mode 100644 index 000000000..37303606f --- /dev/null +++ b/pkg/sentry/syscalls/linux/sys_splice.go @@ -0,0 +1,293 @@ +// Copyright 2019 The gVisor Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package linux + +import ( + "gvisor.googlesource.com/gvisor/pkg/abi/linux" + "gvisor.googlesource.com/gvisor/pkg/sentry/arch" + "gvisor.googlesource.com/gvisor/pkg/sentry/fs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.googlesource.com/gvisor/pkg/syserror" + "gvisor.googlesource.com/gvisor/pkg/waiter" +) + +// doSplice implements a blocking splice operation. +func doSplice(t *kernel.Task, outFile, inFile *fs.File, opts fs.SpliceOpts, nonBlocking bool) (int64, error) { + var ( + total int64 + n int64 + err error + ch chan struct{} + inW bool + outW bool + ) + for opts.Length > 0 { + n, err = fs.Splice(t, outFile, inFile, opts) + opts.Length -= n + total += n + if err != syserror.ErrWouldBlock { + break + } else if err == syserror.ErrWouldBlock && nonBlocking { + break + } + + // Are we a registered waiter? + if ch == nil { + ch = make(chan struct{}, 1) + } + if !inW && inFile.Readiness(EventMaskRead) == 0 && !inFile.Flags().NonBlocking { + w, _ := waiter.NewChannelEntry(ch) + inFile.EventRegister(&w, EventMaskRead) + defer inFile.EventUnregister(&w) + inW = true // Registered. + } else if !outW && outFile.Readiness(EventMaskWrite) == 0 && !outFile.Flags().NonBlocking { + w, _ := waiter.NewChannelEntry(ch) + outFile.EventRegister(&w, EventMaskWrite) + defer outFile.EventUnregister(&w) + outW = true // Registered. + } + + // Was anything registered? If no, everything is non-blocking. + if !inW && !outW { + break + } + + // Block until there's data. + if err = t.Block(ch); err != nil { + break + } + } + + return total, err +} + +// Sendfile implements linux system call sendfile(2). +func Sendfile(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + outFD := kdefs.FD(args[0].Int()) + inFD := kdefs.FD(args[1].Int()) + offsetAddr := args[2].Pointer() + count := int64(args[3].SizeT()) + + // Don't send a negative number of bytes. + if count < 0 { + return 0, nil, syserror.EINVAL + } + + // Get files. + outFile := t.FDMap().GetFile(outFD) + if outFile == nil { + return 0, nil, syserror.EBADF + } + defer outFile.DecRef() + + inFile := t.FDMap().GetFile(inFD) + if inFile == nil { + return 0, nil, syserror.EBADF + } + defer inFile.DecRef() + + // Verify that the outfile Append flag is not set. Note that fs.Splice + // itself validates that the output file is writable. + if outFile.Flags().Append { + return 0, nil, syserror.EBADF + } + + // Verify that we have a regular infile. This is a requirement; the + // same check appears in Linux (fs/splice.c:splice_direct_to_actor). + if !fs.IsRegular(inFile.Dirent.Inode.StableAttr) { + return 0, nil, syserror.EINVAL + } + + var ( + n int64 + err error + ) + if offsetAddr != 0 { + // Verify that when offset address is not null, infile must be + // seekable. The fs.Splice routine itself validates basic read. + if !inFile.Flags().Pread { + return 0, nil, syserror.ESPIPE + } + + // Copy in the offset. + var offset int64 + if _, err := t.CopyIn(offsetAddr, &offset); err != nil { + return 0, nil, err + } + + // The offset must be valid. + if offset < 0 { + return 0, nil, syserror.EINVAL + } + + // Do the splice. + n, err = doSplice(t, outFile, inFile, fs.SpliceOpts{ + Length: count, + SrcOffset: true, + SrcStart: offset, + }, false) + + // Copy out the new offset. + if _, err := t.CopyOut(offsetAddr, n+offset); err != nil { + return 0, nil, err + } + } else { + // Send data using splice. + n, err = doSplice(t, outFile, inFile, fs.SpliceOpts{ + Length: count, + }, false) + } + + // We can only pass a single file to handleIOError, so pick inFile + // arbitrarily. This is used only for debugging purposes. + return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "sendfile", inFile) +} + +// Splice implements splice(2). +func Splice(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + inFD := kdefs.FD(args[0].Int()) + inOffset := args[1].Pointer() + outFD := kdefs.FD(args[2].Int()) + outOffset := args[3].Pointer() + count := int64(args[4].SizeT()) + flags := args[5].Int() + + // Check for invalid flags. + if flags&^(linux.SPLICE_F_MOVE|linux.SPLICE_F_NONBLOCK|linux.SPLICE_F_MORE|linux.SPLICE_F_GIFT) != 0 { + return 0, nil, syserror.EINVAL + } + + // Only non-blocking is meaningful. Note that unlike in Linux, this + // flag is applied consistently. We will have either fully blocking or + // non-blocking behavior below, regardless of the underlying files + // being spliced to. It's unclear if this is a bug or not yet. + nonBlocking := (flags & linux.SPLICE_F_NONBLOCK) != 0 + + // Get files. + outFile := t.FDMap().GetFile(outFD) + if outFile == nil { + return 0, nil, syserror.EBADF + } + defer outFile.DecRef() + + inFile := t.FDMap().GetFile(inFD) + if inFile == nil { + return 0, nil, syserror.EBADF + } + defer inFile.DecRef() + + // Construct our options. + // + // Note that exactly one of the underlying buffers must be a pipe. We + // don't actually have this constraint internally, but we enforce it + // for the semantics of the call. + opts := fs.SpliceOpts{ + Length: count, + } + switch { + case fs.IsPipe(inFile.Dirent.Inode.StableAttr) && !fs.IsPipe(outFile.Dirent.Inode.StableAttr): + if inOffset != 0 { + return 0, nil, syserror.ESPIPE + } + if outOffset != 0 { + var offset int64 + if _, err := t.CopyIn(outOffset, &offset); err != nil { + return 0, nil, err + } + // Use the destination offset. + opts.DstOffset = true + opts.DstStart = offset + } + case !fs.IsPipe(inFile.Dirent.Inode.StableAttr) && fs.IsPipe(outFile.Dirent.Inode.StableAttr): + if outOffset != 0 { + return 0, nil, syserror.ESPIPE + } + if inOffset != 0 { + var offset int64 + if _, err := t.CopyIn(inOffset, &offset); err != nil { + return 0, nil, err + } + // Use the source offset. + opts.SrcOffset = true + opts.SrcStart = offset + } + case fs.IsPipe(inFile.Dirent.Inode.StableAttr) && fs.IsPipe(outFile.Dirent.Inode.StableAttr): + if inOffset != 0 || outOffset != 0 { + return 0, nil, syserror.ESPIPE + } + default: + return 0, nil, syserror.EINVAL + } + + // We may not refer to the same pipe; otherwise it's a continuous loop. + if inFile.Dirent.Inode.StableAttr.InodeID == outFile.Dirent.Inode.StableAttr.InodeID { + return 0, nil, syserror.EINVAL + } + + // Splice data. + n, err := doSplice(t, outFile, inFile, opts, nonBlocking) + + // See above; inFile is chosen arbitrarily here. + return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "splice", inFile) +} + +// Tee imlements tee(2). +func Tee(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.SyscallControl, error) { + inFD := kdefs.FD(args[0].Int()) + outFD := kdefs.FD(args[1].Int()) + count := int64(args[2].SizeT()) + flags := args[3].Int() + + // Check for invalid flags. + if flags&^(linux.SPLICE_F_MOVE|linux.SPLICE_F_NONBLOCK|linux.SPLICE_F_MORE|linux.SPLICE_F_GIFT) != 0 { + return 0, nil, syserror.EINVAL + } + + // Only non-blocking is meaningful. + nonBlocking := (flags & linux.SPLICE_F_NONBLOCK) != 0 + + // Get files. + outFile := t.FDMap().GetFile(outFD) + if outFile == nil { + return 0, nil, syserror.EBADF + } + defer outFile.DecRef() + + inFile := t.FDMap().GetFile(inFD) + if inFile == nil { + return 0, nil, syserror.EBADF + } + defer inFile.DecRef() + + // All files must be pipes. + if !fs.IsPipe(inFile.Dirent.Inode.StableAttr) || !fs.IsPipe(outFile.Dirent.Inode.StableAttr) { + return 0, nil, syserror.EINVAL + } + + // We may not refer to the same pipe; see above. + if inFile.Dirent.Inode.StableAttr.InodeID == outFile.Dirent.Inode.StableAttr.InodeID { + return 0, nil, syserror.EINVAL + } + + // Splice data. + n, err := doSplice(t, outFile, inFile, fs.SpliceOpts{ + Length: count, + Dup: true, + }, nonBlocking) + + // See above; inFile is chosen arbitrarily here. + return uintptr(n), nil, handleIOError(t, n != 0, err, kernel.ERESTARTSYS, "tee", inFile) +} diff --git a/test/syscalls/BUILD b/test/syscalls/BUILD index 79be06494..b531d7629 100644 --- a/test/syscalls/BUILD +++ b/test/syscalls/BUILD @@ -277,6 +277,8 @@ syscall_test(test = "//test/syscalls/linux:sendfile_socket_test") syscall_test(test = "//test/syscalls/linux:sendfile_test") +syscall_test(test = "//test/syscalls/linux:splice_test") + syscall_test(test = "//test/syscalls/linux:sigaction_test") # TODO(b/119826902): Enable once the test passes in runsc. diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index ee40be569..d4e49bb3a 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -1747,6 +1747,22 @@ cc_binary( ], ) +cc_binary( + name = "splice_test", + testonly = 1, + srcs = ["splice.cc"], + linkstatic = 1, + deps = [ + "//test/util:file_descriptor", + "//test/util:temp_path", + "//test/util:test_main", + "//test/util:test_util", + "//test/util:thread_util", + "@com_google_absl//absl/strings", + "@com_google_googletest//:gtest", + ], +) + cc_binary( name = "sigaction_test", testonly = 1, diff --git a/test/syscalls/linux/splice.cc b/test/syscalls/linux/splice.cc new file mode 100644 index 000000000..1875f4533 --- /dev/null +++ b/test/syscalls/linux/splice.cc @@ -0,0 +1,404 @@ +// Copyright 2019 The gVisor Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include +#include + +#include "gmock/gmock.h" +#include "gtest/gtest.h" +#include "absl/strings/string_view.h" +#include "test/util/file_descriptor.h" +#include "test/util/temp_path.h" +#include "test/util/test_util.h" +#include "test/util/thread_util.h" + +namespace gvisor { +namespace testing { + +namespace { + +TEST(SpliceTest, TwoRegularFiles) { + // Create temp files. + const TempPath in_file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + const TempPath out_file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + + // Open the input file as read only. + const FileDescriptor inf = + ASSERT_NO_ERRNO_AND_VALUE(Open(in_file.path(), O_RDONLY)); + + // Open the output file as write only. + const FileDescriptor outf = + ASSERT_NO_ERRNO_AND_VALUE(Open(out_file.path(), O_WRONLY)); + + // Verify that it is rejected as expected; regardless of offsets. + loff_t in_offset = 0; + loff_t out_offset = 0; + EXPECT_THAT(splice(inf.get(), &in_offset, outf.get(), &out_offset, 1, 0), + SyscallFailsWithErrno(EINVAL)); + EXPECT_THAT(splice(inf.get(), nullptr, outf.get(), &out_offset, 1, 0), + SyscallFailsWithErrno(EINVAL)); + EXPECT_THAT(splice(inf.get(), &in_offset, outf.get(), nullptr, 1, 0), + SyscallFailsWithErrno(EINVAL)); + EXPECT_THAT(splice(inf.get(), nullptr, outf.get(), nullptr, 1, 0), + SyscallFailsWithErrno(EINVAL)); +} + +TEST(SpliceTest, SamePipe) { + // Create a new pipe. + int fds[2]; + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + const FileDescriptor rfd(fds[0]); + const FileDescriptor wfd(fds[1]); + + // Fill the pipe. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ASSERT_THAT(write(wfd.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + + // Attempt to splice to itself. + EXPECT_THAT(splice(rfd.get(), nullptr, wfd.get(), nullptr, kPageSize, 0), + SyscallFailsWithErrno(EINVAL)); +} + +TEST(TeeTest, SamePipe) { + SKIP_IF(IsRunningOnGvisor()); + + // Create a new pipe. + int fds[2]; + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + const FileDescriptor rfd(fds[0]); + const FileDescriptor wfd(fds[1]); + + // Fill the pipe. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ASSERT_THAT(write(wfd.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + + // Attempt to tee to itself. + EXPECT_THAT(tee(rfd.get(), wfd.get(), kPageSize, 0), + SyscallFailsWithErrno(EINVAL)); +} + +TEST(TeeTest, RegularFile) { + SKIP_IF(IsRunningOnGvisor()); + + // Open some file. + const TempPath in_file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + const FileDescriptor inf = + ASSERT_NO_ERRNO_AND_VALUE(Open(in_file.path(), O_RDWR)); + + // Create a new pipe. + int fds[2]; + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + const FileDescriptor rfd(fds[0]); + const FileDescriptor wfd(fds[1]); + + // Attempt to tee from the file. + EXPECT_THAT(tee(inf.get(), wfd.get(), kPageSize, 0), + SyscallFailsWithErrno(EINVAL)); + EXPECT_THAT(tee(rfd.get(), inf.get(), kPageSize, 0), + SyscallFailsWithErrno(EINVAL)); +} + +TEST(SpliceTest, PipeOffsets) { + // Create two new pipes. + int first[2], second[2]; + ASSERT_THAT(pipe(first), SyscallSucceeds()); + const FileDescriptor rfd1(first[0]); + const FileDescriptor wfd1(first[1]); + ASSERT_THAT(pipe(second), SyscallSucceeds()); + const FileDescriptor rfd2(second[0]); + const FileDescriptor wfd2(second[1]); + + // All pipe offsets should be rejected. + loff_t in_offset = 0; + loff_t out_offset = 0; + EXPECT_THAT(splice(rfd1.get(), &in_offset, wfd2.get(), &out_offset, 1, 0), + SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(splice(rfd1.get(), nullptr, wfd2.get(), &out_offset, 1, 0), + SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(splice(rfd1.get(), &in_offset, wfd2.get(), nullptr, 1, 0), + SyscallFailsWithErrno(ESPIPE)); +} + +TEST(SpliceTest, ToPipe) { + // Open the input file. + const TempPath in_file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + const FileDescriptor inf = + ASSERT_NO_ERRNO_AND_VALUE(Open(in_file.path(), O_RDWR)); + + // Fill with some random data. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ASSERT_THAT(write(inf.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + ASSERT_THAT(lseek(inf.get(), 0, SEEK_SET), SyscallSucceedsWithValue(0)); + + // Create a new pipe. + int fds[2]; + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + const FileDescriptor rfd(fds[0]); + const FileDescriptor wfd(fds[1]); + + // Splice to the pipe. + EXPECT_THAT(splice(inf.get(), nullptr, wfd.get(), nullptr, kPageSize, 0), + SyscallSucceedsWithValue(kPageSize)); + + // Contents should be equal. + std::vector rbuf(kPageSize); + ASSERT_THAT(read(rfd.get(), rbuf.data(), rbuf.size()), + SyscallSucceedsWithValue(kPageSize)); + EXPECT_EQ(memcmp(rbuf.data(), buf.data(), buf.size()), 0); +} + +TEST(SpliceTest, ToPipeOffset) { + // Open the input file. + const TempPath in_file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + const FileDescriptor inf = + ASSERT_NO_ERRNO_AND_VALUE(Open(in_file.path(), O_RDWR)); + + // Fill with some random data. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ASSERT_THAT(write(inf.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + + // Create a new pipe. + int fds[2]; + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + const FileDescriptor rfd(fds[0]); + const FileDescriptor wfd(fds[1]); + + // Splice to the pipe. + loff_t in_offset = kPageSize / 2; + EXPECT_THAT( + splice(inf.get(), &in_offset, wfd.get(), nullptr, kPageSize / 2, 0), + SyscallSucceedsWithValue(kPageSize / 2)); + + // Contents should be equal to only the second part. + std::vector rbuf(kPageSize / 2); + ASSERT_THAT(read(rfd.get(), rbuf.data(), rbuf.size()), + SyscallSucceedsWithValue(kPageSize / 2)); + EXPECT_EQ(memcmp(rbuf.data(), buf.data() + (kPageSize / 2), rbuf.size()), 0); +} + +TEST(SpliceTest, FromPipe) { + // Create a new pipe. + int fds[2]; + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + const FileDescriptor rfd(fds[0]); + const FileDescriptor wfd(fds[1]); + + // Fill with some random data. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ASSERT_THAT(write(wfd.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + + // Open the input file. + const TempPath out_file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + const FileDescriptor outf = + ASSERT_NO_ERRNO_AND_VALUE(Open(out_file.path(), O_RDWR)); + + // Splice to the output file. + EXPECT_THAT(splice(rfd.get(), nullptr, outf.get(), nullptr, kPageSize, 0), + SyscallSucceedsWithValue(kPageSize)); + + // The offset of the output should be equal to kPageSize. We assert that and + // reset to zero so that we can read the contents and ensure they match. + EXPECT_THAT(lseek(outf.get(), 0, SEEK_CUR), + SyscallSucceedsWithValue(kPageSize)); + ASSERT_THAT(lseek(outf.get(), 0, SEEK_SET), SyscallSucceedsWithValue(0)); + + // Contents should be equal. + std::vector rbuf(kPageSize); + ASSERT_THAT(read(outf.get(), rbuf.data(), rbuf.size()), + SyscallSucceedsWithValue(kPageSize)); + EXPECT_EQ(memcmp(rbuf.data(), buf.data(), buf.size()), 0); +} + +TEST(SpliceTest, FromPipeOffset) { + // Create a new pipe. + int fds[2]; + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + const FileDescriptor rfd(fds[0]); + const FileDescriptor wfd(fds[1]); + + // Fill with some random data. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ASSERT_THAT(write(wfd.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + + // Open the input file. + const TempPath out_file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + const FileDescriptor outf = + ASSERT_NO_ERRNO_AND_VALUE(Open(out_file.path(), O_RDWR)); + + // Splice to the output file. + loff_t out_offset = kPageSize / 2; + EXPECT_THAT(splice(rfd.get(), nullptr, outf.get(), &out_offset, kPageSize, 0), + SyscallSucceedsWithValue(kPageSize)); + + // Content should reflect the splice. We write to a specific offset in the + // file, so the internals should now be allocated sparsely. + std::vector rbuf(kPageSize); + ASSERT_THAT(read(outf.get(), rbuf.data(), rbuf.size()), + SyscallSucceedsWithValue(kPageSize)); + std::vector zbuf(kPageSize / 2); + memset(zbuf.data(), 0, zbuf.size()); + EXPECT_EQ(memcmp(rbuf.data(), zbuf.data(), zbuf.size()), 0); + EXPECT_EQ(memcmp(rbuf.data() + kPageSize / 2, buf.data(), kPageSize / 2), 0); +} + +TEST(SpliceTest, TwoPipes) { + // Create two new pipes. + int first[2], second[2]; + ASSERT_THAT(pipe(first), SyscallSucceeds()); + const FileDescriptor rfd1(first[0]); + const FileDescriptor wfd1(first[1]); + ASSERT_THAT(pipe(second), SyscallSucceeds()); + const FileDescriptor rfd2(second[0]); + const FileDescriptor wfd2(second[1]); + + // Fill with some random data. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ASSERT_THAT(write(wfd1.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + + // Splice to the second pipe, using two operations. + EXPECT_THAT( + splice(rfd1.get(), nullptr, wfd2.get(), nullptr, kPageSize / 2, 0), + SyscallSucceedsWithValue(kPageSize / 2)); + EXPECT_THAT( + splice(rfd1.get(), nullptr, wfd2.get(), nullptr, kPageSize / 2, 0), + SyscallSucceedsWithValue(kPageSize / 2)); + + // Content should reflect the splice. + std::vector rbuf(kPageSize); + ASSERT_THAT(read(rfd2.get(), rbuf.data(), rbuf.size()), + SyscallSucceedsWithValue(kPageSize)); + EXPECT_EQ(memcmp(rbuf.data(), buf.data(), kPageSize), 0); +} + +TEST(SpliceTest, Blocking) { + // Create two new pipes. + int first[2], second[2]; + ASSERT_THAT(pipe(first), SyscallSucceeds()); + const FileDescriptor rfd1(first[0]); + const FileDescriptor wfd1(first[1]); + ASSERT_THAT(pipe(second), SyscallSucceeds()); + const FileDescriptor rfd2(second[0]); + const FileDescriptor wfd2(second[1]); + + // This thread writes to the main pipe. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ScopedThread t([&]() { + ASSERT_THAT(write(wfd1.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + }); + + // Attempt a splice immediately; it should block. + EXPECT_THAT(splice(rfd1.get(), nullptr, wfd2.get(), nullptr, kPageSize, 0), + SyscallSucceedsWithValue(kPageSize)); + + // Thread should be joinable. + t.Join(); + + // Content should reflect the splice. + std::vector rbuf(kPageSize); + ASSERT_THAT(read(rfd2.get(), rbuf.data(), rbuf.size()), + SyscallSucceedsWithValue(kPageSize)); + EXPECT_EQ(memcmp(rbuf.data(), buf.data(), kPageSize), 0); +} + +TEST(TeeTest, Blocking) { + SKIP_IF(IsRunningOnGvisor()); + + // Create two new pipes. + int first[2], second[2]; + ASSERT_THAT(pipe(first), SyscallSucceeds()); + const FileDescriptor rfd1(first[0]); + const FileDescriptor wfd1(first[1]); + ASSERT_THAT(pipe(second), SyscallSucceeds()); + const FileDescriptor rfd2(second[0]); + const FileDescriptor wfd2(second[1]); + + // This thread writes to the main pipe. + std::vector buf(kPageSize); + RandomizeBuffer(buf.data(), buf.size()); + ScopedThread t([&]() { + ASSERT_THAT(write(wfd1.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(kPageSize)); + }); + + // Attempt a tee immediately; it should block. + EXPECT_THAT(tee(rfd1.get(), wfd2.get(), kPageSize, 0), + SyscallSucceedsWithValue(kPageSize)); + + // Thread should be joinable. + t.Join(); + + // Content should reflect the splice, in both pipes. + std::vector rbuf(kPageSize); + ASSERT_THAT(read(rfd2.get(), rbuf.data(), rbuf.size()), + SyscallSucceedsWithValue(kPageSize)); + EXPECT_EQ(memcmp(rbuf.data(), buf.data(), kPageSize), 0); + ASSERT_THAT(read(rfd1.get(), rbuf.data(), rbuf.size()), + SyscallSucceedsWithValue(kPageSize)); + EXPECT_EQ(memcmp(rbuf.data(), buf.data(), kPageSize), 0); +} + +TEST(SpliceTest, NonBlocking) { + // Create two new pipes. + int first[2], second[2]; + ASSERT_THAT(pipe(first), SyscallSucceeds()); + const FileDescriptor rfd1(first[0]); + const FileDescriptor wfd1(first[1]); + ASSERT_THAT(pipe(second), SyscallSucceeds()); + const FileDescriptor rfd2(second[0]); + const FileDescriptor wfd2(second[1]); + + // Splice with no data to back it. + EXPECT_THAT(splice(rfd1.get(), nullptr, wfd2.get(), nullptr, kPageSize, + SPLICE_F_NONBLOCK), + SyscallFailsWithErrno(EAGAIN)); +} + +TEST(TeeTest, NonBlocking) { + SKIP_IF(IsRunningOnGvisor()); + + // Create two new pipes. + int first[2], second[2]; + ASSERT_THAT(pipe(first), SyscallSucceeds()); + const FileDescriptor rfd1(first[0]); + const FileDescriptor wfd1(first[1]); + ASSERT_THAT(pipe(second), SyscallSucceeds()); + const FileDescriptor rfd2(second[0]); + const FileDescriptor wfd2(second[1]); + + // Splice with no data to back it. + EXPECT_THAT(tee(rfd1.get(), wfd2.get(), kPageSize, SPLICE_F_NONBLOCK), + SyscallFailsWithErrno(EAGAIN)); +} + +} // namespace + +} // namespace testing +} // namespace gvisor -- cgit v1.2.3 From ae1bb08871758844fa23fc7255ffeb0392f9dee6 Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Tue, 21 May 2019 20:11:26 -0700 Subject: Clean up pipe internals and add fcntl support Pipe internals are made more efficient by avoiding garbage collection. A pool is now used that can be shared by all pipes, and buffers are chained via an intrusive list. The documentation for pipe structures and methods is also simplified and clarified. The pipe tests are now parameterized, so that they are run on all different variants (named pipes, small buffers, default buffers). The pipe buffer sizes are exposed by fcntl, which is now supported by this change. A size change test has been added to the suite. These new tests uncovered a bug regarding the semantics of open named pipes with O_NONBLOCK, which is also fixed by this CL. This fix also addresses the lack of the O_LARGEFILE flag for named pipes. PiperOrigin-RevId: 249375888 Change-Id: I48e61e9c868aedb0cadda2dff33f09a560dee773 --- pkg/abi/linux/fcntl.go | 4 +- pkg/sentry/kernel/pipe/BUILD | 8 +- pkg/sentry/kernel/pipe/buffer.go | 90 ++++ pkg/sentry/kernel/pipe/buffer_test.go | 32 ++ pkg/sentry/kernel/pipe/buffers.go | 48 -- pkg/sentry/kernel/pipe/node.go | 7 +- pkg/sentry/kernel/pipe/pipe.go | 315 ++++++++----- pkg/sentry/kernel/pipe/pipe_test.go | 7 +- pkg/sentry/kernel/pipe/reader.go | 3 + pkg/sentry/kernel/pipe/reader_writer.go | 6 +- pkg/sentry/kernel/pipe/writer.go | 3 + pkg/sentry/syscalls/linux/sys_file.go | 14 + test/syscalls/BUILD | 6 +- test/syscalls/linux/BUILD | 2 + test/syscalls/linux/pipe.cc | 761 +++++++++++++++++++------------- 15 files changed, 828 insertions(+), 478 deletions(-) create mode 100644 pkg/sentry/kernel/pipe/buffer.go create mode 100644 pkg/sentry/kernel/pipe/buffer_test.go delete mode 100644 pkg/sentry/kernel/pipe/buffers.go (limited to 'pkg/sentry/kernel') diff --git a/pkg/abi/linux/fcntl.go b/pkg/abi/linux/fcntl.go index cc8f2702d..b30350193 100644 --- a/pkg/abi/linux/fcntl.go +++ b/pkg/abi/linux/fcntl.go @@ -17,7 +17,6 @@ package linux // Comands from linux/fcntl.h. const ( F_DUPFD = 0 - F_DUPFD_CLOEXEC = 1030 F_GETFD = 1 F_GETFL = 3 F_GETOWN = 9 @@ -26,6 +25,9 @@ const ( F_SETLK = 6 F_SETLKW = 7 F_SETOWN = 8 + F_DUPFD_CLOEXEC = 1024 + 6 + F_SETPIPE_SZ = 1024 + 7 + F_GETPIPE_SZ = 1024 + 8 ) // Flags for fcntl. diff --git a/pkg/sentry/kernel/pipe/BUILD b/pkg/sentry/kernel/pipe/BUILD index 6b23117d9..b07d15a2a 100644 --- a/pkg/sentry/kernel/pipe/BUILD +++ b/pkg/sentry/kernel/pipe/BUILD @@ -10,16 +10,16 @@ go_template_instance( prefix = "buffer", template = "//pkg/ilist:generic_list", types = { - "Element": "*Buffer", - "Linker": "*Buffer", + "Element": "*buffer", + "Linker": "*buffer", }, ) go_library( name = "pipe", srcs = [ + "buffer.go", "buffer_list.go", - "buffers.go", "device.go", "node.go", "pipe.go", @@ -37,6 +37,7 @@ go_library( "//pkg/sentry/device", "//pkg/sentry/fs", "//pkg/sentry/fs/fsutil", + "//pkg/sentry/safemem", "//pkg/sentry/usermem", "//pkg/syserror", "//pkg/waiter", @@ -47,6 +48,7 @@ go_test( name = "pipe_test", size = "small", srcs = [ + "buffer_test.go", "node_test.go", "pipe_test.go", ], diff --git a/pkg/sentry/kernel/pipe/buffer.go b/pkg/sentry/kernel/pipe/buffer.go new file mode 100644 index 000000000..4360dc44f --- /dev/null +++ b/pkg/sentry/kernel/pipe/buffer.go @@ -0,0 +1,90 @@ +// Copyright 2018 The gVisor Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "sync" + + "gvisor.googlesource.com/gvisor/pkg/sentry/safemem" +) + +// buffer encapsulates a queueable byte buffer. +// +// Note that the total size is slightly less than two pages. This +// is done intentionally to ensure that the buffer object aligns +// with runtime internals. We have no hard size or alignment +// requirements. This two page size will effectively minimize +// internal fragmentation, but still have a large enough chunk +// to limit excessive segmentation. +// +// +stateify savable +type buffer struct { + data [8144]byte + read int + write int + bufferEntry +} + +// Reset resets internal data. +// +// This must be called before use. +func (b *buffer) Reset() { + b.read = 0 + b.write = 0 +} + +// Empty indicates the buffer is empty. +// +// This indicates there is no data left to read. +func (b *buffer) Empty() bool { + return b.read == b.write +} + +// Full indicates the buffer is full. +// +// This indicates there is no capacity left to write. +func (b *buffer) Full() bool { + return b.write == len(b.data) +} + +// WriteFromBlocks implements safemem.Writer.WriteFromBlocks. +func (b *buffer) WriteFromBlocks(srcs safemem.BlockSeq) (uint64, error) { + dst := safemem.BlockSeqOf(safemem.BlockFromSafeSlice(b.data[b.write:])) + n, err := safemem.CopySeq(dst, srcs) + b.write += int(n) + return n, err +} + +// ReadToBlocks implements safemem.Reader.ReadToBlocks. +func (b *buffer) ReadToBlocks(dsts safemem.BlockSeq) (uint64, error) { + src := safemem.BlockSeqOf(safemem.BlockFromSafeSlice(b.data[b.read:b.write])) + n, err := safemem.CopySeq(dsts, src) + b.read += int(n) + return n, err +} + +// bufferPool is a pool for buffers. +var bufferPool = sync.Pool{ + New: func() interface{} { + return new(buffer) + }, +} + +// newBuffer grabs a new buffer from the pool. +func newBuffer() *buffer { + b := bufferPool.Get().(*buffer) + b.Reset() + return b +} diff --git a/pkg/sentry/kernel/pipe/buffer_test.go b/pkg/sentry/kernel/pipe/buffer_test.go new file mode 100644 index 000000000..4b7dbc43f --- /dev/null +++ b/pkg/sentry/kernel/pipe/buffer_test.go @@ -0,0 +1,32 @@ +// Copyright 2019 The gVisor Authors. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package pipe + +import ( + "testing" + "unsafe" + + "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" +) + +func TestBufferSize(t *testing.T) { + bufferSize := unsafe.Sizeof(buffer{}) + if bufferSize < usermem.PageSize { + t.Errorf("buffer is less than a page") + } + if bufferSize > (2 * usermem.PageSize) { + t.Errorf("buffer is greater than two pages") + } +} diff --git a/pkg/sentry/kernel/pipe/buffers.go b/pkg/sentry/kernel/pipe/buffers.go deleted file mode 100644 index ba53fd482..000000000 --- a/pkg/sentry/kernel/pipe/buffers.go +++ /dev/null @@ -1,48 +0,0 @@ -// Copyright 2018 The gVisor Authors. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -package pipe - -// Buffer encapsulates a queueable byte buffer that can -// easily be truncated. It is designed only for use with pipes. -// -// +stateify savable -type Buffer struct { - bufferEntry - data []byte -} - -// newBuffer initializes a Buffer. -func newBuffer(buf []byte) *Buffer { - return &Buffer{data: buf} -} - -// bytes returns the bytes contained in the buffer. -func (b *Buffer) bytes() []byte { - return b.data -} - -// size returns the number of bytes contained in the buffer. -func (b *Buffer) size() int { - return len(b.data) -} - -// truncate removes the first n bytes from the buffer. -func (b *Buffer) truncate(n int) int { - if n > len(b.data) { - panic("Trying to truncate past end of array.") - } - b.data = b.data[n:] - return len(b.data) -} diff --git a/pkg/sentry/kernel/pipe/node.go b/pkg/sentry/kernel/pipe/node.go index 7c3739360..926c4c623 100644 --- a/pkg/sentry/kernel/pipe/node.go +++ b/pkg/sentry/kernel/pipe/node.go @@ -67,7 +67,6 @@ func NewInodeOperations(ctx context.Context, perms fs.FilePermissions, p *Pipe) InodeSimpleAttributes: fsutil.NewInodeSimpleAttributes(ctx, fs.FileOwnerFromContext(ctx), perms, linux.PIPEFS_MAGIC), p: p, } - } // GetFile implements fs.InodeOperations.GetFile. Named pipes have special blocking @@ -87,7 +86,7 @@ func (i *inodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags fs.Fi switch { case flags.Read && !flags.Write: // O_RDONLY. - r := i.p.ROpen(ctx) + r := i.p.Open(ctx, flags) i.newHandleLocked(&i.rWakeup) if i.p.isNamed && !flags.NonBlocking && !i.p.HasWriters() { @@ -103,7 +102,7 @@ func (i *inodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags fs.Fi return r, nil case flags.Write && !flags.Read: // O_WRONLY. - w := i.p.WOpen(ctx) + w := i.p.Open(ctx, flags) i.newHandleLocked(&i.wWakeup) if i.p.isNamed && !i.p.HasReaders() { @@ -123,7 +122,7 @@ func (i *inodeOperations) GetFile(ctx context.Context, d *fs.Dirent, flags fs.Fi case flags.Read && flags.Write: // O_RDWR. // Pipes opened for read-write always succeeds without blocking. - rw := i.p.RWOpen(ctx) + rw := i.p.Open(ctx, flags) i.newHandleLocked(&i.rWakeup) i.newHandleLocked(&i.wWakeup) return rw, nil diff --git a/pkg/sentry/kernel/pipe/pipe.go b/pkg/sentry/kernel/pipe/pipe.go index bd7649d2f..b65204492 100644 --- a/pkg/sentry/kernel/pipe/pipe.go +++ b/pkg/sentry/kernel/pipe/pipe.go @@ -12,11 +12,7 @@ // See the License for the specific language governing permissions and // limitations under the License. -// Package pipe provides an in-memory implementation of a unidirectional -// pipe. -// -// The goal of this pipe is to emulate the pipe syscall in all of its -// edge cases and guarantees of atomic IO. +// Package pipe provides a pipe implementation. package pipe import ( @@ -32,8 +28,29 @@ import ( "gvisor.googlesource.com/gvisor/pkg/waiter" ) -// DefaultPipeSize is the system-wide default size of a pipe in bytes. -const DefaultPipeSize = 65536 +const ( + // MinimumPipeSize is a hard limit of the minimum size of a pipe. + MinimumPipeSize = 64 << 10 + + // DefaultPipeSize is the system-wide default size of a pipe in bytes. + DefaultPipeSize = MinimumPipeSize + + // MaximumPipeSize is a hard limit on the maximum size of a pipe. + MaximumPipeSize = 8 << 20 +) + +// Sizer is an interface for setting and getting the size of a pipe. +// +// It is implemented by Pipe and, through embedding, all other types. +type Sizer interface { + // PipeSize returns the pipe capacity in bytes. + PipeSize() int64 + + // SetPipeSize sets the new pipe capacity in bytes. + // + // The new size is returned (which may be capped). + SetPipeSize(int64) (int64, error) +} // Pipe is an encapsulation of a platform-independent pipe. // It manages a buffered byte queue shared between a reader/writer @@ -43,49 +60,76 @@ const DefaultPipeSize = 65536 type Pipe struct { waiter.Queue `state:"nosave"` - // Whether this is a named or anonymous pipe. + // isNamed indicates whether this is a named pipe. + // + // This value is immutable. isNamed bool + // atomicIOBytes is the maximum number of bytes that the pipe will + // guarantee atomic reads or writes atomically. + // + // This value is immutable. + atomicIOBytes int64 + // The dirent backing this pipe. Shared by all readers and writers. + // + // This value is immutable. Dirent *fs.Dirent - // The buffered byte queue. - data bufferList + // The number of active readers for this pipe. + // + // Access atomically. + readers int32 - // Max size of the pipe in bytes. When this max has been reached, - // writers will get EWOULDBLOCK. - max int + // The number of active writes for this pipe. + // + // Access atomically. + writers int32 - // Current size of the pipe in bytes. - size int + // mu protects all pipe internal state below. + mu sync.Mutex `state:"nosave"` - // Max number of bytes the pipe can guarantee to read or write - // atomically. - atomicIOBytes int + // data is the buffer queue of pipe contents. + // + // This is protected by mu. + data bufferList - // The number of active readers for this pipe. Load/store atomically. - readers int32 + // max is the maximum size of the pipe in bytes. When this max has been + // reached, writers will get EWOULDBLOCK. + // + // This is protected by mu. + max int64 - // The number of active writes for this pipe. Load/store atomically. - writers int32 + // size is the current size of the pipe in bytes. + // + // This is protected by mu. + size int64 - // This flag indicates if this pipe ever had a writer. Note that this does - // not necessarily indicate there is *currently* a writer, just that there - // has been a writer at some point since the pipe was created. + // hadWriter indicates if this pipe ever had a writer. Note that this + // does not necessarily indicate there is *currently* a writer, just + // that there has been a writer at some point since the pipe was + // created. // - // Protected by mu. + // This is protected by mu. hadWriter bool - - // Lock protecting all pipe internal state. - mu sync.Mutex `state:"nosave"` } -// NewPipe initializes and returns a pipe. A pipe created by this function is -// persistent, and will remain valid even without any open fds to it. Named -// pipes for mknod(2) are created via this function. Note that the -// implementation of blocking semantics for opening the read and write ends of a -// named pipe are left to filesystems. -func NewPipe(ctx context.Context, isNamed bool, sizeBytes, atomicIOBytes int) *Pipe { +// NewPipe initializes and returns a pipe. +// +// N.B. The size and atomicIOBytes will be bounded. +func NewPipe(ctx context.Context, isNamed bool, sizeBytes, atomicIOBytes int64) *Pipe { + if sizeBytes < MinimumPipeSize { + sizeBytes = MinimumPipeSize + } + if sizeBytes > MaximumPipeSize { + sizeBytes = MaximumPipeSize + } + if atomicIOBytes <= 0 { + atomicIOBytes = 1 + } + if atomicIOBytes > sizeBytes { + atomicIOBytes = sizeBytes + } p := &Pipe{ isNamed: isNamed, max: sizeBytes, @@ -110,48 +154,45 @@ func NewPipe(ctx context.Context, isNamed bool, sizeBytes, atomicIOBytes int) *P return p } -// NewConnectedPipe initializes a pipe and returns a pair of objects (which -// implement kio.File) representing the read and write ends of the pipe. A pipe -// created by this function becomes invalid as soon as either the read or write -// end is closed, and errors on subsequent operations on either end. Pipes -// for pipe(2) and pipe2(2) are generally created this way. -func NewConnectedPipe(ctx context.Context, sizeBytes int, atomicIOBytes int) (*fs.File, *fs.File) { +// NewConnectedPipe initializes a pipe and returns a pair of objects +// representing the read and write ends of the pipe. +func NewConnectedPipe(ctx context.Context, sizeBytes, atomicIOBytes int64) (*fs.File, *fs.File) { p := NewPipe(ctx, false /* isNamed */, sizeBytes, atomicIOBytes) - return p.ROpen(ctx), p.WOpen(ctx) -} - -// ROpen opens the pipe for reading. -func (p *Pipe) ROpen(ctx context.Context) *fs.File { - p.rOpen() - return fs.NewFile(ctx, p.Dirent, fs.FileFlags{Read: true}, &Reader{ - ReaderWriter: ReaderWriter{Pipe: p}, - }) -} - -// WOpen opens the pipe for writing. -func (p *Pipe) WOpen(ctx context.Context) *fs.File { - p.wOpen() - return fs.NewFile(ctx, p.Dirent, fs.FileFlags{Write: true}, &Writer{ - ReaderWriter: ReaderWriter{Pipe: p}, - }) + return p.Open(ctx, fs.FileFlags{Read: true}), p.Open(ctx, fs.FileFlags{Write: true}) } -// RWOpen opens the pipe for both reading and writing. -func (p *Pipe) RWOpen(ctx context.Context) *fs.File { - p.rOpen() - p.wOpen() - return fs.NewFile(ctx, p.Dirent, fs.FileFlags{Read: true, Write: true}, &ReaderWriter{ - Pipe: p, - }) +// Open opens the pipe and returns a new file. +// +// Precondition: at least one of flags.Read or flags.Write must be set. +func (p *Pipe) Open(ctx context.Context, flags fs.FileFlags) *fs.File { + switch { + case flags.Read && flags.Write: + p.rOpen() + p.wOpen() + return fs.NewFile(ctx, p.Dirent, flags, &ReaderWriter{ + Pipe: p, + }) + case flags.Read: + p.rOpen() + return fs.NewFile(ctx, p.Dirent, flags, &Reader{ + ReaderWriter: ReaderWriter{Pipe: p}, + }) + case flags.Write: + p.wOpen() + return fs.NewFile(ctx, p.Dirent, flags, &Writer{ + ReaderWriter: ReaderWriter{Pipe: p}, + }) + default: + // Precondition violated. + panic("invalid pipe flags") + } } // read reads data from the pipe into dst and returns the number of bytes // read, or returns ErrWouldBlock if the pipe is empty. +// +// Precondition: this pipe must have readers. func (p *Pipe) read(ctx context.Context, dst usermem.IOSequence) (int64, error) { - if !p.HasReaders() { - return 0, syscall.EBADF - } - // Don't block for a zero-length read even if the pipe is empty. if dst.NumBytes() == 0 { return 0, nil @@ -159,8 +200,8 @@ func (p *Pipe) read(ctx context.Context, dst usermem.IOSequence) (int64, error) p.mu.Lock() defer p.mu.Unlock() - // If there is nothing to read at the moment but there is a writer, tell the - // caller to block. + + // Is the pipe empty? if p.size == 0 { if !p.HasWriters() { // There are no writers, return EOF. @@ -168,64 +209,94 @@ func (p *Pipe) read(ctx context.Context, dst usermem.IOSequence) (int64, error) } return 0, syserror.ErrWouldBlock } - var n int64 - for buffer := p.data.Front(); buffer != nil; buffer = p.data.Front() { - n0, err := dst.CopyOut(ctx, buffer.bytes()) - n += int64(n0) - p.size -= n0 - if buffer.truncate(n0) == 0 { - p.data.Remove(buffer) + + // Limit how much we consume. + if dst.NumBytes() > p.size { + dst = dst.TakeFirst64(p.size) + } + + done := int64(0) + for dst.NumBytes() > 0 { + // Pop the first buffer. + first := p.data.Front() + if first == nil { + break } - dst = dst.DropFirst(n0) - if dst.NumBytes() == 0 || err != nil { - return n, err + + // Copy user data. + n, err := dst.CopyOutFrom(ctx, first) + done += int64(n) + p.size -= n + dst = dst.DropFirst64(n) + + // Empty buffer? + if first.Empty() { + // Push to the free list. + p.data.Remove(first) + bufferPool.Put(first) + } + + // Handle errors. + if err != nil { + return done, err } } - return n, nil + + return done, nil } // write writes data from sv into the pipe and returns the number of bytes // written. If no bytes are written because the pipe is full (or has less than // atomicIOBytes free capacity), write returns ErrWouldBlock. +// +// Precondition: this pipe must have writers. func (p *Pipe) write(ctx context.Context, src usermem.IOSequence) (int64, error) { p.mu.Lock() defer p.mu.Unlock() - if !p.HasWriters() { - return 0, syscall.EBADF - } + // Can't write to a pipe with no readers. if !p.HasReaders() { return 0, syscall.EPIPE } // POSIX requires that a write smaller than atomicIOBytes (PIPE_BUF) be - // atomic, but requires no atomicity for writes larger than this. However, - // Linux appears to provide stronger semantics than this in practice: - // unmerged writes are done one PAGE_SIZE buffer at a time, so for larger - // writes, the writing of each PIPE_BUF-sized chunk is atomic. We implement - // this by writing at most atomicIOBytes at a time if we can't service the - // write in its entirety. - canWrite := src.NumBytes() - if canWrite > int64(p.max-p.size) { - if p.max-p.size >= p.atomicIOBytes { - canWrite = int64(p.atomicIOBytes) - } else { + // atomic, but requires no atomicity for writes larger than this. + wanted := src.NumBytes() + if avail := p.max - p.size; wanted > avail { + if wanted <= p.atomicIOBytes { return 0, syserror.ErrWouldBlock } + // Limit to the available capacity. + src = src.TakeFirst64(avail) } - // Copy data from user memory into a pipe-owned buffer. - buf := make([]byte, canWrite) - n, err := src.CopyIn(ctx, buf) - if n > 0 { - p.data.PushBack(newBuffer(buf[:n])) + done := int64(0) + for src.NumBytes() > 0 { + // Need a new buffer? + last := p.data.Back() + if last == nil || last.Full() { + // Add a new buffer to the data list. + last = newBuffer() + p.data.PushBack(last) + } + + // Copy user data. + n, err := src.CopyInTo(ctx, last) + done += int64(n) p.size += n + src = src.DropFirst64(n) + + // Handle errors. + if err != nil { + return done, err + } } - if int64(n) < src.NumBytes() && err == nil { + if wanted > done { // Partial write due to full pipe. - err = syserror.ErrWouldBlock + return done, syserror.ErrWouldBlock } - return int64(n), err + + return done, nil } // rOpen signals a new reader of the pipe. @@ -267,6 +338,9 @@ func (p *Pipe) HasWriters() bool { return atomic.LoadInt32(&p.writers) > 0 } +// rReadinessLocked calculates the read readiness. +// +// Precondition: mu must be held. func (p *Pipe) rReadinessLocked() waiter.EventMask { ready := waiter.EventMask(0) if p.HasReaders() && p.data.Front() != nil { @@ -290,6 +364,9 @@ func (p *Pipe) rReadiness() waiter.EventMask { return p.rReadinessLocked() } +// wReadinessLocked calculates the write readiness. +// +// Precondition: mu must be held. func (p *Pipe) wReadinessLocked() waiter.EventMask { ready := waiter.EventMask(0) if p.HasWriters() && p.size < p.max { @@ -317,8 +394,36 @@ func (p *Pipe) rwReadiness() waiter.EventMask { return p.rReadinessLocked() | p.wReadinessLocked() } -func (p *Pipe) queuedSize() int { +// queued returns the amount of queued data. +func (p *Pipe) queued() int64 { p.mu.Lock() defer p.mu.Unlock() return p.size } + +// PipeSize implements PipeSizer.PipeSize. +func (p *Pipe) PipeSize() int64 { + p.mu.Lock() + defer p.mu.Unlock() + return p.max +} + +// SetPipeSize implements PipeSize.SetPipeSize. +func (p *Pipe) SetPipeSize(size int64) (int64, error) { + if size < 0 { + return 0, syserror.EINVAL + } + if size < MinimumPipeSize { + size = MinimumPipeSize // Per spec. + } + if size > MaximumPipeSize { + return 0, syserror.EPERM + } + p.mu.Lock() + defer p.mu.Unlock() + if size < p.size { + return 0, syserror.EBUSY + } + p.max = size + return size, nil +} diff --git a/pkg/sentry/kernel/pipe/pipe_test.go b/pkg/sentry/kernel/pipe/pipe_test.go index de340c40c..298c6587b 100644 --- a/pkg/sentry/kernel/pipe/pipe_test.go +++ b/pkg/sentry/kernel/pipe/pipe_test.go @@ -58,15 +58,16 @@ func TestPipeReadBlock(t *testing.T) { func TestPipeWriteBlock(t *testing.T) { const atomicIOBytes = 2 + const capacity = MinimumPipeSize ctx := contexttest.Context(t) - r, w := NewConnectedPipe(ctx, 10, atomicIOBytes) + r, w := NewConnectedPipe(ctx, capacity, atomicIOBytes) defer r.DecRef() defer w.DecRef() - msg := []byte("here's some bytes") + msg := make([]byte, capacity+1) n, err := w.Writev(ctx, usermem.BytesIOSequence(msg)) - if wantN, wantErr := int64(atomicIOBytes), syserror.ErrWouldBlock; n != wantN || err != wantErr { + if wantN, wantErr := int64(capacity), syserror.ErrWouldBlock; n != wantN || err != wantErr { t.Fatalf("Writev: got (%d, %v), wanted (%d, %v)", n, err, wantN, wantErr) } } diff --git a/pkg/sentry/kernel/pipe/reader.go b/pkg/sentry/kernel/pipe/reader.go index 48fab45d1..656be824d 100644 --- a/pkg/sentry/kernel/pipe/reader.go +++ b/pkg/sentry/kernel/pipe/reader.go @@ -27,8 +27,11 @@ type Reader struct { } // Release implements fs.FileOperations.Release. +// +// This overrides ReaderWriter.Release. func (r *Reader) Release() { r.Pipe.rClose() + // Wake up writers. r.Pipe.Notify(waiter.EventOut) } diff --git a/pkg/sentry/kernel/pipe/reader_writer.go b/pkg/sentry/kernel/pipe/reader_writer.go index 59899be49..e560b9be9 100644 --- a/pkg/sentry/kernel/pipe/reader_writer.go +++ b/pkg/sentry/kernel/pipe/reader_writer.go @@ -15,7 +15,6 @@ package pipe import ( - "fmt" "math" "syscall" @@ -49,6 +48,7 @@ type ReaderWriter struct { func (rw *ReaderWriter) Release() { rw.Pipe.rClose() rw.Pipe.wClose() + // Wake up readers and writers. rw.Pipe.Notify(waiter.EventIn | waiter.EventOut) } @@ -81,9 +81,9 @@ func (rw *ReaderWriter) Ioctl(ctx context.Context, io usermem.IO, args arch.Sysc // Switch on ioctl request. switch int(args[1].Int()) { case linux.FIONREAD: - v := rw.queuedSize() + v := rw.queued() if v > math.MaxInt32 { - panic(fmt.Sprintf("Impossibly large pipe queued size: %d", v)) + v = math.MaxInt32 // Silently truncate. } // Copy result to user-space. _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), int32(v), usermem.IOOpts{ diff --git a/pkg/sentry/kernel/pipe/writer.go b/pkg/sentry/kernel/pipe/writer.go index 0f29fbc43..8d5b68541 100644 --- a/pkg/sentry/kernel/pipe/writer.go +++ b/pkg/sentry/kernel/pipe/writer.go @@ -27,8 +27,11 @@ type Writer struct { } // Release implements fs.FileOperations.Release. +// +// This overrides ReaderWriter.Release. func (w *Writer) Release() { w.Pipe.wClose() + // Wake up readers. w.Pipe.Notify(waiter.EventHUp) } diff --git a/pkg/sentry/syscalls/linux/sys_file.go b/pkg/sentry/syscalls/linux/sys_file.go index 8a80cd430..19f579930 100644 --- a/pkg/sentry/syscalls/linux/sys_file.go +++ b/pkg/sentry/syscalls/linux/sys_file.go @@ -27,6 +27,7 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/auth" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/fasync" "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/kdefs" + "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/pipe" ktime "gvisor.googlesource.com/gvisor/pkg/sentry/kernel/time" "gvisor.googlesource.com/gvisor/pkg/sentry/limits" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" @@ -943,6 +944,19 @@ func Fcntl(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall } err := tmpfs.AddSeals(file.Dirent.Inode, args[2].Uint()) return 0, nil, err + case linux.F_GETPIPE_SZ: + sz, ok := file.FileOperations.(pipe.Sizer) + if !ok { + return 0, nil, syserror.EINVAL + } + return uintptr(sz.PipeSize()), nil, nil + case linux.F_SETPIPE_SZ: + sz, ok := file.FileOperations.(pipe.Sizer) + if !ok { + return 0, nil, syserror.EINVAL + } + n, err := sz.SetPipeSize(int64(args[2].Int())) + return uintptr(n), nil, err default: // Everything else is not yet supported. return 0, nil, syserror.EINVAL diff --git a/test/syscalls/BUILD b/test/syscalls/BUILD index b531d7629..0d6b6ccc7 100644 --- a/test/syscalls/BUILD +++ b/test/syscalls/BUILD @@ -191,7 +191,11 @@ syscall_test(test = "//test/syscalls/linux:partial_bad_buffer_test") syscall_test(test = "//test/syscalls/linux:pause_test") -syscall_test(test = "//test/syscalls/linux:pipe_test") +syscall_test( + size = "large", + shard_count = 5, + test = "//test/syscalls/linux:pipe_test", +) syscall_test(test = "//test/syscalls/linux:poll_test") diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index d4e49bb3a..4e239617b 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -1237,6 +1237,8 @@ cc_binary( linkstatic = 1, deps = [ "//test/util:file_descriptor", + "//test/util:posix_error", + "//test/util:temp_path", "//test/util:test_main", "//test/util:test_util", "//test/util:thread_util", diff --git a/test/syscalls/linux/pipe.cc b/test/syscalls/linux/pipe.cc index 8698295b3..bce351e08 100644 --- a/test/syscalls/linux/pipe.cc +++ b/test/syscalls/linux/pipe.cc @@ -26,6 +26,8 @@ #include "absl/time/clock.h" #include "absl/time/time.h" #include "test/util/file_descriptor.h" +#include "test/util/posix_error.h" +#include "test/util/temp_path.h" #include "test/util/test_util.h" #include "test/util/thread_util.h" @@ -34,449 +36,588 @@ namespace testing { namespace { -// Buffer size of a pipe. -// -// TODO(b/35762278): Get this from F_GETPIPE_SZ. -constexpr int kPipeSize = 65536; +// Used as a non-zero sentinel value, below. +constexpr int kTestValue = 0x12345678; + +// Used for synchronization in race tests. +const absl::Duration syncDelay = absl::Seconds(2); + +struct PipeCreator { + std::string name_; + + // void (fds, is_blocking, is_namedpipe). + std::function create_; +}; + +class PipeTest : public ::testing::TestWithParam { + protected: + FileDescriptor rfd; + FileDescriptor wfd; -class PipeTest : public ::testing::Test { public: static void SetUpTestCase() { // Tests intentionally generate SIGPIPE. TEST_PCHECK(signal(SIGPIPE, SIG_IGN) != SIG_ERR); } + // Initializes rfd and wfd as a blocking pipe. + // + // The return value indicates success: the test should be skipped otherwise. + bool CreateBlocking() { return create(true); } + + // Initializes rfd and wfd as a non-blocking pipe. + // + // The return value is per CreateBlocking. + bool CreateNonBlocking() { return create(false); } + + // Returns true iff the pipe represents a named pipe. + bool IsNamedPipe() { return namedpipe_; } + + int Size() { + int s1 = fcntl(rfd.get(), F_GETPIPE_SZ); + int s2 = fcntl(wfd.get(), F_GETPIPE_SZ); + EXPECT_GT(s1, 0); + EXPECT_GT(s2, 0); + EXPECT_EQ(s1, s2); + return s1; + } + static void TearDownTestCase() { TEST_PCHECK(signal(SIGPIPE, SIG_DFL) != SIG_ERR); } + + private: + bool namedpipe_ = false; + + bool create(bool wants_blocking) { + // Generate the pipe. + int fds[2] = {-1, -1}; + bool is_blocking = false; + GetParam().create_(fds, &is_blocking, &namedpipe_); + if (fds[0] < 0 || fds[1] < 0) { + return false; + } + + // Save descriptors. + rfd.reset(fds[0]); + wfd.reset(fds[1]); + + // Adjust blocking, if needed. + if (!is_blocking && wants_blocking) { + // Clear the blocking flag. + EXPECT_THAT(fcntl(fds[0], F_SETFL, 0), SyscallSucceeds()); + EXPECT_THAT(fcntl(fds[1], F_SETFL, 0), SyscallSucceeds()); + } else if (is_blocking && !wants_blocking) { + // Set the descriptors to blocking. + EXPECT_THAT(fcntl(fds[0], F_SETFL, O_NONBLOCK), SyscallSucceeds()); + EXPECT_THAT(fcntl(fds[1], F_SETFL, O_NONBLOCK), SyscallSucceeds()); + } + + return true; + } }; -TEST_F(PipeTest, Basic) { - // fds[0] is read end, fds[1] is write end. - int fds[2]; - int i = 0x12345678; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); +TEST_P(PipeTest, Inode) { + SKIP_IF(!CreateBlocking()); // Ensure that the inode number is the same for each end. struct stat rst; - ASSERT_THAT(fstat(fds[0], &rst), SyscallSucceeds()); + ASSERT_THAT(fstat(rfd.get(), &rst), SyscallSucceeds()); struct stat wst; - ASSERT_THAT(fstat(fds[1], &wst), SyscallSucceeds()); + ASSERT_THAT(fstat(wfd.get(), &wst), SyscallSucceeds()); EXPECT_EQ(rst.st_ino, wst.st_ino); - - ASSERT_THAT(write(fds[0], &i, sizeof(i)), SyscallFailsWithErrno(EBADF)); - ASSERT_THAT(read(fds[1], &i, sizeof(i)), SyscallFailsWithErrno(EBADF)); - - ASSERT_THAT(write(fds[1], &i, sizeof(i)), - SyscallSucceedsWithValue(sizeof(i))); - int j; - ASSERT_THAT(read(fds[0], &j, sizeof(j)), SyscallSucceedsWithValue(sizeof(j))); - EXPECT_EQ(i, j); - - ASSERT_THAT(fcntl(fds[0], F_GETFL), SyscallSucceeds()); - ASSERT_THAT(fcntl(fds[1], F_GETFL), SyscallSucceedsWithValue(O_WRONLY)); - - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); } -TEST_F(PipeTest, BasicCloExec) { - // fds[0] is read end, fds[1] is write end. - int fds[2]; - int i = 0x12345678; - ASSERT_THAT(pipe2(fds, O_CLOEXEC), SyscallSucceeds()); +TEST_P(PipeTest, Permissions) { + SKIP_IF(!CreateBlocking()); - ASSERT_THAT(write(fds[0], &i, sizeof(i)), SyscallFailsWithErrno(EBADF)); - ASSERT_THAT(read(fds[1], &i, sizeof(i)), SyscallFailsWithErrno(EBADF)); + // Attempt bad operations. + int buf = kTestValue; + ASSERT_THAT(write(rfd.get(), &buf, sizeof(buf)), + SyscallFailsWithErrno(EBADF)); + EXPECT_THAT(read(wfd.get(), &buf, sizeof(buf)), SyscallFailsWithErrno(EBADF)); +} - ASSERT_THAT(write(fds[1], &i, sizeof(i)), - SyscallSucceedsWithValue(sizeof(i))); - int j; - ASSERT_THAT(read(fds[0], &j, sizeof(j)), SyscallSucceedsWithValue(sizeof(j))); - EXPECT_EQ(i, j); +TEST_P(PipeTest, Flags) { + SKIP_IF(!CreateBlocking()); + + if (IsNamedPipe()) { + // May be stubbed to zero; define locally. + constexpr int kLargefile = 0100000; + EXPECT_THAT(fcntl(rfd.get(), F_GETFL), + SyscallSucceedsWithValue(kLargefile | O_RDONLY)); + EXPECT_THAT(fcntl(wfd.get(), F_GETFL), + SyscallSucceedsWithValue(kLargefile | O_WRONLY)); + } else { + EXPECT_THAT(fcntl(rfd.get(), F_GETFL), SyscallSucceedsWithValue(O_RDONLY)); + EXPECT_THAT(fcntl(wfd.get(), F_GETFL), SyscallSucceedsWithValue(O_WRONLY)); + } +} - ASSERT_THAT(fcntl(fds[0], F_GETFL), SyscallSucceeds()); - ASSERT_THAT(fcntl(fds[1], F_GETFL), SyscallSucceeds()); +TEST_P(PipeTest, Write) { + SKIP_IF(!CreateBlocking()); - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); + int wbuf = kTestValue; + int rbuf = ~kTestValue; + ASSERT_THAT(write(wfd.get(), &wbuf, sizeof(wbuf)), + SyscallSucceedsWithValue(sizeof(wbuf))); + ASSERT_THAT(read(rfd.get(), &rbuf, sizeof(rbuf)), + SyscallSucceedsWithValue(sizeof(rbuf))); + EXPECT_EQ(wbuf, rbuf); } -TEST_F(PipeTest, BasicNoBlock) { - // fds[0] is read end, fds[1] is write end. - int fds[2]; - int i = 0x12345678; - ASSERT_THAT(pipe2(fds, O_NONBLOCK), SyscallSucceeds()); - - ASSERT_THAT(write(fds[0], &i, sizeof(i)), SyscallFailsWithErrno(EBADF)); - ASSERT_THAT(read(fds[1], &i, sizeof(i)), SyscallFailsWithErrno(EBADF)); - - ASSERT_THAT(read(fds[0], &i, sizeof(i)), SyscallFailsWithErrno(EWOULDBLOCK)); - ASSERT_THAT(write(fds[1], &i, sizeof(i)), - SyscallSucceedsWithValue(sizeof(i))); - int j; - ASSERT_THAT(read(fds[0], &j, sizeof(j)), SyscallSucceedsWithValue(sizeof(j))); - EXPECT_EQ(i, j); - ASSERT_THAT(read(fds[0], &i, sizeof(i)), SyscallFailsWithErrno(EWOULDBLOCK)); - - ASSERT_THAT(fcntl(fds[0], F_GETFL), SyscallSucceedsWithValue(O_NONBLOCK)); - ASSERT_THAT(fcntl(fds[1], F_GETFL), - SyscallSucceedsWithValue(O_NONBLOCK | O_WRONLY)); - - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); +TEST_P(PipeTest, NonBlocking) { + SKIP_IF(!CreateNonBlocking()); + + int wbuf = kTestValue; + int rbuf = ~kTestValue; + EXPECT_THAT(read(rfd.get(), &rbuf, sizeof(rbuf)), + SyscallFailsWithErrno(EWOULDBLOCK)); + ASSERT_THAT(write(wfd.get(), &wbuf, sizeof(wbuf)), + SyscallSucceedsWithValue(sizeof(wbuf))); + + ASSERT_THAT(read(rfd.get(), &rbuf, sizeof(rbuf)), + SyscallSucceedsWithValue(sizeof(rbuf))); + EXPECT_EQ(wbuf, rbuf); + EXPECT_THAT(read(rfd.get(), &rbuf, sizeof(rbuf)), + SyscallFailsWithErrno(EWOULDBLOCK)); } -TEST_F(PipeTest, BasicBothOptions) { - // fds[0] is read end, fds[1] is write end. +TEST(Pipe2Test, CloExec) { int fds[2]; - int i = 0x12345678; - ASSERT_THAT(pipe2(fds, O_NONBLOCK | O_CLOEXEC), SyscallSucceeds()); - - ASSERT_THAT(write(fds[0], &i, sizeof(i)), SyscallFailsWithErrno(EBADF)); - ASSERT_THAT(read(fds[1], &i, sizeof(i)), SyscallFailsWithErrno(EBADF)); - - ASSERT_THAT(read(fds[0], &i, sizeof(i)), SyscallFailsWithErrno(EWOULDBLOCK)); - ASSERT_THAT(write(fds[1], &i, sizeof(i)), - SyscallSucceedsWithValue(sizeof(i))); - int j; - ASSERT_THAT(read(fds[0], &j, sizeof(j)), SyscallSucceedsWithValue(sizeof(j))); - EXPECT_EQ(i, j); - ASSERT_THAT(read(fds[0], &i, sizeof(i)), SyscallFailsWithErrno(EWOULDBLOCK)); - - ASSERT_THAT(fcntl(fds[0], F_GETFL), SyscallSucceedsWithValue(O_NONBLOCK)); - ASSERT_THAT(fcntl(fds[1], F_GETFL), - SyscallSucceedsWithValue(O_NONBLOCK | O_WRONLY)); - - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); + ASSERT_THAT(pipe2(fds, O_CLOEXEC), SyscallSucceeds()); + EXPECT_THAT(fcntl(fds[0], F_GETFD), SyscallSucceedsWithValue(FD_CLOEXEC)); + EXPECT_THAT(fcntl(fds[1], F_GETFD), SyscallSucceedsWithValue(FD_CLOEXEC)); + EXPECT_THAT(close(fds[0]), SyscallSucceeds()); + EXPECT_THAT(close(fds[1]), SyscallSucceeds()); } -TEST_F(PipeTest, BasicBadOptions) { +TEST(Pipe2Test, BadOptions) { int fds[2]; - ASSERT_THAT(pipe2(fds, 0xDEAD), SyscallFailsWithErrno(EINVAL)); + EXPECT_THAT(pipe2(fds, 0xDEAD), SyscallFailsWithErrno(EINVAL)); } -TEST_F(PipeTest, Seek) { - // fds[0] is read end, fds[1] is write end. - int fds[2]; - int i = 0x12345678; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - - ASSERT_THAT(lseek(fds[0], 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[0], 0, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[0], 4, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 0, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 4, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); - - ASSERT_THAT(lseek(fds[0], 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[0], 4, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 4, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - - ASSERT_THAT(write(fds[1], &i, sizeof(i)), - SyscallSucceedsWithValue(sizeof(i))); - int j; - - ASSERT_THAT(lseek(fds[0], 0, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[0], 4, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 0, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 4, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); - - ASSERT_THAT(lseek(fds[0], 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[0], 4, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - ASSERT_THAT(lseek(fds[1], 4, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); - - ASSERT_THAT(read(fds[0], &j, sizeof(j)), SyscallSucceedsWithValue(sizeof(j))); - EXPECT_EQ(i, j); - - ASSERT_THAT(fcntl(fds[0], F_GETFL), SyscallSucceeds()); - ASSERT_THAT(fcntl(fds[1], F_GETFL), SyscallSucceedsWithValue(O_WRONLY)); - - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); +TEST_P(PipeTest, Seek) { + SKIP_IF(!CreateBlocking()); + + for (int i = 0; i < 4; i++) { + // Attempt absolute seeks. + EXPECT_THAT(lseek(rfd.get(), 0, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(rfd.get(), 4, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(wfd.get(), 0, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(wfd.get(), 4, SEEK_SET), SyscallFailsWithErrno(ESPIPE)); + + // Attempt relative seeks. + EXPECT_THAT(lseek(rfd.get(), 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(rfd.get(), 4, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(wfd.get(), 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(wfd.get(), 4, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); + + // Attempt end-of-file seeks. + EXPECT_THAT(lseek(rfd.get(), 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(rfd.get(), -4, SEEK_END), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(wfd.get(), 0, SEEK_CUR), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(lseek(wfd.get(), -4, SEEK_END), SyscallFailsWithErrno(ESPIPE)); + + // Add some more data to the pipe. + int buf = kTestValue; + ASSERT_THAT(write(wfd.get(), &buf, sizeof(buf)), + SyscallSucceedsWithValue(sizeof(buf))); + } } -TEST_F(PipeTest, AbsoluteOffsetSyscallsFail) { - // Syscalls for IO at absolute offsets fail because pipes are not seekable. - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - - std::vector buf(4096); - struct iovec iov; +TEST_P(PipeTest, OffsetCalls) { + SKIP_IF(!CreateBlocking()); - EXPECT_THAT(pread(fds[1], buf.data(), buf.size(), 0), + int buf; + EXPECT_THAT(pread(wfd.get(), &buf, sizeof(buf), 0), SyscallFailsWithErrno(ESPIPE)); - EXPECT_THAT(pwrite(fds[0], buf.data(), buf.size(), 0), + EXPECT_THAT(pwrite(rfd.get(), &buf, sizeof(buf), 0), SyscallFailsWithErrno(ESPIPE)); - EXPECT_THAT(preadv(fds[1], &iov, 1, 0), SyscallFailsWithErrno(ESPIPE)); - EXPECT_THAT(pwritev(fds[0], &iov, 1, 0), SyscallFailsWithErrno(ESPIPE)); - EXPECT_THAT(close(fds[0]), SyscallSucceeds()); - EXPECT_THAT(close(fds[1]), SyscallSucceeds()); + struct iovec iov; + EXPECT_THAT(preadv(wfd.get(), &iov, 1, 0), SyscallFailsWithErrno(ESPIPE)); + EXPECT_THAT(pwritev(rfd.get(), &iov, 1, 0), SyscallFailsWithErrno(ESPIPE)); } -TEST_F(PipeTest, WriterSideCloses) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - int rfd = fds[0]; - int i = 123; - ScopedThread t([rfd]() { - int j; - ASSERT_THAT(read(rfd, &j, sizeof(j)), SyscallSucceedsWithValue(sizeof(j))); +TEST_P(PipeTest, WriterSideCloses) { + SKIP_IF(!CreateBlocking()); + + ScopedThread t([this]() { + int buf = ~kTestValue; + ASSERT_THAT(read(rfd.get(), &buf, sizeof(buf)), + SyscallSucceedsWithValue(sizeof(buf))); + EXPECT_EQ(buf, kTestValue); // This will return when the close() completes. - ASSERT_THAT(read(rfd, &j, sizeof(j)), SyscallSucceeds()); + ASSERT_THAT(read(rfd.get(), &buf, sizeof(buf)), SyscallSucceeds()); // This will return straight away. - ASSERT_THAT(read(rfd, &j, sizeof(j)), SyscallSucceeds()); + ASSERT_THAT(read(rfd.get(), &buf, sizeof(buf)), + SyscallSucceedsWithValue(0)); }); + // Sleep a bit so the thread can block. - absl::SleepFor(absl::Seconds(1.0)); - ASSERT_THAT(write(fds[1], &i, sizeof(i)), - SyscallSucceedsWithValue(sizeof(i))); + absl::SleepFor(syncDelay); + + // Write to unblock. + int buf = kTestValue; + ASSERT_THAT(write(wfd.get(), &buf, sizeof(buf)), + SyscallSucceedsWithValue(sizeof(buf))); + // Sleep a bit so the thread can block again. - absl::SleepFor(absl::Seconds(3.0)); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); - t.Join(); + absl::SleepFor(syncDelay); - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); + // Allow the thread to complete. + ASSERT_THAT(close(wfd.release()), SyscallSucceeds()); + t.Join(); } -TEST_F(PipeTest, WriterSideClosesReadDataFirst) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - int i = 123; - ASSERT_THAT(write(fds[1], &i, sizeof(i)), - SyscallSucceedsWithValue(sizeof(i))); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); - int j; - ASSERT_THAT(read(fds[0], &j, sizeof(j)), SyscallSucceedsWithValue(sizeof(j))); - ASSERT_EQ(j, i); - ASSERT_THAT(read(fds[0], &j, sizeof(j)), SyscallSucceeds()); - - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); +TEST_P(PipeTest, WriterSideClosesReadDataFirst) { + SKIP_IF(!CreateBlocking()); + + int wbuf = kTestValue; + ASSERT_THAT(write(wfd.get(), &wbuf, sizeof(wbuf)), + SyscallSucceedsWithValue(sizeof(wbuf))); + ASSERT_THAT(close(wfd.release()), SyscallSucceeds()); + + int rbuf; + ASSERT_THAT(read(rfd.get(), &rbuf, sizeof(rbuf)), + SyscallSucceedsWithValue(sizeof(rbuf))); + EXPECT_EQ(wbuf, rbuf); + EXPECT_THAT(read(rfd.get(), &rbuf, sizeof(rbuf)), + SyscallSucceedsWithValue(0)); } -TEST_F(PipeTest, ReaderSideCloses) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - int i = 123; - ASSERT_THAT(write(fds[1], &i, sizeof(i)), SyscallFailsWithErrno(EPIPE)); +TEST_P(PipeTest, ReaderSideCloses) { + SKIP_IF(!CreateBlocking()); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); + ASSERT_THAT(close(rfd.release()), SyscallSucceeds()); + int buf = kTestValue; + EXPECT_THAT(write(wfd.get(), &buf, sizeof(buf)), + SyscallFailsWithErrno(EPIPE)); } -TEST_F(PipeTest, CloseTwice) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); - ASSERT_THAT(close(fds[0]), SyscallFailsWithErrno(EBADF)); - ASSERT_THAT(close(fds[1]), SyscallFailsWithErrno(EBADF)); - - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - ASSERT_THAT(close(fds[0]), SyscallFailsWithErrno(EBADF)); - ASSERT_THAT(close(fds[1]), SyscallFailsWithErrno(EBADF)); +TEST_P(PipeTest, CloseTwice) { + SKIP_IF(!CreateBlocking()); + + int _rfd = rfd.release(); + int _wfd = wfd.release(); + ASSERT_THAT(close(_rfd), SyscallSucceeds()); + ASSERT_THAT(close(_wfd), SyscallSucceeds()); + EXPECT_THAT(close(_rfd), SyscallFailsWithErrno(EBADF)); + EXPECT_THAT(close(_wfd), SyscallFailsWithErrno(EBADF)); } // Blocking write returns EPIPE when read end is closed if nothing has been // written. -TEST_F(PipeTest, BlockWriteClosed) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - int wfd = fds[1]; +TEST_P(PipeTest, BlockWriteClosed) { + SKIP_IF(!CreateBlocking()); absl::Notification notify; - ScopedThread t([wfd, ¬ify]() { - std::vector buf(kPipeSize); + ScopedThread t([this, ¬ify]() { + std::vector buf(Size()); // Exactly fill the pipe buffer. - ASSERT_THAT(WriteFd(wfd, buf.data(), buf.size()), + ASSERT_THAT(WriteFd(wfd.get(), buf.data(), buf.size()), SyscallSucceedsWithValue(buf.size())); notify.Notify(); // Attempt to write one more byte. Blocks. // N.B. Don't use WriteFd, we don't want a retry. - ASSERT_THAT(write(wfd, buf.data(), 1), SyscallFailsWithErrno(EPIPE)); + EXPECT_THAT(write(wfd.get(), buf.data(), 1), SyscallFailsWithErrno(EPIPE)); }); notify.WaitForNotification(); - absl::SleepFor(absl::Seconds(1.0)); - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - + ASSERT_THAT(close(rfd.release()), SyscallSucceeds()); t.Join(); - - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); } // Blocking write returns EPIPE when read end is closed even if something has // been written. -// -// FIXME(b/35924046): Pipe writes blocking early allows S/R to interrupt the -// write(2) call before the buffer is full. Then the next call will will return -// non-zero instead of EPIPE. -TEST_F(PipeTest, BlockPartialWriteClosed_NoRandomSave) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - int wfd = fds[1]; +TEST_P(PipeTest, BlockPartialWriteClosed) { + SKIP_IF(!CreateBlocking()); - ScopedThread t([wfd]() { - std::vector buf(2 * kPipeSize); + ScopedThread t([this]() { + std::vector buf(2 * Size()); // Write more than fits in the buffer. Blocks then returns partial write // when the other end is closed. The next call returns EPIPE. - if (IsRunningOnGvisor()) { - // FIXME(b/35924046): Pipe writes block early on gVisor, resulting in a - // shorter than expected partial write. - ASSERT_THAT(write(wfd, buf.data(), buf.size()), - SyscallSucceedsWithValue(::testing::Gt(0))); - } else { - ASSERT_THAT(write(wfd, buf.data(), buf.size()), - SyscallSucceedsWithValue(kPipeSize)); - } - ASSERT_THAT(write(wfd, buf.data(), buf.size()), + ASSERT_THAT(write(wfd.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(Size())); + EXPECT_THAT(write(wfd.get(), buf.data(), buf.size()), SyscallFailsWithErrno(EPIPE)); }); // Leave time for write to become blocked. - absl::SleepFor(absl::Seconds(1.0)); - - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); + absl::SleepFor(syncDelay); + // Unblock the above. + ASSERT_THAT(close(rfd.release()), SyscallSucceeds()); t.Join(); - - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); } -TEST_F(PipeTest, ReadFromClosedFd_NoRandomSave) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - int rfd = fds[0]; +TEST_P(PipeTest, ReadFromClosedFd_NoRandomSave) { + SKIP_IF(!CreateBlocking()); + absl::Notification notify; - ScopedThread t([rfd, ¬ify]() { - int f; + ScopedThread t([this, ¬ify]() { notify.Notify(); - ASSERT_THAT(read(rfd, &f, sizeof(f)), SyscallSucceedsWithValue(sizeof(f))); - ASSERT_EQ(123, f); + int buf; + ASSERT_THAT(read(rfd.get(), &buf, sizeof(buf)), + SyscallSucceedsWithValue(sizeof(buf))); + ASSERT_EQ(kTestValue, buf); }); notify.WaitForNotification(); + // Make sure that the thread gets to read(). - absl::SleepFor(absl::Seconds(5.0)); + absl::SleepFor(syncDelay); + { // We cannot save/restore here as the read end of pipe is closed but there // is ongoing read() above. We will not be able to restart the read() // successfully in restore run since the read fd is closed. const DisableSave ds; - ASSERT_THAT(close(fds[0]), SyscallSucceeds()); - int i = 123; - ASSERT_THAT(write(fds[1], &i, sizeof(i)), - SyscallSucceedsWithValue(sizeof(i))); + ASSERT_THAT(close(rfd.release()), SyscallSucceeds()); + int buf = kTestValue; + ASSERT_THAT(write(wfd.get(), &buf, sizeof(buf)), + SyscallSucceedsWithValue(sizeof(buf))); t.Join(); } - ASSERT_THAT(close(fds[1]), SyscallSucceeds()); } -TEST_F(PipeTest, FionRead) { - // fds[0] is read end, fds[1] is write end. - int fds[2]; - int data[2] = {0x12345678, 0x9101112}; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); +TEST_P(PipeTest, FionRead) { + SKIP_IF(!CreateBlocking()); - int n = -1; - EXPECT_THAT(ioctl(fds[0], FIONREAD, &n), SyscallSucceedsWithValue(0)); + int n; + ASSERT_THAT(ioctl(rfd.get(), FIONREAD, &n), SyscallSucceedsWithValue(0)); EXPECT_EQ(n, 0); - n = -1; - EXPECT_THAT(ioctl(fds[1], FIONREAD, &n), SyscallSucceedsWithValue(0)); + ASSERT_THAT(ioctl(wfd.get(), FIONREAD, &n), SyscallSucceedsWithValue(0)); EXPECT_EQ(n, 0); - EXPECT_THAT(write(fds[1], data, sizeof(data)), - SyscallSucceedsWithValue(sizeof(data))); + std::vector buf(Size()); + ASSERT_THAT(write(wfd.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(buf.size())); - n = -1; - EXPECT_THAT(ioctl(fds[0], FIONREAD, &n), SyscallSucceedsWithValue(0)); - EXPECT_EQ(n, sizeof(data)); - n = -1; - EXPECT_THAT(ioctl(fds[1], FIONREAD, &n), SyscallSucceedsWithValue(0)); - EXPECT_EQ(n, sizeof(data)); + EXPECT_THAT(ioctl(rfd.get(), FIONREAD, &n), SyscallSucceedsWithValue(0)); + EXPECT_EQ(n, buf.size()); + EXPECT_THAT(ioctl(wfd.get(), FIONREAD, &n), SyscallSucceedsWithValue(0)); + EXPECT_EQ(n, buf.size()); } // Test that opening an empty anonymous pipe RDONLY via /proc/self/fd/N does not // block waiting for a writer. -TEST_F(PipeTest, OpenViaProcSelfFD) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - FileDescriptor rfd(fds[0]); - FileDescriptor wfd(fds[1]); +TEST_P(PipeTest, OpenViaProcSelfFD) { + SKIP_IF(!CreateBlocking()); + SKIP_IF(IsNamedPipe()); // Close the write end of the pipe. - wfd.release(); + ASSERT_THAT(close(wfd.release()), SyscallSucceeds()); // Open other side via /proc/self/fd. It should not block. FileDescriptor proc_self_fd = ASSERT_NO_ERRNO_AND_VALUE( - Open(absl::StrCat("/proc/self/fd/", fds[0]), O_RDONLY)); + Open(absl::StrCat("/proc/self/fd/", rfd.get()), O_RDONLY)); } // Test that opening and reading from an anonymous pipe (with existing writes) // RDONLY via /proc/self/fd/N returns the existing data. -TEST_F(PipeTest, OpenViaProcSelfFDWithWrites) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - FileDescriptor rfd(fds[0]); - FileDescriptor wfd(fds[1]); +TEST_P(PipeTest, OpenViaProcSelfFDWithWrites) { + SKIP_IF(!CreateBlocking()); + SKIP_IF(IsNamedPipe()); // Write to the pipe and then close the write fd. - char data = 'x'; - ASSERT_THAT(write(fds[1], &data, 1), SyscallSucceedsWithValue(1)); - wfd.release(); + int wbuf = kTestValue; + ASSERT_THAT(write(wfd.get(), &wbuf, sizeof(wbuf)), + SyscallSucceedsWithValue(sizeof(wbuf))); + ASSERT_THAT(close(wfd.release()), SyscallSucceeds()); // Open read side via /proc/self/fd, and read from it. FileDescriptor proc_self_fd = ASSERT_NO_ERRNO_AND_VALUE( - Open(absl::StrCat("/proc/self/fd/", fds[0]), O_RDONLY)); - char got; - ASSERT_THAT(read(proc_self_fd.get(), &got, 1), SyscallSucceedsWithValue(1)); + Open(absl::StrCat("/proc/self/fd/", rfd.get()), O_RDONLY)); + int rbuf; + ASSERT_THAT(read(proc_self_fd.get(), &rbuf, sizeof(rbuf)), + SyscallSucceedsWithValue(sizeof(rbuf))); + EXPECT_EQ(wbuf, rbuf); +} + +// Test that accesses of /proc//fd correctly decrement the refcount. +TEST_P(PipeTest, ProcFDReleasesFile) { + SKIP_IF(!CreateBlocking()); - // We should get what we sent. - EXPECT_EQ(got, data); + // Stat the pipe FD, which shouldn't alter the refcount. + struct stat wst; + ASSERT_THAT(lstat(absl::StrCat("/proc/self/fd/", wfd.get()).c_str(), &wst), + SyscallSucceeds()); + + // Close the write end and ensure that read indicates EOF. + wfd.reset(); + char buf; + ASSERT_THAT(read(rfd.get(), &buf, 1), SyscallSucceedsWithValue(0)); } -TEST_F(PipeTest, LargeFile) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - FileDescriptor rfd(fds[0]); - FileDescriptor wfd(fds[1]); +// Same for /proc//fdinfo. +TEST_P(PipeTest, ProcFDInfoReleasesFile) { + SKIP_IF(!CreateBlocking()); + + // Stat the pipe FD, which shouldn't alter the refcount. + struct stat wst; + ASSERT_THAT( + lstat(absl::StrCat("/proc/self/fdinfo/", wfd.get()).c_str(), &wst), + SyscallSucceeds()); + + // Close the write end and ensure that read indicates EOF. + wfd.reset(); + char buf; + ASSERT_THAT(read(rfd.get(), &buf, 1), SyscallSucceedsWithValue(0)); +} + +TEST_P(PipeTest, SizeChange) { + SKIP_IF(!CreateBlocking()); + + // Set the minimum possible size. + ASSERT_THAT(fcntl(rfd.get(), F_SETPIPE_SZ, 0), SyscallSucceeds()); + int min = Size(); + EXPECT_GT(min, 0); // Should be rounded up. + + // Set from the read end. + ASSERT_THAT(fcntl(rfd.get(), F_SETPIPE_SZ, min + 1), SyscallSucceeds()); + int med = Size(); + EXPECT_GT(med, min); // Should have grown, may be rounded. + + // Set from the write end. + ASSERT_THAT(fcntl(wfd.get(), F_SETPIPE_SZ, med + 1), SyscallSucceeds()); + int max = Size(); + EXPECT_GT(max, med); // Ditto. +} - int rflags; - EXPECT_THAT(rflags = fcntl(rfd.get(), F_GETFL), SyscallSucceeds()); +TEST_P(PipeTest, SizeChangeMax) { + SKIP_IF(!CreateBlocking()); - // The kernel did *not* set O_LARGEFILE. - EXPECT_EQ(rflags, 0); + // Assert there's some maximum. + EXPECT_THAT(fcntl(rfd.get(), F_SETPIPE_SZ, 0x7fffffffffffffff), + SyscallFailsWithErrno(EINVAL)); + EXPECT_THAT(fcntl(wfd.get(), F_SETPIPE_SZ, 0x7fffffffffffffff), + SyscallFailsWithErrno(EINVAL)); } -// Test that accesses of /proc//fd/ and /proc//fdinfo/ -// correctly decrement the refcount of that file descriptor. -TEST_F(PipeTest, ProcFDReleasesFile) { - std::vector paths = {"/proc/self/fd/", "/proc/self/fdinfo/"}; - for (const std::string& path : paths) { - int fds[2]; - ASSERT_THAT(pipe(fds), SyscallSucceeds()); - FileDescriptor rfd(fds[0]); - FileDescriptor wfd(fds[1]); - - // Stat the pipe FD, which shouldn't alter the refcount of the write end of - // the pipe. - struct stat wst; - ASSERT_THAT(lstat(absl::StrCat(path.c_str(), wfd.get()).c_str(), &wst), - SyscallSucceeds()); - // Close the write end of the pipe and ensure that read indicates EOF. - wfd.reset(); - char buf; - ASSERT_THAT(read(rfd.get(), &buf, 1), SyscallSucceedsWithValue(0)); +TEST_P(PipeTest, SizeChangeFull) { + SKIP_IF(!CreateBlocking()); + + // Ensure that we adjust to a large enough size to avoid rounding when we + // perform the size decrease. If rounding occurs, we may not actually + // adjust the size and the call below will return success. It was found via + // experimentation that this granularity avoids the rounding for Linux. + constexpr int kDelta = 64 * 1024; + ASSERT_THAT(fcntl(wfd.get(), F_SETPIPE_SZ, Size() + kDelta), + SyscallSucceeds()); + + // Fill the buffer and try to change down. + std::vector buf(Size()); + ASSERT_THAT(write(wfd.get(), buf.data(), buf.size()), + SyscallSucceedsWithValue(buf.size())); + EXPECT_THAT(fcntl(wfd.get(), F_SETPIPE_SZ, Size() - kDelta), + SyscallFailsWithErrno(EBUSY)); +} + +TEST_P(PipeTest, Streaming) { + SKIP_IF(!CreateBlocking()); + + // We make too many calls to go through full save cycles. + DisableSave ds; + + absl::Notification notify; + ScopedThread t([this, ¬ify]() { + // Don't start until it's full. + notify.WaitForNotification(); + for (int i = 0; i < 2 * Size(); i++) { + int rbuf; + ASSERT_THAT(read(rfd.get(), &rbuf, sizeof(rbuf)), + SyscallSucceedsWithValue(sizeof(rbuf))); + EXPECT_EQ(rbuf, i); + } + }); + for (int i = 0; i < 2 * Size(); i++) { + int wbuf = i; + ASSERT_THAT(write(wfd.get(), &wbuf, sizeof(wbuf)), + SyscallSucceedsWithValue(sizeof(wbuf))); + // Did that write just fill up the buffer? Wake up the reader. Once only. + if ((i * sizeof(wbuf)) < Size() && ((i + 1) * sizeof(wbuf)) >= Size()) { + notify.Notify(); + } } } +std::string PipeCreatorName(::testing::TestParamInfo info) { + return info.param.name_; // Use the name specified. +} + +INSTANTIATE_TEST_SUITE_P( + Pipes, PipeTest, + ::testing::Values( + PipeCreator{ + "pipe", + [](int fds[2], bool* is_blocking, bool* is_namedpipe) { + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + *is_blocking = true; + *is_namedpipe = false; + }, + }, + PipeCreator{ + "pipe2blocking", + [](int fds[2], bool* is_blocking, bool* is_namedpipe) { + ASSERT_THAT(pipe2(fds, 0), SyscallSucceeds()); + *is_blocking = true; + *is_namedpipe = false; + }, + }, + PipeCreator{ + "pipe2nonblocking", + [](int fds[2], bool* is_blocking, bool* is_namedpipe) { + ASSERT_THAT(pipe2(fds, O_NONBLOCK), SyscallSucceeds()); + *is_blocking = false; + *is_namedpipe = false; + }, + }, + PipeCreator{ + "smallbuffer", + [](int fds[2], bool* is_blocking, bool* is_namedpipe) { + // Set to the minimum available size (will round up). + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + ASSERT_THAT(fcntl(fds[0], F_SETPIPE_SZ, 0), SyscallSucceeds()); + *is_blocking = true; + *is_namedpipe = false; + }, + }, + PipeCreator{ + "namednonblocking", + [](int fds[2], bool* is_blocking, bool* is_namedpipe) { + // Create a new file-based pipe (non-blocking). + auto file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + ASSERT_THAT(unlink(file.path().c_str()), SyscallSucceeds()); + SKIP_IF(mkfifo(file.path().c_str(), 0644) != 0); + fds[0] = open(file.path().c_str(), O_NONBLOCK | O_RDONLY); + fds[1] = open(file.path().c_str(), O_NONBLOCK | O_WRONLY); + MaybeSave(); + *is_blocking = false; + *is_namedpipe = true; + }, + }, + PipeCreator{ + "namedblocking", + [](int fds[2], bool* is_blocking, bool* is_namedpipe) { + // Create a new file-based pipe (blocking). + auto file = ASSERT_NO_ERRNO_AND_VALUE(TempPath::CreateFile()); + ASSERT_THAT(unlink(file.path().c_str()), SyscallSucceeds()); + SKIP_IF(mkfifo(file.path().c_str(), 0644) != 0); + ScopedThread t([&file, &fds]() { + fds[1] = open(file.path().c_str(), O_WRONLY); + }); + fds[0] = open(file.path().c_str(), O_RDONLY); + t.Join(); + MaybeSave(); + *is_blocking = true; + *is_namedpipe = true; + }, + }), + PipeCreatorName); + } // namespace } // namespace testing } // namespace gvisor -- cgit v1.2.3 From 711290a7f6c434ddbfe401e46002afd30df26aa5 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Wed, 22 May 2019 15:53:13 -0700 Subject: Add support for wait(WNOTHREAD) PiperOrigin-RevId: 249537694 Change-Id: Iaa4bca73a2d8341e03064d59a2eb490afc3f80da --- pkg/sentry/kernel/task_exit.go | 164 ++++++++++++++++------------ pkg/sentry/syscalls/linux/sys_thread.go | 10 +- test/syscalls/linux/BUILD | 2 + test/syscalls/linux/wait.cc | 185 ++++++++++++++++++++++---------- 4 files changed, 236 insertions(+), 125 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index 6e9701b01..2e1e46582 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -782,6 +782,10 @@ type WaitOptions struct { // for. CloneTasks bool + // If SiblingChildren is true, events from children tasks of any task + // in the thread group of the waiter are eligible to be waited for. + SiblingChildren bool + // Events is a bitwise combination of the events defined above that specify // what events are of interest to the call to Wait. Events waiter.EventMask @@ -869,87 +873,109 @@ func (t *Task) waitOnce(opts *WaitOptions) (*WaitResult, error) { t.tg.pidns.owner.mu.Lock() defer t.tg.pidns.owner.mu.Unlock() - // Without the (unimplemented) __WNOTHREAD flag, a task can wait on the - // children and tracees of any task in the same thread group. - for parent := t.tg.tasks.Front(); parent != nil; parent = parent.Next() { - for child := range parent.children { - if !opts.matchesTask(child, parent.tg.pidns) { - continue - } - // Non-leaders don't notify parents on exit and aren't eligible to - // be waited on. - if opts.Events&EventExit != 0 && child == child.tg.leader && !child.exitParentAcked { - anyWaitableTasks = true - if wr := t.waitCollectZombieLocked(child, opts, false); wr != nil { - return wr, nil - } - } - // Check for group stops and continues. Tasks that have passed - // TaskExitInitiated can no longer participate in group stops. - if opts.Events&(EventChildGroupStop|EventGroupContinue) == 0 { - continue - } - if child.exitState >= TaskExitInitiated { - continue - } - // If the waiter is in the same thread group as the task's - // tracer, do not report its group stops; they will be reported - // as ptrace stops instead. This also skips checking for group - // continues, but they'll be checked for when scanning tracees - // below. (Per kernel/exit.c:wait_consider_task(): "If a - // ptracer wants to distinguish the two events for its own - // children, it should create a separate process which takes - // the role of real parent.") - if tracer := child.Tracer(); tracer != nil && tracer.tg == parent.tg { - continue + if opts.SiblingChildren { + // We can wait on the children and tracees of any task in the + // same thread group. + for parent := t.tg.tasks.Front(); parent != nil; parent = parent.Next() { + wr, any := t.waitParentLocked(opts, parent) + if wr != nil { + return wr, nil } + anyWaitableTasks = anyWaitableTasks || any + } + } else { + // We can only wait on this task. + var wr *WaitResult + wr, anyWaitableTasks = t.waitParentLocked(opts, t) + if wr != nil { + return wr, nil + } + } + + if anyWaitableTasks { + return nil, ErrNoWaitableEvent + } + return nil, syserror.ECHILD +} + +// Preconditions: The TaskSet mutex must be locked for writing. +func (t *Task) waitParentLocked(opts *WaitOptions, parent *Task) (*WaitResult, bool) { + anyWaitableTasks := false + + for child := range parent.children { + if !opts.matchesTask(child, parent.tg.pidns) { + continue + } + // Non-leaders don't notify parents on exit and aren't eligible to + // be waited on. + if opts.Events&EventExit != 0 && child == child.tg.leader && !child.exitParentAcked { anyWaitableTasks = true - if opts.Events&EventChildGroupStop != 0 { - if wr := t.waitCollectChildGroupStopLocked(child, opts); wr != nil { - return wr, nil - } - } - if opts.Events&EventGroupContinue != 0 { - if wr := t.waitCollectGroupContinueLocked(child, opts); wr != nil { - return wr, nil - } + if wr := t.waitCollectZombieLocked(child, opts, false); wr != nil { + return wr, anyWaitableTasks } } - for tracee := range parent.ptraceTracees { - if !opts.matchesTask(tracee, parent.tg.pidns) { - continue - } - // Non-leaders do notify tracers on exit. - if opts.Events&EventExit != 0 && !tracee.exitTracerAcked { - anyWaitableTasks = true - if wr := t.waitCollectZombieLocked(tracee, opts, true); wr != nil { - return wr, nil - } - } - if opts.Events&(EventTraceeStop|EventGroupContinue) == 0 { - continue + // Check for group stops and continues. Tasks that have passed + // TaskExitInitiated can no longer participate in group stops. + if opts.Events&(EventChildGroupStop|EventGroupContinue) == 0 { + continue + } + if child.exitState >= TaskExitInitiated { + continue + } + // If the waiter is in the same thread group as the task's + // tracer, do not report its group stops; they will be reported + // as ptrace stops instead. This also skips checking for group + // continues, but they'll be checked for when scanning tracees + // below. (Per kernel/exit.c:wait_consider_task(): "If a + // ptracer wants to distinguish the two events for its own + // children, it should create a separate process which takes + // the role of real parent.") + if tracer := child.Tracer(); tracer != nil && tracer.tg == parent.tg { + continue + } + anyWaitableTasks = true + if opts.Events&EventChildGroupStop != 0 { + if wr := t.waitCollectChildGroupStopLocked(child, opts); wr != nil { + return wr, anyWaitableTasks } - if tracee.exitState >= TaskExitInitiated { - continue + } + if opts.Events&EventGroupContinue != 0 { + if wr := t.waitCollectGroupContinueLocked(child, opts); wr != nil { + return wr, anyWaitableTasks } + } + } + for tracee := range parent.ptraceTracees { + if !opts.matchesTask(tracee, parent.tg.pidns) { + continue + } + // Non-leaders do notify tracers on exit. + if opts.Events&EventExit != 0 && !tracee.exitTracerAcked { anyWaitableTasks = true - if opts.Events&EventTraceeStop != 0 { - if wr := t.waitCollectTraceeStopLocked(tracee, opts); wr != nil { - return wr, nil - } + if wr := t.waitCollectZombieLocked(tracee, opts, true); wr != nil { + return wr, anyWaitableTasks } - if opts.Events&EventGroupContinue != 0 { - if wr := t.waitCollectGroupContinueLocked(tracee, opts); wr != nil { - return wr, nil - } + } + if opts.Events&(EventTraceeStop|EventGroupContinue) == 0 { + continue + } + if tracee.exitState >= TaskExitInitiated { + continue + } + anyWaitableTasks = true + if opts.Events&EventTraceeStop != 0 { + if wr := t.waitCollectTraceeStopLocked(tracee, opts); wr != nil { + return wr, anyWaitableTasks + } + } + if opts.Events&EventGroupContinue != 0 { + if wr := t.waitCollectGroupContinueLocked(tracee, opts); wr != nil { + return wr, anyWaitableTasks } } } - if anyWaitableTasks { - return nil, ErrNoWaitableEvent - } - return nil, syserror.ECHILD + return nil, anyWaitableTasks } // Preconditions: The TaskSet mutex must be locked for writing. diff --git a/pkg/sentry/syscalls/linux/sys_thread.go b/pkg/sentry/syscalls/linux/sys_thread.go index cc441460c..14fa7ef92 100644 --- a/pkg/sentry/syscalls/linux/sys_thread.go +++ b/pkg/sentry/syscalls/linux/sys_thread.go @@ -183,7 +183,7 @@ func Vfork(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscall // wait4 waits for the given child process to exit. func wait4(t *kernel.Task, pid int, statusAddr usermem.Addr, options int, rusageAddr usermem.Addr) (uintptr, error) { - if options&^(linux.WNOHANG|linux.WUNTRACED|linux.WCONTINUED|linux.WALL|linux.WCLONE) != 0 { + if options&^(linux.WNOHANG|linux.WUNTRACED|linux.WCONTINUED|linux.WNOTHREAD|linux.WALL|linux.WCLONE) != 0 { return 0, syscall.EINVAL } wopts := kernel.WaitOptions{ @@ -227,6 +227,9 @@ func wait4(t *kernel.Task, pid int, statusAddr usermem.Addr, options int, rusage if options&linux.WNOHANG == 0 { wopts.BlockInterruptErr = kernel.ERESTARTSYS } + if options&linux.WNOTHREAD == 0 { + wopts.SiblingChildren = true + } wr, err := t.Wait(&wopts) if err != nil { @@ -278,7 +281,7 @@ func Waitid(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal options := int(args[3].Uint()) rusageAddr := args[4].Pointer() - if options&^(linux.WNOHANG|linux.WEXITED|linux.WSTOPPED|linux.WCONTINUED|linux.WNOWAIT) != 0 { + if options&^(linux.WNOHANG|linux.WEXITED|linux.WSTOPPED|linux.WCONTINUED|linux.WNOWAIT|linux.WNOTHREAD) != 0 { return 0, nil, syscall.EINVAL } if options&(linux.WEXITED|linux.WSTOPPED|linux.WCONTINUED) == 0 { @@ -310,6 +313,9 @@ func Waitid(t *kernel.Task, args arch.SyscallArguments) (uintptr, *kernel.Syscal if options&linux.WNOHANG == 0 { wopts.BlockInterruptErr = kernel.ERESTARTSYS } + if options&linux.WNOTHREAD == 0 { + wopts.SiblingChildren = true + } wr, err := t.Wait(&wopts) if err != nil { diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index e8caf31fc..014679ec5 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -3179,7 +3179,9 @@ cc_binary( "//test/util:signal_util", "//test/util:test_main", "//test/util:test_util", + "//test/util:thread_util", "@com_google_absl//absl/strings", + "@com_google_absl//absl/synchronization", "@com_google_absl//absl/time", "@com_google_googletest//:gtest", ], diff --git a/test/syscalls/linux/wait.cc b/test/syscalls/linux/wait.cc index 50d0725a7..da3b97828 100644 --- a/test/syscalls/linux/wait.cc +++ b/test/syscalls/linux/wait.cc @@ -21,11 +21,13 @@ #include #include +#include #include #include "gmock/gmock.h" #include "gtest/gtest.h" #include "absl/strings/str_cat.h" +#include "absl/synchronization/mutex.h" #include "absl/time/clock.h" #include "absl/time/time.h" #include "test/util/cleanup.h" @@ -34,6 +36,7 @@ #include "test/util/posix_error.h" #include "test/util/signal_util.h" #include "test/util/test_util.h" +#include "test/util/thread_util.h" using ::testing::UnorderedElementsAre; @@ -42,10 +45,8 @@ using ::testing::UnorderedElementsAre; // // NOTE(b/22640830,b/27680907,b/29049891): Some functionality is not tested as // it is not currently supported by gVisor: -// * UID in waitid(2) siginfo. // * Process groups. // * Core dump status (WCOREDUMP). -// * Linux only option __WNOTHREAD. // // Tests for waiting on stopped/continued children are in sigstop.cc. @@ -357,13 +358,22 @@ INSTANTIATE_TEST_SUITE_P( return static_cast(si.si_pid); })); -// Fixture for tests parameterized by a function that takes the PID of a -// specific child to wait for, waits for it to exit, and checks that it exits -// with the given code. +// Fixture for tests parameterized by a (sysno, function) tuple. The function +// takes the PID of a specific child to wait for, waits for it to exit, and +// checks that it exits with the given code. class WaitSpecificChildTest - : public ::testing::TestWithParam> { + : public ::testing::TestWithParam< + std::tuple>> { protected: - PosixError WaitFor(pid_t pid, int code) { return GetParam()(pid, code); } + int Sysno() { return std::get<0>(GetParam()); } + + PosixError WaitForWithOptions(pid_t pid, int options, int code) { + return std::get<1>(GetParam())(pid, options, code); + } + + PosixError WaitFor(pid_t pid, int code) { + return std::get<1>(GetParam())(pid, 0, code); + } }; // Wait for specific child to exit. @@ -432,6 +442,75 @@ TEST_P(WaitSpecificChildTest, AfterExit) { EXPECT_NO_ERRNO(WaitFor(child, 0)); } +// Wait for child of sibling thread. +TEST_P(WaitSpecificChildTest, SiblingChildren) { + absl::Mutex mu; + pid_t child; + bool ready = false; + bool stop = false; + + ScopedThread t([&] { + absl::MutexLock ml(&mu); + EXPECT_THAT(child = ForkAndExit(0, 0), SyscallSucceeds()); + ready = true; + mu.Await(absl::Condition(&stop)); + }); + + // N.B. This must be declared after ScopedThread, so it is destructed first, + // thus waking the thread. + absl::MutexLock ml(&mu); + mu.Await(absl::Condition(&ready)); + + EXPECT_NO_ERRNO(WaitFor(child, 0)); + + // Keep the sibling alive until after we've waited so the child isn't + // reparented. + stop = true; +} + +// Waiting for child of sibling thread not allowed with WNOTHREAD. +TEST_P(WaitSpecificChildTest, SiblingChildrenWNOTHREAD) { + // Linux added WNOTHREAD support to waitid(2) in + // 91c4e8ea8f05916df0c8a6f383508ac7c9e10dba ("wait: allow sys_waitid() to + // accept __WNOTHREAD/__WCLONE/__WALL"). i.e., Linux 4.7. + // + // Skip the test if it isn't supported yet. + if (Sysno() == SYS_waitid) { + int ret = waitid(P_ALL, 0, nullptr, WEXITED | WNOHANG | __WNOTHREAD); + SKIP_IF(ret < 0 && errno == EINVAL); + } + + absl::Mutex mu; + pid_t child; + bool ready = false; + bool stop = false; + + ScopedThread t([&] { + absl::MutexLock ml(&mu); + EXPECT_THAT(child = ForkAndExit(0, 0), SyscallSucceeds()); + ready = true; + mu.Await(absl::Condition(&stop)); + + // This thread can wait on child. + EXPECT_NO_ERRNO(WaitForWithOptions(child, __WNOTHREAD, 0)); + }); + + // N.B. This must be declared after ScopedThread, so it is destructed first, + // thus waking the thread. + absl::MutexLock ml(&mu); + mu.Await(absl::Condition(&ready)); + + // This thread can't wait on child. + EXPECT_THAT( + WaitForWithOptions(child, __WNOTHREAD, 0), + PosixErrorIs(ECHILD, ::testing::AnyOf(::testing::StrEq("waitid"), + ::testing::StrEq("wait4")))); + + // Keep the sibling alive until after we've waited so the child isn't + // reparented. + stop = true; +} + // Wait for specific child to exit. // A non-CLONE_THREAD child which sends SIGCHLD upon exit behaves much like // a forked process. @@ -551,55 +630,53 @@ TEST_P(WaitSpecificChildTest, AfterChildExecve) { EXPECT_NO_ERRNO(WaitFor(child, 0)); } -INSTANTIATE_TEST_SUITE_P( - Waiters, WaitSpecificChildTest, - ::testing::Values( - [](pid_t pid, int code) -> PosixError { - int status; - auto const rv = Wait4(pid, &status, 0, nullptr); - MaybeSave(); - if (rv < 0) { - return PosixError(errno, "wait4"); - } else if (rv != pid) { - return PosixError(EINVAL, absl::StrCat("unexpected pid: got ", rv, - ", wanted ", pid)); - } - if (!WIFEXITED(status) || WEXITSTATUS(status) != code) { - return PosixError( - EINVAL, absl::StrCat("unexpected wait status: got ", status, - ", wanted ", code)); - } - return NoError(); - }, - [](pid_t pid, int code) -> PosixError { - siginfo_t si; - auto const rv = Waitid(P_PID, pid, &si, WEXITED); - MaybeSave(); - if (rv < 0) { - return PosixError(errno, "waitid"); - } - if (si.si_pid != pid) { - return PosixError(EINVAL, - absl::StrCat("unexpected pid: got ", si.si_pid, +PosixError CheckWait4(pid_t pid, int options, int code) { + int status; + auto const rv = Wait4(pid, &status, options, nullptr); + MaybeSave(); + if (rv < 0) { + return PosixError(errno, "wait4"); + } else if (rv != pid) { + return PosixError( + EINVAL, absl::StrCat("unexpected pid: got ", rv, ", wanted ", pid)); + } + if (!WIFEXITED(status) || WEXITSTATUS(status) != code) { + return PosixError(EINVAL, absl::StrCat("unexpected wait status: got ", + status, ", wanted ", code)); + } + return NoError(); +}; + +PosixError CheckWaitid(pid_t pid, int options, int code) { + siginfo_t si; + auto const rv = Waitid(P_PID, pid, &si, options | WEXITED); + MaybeSave(); + if (rv < 0) { + return PosixError(errno, "waitid"); + } + if (si.si_pid != pid) { + return PosixError(EINVAL, absl::StrCat("unexpected pid: got ", si.si_pid, ", wanted ", pid)); - } - if (si.si_signo != SIGCHLD) { - return PosixError( - EINVAL, absl::StrCat("unexpected signo: got ", si.si_signo, - ", wanted ", SIGCHLD)); - } - if (si.si_status != code) { - return PosixError( - EINVAL, absl::StrCat("unexpected status: got ", si.si_status, - ", wanted ", code)); - } - if (si.si_code != CLD_EXITED) { - return PosixError(EINVAL, - absl::StrCat("unexpected code: got ", si.si_code, + } + if (si.si_signo != SIGCHLD) { + return PosixError(EINVAL, absl::StrCat("unexpected signo: got ", + si.si_signo, ", wanted ", SIGCHLD)); + } + if (si.si_status != code) { + return PosixError(EINVAL, absl::StrCat("unexpected status: got ", + si.si_status, ", wanted ", code)); + } + if (si.si_code != CLD_EXITED) { + return PosixError(EINVAL, absl::StrCat("unexpected code: got ", si.si_code, ", wanted ", CLD_EXITED)); - } - return NoError(); - })); + } + return NoError(); +} + +INSTANTIATE_TEST_SUITE_P( + Waiters, WaitSpecificChildTest, + ::testing::Values(std::make_tuple(SYS_wait4, CheckWait4), + std::make_tuple(SYS_waitid, CheckWaitid))); // WIFEXITED, WIFSIGNALED, WTERMSIG indicate signal exit. TEST(WaitTest, SignalExit) { -- cgit v1.2.3 From 79738d3958a027bcf449cf1bd608f3adec42b72c Mon Sep 17 00:00:00 2001 From: Adin Scannell Date: Wed, 22 May 2019 18:18:01 -0700 Subject: Log unhandled faults only at DEBUG level. PiperOrigin-RevId: 249561399 Change-Id: Ic73c68c8538bdca53068f38f82b7260939addac2 --- pkg/sentry/kernel/task_run.go | 10 ++-------- 1 file changed, 2 insertions(+), 8 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_run.go b/pkg/sentry/kernel/task_run.go index 4549b437e..a79101a18 100644 --- a/pkg/sentry/kernel/task_run.go +++ b/pkg/sentry/kernel/task_run.go @@ -26,7 +26,6 @@ import ( "gvisor.googlesource.com/gvisor/pkg/sentry/memmap" "gvisor.googlesource.com/gvisor/pkg/sentry/platform" "gvisor.googlesource.com/gvisor/pkg/sentry/usermem" - "gvisor.googlesource.com/gvisor/pkg/syserror" ) // A taskRunState is a reified state in the task state machine. See README.md @@ -267,13 +266,8 @@ func (*runApp) execute(t *Task) taskRunState { } } - // The JVM will trigger these errors constantly, so don't - // spam logs with this error. - if err == syserror.EFAULT || err == syserror.EPERM { - t.Debugf("Unhandled user fault: addr=%x ip=%x access=%v err=%v", addr, t.Arch().IP(), at, err) - } else { - t.Warningf("Unhandled user fault: addr=%x ip=%x access=%v err=%v", addr, t.Arch().IP(), at, err) - } + // Faults are common, log only at debug level. + t.Debugf("Unhandled user fault: addr=%x ip=%x access=%v err=%v", addr, t.Arch().IP(), at, err) t.DebugDumpState() // Continue to signal handling. -- cgit v1.2.3 From a949133c4b22a87c79310b2d825f2899028d6088 Mon Sep 17 00:00:00 2001 From: Andrei Vagin Date: Thu, 23 May 2019 23:20:11 -0700 Subject: gvisor: interrupt the sendfile system call if a task has been interrupted sendfile can be called for a big range and it can require significant amount of time to process it, so we need to handle task interrupts in this system call. PiperOrigin-RevId: 249781023 Change-Id: Ifc2ec505d74c06f5ee76f93b8d30d518ec2d4015 --- pkg/amutex/amutex.go | 6 ++++++ pkg/amutex/amutex_test.go | 4 ++++ pkg/sentry/fs/file.go | 6 ++++++ pkg/sentry/kernel/pipe/node_test.go | 4 ++++ pkg/sentry/kernel/task_block.go | 5 +++++ 5 files changed, 25 insertions(+) (limited to 'pkg/sentry/kernel') diff --git a/pkg/amutex/amutex.go b/pkg/amutex/amutex.go index 85e819304..4f7759b87 100644 --- a/pkg/amutex/amutex.go +++ b/pkg/amutex/amutex.go @@ -33,6 +33,9 @@ type Sleeper interface { // SleepFinish is called by AbortableMutex.Lock() once a contended mutex // is acquired or the wait is aborted. SleepFinish(success bool) + + // Interrupted returns true if the wait is aborted. + Interrupted() bool } // NoopSleeper is a stateless no-op implementation of Sleeper for anonymous @@ -47,6 +50,9 @@ func (NoopSleeper) SleepStart() <-chan struct{} { // SleepFinish implements Sleeper.SleepFinish. func (NoopSleeper) SleepFinish(success bool) {} +// Interrupted implements Sleeper.Interrupted. +func (NoopSleeper) Interrupted() bool { return false } + // AbortableMutex is an abortable mutex. It allows Lock() to be aborted while it // waits to acquire the mutex. type AbortableMutex struct { diff --git a/pkg/amutex/amutex_test.go b/pkg/amutex/amutex_test.go index 6a0af006e..211bdda4b 100644 --- a/pkg/amutex/amutex_test.go +++ b/pkg/amutex/amutex_test.go @@ -31,6 +31,10 @@ func (s *sleeper) SleepStart() <-chan struct{} { func (*sleeper) SleepFinish(bool) { } +func (s *sleeper) Interrupted() bool { + return len(s.ch) != 0 +} + func TestMutualExclusion(t *testing.T) { var m AbortableMutex m.Init() diff --git a/pkg/sentry/fs/file.go b/pkg/sentry/fs/file.go index 8f1baca23..8c1307235 100644 --- a/pkg/sentry/fs/file.go +++ b/pkg/sentry/fs/file.go @@ -516,12 +516,18 @@ type lockedReader struct { // Read implements io.Reader.Read. func (r *lockedReader) Read(buf []byte) (int, error) { + if r.Ctx.Interrupted() { + return 0, syserror.ErrInterrupted + } n, err := r.File.FileOperations.Read(r.Ctx, r.File, usermem.BytesIOSequence(buf), r.File.offset) return int(n), err } // ReadAt implements io.Reader.ReadAt. func (r *lockedReader) ReadAt(buf []byte, offset int64) (int, error) { + if r.Ctx.Interrupted() { + return 0, syserror.ErrInterrupted + } n, err := r.File.FileOperations.Read(r.Ctx, r.File, usermem.BytesIOSequence(buf), offset) return int(n), err } diff --git a/pkg/sentry/kernel/pipe/node_test.go b/pkg/sentry/kernel/pipe/node_test.go index 7ddecdad8..31d9b0443 100644 --- a/pkg/sentry/kernel/pipe/node_test.go +++ b/pkg/sentry/kernel/pipe/node_test.go @@ -48,6 +48,10 @@ func (s *sleeper) Cancel() { s.ch <- struct{}{} } +func (s *sleeper) Interrupted() bool { + return len(s.ch) != 0 +} + type openResult struct { *fs.File error diff --git a/pkg/sentry/kernel/task_block.go b/pkg/sentry/kernel/task_block.go index 30a7f6b1e..1c76c4d84 100644 --- a/pkg/sentry/kernel/task_block.go +++ b/pkg/sentry/kernel/task_block.go @@ -158,6 +158,11 @@ func (t *Task) SleepFinish(success bool) { t.Activate() } +// Interrupted implements amutex.Sleeper.Interrupted +func (t *Task) Interrupted() bool { + return len(t.interruptChan) != 0 +} + // UninterruptibleSleepStart implements context.Context.UninterruptibleSleepStart. func (t *Task) UninterruptibleSleepStart(deactivate bool) { if deactivate { -- cgit v1.2.3 From 507a15dce974d0cff18253ba50af29d6579bacc5 Mon Sep 17 00:00:00 2001 From: Michael Pratt Date: Tue, 28 May 2019 18:02:07 -0700 Subject: Always wait on tracee children After bf959931ddb88c4e4366e96dd22e68fa0db9527c ("wait/ptrace: assume __WALL if the child is traced") (Linux 4.7), tracees are always eligible for waiting, regardless of type. PiperOrigin-RevId: 250399527 --- pkg/sentry/kernel/task_exit.go | 10 ++++++--- test/syscalls/linux/BUILD | 1 + test/syscalls/linux/wait.cc | 46 ++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 54 insertions(+), 3 deletions(-) (limited to 'pkg/sentry/kernel') diff --git a/pkg/sentry/kernel/task_exit.go b/pkg/sentry/kernel/task_exit.go index 2e1e46582..158e665d3 100644 --- a/pkg/sentry/kernel/task_exit.go +++ b/pkg/sentry/kernel/task_exit.go @@ -803,13 +803,17 @@ type WaitOptions struct { } // Preconditions: The TaskSet mutex must be locked (for reading or writing). -func (o *WaitOptions) matchesTask(t *Task, pidns *PIDNamespace) bool { +func (o *WaitOptions) matchesTask(t *Task, pidns *PIDNamespace, tracee bool) bool { if o.SpecificTID != 0 && o.SpecificTID != pidns.tids[t] { return false } if o.SpecificPGID != 0 && o.SpecificPGID != pidns.pgids[t.tg.processGroup] { return false } + // Tracees are always eligible. + if tracee { + return true + } if t == t.tg.leader && t.tg.terminationSignal == linux.SIGCHLD { return o.NonCloneTasks } @@ -903,7 +907,7 @@ func (t *Task) waitParentLocked(opts *WaitOptions, parent *Task) (*WaitResult, b anyWaitableTasks := false for child := range parent.children { - if !opts.matchesTask(child, parent.tg.pidns) { + if !opts.matchesTask(child, parent.tg.pidns, false) { continue } // Non-leaders don't notify parents on exit and aren't eligible to @@ -946,7 +950,7 @@ func (t *Task) waitParentLocked(opts *WaitOptions, parent *Task) (*WaitResult, b } } for tracee := range parent.ptraceTracees { - if !opts.matchesTask(tracee, parent.tg.pidns) { + if !opts.matchesTask(tracee, parent.tg.pidns, true) { continue } // Non-leaders do notify tracers on exit. diff --git a/test/syscalls/linux/BUILD b/test/syscalls/linux/BUILD index 750f3a1e2..ec57ec129 100644 --- a/test/syscalls/linux/BUILD +++ b/test/syscalls/linux/BUILD @@ -3179,6 +3179,7 @@ cc_binary( linkstatic = 1, deps = [ "//test/util:cleanup", + "//test/util:file_descriptor", "//test/util:logging", "//test/util:multiprocess_util", "//test/util:posix_error", diff --git a/test/syscalls/linux/wait.cc b/test/syscalls/linux/wait.cc index aa27194cb..944149d5e 100644 --- a/test/syscalls/linux/wait.cc +++ b/test/syscalls/linux/wait.cc @@ -14,6 +14,7 @@ #include #include +#include #include #include #include @@ -31,6 +32,7 @@ #include "absl/time/clock.h" #include "absl/time/time.h" #include "test/util/cleanup.h" +#include "test/util/file_descriptor.h" #include "test/util/logging.h" #include "test/util/multiprocess_util.h" #include "test/util/posix_error.h" @@ -861,6 +863,50 @@ TEST(WaitTest, WaitidRusage) { EXPECT_GE(RusageCpuTime(rusage), kSpin); } +// After bf959931ddb88c4e4366e96dd22e68fa0db9527c ("wait/ptrace: assume __WALL +// if the child is traced") (Linux 4.7), tracees are always eligible for +// waiting, regardless of type. +TEST(WaitTest, TraceeWALL) { + int fds[2]; + ASSERT_THAT(pipe(fds), SyscallSucceeds()); + FileDescriptor rfd(fds[0]); + FileDescriptor wfd(fds[1]); + + pid_t child = fork(); + if (child == 0) { + // Child. + rfd.reset(); + + TEST_PCHECK(ptrace(PTRACE_TRACEME, 0, nullptr, nullptr) == 0); + + // Notify parent that we're now a tracee. + wfd.reset(); + + _exit(0); + } + ASSERT_THAT(child, SyscallSucceeds()); + + wfd.reset(); + + // Wait for child to become tracee. + char c; + EXPECT_THAT(ReadFd(rfd.get(), &c, sizeof(c)), SyscallSucceedsWithValue(0)); + + // We can wait on the fork child with WCLONE, as it is a tracee. + int status; + if (IsRunningOnGvisor()) { + ASSERT_THAT(Wait4(child, &status, __WCLONE, nullptr), + SyscallSucceedsWithValue(child)); + + EXPECT_TRUE(WIFEXITED(status) && WEXITSTATUS(status) == 0) << status; + } else { + // On older versions of Linux, we may get ECHILD. + ASSERT_THAT(Wait4(child, &status, __WCLONE, nullptr), + ::testing::AnyOf(SyscallSucceedsWithValue(child), + SyscallFailsWithErrno(ECHILD))); + } +} + } // namespace } // namespace testing -- cgit v1.2.3